Diss ETH No. 17650
Impact of fire, large herbivores and
N2-fixation on nutrient cycling in
humid savanna, Tanzania
Patrick Cech
2008
Diss ETH No. 17650
Impact of fire, large herbivores and N2-fixation on
nutrient cycling in humid savanna, Tanzania
A dissertation submitted to the
SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH
For the degree of
DOCTOR OF SCIENCEs
presented by
PATRICK GEORGES CECH
MSc Biology I, University of Basel
born April 28th, 1976
citizen of Basel, BS
accepted on the recommendation of
Prof. Peter J. Edwards, examiner
Dr. Harry Olde Venterink, co-examiner
Dr. Michael Scherer-Lorenzen, co-examiner
2008
Ninatabaruku tasinifu hii kwa mke wangu Patrizia na mwanangu Luisa.
Contents
Contents
Summary
1
Zusammenfassung
3
General introduction
5
Chapter 1
Effects of herbivory, fire and N2-fixation on nutrient
limitation in a humid African savanna
17
Chapter 2
Effects of large herbivores, fire, and N2-fixation on
N and P cycling in humid savanna
43
Chapter 3
Why are herbaceous legumes scarce in savanna?
A grass-legume competition experiment
79
Chapter 4
Effects of clipping and fertilization on productivity
and species composition of abandoned cattle grounds
in humid savanna, Tanzania
99
Acknowledgements
123
Curriculum Vitae
125
Summary
Summary
African savannas contain the world’s largest diversity of wild ungulates, and also constitute
an important resource for a growing human population. These fragile ecosystems are at
risk of being overexploited for livestock production, and perhaps in the future for biofuel
production, leading to a loss of ecosystem services and biodiversity. This thesis aimed at
providing a more profound understanding of nitrogen and phosphorus cycling in savannas
by quantifying and testing the impact of large herbivores, fire and N2-fixation upon shortterm and long-term availabilities of these nutrients. The results of this research can be used
to evaluate the consequences of different types of land-use, fire management and nutrient
enrichment in savannas. The study area was the coastal savanna of Saadani National Park,
which includes an abandoned cattle ranch and a former game reserve with high densities
of wild herbivores.
The effects of fire, large herbivores and N2-fixation on the type of nutrient limitation
and N pools in savanna ecosystems were to a large extent predictable from theoretical
considerations. The grass productivity at three kinds of site – frequently burnt, intensively
grazed by wild herbivores, and open Acacia woodland – were all co-limited by N and P,
indicating a very low availability of both nutrients. An area that previously received large
amounts of excreta from livestock was limited by N. Nitrogen pools in soil were low at sites
where fires were frequent, and higher in Acacia woodland and sites with high herbivore
excretion.
Cattle ranching has led to a spatial re-distribution of large quantities of nutrients, with local
accumulation being stronger and more persistent for P than for N. At the scale of the former
cattle ranch, local depletion of nutrients by cattle grazing may have been compensated
for by atmospheric inputs. Lawn patches intensively grazed by wild herbivores tended
to have elevated availabilities of N and P, but because rates of nutrient removal through
consumption were higher than those of return in excreta, the nutrient balances were in
both cases negative. N2-fixation by dense stands of Acacia trees increased N availability
and caused a net annual N input, whereas the N contribution of herbaceous legumes was
negligible. Fire was the major cause for nutrient losses from tallgrass savanna. N inputs
from the atmosphere and symbiotic N2-fixation as well as P input from the atmosphere
were not sufficient to compensate for these losses. Thus, the common assumption that N
and P budgets in annually burned savanna are balanced, appears to be incorrect, and the
use of fire as a management tool has to be evaluated more carefully.
1
Summary
Assuming that N2-fixing plants have a competitive advantage in ecosystems with low
availability of N, the very low abundance of herbaceous legumes in savannas is puzzling
and often attributed to low P availability. Results of a pot experiment with a common
leguminous herb and a common C4 grass of humid African savannas, however, showed that
the low abundance of herbaceous legumes in savannas is due not to low soil P availability
but to a greater ability of C4 grasses to compete for both N and P, and to their higher
nutrient use efficiency.
In vegetation altered by long-term cattle ranching, the simulation of the effects of grazing by
experimental clipping with and without fertilization produced a rapid positive feedback on
the quality of grazed areas. Clipping and fertilization increased productivity and favoured
certain preferred fodder species of wild ungulates. This suggests that savanna restoration
and the attraction of wild herbivores through facilitation may be successful, provided that
the nutrients are returned to the grazed areas and that the numbers of animals are sufficient
to maintain the facilitated areas.
2
Zusammenfassung
Zusammenfassung
Afrikanische Savannen beherbergen die weltweit höchste Vielfalt an wilden Huftieren und
stellen ausserdem eine wichtige natürliche Ressource für eine wachsende Anzahl Menschen
dar. Es besteht die Gefahr, dass diese sensiblen Ökosysteme für die Viehzucht und vielleicht
auch zur Produktion von Biotreibstoff übernutzt werden, was zu einem Verlust von
Biodiversität und Ökosystemfunktionen führen könnte. Das Ziel dieser Arbeit ist es, durch
die Untersuchung der Auswirkungen von grossen Herbivoren, Feuer und N2-Fixierung auf
die kurz- und langfristigen Verfügbarkeiten dieser Nährstoffe, ein verbessertes Verständnis
der Stickstoff- und Phosphorkreisläufe in Savannen zu erlangen. Die Ergebnisse dieser
Arbeit können auch eingesetzt werden, um die Auswirkungen von verschiedenen Arten der
Landnutzung, des Einsatzes von Feuer als Bewirtschaftungsinstrument und von erhöhten
Nährstoffeinträgen auf Savannenökosysteme einzuschätzen. Das Untersuchungsgebiet an
der Pazifikküste von Tansania liegt im Saadani Nationalpark, der sich unter anderem aus
einer ehemaligen Rinderfarm und einem ehemaligen Jagdreservat mit einer hohen Dichte
von wilden Herbivoren zusammensetzt.
Die Auswirkungen von Feuer, grossen Herbivoren und N2-Fixierung auf die Art der
Nährstofflimitierung konnten aufgrund theoretischer Überlegungen grösstenteils
vorausgesagt werden. An drei Standorten – einer häufig brennenden Langgrassavanne,
einer von wilden Huftieren intensiv begrasten Fläche und einem Akazienbuschwald
– war die Produktivität der Grasdecke durch N und P kolimitiert, was auf eine sehr
tiefe Verfügbarkeit beider Nährstoffe hindeutet. Auf einer Fläche, die grössere Einträge
von Viehexkrementen erhalten hatte, war das Pflanzenwachstum N-limitiert. Niedrige
Stickstoffvorräte im Boden fanden sich in Flächen mit hoher Feuerhäufigkeit, während
im Akazienbuschwald und der Fläche mit hohen Dung- und Urineinträgen von Kühen
erhöhte Stickstoffvorräte gemessen wurden.
Die Rinderzucht hat zu einer grossflächigen Umverteilung von Nährstoffen geführt,
wobei die lokale Anreichung von P stärker und dauerhafter war, als jene von N. Auf die
Gesamtheit des Gebietes der ehemaligen Rinderfarm bezogen, sind die lokalen Verluste
von Nährstoffen durch die Beweidung mit Vieh möglicherweise durch Einträge aus der
Atmosphäre kompensiert worden. Die Verfügbarkeiten von N und P auf von wilden Tieren
intensiv beweideten Flächen waren leicht erhöht, da aber den Flächen durch Beweidung
mehr Nährstoffe entzogen als durch Exkremente zurückgeführt wurden, war die Bilanz für
beide Nährstoffe negativ.
3
Zusammenfassung
Während N2-Fixierung in dichten Akazienbeständen zu einer erhöhten Verfügbarkeit
und einem Nettoeintrag von Stickstoff führte, war der Stickstoffeintrag durch krautige
Leguminosen vernachlässigbar. In der Langgrassavanne war Feuer die Hauptursache für
Nährstoffverluste. Einträge aus der Atmosphäre oder durch N2-Fixierung konnten die
Stickstoff- und Phosphorverluste durch Feuer nicht kompensieren. Die übliche Annahme,
dass die Nährstoffbilanzen für N und P in Savannen mit jährlich wiederkehrenden Feuern
ausgeglichen sind, scheint also nicht korrekt zu sein, weshalb die Nutzung von Feuer zur
Bewirtschaftung von Savannengebieten sorgfältig abgeschätzt werden sollte.
Da man vermuten könnte, dass N2-fixierende Pflanzen bei tiefer Stickstoffverfügbarkeit
anderen Pflanzen gegenüber einen Konkurrenzvorteil haben, erstaunt die sehr
geringe Abundanz von krautigen Leguminosen in Savannen und wird oft der tiefen
P-Verfügbarkeit zugeschrieben. Die Ergebnisse eines Topfversuchs mit einer Leguminose
und einer C4-Langgrasart, die beide in vielen afrikanischen Feuchtsavannen vorkommen,
haben jedoch gezeigt, dass die geringe Abundanz von krautigen Leguminosen nicht
durch die tiefe Phosphorverfügbarkeit im Boden bedingt ist, sondern durch die höhere
Konkurrenzfähigkeit der C4-Gräser bei der Aufnahme von N und P aus dem Boden, und
deren höhere Nährstoffeffizienz.
In Vegetationstypen, die durch die langjährige Viehbeweidung verändert worden waren,
konnte mit der Simulation der Auswirkungen von Beweidung durch Mähen mit und
ohne Düngung ein rascher positiver Effekt auf die Qualität der begrasten Fläche erreicht
werden. Durch Mähen und Düngen stiegen die Produktivität der Grasdecke und der
Anteil einiger von wilden Herbivoren bevorzugten Grasarten. Dies weist darauf hin, dass
die Renaturierung von Savannen und der damit bewirkte Anziehungseffekt auf Wildtiere
erfolgreich verlaufen kann, wenn Nährstoffe den beweideten Flächen zurückgeführt
werden, und es genügend Tiere hat, um die meliorierten Flächen aufrecht zu erhalten.
4
General introduction
General introduction
The African savannas contain the world’s largest diversity of wild ungulates, and extensive
areas are managed as national parks and game reserves. The savanna biome, in Africa
and other tropical areas, also constitutes an important resource to a large and growing
proportion of the human population, mostly for livestock production (Augustine et al.
2003; Kauffman et al. 1995; Scholes and Archer 1997). Recently, the idea of harvesting
herbaceous biomass from savannas for biofuel production has been proposed (Samson et
al. 2005), which would put further pressure on these ecosystems. There is a serious threat
that savanna ecosystems are getting overexploited leading to impoverishment of resources
and biodiversity. It is is therefore important to understand the processes controlling and
shaping savanna structure in order to assess the impact of environmental changes as well
as land-use and management by humans on these ecosystems.
Savanna ecosystems are characterized by the co-existence of grasses and trees, forming a
mosaic that is neither grassland nor forest. The main characteristic of savanna climates is
a strong seasonality in water availability. In drier climates the balance between woody and
grass cover appear to be mainly determined by water, whereas in more humid climates
woody plants could potentially achieve full cover (Sankaran et al. 2005). The vast majority
of savannas owe their existence to recurrent natural and anthropogenic fires in the dry
seasons (Bond et al. 2005), but herbivores may also be an important factor affecting the
tree-grass balance (Augustine and McNaughton 2004). Soils in tropical savannas are often
of low fertility (Medina 1987), although a distinction is made between dry and humid
savannas, with the former being relatively more nutrient-rich (Bell and Koch 1980; Huntley
1982). In both dry and humid savannas the productivity of the herbaceous stratum
outside tree canopies has been found to be limited by nutrients, particularly nitrogen and/
or phosphorus are thought to be important (Augustine et al. 2003; Barger et al. 2002;
Brockington 1961; Norman 1966; Sarmiento et al. 2006; Snyman 2002).
Thus, the main determinants of tropical savannas are the interaction of seasonal water and
low nutrient availability, the recurrence of fire and herbivory. Availabilities of nutrients
can be modified by fire and herbivores, but also by leguminous plants, which in symbiosis
with bacteria are able to fix atmospheric nitrogen. Our understanding of the various factors
affecting nutrient cycling in tropical savannas is still very poor, especially with regards to
5
General introduction
phosphorus. Moreover, only a few studies have investigated the importance of the relative
availabilities of nitrogen and phosphorus in savanna ecosystems, despite the increasing
recognition of the significance of N:P stoichiometry for plant growth and vegetation
composition (cf. Güsewell 2004). In the following paragraphs the effects of fire, large
herbivores and N2-fixation on nutrient availability will be described in more detail.
Fire
The humid savannas of Africa are the most frequently burnt ecosystem in the world, with
half of the total surface being estimated to burn every year, and fire frequencies of up
to twice a year in some areas (Bond and Keeley 2005; Hao 1994). The most fire-prone
areas are humid savannas with high productivity and therefore large fuel loads (Frost and
Robertson 1987).
Fire causes a significant loss of nutrients to the atmosphere (Medina 1987; Pivello and
Coutinho 1992); it can be considered as a non-selective herbivore that ‘feeds’ uniformly on
vegetation (Bond and Keeley 2005). Fire therefore tends to reduce heterogeneity in nutrient
distribution. Given the low fertility of savanna soils, the losses of N and P from tropical
savannas reported in literature may be ecologically very significant, and could lead to longterm declines of N and P stocks in soils.
Nutrient losses through fire are proportionally much larger for nitrogen than for phosphorus
and other nutrients (Cook 1994; Kauffman et al. 1995; Van de Vijver et al. 1999). Therefore,
fires are thought to reduce the relative availability of nitrogen compared to phosphorus and
other nutrients, and thus biomass production in frequently burnt savannas is thought to
be primarily limited by nitrogen (Medina 1987; Vitousek and Howarth 1991). However,
for such conclusions insight is required not only in relative loss of N and P in burned
vegetation, but also in the contributions of N and P outputs through fire on N and P
balances of the ecosystem.
Some authors presume neutral N balances for savanna ecosystems (Bate 1981; Laclau et
al. 2005; Sanhueza and Crutzen 1998); hence, they assume that nitrogen losses through
fires are compensated by nitrogen inputs, but there is little evidence for this. The impact of
fire on the P balance is even less clear, as there have been only very few studies assessing
P balance in savannas. Hence, there is not only a need to quantify the outputs of N and P
through fire, but also to evaluate their relative importance for N and P balances.
6
General introduction
Herbivory
After fire, large ungulate herbivores are the most conspicuous consumers of biomass in
savannas. Large herbivores may directly affect nutrient turnover and availability by creating
a ‘short-cut’ to the decomposition of biomass, since they return nutrients in a form that is
more rapidly mineralized (Bakker et al. 2004). This effect is likely to differ spatially and
among animal species. Some domestic herbivores like cows and horses are known to redistribute nutrients by feeding in certain areas while distributing their excreta in others,
thus causing local depletion or enrichment of nutrients (Augustine 2003; Edwards and
Hollis 1982; Jewell et al. 2007). Other herbivores feed and return their excreta in the same,
intensively used areas, thereby maintaining nutrient-rich patches in otherwise nutrient-poor
vegetation; the grazing lawns of hippos are well known examples of this process.
Thus, in contrast to fire, large herbivores may increase heterogeneity in nutrient availabilities,
and this effect is likely to differ among domestic and wild herbivores. While patterns in
nutrient re-distribution by domestic herbivores have been qualitatively and quantitatively
documented for some semi-arid savanna ecosystems (Augustine 2003; Turner 1998), the
effects of wild herbivores have not yet been investigated.
Large herbivores not only re-distribute nutrients, but also cause losses of N to the atmosphere,
because ammonia volatilizes from urine and dung (Frank and Zhang 1997; Ruess and
McNaughton 1988). A consequence of this process is an increased availability of P relative
to N in areas where animal excreta are deposited (Augustine 2003).
Consumption of biomass by large herbivores reduces the fuel load for fires and may thus
result in reduced losses through fire (Hobbs et al. 1991). This effect is likely to be significant
only in mesic savannas where biomass production is high enough to result in frequent fires
in the absence of grazing, but not too high so that herbivores may significantly reduce fuel
loads (Hobbs 1996).
By enhancing plant biomass production through grazing, large herbivores may function
as ecological engineers (Jones et al. 1994). Grazing optimization theory suggesting that
net primary production is maximized at an intermediate grazing intensity is supported by
model simulations (de Mazancourt et al. 1998; 1999; Hilbert et al. 1981; Holdo et al. 2007),
but for savannas experimental evidence for overcompensatory growth at moderate grazing
frequencies under field conditions has been reported only for Serengeti (McNaughton
1983a). Ungulate herbivores can also alter the species composition and structure of
grassland vegetation (Belsky 1986; McNaughton 1983b; Thornton 1971), through which
they may affect quality of their forage.
7
General introduction
Overall, the net effect of large herbivores on nutrient balance of tropical savannas is fairly
unknown, and domestic and wild herbivores may differ in this respect. There is a need to
improve insight in the effects of defoliation as well as nutrient return through excretion of
different herbivores on N and P balances of savanna ecosystems. Also, facilitation effects
of herbivores on properties of savanna vegetation and ecosystems require further study.
N2-fixation
The Leguminosae are the third largest family of flowering plants, of which many species
are able to form a symbiosis with nitrogen-fixing bacteria. These bacteria, which establish
in root nodules of the host plant, reduce atmospheric N2 to ammonia and transfer it to the
host plant. In return the bacteria are supplied with sugars from the host plant. Symbiotic
N2-fixation thus represents a net N input into the ecosystem, and increases soil N pools
and the availability of N relative to other nutrients (Ludwig et al. 2001; Yelenik et al. 2004).
The extent of symbiotic N2-fixation not only differs among plant species, but also among
environmental conditions such as soil N and P availabilities (Perreijn 2002; Sprent 1999;
Vitousek et al. 2002).
Assuming that particularly nitrogen is lost through fires, it is speculated that legumes
and other plants with access to atmospheric N through symbiotic fixation, would have
a competitive advantage under conditions of frequent fires (Vitousek and Field 1999;
Vitousek and Howarth 1991). However, even in tropical savannas experiencing a very high
fire frequency, abundance of herbaceous legume species is generally very low; i.e. often
less than 1% of the biomass (Ezedinma et al. 1979; Huntley 1982; Isichei 1995; Laclau et
al. 2002; Medina 1987; Menaut and Cesar 1979; San Jose et al. 1985). In contrast, woody
legume species, especially of the genus Acacia, are an important component of African and
Australian savannas and they are also predominant among species responsible for bush
encroachment (Archer 1995), which is largely attributed to overgrazing by cattle (Hudak
1999; Tobler et al. 2003).
There is a large body of literature about the contribution of N2-fixation by herbaceous
and woody legumes in agricultural systems. However, the input of N through symbiotic N
fixation in natural savanna ecosystems has not yet been quantitatively assessed, although
symbiotic N2-fixation is sometimes assumed to balance N losses through fire (Laclau et al.
2005; Sanhueza and Crutzen 1998). To improve understanding of the N cycle in savanna
ecosystems, there is a need to assess N input fluxes through symbiotic N2-fixation by trees
and herbs, and to compare them to other N fluxes. Finally, the cause for legume herbs
being so scarce in savannas needs to be elucidated.
8
General introduction
Aim of this thesis
The central aim of this thesis is to investigate the influence of large herbivores, symbiotic N2fixation and fire on the availabilities of nitrogen and phosphorus, as well as on the relative
availabilities of these nutrients for plants. Besides gaining a more profound understanding
of nutrient cycling in savanna ecosystems, this thesis intends to provide information useful
for managing savanna ecosystems sustainably and for assessing the sensitivity of savannas
to changes in land-use. These aims were achieved by means of experiments and field
measurements addressing the following questions:
1) Which is/are the main limiting nutrient(s) in different savanna vegetation types:
nitrogen, phosphorus, or another nutrient?
2) Can the type of nutrient limitation be explained by the relative impact of fire, herbivory
and N2-fixation?
3) What are the absolute and relative availabilities of N and P along vegetation gradients
created by cattle ranching, and how do they compare with habitats used exclusively by
wild herbivores?
4) How do large herbivores, fire and N2-fixation affect N and P balances in various savanna
vegetation types, and hence the availabilities of N and P in the long-term?
5) What are the reasons for the low abundance of herbaceous legumes in African
savannas?
6) Is the re-colonization of former cattle grounds by wild herbivores limited by deteriorated
habitat quality, or is facilitation of high quality grazing areas possible?
Study area
Saadani National Park lies on the northern Tanzanian coast (5° 43′ S, 38° 47′ E) and is
Tanzania’s newest national park. Between 1954 and 2000, the northern part of the area
was used as a cattle ranch, with up to 13’000 head of cattle on ~460 km2 (Figure 1).
Much of the southern part of Saadani National Park was a game reserve from 1969 until
it became national park in 2002. The entire national park is grazed by wild herbivores
such as warthog, waterbuck, reedbuck, buffalo, wildebeest, giraffe and elephant. Densities
and diversity of wild herbivores are much higher in the southern part of the park, and no
increase in ungulate abundance was observed in the former ranch area during the first three
years after abandonment of the ranch (Treydte et al. 2005).
9
General introduction
Msangazi River
Tanzania
former
Mkwaja Ranch
Mkwaja
Mlig
a
ji Riv
er
Indian Ocean
Saadani
former
Saadani Game
Reserve
N
0
3
6
km
9
ive
R
mi
Wa
r
Figure 1 Map of Saadani National Park on the Tanzanian coast. The former Mkwaja Ranch
and the former Saadani Game Reserve are indicated by the light grey and dark areas, respectively.
Filled circles indicate the location of paddocks where cattle were herded overnight, filled triangles
are villages.
The area can be classified as humid dystrophic savanna (Huntley 1982), since annual
precipitation averages 1000 mm and soils are relatively nutrient-poor (Klötzli 1980). The
driest months are January and February and August and September, and during these
periods fires are common, many of them being started deliberately by the local people,
poachers or by the park management.
The original vegetation of the area was probably moist and dry evergreen coastal forests,
but these were replaced by savanna vegetation in quaternary times (Klötzli 1980). At the
time when Mkwaja Ranch was established, the region was an open savanna relatively
10
General introduction
rich in wildlife, and with a low number of human settlements. All activities on the ranch,
as well as detailed information on size and composition of cattle stocks, economic and
meteorological data were meticulously documented in monthly and yearly reports, which
probably represent the best dataset on a large-scale cattle ranching enterprise in Africa.
Cattle were kept in paddocks with up to 1500 animals at night, and during the day were
driven to pasture areas and dams in herds of 200-400 animals.
The effect of the former ranch management is still reflected in the current vegetation, with
five concentric vegetation zones around each paddock (Tobler et al. 2003; Figure 2): (1) the
paddock centres are dominated by the stoloniferous grass Cynodon dactylon, which forms
dense mats; (2) the margins of paddocks are characterized by the grasses Digitaria milanjiana,
Eragrostis superba and the sedge Cyperus bulbipes; (3) Heteropogon savanna is dominated by
the grass Heteropogon contortus, accompanied by Panicum infestum; (4) Acacia woodlands
are dominated by Acacia zanzibarica and Terminalia spinosa reaching heights up to c. 8 m
- Panicum infestum and Heteropogon contortus are most abundant in the herbaceous layer; (5)
tallgrass savanna is dominated by the tall grasses Hyperthelia dissoluta and Diheteropogon
amplectens with culms up to ~2 m height.
A
Paddock
Centre
(PC)
Paddock
Margin
(PM)
Heteropogon Acacia
Woodland
Savanna
(AW)
(HS)
Tallgrass
Savanna
(TG)
Distance from Paddock Centre
B
PC
PM
HS
AW
TG
Figure 2 Schematic cross-section (A) and top view (B) of the vegetation types found along
paddock gradients in the former Mkwaja Ranch area.
11
General introduction
Despite large efforts to fight bush encroachment large amounts of pasture area were lost,
and nowadays vegetation in the former Mkwaja Ranch area has a much higher proportion
of bushland than in 1954.
Because of its recent history, Saadani National Park offers a unique opportunity to study the
effects of former intensive cattle ranching, fire and wild herbivores in a savanna ecosystem,
and how the interaction of these effects may affect the process of re-colonization of
abandoned cattle grounds by wild herbivores.
Thesis outline
In Chapter 1 – ‘Effects of herbivory, fire and N2-fixation on nutrient limitation in a humid African
savanna’ short-term fertilization experiments are used to test whether N, P and/or other
nutrients are limiting growth of the herbaceous vegetation in different savanna vegetation
types. The results are also used to evaluate whether the N:P ratio in aboveground biomass
is a good predictor for the type of nutrient limitation, and whether the type of nutrient
limitation can be explained by the differential impact of large herbivores, fire and N2fixation.
In Chapter 2 – ‘Effects of large herbivores, fire, and N2-fixation on N and P cycling in humid
savanna’ various measures of the short-term availabilities of N and P are determined along
vegetation gradients to evaluate how cattle ranching, wild herbivores, fire and N2-fixation
affect the absolute and relative availabilities of N and P. Based on measured and estimated
input and output fluxes, annual nutrient balances are calculated in order to assess the
relative importance of these fluxes for N and P availabilities in the long term.
Chapter 3 – ‘Why are herbaceous legumes so scarce in savanna? A competition experiment’ describes
an experiment to investigate whether the abundance of herbaceous legumes in savanna is
limited by the availability of nitrogen and/or phosphorus, water, or by the competition
with C4-grasses.
The experiment described in Chapter 4 – ‘Effects of clipping and fertilization on productivity and
species composition on abandoned cattle grounds in humid savanna, Tanzania’ aims at evaluating
whether the slow re-colonization of an abandoned ranch by wild herbivores may be due
poor quality of the habitat after more than 40 years of cattle grazing. By experimentally
simulating the effects of grazing through clipping with and without fertilization, and
analyzing the effects on productivity and species composition of the vegetation, it is
investigated whether grazing by wild herbivores can promote high quality grazing areas in
different types of savanna vegetation used formerly as cattle pastures.
12
General introduction
Literature cited
Archer S (1995) Tree-grass dynamics in a Prosopis-thornscrub savanna parkland: Reconstructing
the past and predicting the future. Ecoscience 2:83-99
Augustine DJ (2003) Long-term, livestock-mediated redistribution of nitrogen and phosphorus in
an East African savanna. Journal of Applied Ecology 40:137-149
Augustine DJ, McNaughton SJ (2004) Regulation of shrub dynamics by native browsing ungulates
on East African rangeland. Journal of Applied Ecology 41:45-58
Augustine DJ, McNaughton SJ, Frank DA (2003) Feedbacks between soil nutrients and large
herbivores in a managed savanna ecosystem. Ecological Applications 13:1325-1337
Bakker ES, Olff H, Boekhoff M, Gleichmann JM, Berendse F (2004) Impact of herbivores on
nitrogen cycling: contrasting effect of small and large species. Oecologia 138:91-101
Barger NN, D’Antonio CM, Ghneim T, Brink K, Cuevas E (2002) Nutrient limitation to primary
productivity in a secondary savanna in Venezuela. Biotropica 34:493-501
Bate GC (1981) Nitrogen cycling in savanna ecosystems. In: Clark FE & Rosswall T (eds) Terrestrial
nitrogen cycles, Ecol Bull 33. Royal Swedish Academy of Sciences, Stockholm, pp 463-475
Bell DT, Koch JM (1980) Post-fire succession in the northern jarrah Forest of Western Australia.
Australian Journal of Ecology 5:9-14
Belsky AJ (1986) Population and community processes in a mosaic grassland in the Serengeti,
Tanzania. Journal of Ecology 74:841-856
Bond W, Keeley JE (2005) Fire as a ‘global’ herbivore: the ecology and evolution of flammable
ecosystems. Trends in Ecology and Evolution 20:387-393
Bond WJ, Woodward FI, Midgley GF (2005) The global distribution of ecosystems in a world
without fire. New Phytologist 165:525-537
Brockington NR (1961) Studies of the growth of a Hyparrhenia-dominant grassland in Northern
Rhodesia. II. Fertilizer responses. III. The effect of fire. Journal of the British Grassland Society
16:54-65
Cook GD (1994) The fate of nutrients during fires in a tropical savanna. Australian Journal of
Ecology 19:359-365
de Mazancourt C, Loreau M, Abbadie L (1998) Grazing optimization and nutrient cycling: when
do herbivores enhance plant production? Ecology 79:2242-2252
de Mazancourt C, Loreau M, Abbadie L (1999) Grazing optimization and nutrient cycling: potential
impact of large herbivores in a savanna system. Ecological Applications 9:784-797
Edwards PJ, Hollis S (1982) The distribution of excreta on New Forest grassland used by cattle,
ponies and deer. Journal of Applied Ecology 19:953-964
Ezedinma FOC, Agbim NN, Onyekwelu SSC (1979) Contribution of symbiotic nitrogen fixation
to the nitrogen economy of natural ecosystems: 1. Occurrence of herbaceous legumes in derived
savanna fallow and their nodulation in pot culture. Plant and Soil 51:503-513
Frank DA, Zhang YM (1997) Ammonia volatilization from a seasonally and spatially variable
grazed grassland: Yellowstone National Park. Biogeochemistry 36:189-203
Frost PGH, Robertson F (1987) The ecological effects of fire in savannas. In: Walker BH (ed)
Determinants of tropical savannas. IRL Press, Oxford, UK, pp 93-140
Güsewell S (2004) N:P ratios in terrestrial plants: variation and functional significance. New
Phytologist 164:243-266
13
General introduction
Hao WM (1994) Spatial and temporal distribution of tropical biomass burning. Global
Biogeochemical Cycles 8:495-503
Hilbert DW, Swift DM, Detling JK, Dyer MI (1981) Relative growth rates and the grazing
optimization hypothesis. Oecologia 51:14-18
Hobbs NT (1996) Modification of ecosystems by ungulates. Journal of Wildlife Management
60:695-713
Hobbs NT, Schimel DS, Owensby CE, Ojima DS (1991) Fire and grazing in the tallgrass prairie:
contingent effects on nitrogen budgets. Ecology 72:1374-1382
Holdo RM, Holt RD, Coughenour MB, Ritchie ME (2007) Plant productivity and soil nitrogen as
a function of grazing, migration and fire in an African savanna. Journal of Ecology 95:115-128
Hudak AT (1999) Rangeland mismanagement in South Africa: Failure to apply ecological
knowledge. Human Ecology 27:55-78
Huntley BJ (1982) Southern African savannas. In: Huntley BJ & Walker BH (eds) Ecology of
tropical savannas. Springer, Berlin, pp 101-119
Isichei AO (1995) Stocks of nitrogen in vegetation and soil in West African moist savannas and
potential effects of climate change and land use on these stocks. Journal of Biogeography
22:393-399
Jewell PL, Kauferle D, Gusewell S, Berry NR, Kreuzer M, Edwards PJ (2007) Redistribution of
phosphorus by mountain pasture in cattle on a traditional the Alps. Agriculture Ecosystems &
Environment 122:377-386
Jones CG, Lawton JH, Shachak M (1994) Organisms as ecosystem engineers. Oikos 69:373-386
Kauffman JB, Cummings DL, Ward DE, Babbitt R (1995) Fire in the Brazilian Amazon: 1.
Biomass, nutrient pools, and losses in slashed primary forests. Oecologia 104:397-408
Klötzli F (1980) Analysis of species oscillations in tropical grasslands in Tanzania due to
management and weather conditions. Phytocoenologia 8:13-33
Laclau JP, Ranger J, Deleporte P, Nouvellon Y, Saint-Andre L, Marlet S, Bouillet JP (2005) Nutrient
cycling in a clonal stand of Eucalyptus and an adjacent savanna ecosystem in Congo 3. Inputoutput budgets and consequences for the sustainability of the plantations. Forest Ecology And
Management 210:375-391
Laclau JP, Sama-Poumba W, Nzila JD, Bouillet JP, Ranger J (2002) Biomass and nutrient dynamics
in a littoral savanna subjected to annual fires in Congo. Acta Oecologica 23:41-50
Ludwig F, de Kroon H, Prins HHT, Berendse F (2001) Effects of nutrients and shade on tree-grass
interactions in an East African savanna. Journal of Vegetation Science 12:579-588
McNaughton SJ (1983a) Compensatory plant growth as a response to herbivory. Oikos
40:329-336
McNaughton SJ (1983b) Serengeti grassland ecology: the role of composite environmental factors
and contingency in community organization. Ecological Monographs 53:291-320
Medina E (1987) Nutrients: requirements, conservation and cycles in the herbaceous layer. In:
Walker BH (ed) Determinants of tropical savannas. IRL Press, Oxford, pp 39-65
Menaut JC, Cesar J (1979) Structure and primary productivity of Lamto savannas, Ivory Coast.
Ecology 60:1197-1210
Norman MJT (1966) Katherine Research Station 1956-64: A review of published work. CSIRO,
Melbourne
Perreijn K (2002) Symbiotic nitrogen fixation by leguminous trees in tropical rain forest in Guyana.
Tropenbos-Guyana Series 11, Tropenbos, Georgetown, Guyana
14
General introduction
Pivello VR, Coutinho LM (1992) Transfer of macro-nutrients to the atmosphere during experimental
burnings in an open cerrado (Brazilian savanna). Journal of Tropical Ecology 8:487-497
Ruess RW, McNaughton SJ (1988) Ammonia volatilization and the effects of large grazing
mammals on nutrient loss from East African grasslands. Oecologia 77:382-386
Samson R, Mani S, Boddey R, Sokhansanj S, Quesada D, Urquiaga S, Reis V, Ho Lem C (2005) The
potential of C4 perennial grasses for developing a global BIOHEAT industry. Critical Reviews in
Plant Sciences 24:461-495
San Jose JJ, Montes R, Garcia-Miragaya J, Orihuela BE (1985) Bio-production of Trachypogon
savannas in a latitudinal cross-section of the Orinoco Llanos, Venezuela. Acta Oecologica
6:25-43
Sanhueza E, Crutzen PJ (1998) Budgets of fixed nitrogen in the Orinoco savannah region: Role of
pyrodenitrification. Global Biogeochemical Cycles 12:653-666
Sankaran M, Hanan NP, Scholes RJ, Ratnam J, Augustine DJ, Cade BS, Gignoux J, Higgins SI,
Le Roux X, Ludwig F, Ardo J, Banyikwa F, Bronn A, Bucini G, Caylor KK, Coughenour MB,
Diouf A, Ekaya W, Feral CJ, February EC, Frost PGH, Hiernaux P, Hrabar H, Metzger KL,
Prins HHT, Ringrose S, Sea W, Tews J, Worden J, Zambatis N (2005) Determinants of woody
cover in African savannas. Nature 438:846-849
Sarmiento G, da Silva MP, Naranjo ME, Pinillos M (2006) Nitrogen and phosphorus as limiting
factors for growth and primary production in a flooded savanna in the Venezuelan Llanos. Journal
of Tropical Ecology 22:203-212
Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annual Review of Ecology and
Systematics 28:517-544
Snyman HA (2002) Short-term response of rangeland botanical composition and productivity to
fertilization (N and P) in a semi-arid climate of South Africa. Journal of Arid Environments
50:167-183
Sprent JI (1999) Nitrogen fixation and growth of non-crop legume species in diverse environments.
Perspectives in Plant Ecology, Evolution and Systematics 2:149-162
Thornton DD (1971) The effect of complete removal of Hippopotamus on grassland in the Queen
Elizabeth National Park, Uganda. East African Wildlife Journal 9:47-55
Tobler MW, Cochard R, Edwards PJ (2003) The impact of cattle ranching on large-scale vegetation
patterns in a coastal savanna in Tanzania. Journal of Applied Ecology 40:430-444
Treydte AC, Edwards PJ, Suter W (2005) Shifts in native ungulate communities on a former cattle
ranch in Tanzania. African Journal of Ecology 43:302-311
Turner MD (1998) Long-term effects of daily grazing orbits on nutrient availability in Sahelian
West Africa: I. Gradients in the chemical composition of rangeland soils and vegetation. Journal
of Biogeography 25:669-682
Van de Vijver C, Poot P, Prins HHT (1999) Causes of increased nutrient concentrations in post-fire
regrowth in an East African savanna. Plant-and-Soil. 1999; 214:173-185.
Vitousek PM, Cassman K, Cleveland C, Crews T, Field CB, Grimm NB, Howarth RW, Marino R,
Martinelli L, Rastetter EB, Sprent JI (2002) Towards an ecological understanding of biological
nitrogen fixation. Biogeochemistry 57:1-45
Vitousek PM, Field CB (1999) Ecosystem constraints to symbiotic nitrogen fixers: a simple model
and its implications. Biogeochemistry 46:179-202
Vitousek PM, Howarth RW (1991) Nitrogen limitation on land and in the sea: How can it occur?
Biogeochemistry 13:87-115
Yelenik SG, Stock WD, Richardson DM (2004) Ecosystem level impacts of invasive Acacia saligna
in the South African fynbos. Restoration Ecology 12:44-51
15
Nutrient limitation
Chapter 1
Effects of herbivory, fire and N2-fixation on nutrient
limitation in a humid African savanna
Patrick G. Cech, Thomas Kuster, Peter J. Edwards, and Harry Olde Venterink
Institute of Integrative Biology, ETH Zürich, Universitätsstrasse 16, 8092, Zürich, Switzerland
17
Chapter 1
Abstract
The quantities and spatial distribution of nutrients in savanna ecosystems are affected by
many factors, of which fire, herbivory and symbiotic N2-fixation are particularly important.
We measured soil nitrogen (N) pools and the relative abundance of N and phosphorus
(P) in herbaceous vegetation in five vegetation types in a humid savanna in Tanzania.
We also performed a factorial fertilization experiment to investigate which nutrients most
limit herbaceous production. N pools in the top 10 cm of soil were low at sites where fires
were frequent, and higher in areas with woody legume encroachment, or high herbivore
excretion. Biomass production was co-limited by N and P at sites that were frequently
burnt or heavily grazed by native herbivores. In contrast, aboveground production was
limited by N in areas receiving large amounts of excreta from livestock. N2-fixation by
woody legumes did not lead to P-limitation, but did increase the availability of N relative
to P. We conclude that the effects of fire, herbivory and N2-fixation upon soil N pools and
N:P-stoichiometry in savanna ecosystems are, to a large extent, predictable.
Keywords: Acacia woodland, cattle, grassland, grazing, N:P ratios, nitrogen fixation,
phosphorus, stoichiometry, tallgrass
18
Nutrient limitation
Introduction
In savanna ecosystems grasses and trees co-exist, forming a mosaic that is neither grassland
nor forest. The vast majority of savannas owe their existence to fire, and without it would
eventually be replaced by forests (Bond et al. 2005). However, in drier climates the balance
between woody and grass cover and productivity appears to be mainly determined by water
availability (Aranibar et al. 2004; Sankaran et al. 2005). Compared to humid savannas,
these dry savannas are nutrient-rich (Bell 1982; Huntley 1982), though at some sites the
productivity of herbaceous vegetation is limited by the availability of N (Augustine et
al. 2003; Ludwig et al. 2001; Snyman 2002). In contrast, humid savannas are often on
heavily leached soils and are characterized by low P availability (Sanchez 1976); plant
growth in these ecosystems may be limited by shortages of N, P, or both (Högberg 1986;
Medina 1987), though which nutrient is limiting in any particular case has rarely been
investigated.
Many factors affect the quantities and spatial distribution of nutrients in savanna soils,
with fire, herbivory, and atmospheric N2-fixation being particularly important. In Figure
1 we present a conceptual framework to show how these factors are likely to affect soil N
pools and N:P-stoichiometry.
(1) Fire can cause considerable losses of all nutrients to the atmosphere, with losses of N
being much greater than those of other nutrients (Cook 1994; Kauffman et al. 1995); as a
result, repeated fires are likely to promote a shortage of N, and many savannas are thought
to be primarily N-limited.
(2) Herbivores can affect nutrient conditions in various ways. The first relates to the spatial
pattern of habitat use for feeding and excretion. Some herbivores re-distribute nutrients by
feeding in certain areas while depositing their excreta in others (Edwards and Hollis 1982;
Jewell et al. 2007), thus causing local depletion or enrichment of nutrients. In contrast,
others feed and return their excreta in the same, intensively used areas, thereby potentially
maintaining nutrient rich patches in otherwise nutrient-poor vegetation (Lamoot et al.
2004). Second, a significant proportion of the N in dung and urine may be lost through
ammonia volatilization and leaching (Augustine 2003; Frank and Zhang 1997; Ruess and
McNaughton 1988), leading to a local excess of P relative to N (providing that the dung
and urine are excreted in the same areas, which is commonly the case; Edwards and Hollis
1982). Third, herbivores may have a positive net effect on the soil N pool by reducing
the amount of biomass exposed to fire, and thereby reducing the associated losses to the
19
Chapter 1
atmosphere (Hobbs et al. 1991; Holdo et al. 2007). Hence, depending on patterns of grazing
and excretion, herbivory has a neutral effect on the ratio of N:P availabilities (defoliation)
or potentially shifts the nutrient balance towards N-limitation (excretion) (Figure 1).
A
N2-fixation
P-limitation
N:P
(3) Woody legumes, especially of the genus Acacia, are conspicuous elements of many
savannas, and the majority are thought to fix N2 symbiotically. The abundance and
distribution of these trees is significantly affected by the activities of herbivores. For
example, grazing by domestic livestock is often responsible for extensive encroachment of
savanna grasslands by woody species (Brown and Archer 1989; Hudak 1999; Tobler et al.
2003), of which legumes are a dominant component (Archer 1995; Cramer et al. 2007).
It has been shown that much greater amounts of N accumulate in the bushland than in
the open grassland it replaces; this accumulation could in turn lead to P-limitation under
Acacia canopies (Figure 1).
excretion from
herbivores
fire
N-limitation
defoliation by
herbivores
forest
2° Acacia
woodland
tallgrass
savanna
grazed
patches
paddock
margin
forest
gaps
N-limitation
B
P-limitation
N:P
soil N pool
soil N pool
Figure 1 Conceptual framework of the effects of fire, herbivory and N2-fixation on soil nitrogen
pool, N:P-stoichiometry and the type of nutrient limitation (A), and hypothesized status of nitrogen
pools and N:P ratios of five vegetation types differentially affected by fire, herbivory and N2-fixation
in relation to the presumed climax forest vegetation (B).
20
Nutrient limitation
There have been many agricultural studies of nutrient limitation in tropical savanna
ecosystems (Brockington 1961; Norman 1966; e.g. Weinmann 1938), including several
that investigated the effects of either fire or domestic herbivory (Augustine 2003; Barger
et al. 2002). However, we are unaware of any study of a savanna ecosystem designed to
investigate the interactive effects of fire and native and domestic herbivores. Also, most
studies have focussed on N and sometimes P, and the extent to which cations such as
potassium may also limit plant production is not known. However, the concentrations of
cations in savanna vegetation are typically low, and are known to be a factor influencing
food selection by large herbivores (McNaughton 1988; McNaughton 1990).
The aim of this study was to investigate the type of nutrient limitation in different types of
humid savanna vegetation in Saadani National Park, Tanzania. We selected five contrasting
types of vegetation which from our conceptual framework (cf. Figure 1B) we predicted
would differ with respect to soil N pool and N:P-stoichiometry. These vegetation types were:
forest gaps, tallgrass savanna, Acacia woodland, surroundings of former cattle paddocks
and lawns grazed by wild ungulates. In each of these vegetation types, we measured the
soil N pools and performed a factorial fertilization experiment to determine patterns of
nutrient limitation.
Methods
Study area
The study area is located in Saadani National Park on the Tanzanian coast (5° 43′ S, 38° 47′ E).
The northern part of the national park was operated as cattle ranch from 1954 to 2000 with
up to 13‘000 head of cattle on ~460 km2. The southern part was a game reserve from 1969
until it became national park in 2002. The area is grazed by wild herbivores including
warthog, waterbuck, reedbuck, buffalo, wildebeest, giraffe and elephant. Densities and
diversity of wild herbivores are much higher in the southern part of the park compared
to the former ranch area (Treydte et al. 2005). Mean annual temperature recorded at the
former ranch complex is 25° C (1973–98). Annual precipitation from 1957-98 has ranged
from 610 to 1700 mm, with a mean of 1040 mm. The wet season lasts from March until
June, and there is a short rainy season from mid-October to mid-November. The driest
months are January and February and August and September, and during these periods fires
are common, many of them being started deliberately by the local people. The relatively
nutrient-poor soils consist of greyish, fine or loamy sand in the flats, and reddish, loamy
sand over clay on slopes and hilltops (Klötzli 1980). For a more detailed description of the
study area see (Tobler et al. 2003).
21
Chapter 1
The vegetation of the area is dominated mainly by bushland and grassland, but there are
extensive areas of evergreen forest. The annual precipitation is sufficient to allow for 100%
woody cover (Sankaran et al. 2005), and the fact that the vegetation is now mainly savanna
is probably due to recurrent fires of anthropogenic origin (Bond et al. 2005). We selected a
representative site in each of the following five vegetation types:
(I) Gaps in remnant forest. These grass-dominated patches are surrounded by remnant
forest and connected to the open savanna by narrow grass corridors. Since fires rarely reach
these areas (a conclusion based on aerial photographs and personal observations), they
served as a reference for the more frequently burnt vegetation types. The site selected was
dominated by the grasses Sporobolus pyramidalis and Schoenefeldia transiens, and the sedge
Abildgaardia triflora. The vegetation was 50 cm tall with flowering tillers reaching to 1.5 m;
the average basal cover was 85%.
(II) Tallgrass vegetation. This type of vegetation is known to be the most frequently burnt
due to its high biomass (Frost and Robertson 1987), and we predicted that it would have the
greatest cumulative nutrient losses, especially of N. The site chosen was dominated by the
grasses Hyperthelia dissoluta, Diheteropogon amplectens and Andropogon schirensis. Vegetation
was 2 m tall, and basal cover was 70%.
(III) Acacia woodlands. This open woodland has developed from tall grass communities
as a result of cattle grazing (Tobler et al. 2003); compared to the tall grass savanna, we
expected it to be relatively N-rich because of symbiotic N2-fixation At the chosen site, there
were scattered trees of Acacia zanzibarica 6 m high and with a basal area of 1.3 m2 ha-1. A.
zanzibarica trees were nodulated (Cech, personal observation), and δ15N data indicated that
they derived ca. 58% of their N from fixation (Chapter 2). The understory was dominated
by the grasses Heteropogon contortus, Panicum infestum and Digitaria milanjiana, and the sedge
Abildgaardia triflora. This layer was 60 cm tall, and basal cover reached 90%.
(IV) Paddock margins. The former paddocks (‘bomas’), which were used until 2000 to
hold the cattle overnight, and their surroundings were very nutrient enriched. We predicted
these areas would be relatively P-rich due to large N-losses from ammonia volatilization.
The site selected was dominated by the grasses Eragrostis superba and Digitaria milanjiana
and the sedge Cyperus bulbipes. Vegetation height was 60 cm and basal cover 100%.
(V) Grazed lawns. Grazed patches (sensu Archibald et al. 2005) are localized areas of
intensive grazing that are not continuously maintained by wild herbivores. We explicitly
assumed that wild herbivores not only eat the vegetation, but also deposit dung and urine
in these patches. In support of this assumption, we found much more dung in grazed lawns
22
Nutrient limitation
than in the other vegetation types where dung counts were made (tallgrass savanna, Acacia
woodland, medium height savanna) (Chapter 2). Therefore we expected grazed patches
to be generally nutrient-rich (though less so than the paddock margins) and with relatively
high levels of P compared to N. We selected an area where vegetation - dominated by
the grasses Digitaria milanjiana, Diheteropogon amplectens and Panicum infestum - had been
maintained at less than 10 cm height by wild herbivores for at least 12 months (Cech et
al., personal observation). During the rainy season prior to the start of our experiment,
however, grazing pressure decreased, and at the start of the experiment the vegetation was
40 cm tall and had a basal cover of 95%.
The first four sites were located on sandy soil in the area of the former cattle ranch. The
grazed lawn was on a similar sandy soil in the southern part of the national park where
wild herbivores were much more abundant.
Experimental design and treatments
The experiment was started at the end of the wet season in June 2006. At each site we
set up an experimental plot in a patch of homogeneous vegetation. Plots comprised 30
blocks of 1 m2 separated from each other by 1 m buffer zones, and were surrounded by
a 1.5 m tall fence to exclude small and large mammal herbivores. Blocks were randomly
assigned to one of six nutrient treatments: control, nitrogen (N), phosphorus (P), nitrogen
and phosphorus (N+P), cations (+cations), and all nutrients (N+P+cations). ‘Cations’ was
a combination of the macronutrients K, Ca and Mg, and the micronutrients Fe, Mn, Zn,
B, Cu, and Mo. The rationale for this treatment was to determine whether production was
limited by a nutrient other than N or P. Nitrogen was supplied as NH4NO3 at 20 g m-2,
phosphorus as Na2HPO4 at 5 g m-2, potassium as KCl at 5 g m-2, calcium as CaCl2 at 5 g m-2,
and magnesium as MgSO4 at 1.4 g m-2. These supply levels were sufficiently high to offset
growth limitation without being toxic (Augustine et al. 2003) and are similar to those used
in other fertilization experiments in comparable ecosystems (reported in Table 3). The
micronutrients, in the form of oxides or mineral salts of chloride or sulphate, were supplied
in the same proportions relative to N as are used in Hoagland’s solution. Nutrient addition
was split into two applications of 1.5 l of aqueous solution; one half was added at the
start of the experiment and the second two weeks later. Since the topsoil had become
visibly drier by the second application of nutrients, infiltration was improved by wetting
the soil with 1.5 l of water before adding the nutrient solution. Control blocks received an
equivalent amount of water only.
23
Chapter 1
Harvest and chemical analyses
Before the first application of nutrients, blocks were clipped to 3 cm height and the harvested
biomass was weighed. Regrowth was harvested after 46 days and again after 172 days, and
dried to constant weight.
Biomass samples from control blocks and those receiving the complete nutrient dressing
were ground, and analysed for N and P concentrations after Kjeldahl digestion. Total N
and P concentrations in the digests was measured by means of a continuous flow injection
analyser (FIAStar, Foss Tecator, Höganäs, Sweden). A second subsample was extracted
with 0.5 M HCl and analysed on K, Ca and Mg concentrations, using atomic absorption
spectrometry (Hunt 1982) (SPECTRAA 240 FS, Varian AG, Zug, Switzerland).
Soil samples (2.8 cm diameter cores, top 10 cm soil) were taken at the start of the
experiments (one mixed sample of five cores per site), and at the second harvest (one core
per unfertilized control block; five replicates per site). The mixed samples were used for
K, Ca and Mg analyses; total C and N as well as extractable N and P were determined on
soil cores taken at the second harvest. Total C and N were measured on a dry combustion
analyzer (CN-2000, LECO Corp., St. Joseph, Minnesota, USA). Soil extractable PO42- was
determined by extraction of 5 g of fresh soil with 50 ml Bray-2 solution (Bray and Kurtz
1945). Exchangeable NH4+ and NO3- were determined by extraction of 10 g fresh soil in
50 ml 0.2 M KCl solution. Extraction was done within 12 h of collection of the soil cores.
KCl-extracts were acidified with 5% H2SO4 for conservation until analysis. Concentrations
of PO42-, NO3- and NH4+ in the extracts were measured colorimetrically using a continuous
flow injection analyser (FIAStar, Foss Tecator, Höganäs, Sweden). K, Ca and Mg were
measured by atomic absorption spectrometry from 1 M ammonium acetate extracts (Carter
1993) (SPECTRAA 240 FS, Varian AG, Zug, Switzerland).
Rainfall, temperature and relative humidity were recorded with tipping bucket rain gauges
and data loggers (HOBO RG3-M, HOBO H8 Pro, Onset Computer Corp., Bourne, MA,
USA) at two locations: one in the northern and one the southern part of the national park
at proximity of the experimental plots. During the wet season prior to the experiment, the
northern part had received approximately 630 mm of precipitation and the southern part
430 mm. From the start to the experiment until the first harvest on day 46, cumulative
rainfall was 30 mm in both parts of the national park. Between the first and the second
harvest on day 172, precipitation amounted to 480 mm in the north and 280 mm in the
south. During the experiment, mean daily temperature and mean daily relative humidity
were 24.9° C and 88% in the north, and 25.7° C and 79% in the south.
24
Nutrient limitation
Statistical analysis
Statistical analysis was carried out with JMP 6.0.3 (SAS Institute, Cary, USA). Soil and
vegetation characteristics (including concentrations and ratios of N, P, K, Ca and Mg) of
the five study sites were compared using one-way ANOVAs, with site as the fixed effect.
Differences in aboveground biomass production between nutrient treatments were tested
with one-way ANOVAs, with nutrient treatment as a fixed effect with 6 levels. Betweensite differences in concentrations and ratios of N, P, K, Ca and Mg in aboveground
biomass of control plots were tested with one-way ANOVAs, with site as a fixed effect. If
assumptions of normality and homoscedascity were not fulfilled, data were log or square
root transformed. Multiple comparisons were made with the Tukey-Kramer HSD test
(P < 0.05). Additionally, the effect of fertilizer treatment on nutrient concentrations in
aboveground biomass was tested for each site by comparing concentrations and ratios of
N, P, K, Ca and Mg between control plots and plots receiving N+P+cations, using a t-test
assuming unequal variances.
Results
Soil nutrients
Mean values of total N in the top 10 cm soil varied among sites by a factor of more than
four, from 0.24 mg g-1 in the tallgrass savanna plot to 1.07 mg g-1 in the forest gap (Table
1). The second highest N pool was in the secondary Acacia woodland, but this was only
60% of that in the forest gap. Variation among sites in the N pool paralleled variation
in the topsoil C pools (Table 1), and C:N ratios were similar at all sites (mean values
ranged from 14.0 to 14.7, no significant differences) except the paddock margin where they
were significantly lower (mean 11.4). Mean extractable inorganic N varied by a factor of
around two, with high values in the forest gap and the paddock margin (0.34 and 0.36 mg
g-1, respectively) and lowest values in the grazed patch (0.17 mg g-1; Table 1). The mean
concentrations of extractable P varied fivefold amongst sites, being highest in the paddock
margin (0.88 mg g-1), and 40% lower in the grazed patch; this in turn had significantly
higher P concentrations than the other three types of site (0.15-0.18 mg g-1).
25
Chapter 1
Table 1
Soil and vegetation characteristics of the five study sites in Saadani National Park
n
Forest gap
Tallgrass
savanna
Acacia
woodland
Grazed patch
Paddock
margin
Total N (mg g-1)
5
1.07 ± 0.10A
0.24 ± 0.01D
0.56 ± 0.03B
0.35 ± 0.03C
0.47 ± 0.03BC
Topsoil N pool (g m-2)
5
134 ± 13A
33 ± 1D
78 ± 5B
54 ± 6C
67 ± 4BC
Topsoil C pool (kg m-2)
5
1.98 ± 0.19A
0.46 ± 0.02D
1.11 ± 0.05B
0.75 ± 0.05C
0.76 ± 0.05C
Extractable N (g m-2)
5
0.34 ± 0.04AB
0.21 ± 0.01BC
0.24 ± 0.04ABC
0.17 ± 0.04C
0.36 ± 0.04A
Extractable P (g m-2)
5
0.18 ± 0.03C
0.17 ± 0.02C
0.15 ± 0.02C
0.57 ± 0.02B
0.88 ± 0.10A
Extractable N:P
5
2.18 ± 0.66A
1.26 ± 0.07A
1.57 ± 0.18A
0.29 ± 0.07B
0.43 ± 0.03B
Extractable K (g m-2)
1
10
5
5
13
12
Extractable Ca (g m-2)
1
743
31
130
49
53
Extractable Mg (g m-2)
1
36
-3
5
B
18
AB
12
AB
7
A
Bulk density (g cm )
10
1.26 ± 0.09
1.38 ± 0.12
1.38 ± 0.15
1.53 ± 0.15
1.42 ± 0.11A
Soil water content at day
0 (g g-1)
5
0.14 ± 0.02A
0.17 ± 0.01A
0.17 ± 0.04A
0.07 ± 0.01C
0.11 ± 0.01B
Soil water content at day
172 (g g-1)
5
0.09 ± 0.01A
0.11 ± 0.01A
0.09 ± 0.02A
0.02 ± 0.004C
0.06 ± 0.01B
Total aboveground
biomass at day 0 (g m-2)
30
616 ± 39AB
543 ± 22BC
499 ± 19C
484 ± 18C
671 ± 23A
Proportion of dead
biomass at day 0
5
0.76 ± 0.08A
0.34 ± 0.10BC
0.37 ± 0.08BC
0.28 ± 0.15C
0.52 ± 0.05B
Values are means ± standard errors.
Extractable N: extractable nitrate and ammonium (0.2 M KCl)
Extractable P: extractable phosphorus (Bray-2)
Extractable K, Ca, and Mg: extractable potassium, calcium and magnesium determined from a mixed sample
of 5 pooled soil cores (1 M ammonium acetate)
Values not sharing the same letter indicate significant differences between sites (Tukey-HSD, P < 0.05).
Nutrient limitation of biomass production
Biomass production in the control treatments ranged from 23 g m-2 in the grazed patches
to 100 g m-2 in the paddock margin at the first harvest (46 days), while the corresponding
yields at the second harvest (172 days) were 94 and 286 g m-2, respectively (Figure 2). At
the second harvest, when there was a marked growth response at all sites, the lowest mean
N concentrations in regrowth were in the tallgrass savanna, while the highest values were
in the forest gap and Acacia woodland (Table 2). Mean P concentrations ranged more than
threefold, being lowest in the Acacia woodland and highest in the paddock margin. As a
result, N:P ratios in the regrowth on control plots ranged widely, from 4.8 in the paddock
margin to 16.7 in Acacia woodland (2nd harvest, Table 2).
26
Nutrient limitation
1st harvest (day 46)
Forest gap
600
200
100
a
a
a a
a
200
a
0
300
b
cd
c
cd
a
400
200
d
a
a
b
b
b
0
Acacia woodland
600
a
200
100
c
abc bc
400
a
ab
b
b
0
Grazed patch
600
200
400
100
200
a
a
a
a
a
a
0
300
a
200
bc
0
300
b
b
a
cd
c
b
cd
d
0
Paddock margin
a
a
a
a
200
b
ab
600
400
b
c
bc a
c
P
N+P
N
N+P+cations
cations
N+P
0
P
0
N
200
control
100
control
b
cations
Regrowth (g m-2)
b
600
0
300
b ab ab
a
0
Tallgrass savanna
a
200
100
ab
400
N+P+cations
300
2nd harvest (day 172)
Figure 2 Aboveground biomass production of five sites in Saadani National Park as affected by
the addition of N, P, cations (= all nutrients other than N and P) and combinations thereof. Bars
show mean aboveground biomass harvested after 46 and again after 172 days (± SE; n = 5). Bars
not sharing the same letters are significantly different from each other (Tukey-HSD, P < 0.05).
27
Chapter 1
In the forest gap plot, none of the nutrient treatments had a significant effect upon regrowth
at the first harvest but production at the second harvest was increased by the combined
addition of N, P and cations (Figure 2). The increase in biomass production caused by
the addition of N+P was on average equal to the increase observed by the N+P+cations
treatment, but it was statistically not significantly different from the control due to the large
variation (Figure 2). Strictly speaking, these results therefore lead to the conclusion that
growth was co-limited by N, P and cations; however, in view of the high average response to
N+P we interpret them as indicating N and P co-limitation. Tissue N and P concentrations
both significantly increased at the first harvest in response to the full nutrient treatment, but
not at the second harvest (Table 2).
In the tallgrass savanna site, production at the first harvest was increased in the N+P
treatment (Figure 2), while N+P+cations increased growth even more. In contrast, at the
second harvest N addition alone was sufficient to increase production significantly, and
there was no additional effect of adding either P or P+cations. Thus, production in the first
period was co-limited by N and P but in the second period only by N (Figure 2).
In the Acacia woodland plot, only the addition of both N and P (i.e. N+P and N+P+cations)
increased herbaceous biomass production significantly, indicating that growth was colimited by these nutrients (Figure 2). Nutrient analyses of aboveground biomass show that
only P concentrations increased significantly under full nutrient supply compared to the
controls (Table 2). The Acacia woodland had the highest N:P ratio in biomass (Table 2).
In the grazed patch, none of the nutrient treatments had any effect upon yield at the first
harvest; however, at the second harvest there was a significant increase in aboveground
biomass in the N+P treatment, and an even greater increase in the N+P+cations treatment
(Figure 2). These results suggest that regrowth was co-limited by N and P, but only at the
second harvest. Concentrations of N and K in the biomass were higher in the N+P+cations
treatment at the first harvest, while N and P concentrations were higher at the second
harvest (Table 2).
In the paddock margin plot, N addition significantly increased productivity, but there was
no extra effect when P or P+cations were also added; thus, growth was clearly N-limited
(Figure 2). The full nutrient addition did increase tissue N concentrations, though the effect
was only significant at the first harvest (Table 2).
28
N+P+cations
N+P+cations
10.9 ± 0.6B 16.7 ± 1.2**
Control
Tallgrass savanna
14.0 ± 0.8A
Control
16.2 ± 0.8
N+P+cations
Acacia woodland
N+P+ cations
Control N+P+cations
Paddock margin
14.8 ± 1.3A 22.8 ± 0.6*** 10.2 ± 0.6B 14.8 ± 0.5**
Control
Grazed patch
1.4 ± 0.1
B
29
5.0 ± 0.8
2.1 ± 0.3A
B
N:K
K:P
10.2 ± 0.8
0.97 ± 0.23
12.7 ± 2.5
10.6 ± 0.4A
B
AB
12.5 ± 2.3
8.6 ± 1.7
BC
0.91 ± 0.16
BC
6.8 ± 0.3C
12.5 ± 1.4
AB
0.8 ± 0.0CD
B
9.7 ± 1.2
2.7 ± 0.0
A
4.1 ± 0.2B
8.8 ± 0.3B
12.8 ± 1.4
C
A
1.4 ± 0.1AB
A
16.9 ± 0.6
2.3 ± 0.1
A
6.5 ± 0.4A
10.4 ± 0.9BC
8.4 ± 0.4
9.2 ± 0.2
1.5 ± 0.1
13.6 ± 0.6**
2.7 ± 0.2
5.2 ± 0.4
10.1 ± 0.7
10.6 ± 0.2
1.3 ± 0.0
13.6 ± 0.4***
2.8 ± 0.1
4.2 ± 0.3
7.7 ± 1.9
16.7 ± 0.5A 8.8 ± 0.5***
10.0 ± 0.4BC 7.1 ± 0.1**
B
9.7 ± 0.3AB 12.2 ± 0.3***
8.1 ± 1.0
AB
1.2 ± 0.1BC
B
9.1 ± 0.1
2.7 ± 0.1
A
4.7 ± 0.2B
13.0 ± 1.4AB 17.8 ± 0.3*
8.8 ± 0.5B
9.5 ± 1.2
A
AB
0.6 ± 0.1D
5.5 ± 0.5
C
1.5 ± 0.1
B
2.2 ± 0.2C
9.6 ± 0.4
9.7 ± 1.0
0.8 ± 0.1
7.9 ± 0.2*
2.0 ± 0.2
2.6 ± 0.4
17.2 ± 1.4A 18.1 ± 1.5
4.8 ± 0.5C
4.8 ± 0.5
1.26 ± 0.34 0.53 ± 0.02 0.96 ± 0.05*** 0.98 ± 0.07 1.72 ± 0.04*** 1.90 ± 0.11 2.01 ± 0.18
7.4 ± 0.5
7.9 ± 0.5*
1.0 ± 0.0**
8.3 ± 0.8
2.8 ± 0.2
3.6 ± 0.4
14.1 ± 0.5AB 16.0 ± 0.5*
Values are means (± SE) of five plots.
*P < 0.05, **P < 0.01, ***P < 0.001 indicate significant differences between control plots and those receiving N+P+cations.
Values of controls not sharing the same letter indicate significant differences between sites (Tukey-HSD, P < 0.05)
N:P
Phosphorus (mg g ) 0.94 ± 0.13
Nitrogen (mg g-1)
Day 172
3.2 ± 0.3*
15.0 ± 0.8
N:P
7.5 ± 1.1
15.1 ± 1.0
A
-1
Magnesium (mg g )
1.9 ± 0.2
3.2 ± 0.5
3.6 ± 0.4B
Calcium (mg g-1)
-1
5.9 ± 0.8
6.5 ± 0.9C
Potassium (mg g-1)
Phosphorus (mg g-1) 0.86 ± 0.02C 1.19 ± 0.03*** 1.21 ± 0.18BC 2.04 ± 0.10** 0.83 ± 0.04C 1.20 ± 0.07** 1.61 ± 0.12AB 1.68 ± 0.04 1.89 ± 0.16A 1.88 ± 0.06
12.8 ± 0.5AB 17.9 ± 0.1**
Control
Forest gap
Nutrient concentrations and ratios in aboveground biomass at the five study sites
Nitrogen (mg g-1)
Day 46
Table 2
Nutrient limitation
Chapter 1
Table 3
Types of nutrient limitation in tropical savanna and forest ecosystems
Vegetation type
Tropical forests
Secondary tropical forest
Cerrado sensu stricto
Old secondary dry tropical
forest
Gap in remnant forest (FG)
Young secondary dry tropical
forest
Succession after removal of
mature tropical forest
Herbaceous cover in secondary
dry forest
Humid tallgrass savanna (TG)
Mesic savanna, natural pastures
Semi-arid savanna
Hyparrhenia dominant
grassland
Hyparrhenia dominant
grassland
Sahelian C4 grasslands
Humid savanna, native pastures
Flooded savanna
Secondary savanna
Trachypogon savanna
Trachypogon savanna
Rangeland
Humid savanna, secondary
Acacia woodland (AW)
Semi-arid savanna under Acacia
tortilis canopy
Overgrazed Sahelian pasture
dominated by legume Zornia sp.
Humid savanna, grazed patches
(GP)
Grazing lawns, humid Guinea
savanna
Humid savanna, soils derived
from bomas (PM)
Semi-arid savanna, soils derived
from bomas
Nutrient
limitation
P
Study
type
litter NUE
N:P
ratio
Compares
best to
Forest
Source
(1)
P
N:P ratios
28.8
Forest
(2)
P
P
litter NUE
F
18
Forest
Forest
(3)
(4)+(5)
NPcations
F
12.515.0
Yucatan
NP
F
FG
(4)+(5)
Amazon,
Brazil
Amazon,
Brazil
Tanzania
N or NP
modelling
FG
(6)
FG
(7)
Location
62 sites
worldwide
Amazon,
Brazil
Brazil
Yucatan
Tanzania
TS
NP
F
10.2
F
8.6-9.7
Tanzania
Tanzania
Zambia
N and NP
(NPcations)
N (NP)
N (NP)
N (NP)
F
F
F
6
D
TG
TG
TG
(8)
(9)
(10)
Kenya
N (NP)
F
8.2
TG
(11)
TG
TG
TG
TG
TG
(12)
(13)
(14)
(15)
(16)
TG
TG
(17)
(18)
TS
AW
(9)
AW
(12)
N? (NP)A
NP
NP (NPKS)
N (NPK)
N and P
(NP, NPK)
Venezuela
NP?B
South Africa N (NP)
Tanzania
NP
F
F
F
Tanzania
P
F
16.716.9
12D
Mali
PC
F
~20C
Tanzania
9.1-9.7
Cameroon
NP
F
(NPcations)
N
N:P ratios
Tanzania
N
F
4.8-5.5
Kenya
N (NP)
F
5.2
Mali
Australia
Venezuela
Venezuela
Venezuela
30
F
F
F
F
F
14.8
9.8
11.0
5.8
TS
TS
GP/PM
(19)
TS
PM
(20)
Nutrient limitation
Legend Table 3:
Nutrient limitation: the main limiting nutrient(s) is/are reported, in brackets are nutrient combinations which further increased biomass
production compared to the main limiting nutrient(s).
Study type: F, factorial fertilization experiment; litter NUE, nutrient use efficiencies as calculated from litterfall; N:P ratios, limitation
assessed according to relative abundance of N and P in plant tissue; modelling, model of changes in nutrient pools upon forest clearing
based on the dataset from a study site
N:P ratios: calculated from N and P concentrations in foliage of several tree species in the case of forests, from concentrations in total
aboveground herbaceous biomass, or as weighed average from concentrations in aboveground biomass of one or several herbaceous
species if they represent at least 75% of total biomass and their relative abundance is known in order to prevent a bias due to the large
variation of N:P ratios among plant species (cf. Güsewell and Koerselman 2002).
A
effect of N addition alone was not tested, since all plots receiving varying levels of N fertilizer were given a basic P dressing first
B
neither N nor P addition increased aboveground biomass in cut plots, joint addition of both nutrients was not tested
C
native vegetation was removed and pure stands of two grass and four legume species were sown, N:P ratios of legume species on
control plots averaged 19.9 and the N:P ratio of one grass species was 20.7 (no data available for the second grass species)
D
N:P ratios were determined from young fully expanded leaves of the dominant grass species
Compares best to: indicates the vegetation types from the conceptual model in Figure 1B to which the respective study site is most similar
(based on the influence of herbivory, fire and N2-fixation). FG: forest gaps, TG: tallgrass savanna, AW: Acacia woodland, GP: grazed
patches, PM: paddock margin.
Source: (TS) this study; (1) Vitousek (1984), (2) Markewitz et al. (2004), (3) Nardoto et al. (2006), (4) Campo and Vazquez-Yanes
(2004), (5) Solis and Campo (2004), (6) Herbert et al. (2003), (7) Davidson et al. (2004), (8) Walker (1969), (9) Ludwig et al. (2001),
(10) Brockington (1961), (11) Keya (1973), (12) Penning de Vries et al. (1980), (13) Norman (1966), (14) Sarmiento et al. (2006),
(15) Barger et al. (2002), (16) Garcia-Miragaya et al. (1983), (17) Medina (1978), (18) Snyman (2002), (19) Verweij et al. (2006),
(20) Augustine et al. (2003)
Discussion
This study was designed to test predictions about long-term effects of fire, herbivory and N2fixation on soil nitrogen, N:P-stoichiometry and the type of nutrient limitation (Figure 1).
This research was conducted at five sites representative for the studied vegetation types,
with only within-site replication. To get a more general impression of the type of nutrient
limitation in these savanna vegetation types, we compared our results with data from studies
in other tropical savannas and forested areas which were similar to our sites with regards to
the influence of fire, herbivory and N2-fixation (Table 3).
The main trends that emerge from these comparisons are generally consistent with our
conceptual model (Figure 1B). They include: (1) tallgrass savanna or comparable vegetation
types are N-limited or NP-co-limited; (2) areas receiving large amounts of excreta from
wild and domestic herbivores (grazing lawns and bomas) are N-limited; (3) vegetation
growing in areas of cleared and burned tropical forest is mainly NP-co-limited; (4) mature
tropical forests appear to be primarily limited by P (Table 3); (5) N2-fixation tends to cause
P-limitation (Table 3). This last point was supported by two published studies, but we
31
Chapter 1
found co-limitation by N and P in the Acacia woodland plot. However, the values for N:
P ratios suggest that N2-fixation by Acacia did cause a shift of N:P-stoichiometry in the
direction of P-limitation (Table 2).
In all our sites, aboveground production was limited by N or co-limited by N and P, but
there was no evidence of primary limitation by any other nutrients (Figure 2). Indeed, the
majority of studies on nutrient limitation in savanna ecosystems have assumed that no
nutrient other than N and P limits productivity, though without testing this assumption
explicitly. It is worth noting that nutrients were not always the factor limiting growth
in our experiment; thus, in the first period the nutrient treatments had no effect upon
aboveground production in the forest gap nor in the grazed patch, presumably because
of low water availability. The herbaceous vegetation in the forest gap is likely to have
experienced drought because of water uptake by surrounding trees (Bourliere and Hadley
1983), while the grazed patch in the southern part of the study area received significantly
less rain than the other sites (reflected in a lower soil water content at the start of the
experiment; Table 1). Thus, although nutrients have been shown to limit plant production
at annual levels of precipitation as low as 200 mm yr-1 (Penning de Vries et al. 1980), the
large seasonal fluctuations in water availability characteristic to savannas may temporally
result in water limitation.
Various lines of evidence suggest that our tallgrass site, as well as many other tallgrass
savannas, are at the boundary between N-limitation and NP-co-limitation. First, consistent
with some other studies (Table 3), production at our tallgrass site was limited by N during
the second period, but by both N and P during the first period. Second, in comparable types
of vegetation where N alone stimulated production, the joint addition of N and P always
caused a further increase in growth (Table 3). These findings demonstrate that besides N
availability also P availability is low in tallgrass savanna and comparable vegetation. There
are various possible explanations for low levels of P in this habitat: e.g. losses of P through
fire (though being clearly less than for N and mainly occurring in particulate form; Cook
1994; Laclau et al. 2002), or the leaching of P from ancient highly weathered tropical soils
(Lambers et al. 2008).
At the tallgrass site, the biomass at the first harvest was greater in the N+P+cations
treatment than in the N+P treatment. The fact that the N+P+cations treatment increased
concentrations of potassium in regrowth but not of calcium and magnesium (Table 2)
suggests that K became limiting when N and P were supplied. Indeed, the losses of K
during combustion are usually found to be higher than those of Ca and Mg (Cook 1994;
32
Nutrient limitation
Laclau et al. 2002; Van de Vijver et al. 1999). The use of the tallgrass site as a cattle pasture
between 1954 and 2000 could have contrasting effects upon N and P conditions: on the
one hand it could have increased losses of N and P because more nutrients were ingested
in this area than were returned in excreta; on the other hand, by reducing the intensity and
frequency of fires, grazing could also have reduced N losses relative to those before the area
was grazed (Hobbs et al. 1991). However, since concentrations of total N and extractable
P in the soil of tallgrass sites in the former ranch area were similar to those in an area to
the south of our study sites that had never been grazed by cattle (Cech et al., unpublished
data), we conclude that any such effects must have been rather small.
In contrast to the tallgrass pastures, cattle had a large effect upon soils in the vicinity of
paddocks, which received high amounts of dung and urine for several decades. Soil N
pool, extractable N and extractable P were higher in the paddock margin samples than in
the tallgrass savanna samples, the difference being particularly pronounced for extractable
P (Table 1). As predicted, biomass production in the paddock margin plot was limited
by N (Figure 1), with additional P producing no extra effect. A similar concentration of
nutrients by livestock has been reported for a semi-arid savanna in Kenya, with the effects
persisting for decades (Augustine 2003).
The grazed patches in our study area were not used continuously by wild herbivores,
but were abandoned if the grass grew above a critical height; such an area would then
not be grazed again until the accumulated biomass was removed by fire (Cech, personal
observation). This pattern of habitat use by wild herbivores, which has also been observed
elsewhere (Archibald et al. 2005), might explain why the grazed patch was only moderately
enriched compared to tallgrass savanna of the kind from which our grazed patch probably
developed (based on similarity in plant species). In the experiment, biomass production
in the grazed patch was co-limited by N and P at the second harvest. There is rather little
information on nutrient availabilities in grazing lawns (as defined by McNaugthon 1984)
and in temporarily grazed patches, and we know of no other studies investigating nutrient
limitation in such areas. However, McIvor et al. (2005) report that the soil in patches of
lawn formed by cattle was enriched in P but less so in N. This is consistent with our finding
that soil extractable P was enriched more in the grazed patch relative to the tallgrass site
than were total and extractable N (Table 1). Data from a recent study of grazing lawns
in Cameroon indicate N:P ratios in aboveground biomass around 6 (Verweij et al. 2006),
which is lower than what we observed in the grazed patch (9-10, Table 2, Table 3). However,
those grazing lawns are continuously maintained by herbivores and may thus be expected
to be more heavily impacted than our grazed patches.
33
Chapter 1
In the Acacia woodland plot, biomass production was co-limited by N and P during both
growth periods. Although we did not observe P limitation, as reported for two other
savannas with N2-fixing legumes (Table 3), the N:P ratio in the vegetation of the control
plots suggests that growth limitation in our Acacia woodland was closer to P-limitation
than in our other NP-co-limited sites (Table 2). N2-fixing plants often have a higher P
requirement than other plants (Pate 1986), and several studies have shown N2-fixation in
legumes to be limited by P-availability (Binkley et al. 2003; Crews 1993; Israel 1987; Perreijn
2002). N2-fixing plants increase the overall availability of N, which probably increases the
demand and uptake of P by other plants as well. The high P demand of Acacia zanzibarica
trees and the increased P demand by other plants may thus have reduced soil P availability,
as has already been reported for another woody legume (Binkley 1997). Thus, the high N:P
ratio in herbaceous vegetation in the Acacia woodland (Table 2) was probably due to the
combined effect of increased N and reduced P. N2-fixation may thus have a stronger effect
on N:P-stoichiometry relative to its effect on N-accumulation than was hypothesized in
Figure 1. Additionally, the herbaceous cover of the Acacia woodland is burnt occasionally
(Cech, personal observation), which may slow down the accumulation of N in the soil.
In the forest gap, which was selected as the reference for the other vegetation types, biomass
production tended to be co-limited by N and P during the second growth period (Figure
2). Undisturbed tropical forests are thought to be most commonly limited by P because of
the advanced weathering of soils (Vitousek 1984), and there is some experimental evidence
supporting this view (Table 3; Elser et al. 2007), although at the level of individual trees
the picture might be more complex with some trees limited by N and others by P (Perreijn
2002). We expected that, starting from P-limited undisturbed forest conditions, burning
would lead in the direction of N-limitation (cf. Bustamante et al. 2006; Kauffman et al.
1995) and hence frequently burned sites (like tall grass savanna) would be N-limited, and
less requently burned sites (like forest gaps) would have an intermediate position between
P-limited and N-limited conditions. The responses of re-growth to nutrient addition in the
forest gap were thus in line with our hypothesis that where fire is infrequent the nutrient
balance lies in the range of NP-co-limitation. A relatively high total-N:total-P ratio of 19
in the soil of a remnant forest near our forest gaps indicated that the original forest might
indeed have been P-limited (Cech, unpublished data). The shift from P-limitation towards
N-limitation following removal of tropical forest is also supported by other studies (see
Table 3).
34
Nutrient limitation
20
a
FG
AW
16
N:P ratio in vegetation
12
TG
GP
8
PM
4
b
AW
16
12
FG
GP
TG
8
PM
4
0
30
60
90
120
150
-2
Soil N pool (g m )
Figure 3 Total soil N in topsoil and N:P ratios in aboveground biomass of five sites in Saadani
National Park: forest gap (FG), tallgrass savanna (TG), secondary Acacia woodland (AW), grazed
patch (GP), paddock margin (PM), at the first harvest (a) and the second harvest (b). Open symbols
indicate that N:P ratio in vegetation may not be a good indicator of nutrient limitation, because
water was likely the limiting resource. Data is shown as mean per site and error bars represent
standard errors (n = 5).
Our conceptual framework of the effects of fire, herbivory and N2-fixation on soil N pool
and N:P-stoichiometry (Figure 1) was supported by a good match between the hypothesized
arrangement of the investigated vegetation types, and measured soil N pools and N:Pstoichiometry (represented by the N:P ratio in the aboveground vegetation) (compare
Figures 1B and 3). In the case of the Acacia woodland and the grazed patch measured N:P
ratios deviated from our predictions; we have discussed possible causes for these deviations
above.
The data on soil N pools also support the conceptual framework in various ways. First, the
site that has probably burnt most frequently - the tallgrass savanna (Frost and Robertson
1987; Cech, personal observation) - had the lowest soil N pool (Figure 3, Table 1). This
35
Chapter 1
20
N:P ratio in vegetation
18
16
14
12
10
8
6
4
N
N&P
P
Limiting nutrient
Figure 4 N:P ratios versus type of limitation in aboveground herbaceous vegetation of tropical
savannas. N:P ratios are from unfertilized control plots and type of limitation was determined by
fertilization experiments (data from Table 3). Open circles show data for which N:P ratios represents
at least 75% of aboveground herbaceous biomass, solid circles show data from Ludwig et al. (2001)
who measured N:P ratios in young fully expanded leaves of the dominant grass species.
result is in line with lower soil N reported for long-term burning experiments (Fynn et al.
2003; Ojima et al. 1994). Secondly, the localized return of dung and urine by domestic
and wild herbivores increased soil N in both the grazed patch and the paddock margin.
Enrichment of soil N by domestic herbivores has also been demonstrated in semi-arid
savanna (Augustine 2003). Thirdly, the secondary Acacia woodland had significantly higher
amounts of soil N than the tallgrass savanna, the grazed patch or the paddock margin
(Table 1, Figure 3). Several studies report that woody legumes invading grasslands increase
total soil N and N availability, these effects being attributed at least partly to their ability
to fix atmospheric N2 (Archer 2001; Geesing et al. 2000; Hagos and Smit 2005; SchererLorenzen et al. 2007; Stock et al. 1995). Finally, the forest gap had the largest soil N pool
(Table 1), which may be explained by low fire frequency and possibly by the N input
from tree litter (Hudak et al. 2003). The correlation between soil N and C pools (Table 1)
suggests that the loss and accumulation of N is closely bound to the loss and accumulation
of organic matter.
36
Nutrient limitation
The ratio of N:P in the above-ground herbaceous vegetation has been used to assess
nutrient limitation in temperate ecosystems (Güsewell 2004; Koerselman and Meuleman
1996; Olde Venterink et al. 2003). Koerselman and Meuleman (1996) proposed that at N:P
ratios below 14, productivity was limited by N, between 14-16 production was co-limited
by N and P, and above 16 limited by P. Whether NP-co-limitation can be separated from
P-limitation based on N:P ratios is not fully clear (see Olde Venterink et al. 2003), but the
level of < c. 14 for N-limitation is consistent and supported by other studies in temperate
regions (Güsewell 2004; Olde Venterink et al. 2003). For a dry savanna, Ludwig et al. (2001)
reported variation in N:P ratios among growing seasons, and the results of a fertilization
experiment did not support the predicted type of limitation based on critical values of N:P
from temperate regions. They suggested that the critical N:P values for the boundaries
between N-limitation and NP-co-limitation should be at 6 for savanna vegetations, and that
between NP-co-limitation and P-limitation at 12. As a possible explanation, they suggested
the higher N use efficiency of C4 grasses compared to C3 plants. However, it should be
noted that N:P ratios reported in the study of Ludwig et al. (2001) were measured from
young, fully expanded leaves of the dominant grass species and not from aboveground
biomass. From our data, we suggest that the critical N:P ratio reflecting the transition from
N-limitation to NP-co-limitation lies between 8.6-10.0 (cf. type of nutrient limitation as
determined by fertilization, and N:P ratios of the control plots for the plots not limited by
water; Figure 2 and Table 2). If P-limitation and NP-co-limitation can be separated by N:P
ratios, the critical boundary value should be higher than 16.9. These ranges come close to
the boundaries values of 6.7 and 20.4 given by Penning de Vries et al. (1980) for sahelian
grasslands. Considering the data available from literature, we conclude that the critical N:P
value indicating N-limitation in tropical savannas is probably < c. 9; this value is less than
the < 14 determined for temperate regions, but more than the < 6 mentioned by Ludwig
et al. (2001) (Figure 4). We agree with Ludwig et al. (2001) that the difference compared
with temperate ecosystems is probably due to the very high N use efficiency of tropical
C4 grasses (Le Roux and Mordelet 1995; Sage and Pearcy 1987).
Acknowledgements
This study was financed by the Swiss National Science Foundation grant No. 2-7750204. We thank Prof. S. L. S. Maganga, Markus Schneider-Mmary, and the authorities of
Saadani National Park for their logistical support in Tanzania, and Benjamin Donald,
John Williams and Hamis Williams for their assistance in the field. We also acknowledge
comments provided by D. A. Frank and two anonymous reviewers that helped to improve
this manuscript.
37
Chapter 1
Literature cited
Aranibar JN, Otter L, Macko SA, Feral CJW, Epstein HE, Dowty PR, Eckardt F, Shugart HH,
Swap RJ (2004) Nitrogen cycling in the soil-plant system along a precipitation gradient in the
Kalahari sands. Global Change Biology 10:359-373
Archer S (1995) Tree-grass dynamics in a Prosopis-thornscrub savanna parkland: Reconstructing
the past and predicting the future. Ecoscience 2:83-99
Archer S (2001) Trees in grasslands: biogeochemical consequences of woody plant expansion.
In: Schulze E-D, Harrison SP, Heimann M, Holland EA, Lloyd J, Prentice IC, Schimel D (eds)
Global biogeochemical cycles in the climate system. Academic Press, San Diego, pp 115-137
Archibald S, Bond WJ, Stock WD, Fairbanks DHK (2005) Shaping the landscape: fire-grazer
interactions in an African savanna. Ecological Applications 15:96-109
Augustine DJ (2003) Long-term, livestock-mediated redistribution of nitrogen and phosphorus in
an East African savanna. Journal of Applied Ecology 40:137-149
Augustine DJ, McNaughton SJ, Frank DA (2003) Feedbacks between soil nutrients and large
herbivores in a managed savanna ecosystem. Ecological Applications 13:1325-1337
Barger NN, D‘Antonio CM, Ghneim T, Brink K, Cuevas E (2002) Nutrient limitation to primary
productivity in a secondary savanna in Venezuela. Biotropica 34:493-501
Bell RHV (1982) The effect of soil nutrient availability on community structure in African
ecosystems. In: Huntley BJ, Walker BH (eds) Ecology of tropical savannas. Springer, Berlin,
pp 192-216
Binkley D (1997) Bioassays of the influence of Eucalyptus saligna and Albizia falcataria on soil
nutrient supply and limitation. Forest Ecology and Management 91:229-234
Binkley D, Senock R, Cromack K (2003) Phosphorus limitation on nitrogen fixation by Facaltaria
seedlings. Forest Ecology and Management 186:171-176
Bond WJ, Woodward FI, Midgley GF (2005) The global distribution of ecosystems in a world
without fire. New Phytologist 165:525-537
Bourliere F, Hadley M (1983) Present day savannas: an overview. In: Bourliere F (ed) Tropical
Savannas. Elsevier, Amsterdam, pp 1-18
Bray RH, Kurtz LT (1945) Determination of total, organic and available forms of phosphorus in
soils. Soil Science 59:39-45
Brockington NR (1961) Studies of the growth of a Hyparrhenia-dominant grassland in Northern
Rhodesia. II. Fertilizer responses. III. The effect of fire. Journal of the British Grassland Society
16:54-65
Brown JR, Archer S (1989) Woody plant invasion of grasslands: establishment of honey mesquite
(Prosopis glandulosa var. glandulosa) on sites differing in herbaceous biomass and grazing history.
Oecologia 80:19-26
Bustamante MMC, Medina E, Asner GP, Nardoto GB, Garcia-Montiel DC (2006) Nitrogen cycling
in tropical and temperate savannas. Biogeochemistry 79:209-237
Campo J, Vazquez-Yanes C (2004) Effects of nutrient limitation on aboveground carbon dynamics
during tropical dry forest regeneration in Yucatan, Mexico. Ecosystems 7:311-319
Carter MR (1993) Soil sampling and methods of analysis. Lewis, Boca Raton
Cook GD (1994) The fate of nutrients during fires in a tropical savanna. Australian Journal of
Ecology 19:359-365
Cramer MD, Chimphango SBM, Van Cauter A, Waldram MS, Bond W (2007) Grass competition
induces N2 fixation in some species of African Acacia. Journal of Ecology 95:1123-1133
38
Nutrient limitation
Crews TE (1993) Phosphorus regulation of nitrogen fixation in a traditional Mexican agroecosystem.
Biogeochemistry 21:141-166
Davidson EA, de Carvalho CJR, Vieira ICG, Figueiredo RD, Moutinho P, Ishida FY, dos Santos
MTP, Guerrero JB, Kalif K, Saba RT (2004) Nitrogen and phosphorus limitation of biomass
growth in a tropical secondary forest. Ecological Applications 14:S150-S163
Edwards PJ, Hollis S (1982) The distribution of excreta on New Forest grassland used by cattle,
ponies and deer. Journal of Applied Ecology 19:953-964
Elser JJ, Bracken MES, Cleland EE, Gruner DS, Harpole WS, Hillebrand H, Ngai JT, Seabloom
EW, Shurin JB, Smith JE (2007) Global analysis of nitrogen and phosphorus limitation of primary
producers in freshwater, marine and terrestrial ecosystems. Ecology Letters 10:1135-1142
Frank DA, Zhang YM (1997) Ammonia volatilization from a seasonally and spatially variable
grazed grassland: Yellowstone National Park. Biogeochemistry 36:189-203
Frost PGH, Robertson F (1987) The ecological effects of fire in savannas. In: Walker BH (ed)
Determinants of tropical savannas. IRL Press, Oxford, pp 93-140
Fynn RWS, Haynes RJ, O‘Connor TG (2003) Burning causes long-term changes in soil organic
matter content of a South African grassland. Soil Biology & Biochemistry 35:677-687
Garcia-Miragaya J, San Jose JJ, Hernandez JI (1983) Effect of added nitrogen, phosphorus and
potassium on above-ground biomass production and nutrient content of Trachypogon savanna
grasses. Tropical Ecology 24:33-42
Geesing D, Felker P, Bingham RL (2000) Influence of mesquite (Prosopis glandulosa) on soil
nitrogen and carbon development: Implications for global carbon sequestration. Journal of Arid
Environments 46:157-180
Güsewell S (2004) N:P ratios in terrestrial plants: variation and functional significance. New
Phytologist 164:243-266
Güsewell S, Koerselman W (2002) Variation in nitrogen and phosphorus concentrations of wetland
plants. Perspectives in Plant Ecology, Evolution and Systematics 5:37-61
Hagos MG, Smit GN (2005) Soil enrichment by Acacia mellifera subsp. detinens on nutrient poor
sandy soil in a semi-arid southern African savanna. Journal of Arid Environments 61:47-59
Herbert DA, Williams M, Rastetter EB (2003) A model analysis of N and P limitation on
carbon accumulation in Amazonian secondary forest after alternate land-use abandonment.
Biogeochemistry 65:121-150
Hobbs NT, Schimel DS, Owensby CE, Ojima DS (1991) Fire and grazing in the tallgrass prairie:
contingent effects on nitrogen budgets. Ecology 72:1374-1382
Högberg P (1986) Nitrogen-fixation and nutrient relations in savanna woodland trees (Tanzania).
Journal of Applied Ecology 23:675-688
Holdo RM, Holt RD, Coughenour MB, Ritchie ME (2007) Plant productivity and soil nitrogen as
a function of grazing, migration and fire in an African savanna. Journal of Ecology 95:115-128
Hudak AT (1999) Rangeland mismanagement in South Africa: failure to apply ecological knowledge.
Human Ecology 27:55-78
Hudak AT, Wessman CA, Seastedt TR (2003) Woody overstorey effects on soil carbon and nitrogen
pools in South African savanna. Austral Ecology 28:173-181
Hunt J (1982) Dilute hydrochloric acid extraction of plant material for routine cation analysis.
Communications in Soil Science and Plant Analysis 13:49-55
Huntley BJ (1982) Southern African savannas. In: Huntley BJ, Walker BH (eds) Ecology of tropical
savannas. Springer, Berlin, pp 101-119
Israel DW (1987) Investigation of the role of phosphorus in symbiotic dinitrogen fixation. Plant
Physiology 84:835-840
39
Chapter 1
Jewell PL, Kauferle D, Gusewell S, Berry NR, Kreuzer M, Edwards PJ (2007) Redistribution of
phosphorus by mountain pasture in cattle on a traditional the Alps. Agriculture Ecosystems &
Environment 122:377-386
Kauffman JB, Cummings DL, Ward DE, Babbitt R (1995) Fire in the Brazilian Amazon: 1.
Biomass, nutrient pools, and losses in slashed primary forests. Oecologia 104:397-408
Keya N (1973) The effect of NP fertilizers on the production of Hyparrhenia grassland. East African
Agricultural and Forestry Journal 39:195-200
Klötzli F (1980) Analysis of species oscillations in tropical grasslands in Tanzania due to
management and weather conditions. Phytocoenologia 8:13-33
Koerselman W, Meuleman AFM (1996) The vegetation N:P ratio: a new tool to detect the nature
of nutrient limitation. Journal of Applied Ecology 33:1441-1450
Laclau JP, Sama-Poumba W, Nzila JD, Bouillet JP, Ranger J (2002) Biomass and nutrient dynamics
in a littoral savanna subjected to annual fires in Congo. Acta Oecologica 23:41-50
Lambers H, Raven JA, Shaver GR, Smith SE (2008) Plant nutrient-acquisition strategies change
with soil age. Trends in Ecology and Evolution 23:95-103
Lamoot I, Callebaut J, Degezelle T, Demeulnaere E, Laquière J, Vandenberghe C, Hoffmann M
(2004) Eliminative behaviour of free-ranging horses: do they show latrine behaviour or do they
defecate where they graze? Applied Animal Behaviour Science 86:105-121
Le Roux X, Mordelet P (1995) Leaf and canopy CO2 assimilation in a West African humid savanna
during the early growing season. Journal of Tropical Ecology 11:529-545
Ludwig F, de Kroon H, Prins HHT, Berendse F (2001) Effects of nutrients and shade on tree-grass
interactions in an East African savanna. Journal of Vegetation Science 12:579-588
Markewitz D, Davidson E, Moutinho P, Nepstad D (2004) Nutrient loss and redistribution after
forest clearing on a highly weathered soil in Amazonia. Ecological Applications 14:S177-S199
McIvor JG, McIntyre S, Saeli I, Hodgkinson JJ (2005) Patch dynamics in grazed subtropical native
pastures in south-east Queensland. Austral Ecology 30:445-464
McNaughton SJ (1988) Mineral nutrition and spatial concentrations of African ungulates. Nature
334:343-345
McNaughton SJ (1990) Mineral nutrition and seasonal movements of African migratory ungulates.
Nature 345:613-615
McNaugthon SJ (1984) Grazing lawns: Animals in herds, plant form and coevolution. The
American Naturalist 124:863-886
Medina E (1978) Nutrient balance and organic matter production in the Trachypogon savannas of
Venezuela. Tropical Agriculture 55:243-253
Medina E (1987) Nutrients: requirements, conservation and cycles in the herbaceous layer. In:
Walker BH (ed) Determinants of tropical savannas. IRL Press, Oxford, pp 39-65
Nardoto GB, Bustamante MMD, Pinto AS, Klink CA (2006) Nutrient use efficiency at ecosystem
and species level in savanna areas of Central Brazil and impacts of fire. Journal of Tropical
Ecology 22:191-201
Norman MJT (1966) Katherine Research Station 1956-64: A review of published work. CSIRO,
Melbourne
Ojima DS, Schimel DS, Parton WJ, Owensby CE (1994) Long-term and short-term effects of fire
on nitrogen cycling in tallgrass prairie. Biogeochemistry 24:67-84
Olde Venterink H, Wassen MJ, Verkroost AWM, de Ruiter PC (2003) Species richness-productivity
patterns differ between N-, P-, and K-limited wetlands. Ecology 84:2191-2199
Pate JS (1986) Economy of symbiotic nitrogen fixation. In: Givinish TJ (ed) On the Economy of
Plant Form and Function. Cambridge University Press, Cambridge, pp 299-325
40
Nutrient limitation
Penning de Vries FWT, Krul JM, Van Keulen H (1980) Productivity of Sahelian rangelands
in relation to the availability of nitrogen and phosphorus from the soil. In: Rosswall T (ed)
Nitrogen cycling in West African ecosystems. Royal Swedish Academy of Sciences, Stockholm,
pp 95-113
Perreijn K (2002) Symbiotic nitrogen fixation by leguminous trees in tropical rain forest in Guyana.
Tropenbos-Guyana Series 11, Tropenbos, Georgetown, Guyana
Ruess RW, McNaughton SJ (1988) Ammonia volatilization and the effects of large grazing
mammals on nutrient loss from East African grasslands. Oecologia 77:382-386
Sage RF, Pearcy RW (1987) The nitrogen use efficiency of C3 and C4 plants. 2. Leaf nitrogen effects
on the gas-exchange characteristics of Chenopodium album (L.) and Amaranthus retroflexus (L.).
Plant Physiology 84:959-963
Sanchez PA (1976) Properties and management of soils in the tropics. Wiley, New York
Sankaran M, Hanan NP, Scholes RJ, Ratnam J, Augustine DJ, Cade BS, Gignoux J, Higgins SI,
Le Roux X, Ludwig F, Ardo J, Banyikwa F, Bronn A, Bucini G, Caylor KK, Coughenour MB,
Diouf A, Ekaya W, Feral CJ, February EC, Frost PGH, Hiernaux P, Hrabar H, Metzger KL,
Prins HHT, Ringrose S, Sea W, Tews J, Worden J, Zambatis N (2005) Determinants of woody
cover in African savannas. Nature 438:846-849
Sarmiento G, da Silva MP, Naranjo ME, Pinillos M (2006) Nitrogen and phosphorus as limiting
factors for growth and primary production in a flooded savanna in the Venezuelan Llanos. Journal
of Tropical Ecology 22:203-212
Scherer-Lorenzen M, Olde Venterink H, Buschmann H (2007) Nitrogen enrichment and plant
invasions: the importance of nitrogen fixing plants and anthropogenic eutrophication. In:
Nentwig W (ed) Biological invasions, Ecological Studies. Springer, Berlin, pp 163-180
Snyman HA (2002) Short-term response of rangeland botanical composition and productivity to
fertilization (N and P) in a semi-arid climate of South Africa. Journal Of Arid Environments
50:167-183
Solis E, Campo J (2004) Soil N and P dynamics in two secondary tropical dry forests after
fertilization. Forest Ecology and Management 195:409-418
Stock WD, Wienand KT, Baker AC (1995) Impacts of invading N2-fixing Acacia species on patterns
of nutrient cycling in 2 Cape ecosystems: evidence from soil incubation studies and 15N natural
abundance values. Oecologia 101:375-382
Tobler MW, Cochard R, Edwards PJ (2003) The impact of cattle ranching on large-scale vegetation
patterns in a coastal savanna in Tanzania. Journal of Applied Ecology 40:430-444
Treydte AC, Edwards PJ, Suter W (2005) Shifts in native ungulate communities on a former cattle
ranch in Tanzania. African Journal of Ecology 43:302-311
Van de Vijver CADM, Poot P, Prins HHT (1999) Causes of increased nutrient concentrations in
post-fire regrowth in an East African savanna. Plant and Soil 214:173-185
Verweij RJT, Verrelst J, Loth PE, Heitkonig IMA, Brunsting AMH (2006) Grazing lawns contribute
to the subsistence of mesoherbivores on dystrophic savannas. Oikos 114:108-116
Vitousek PM (1984) Litterfall, nutrient cycling, and nutrient limitation in tropical forests. Ecology
65:285-298
Walker B (1969) Effects of nitrogen fertilizer on natural pastures in Western Tanzania. Experimental
Agriculture 5:215-222
Weinmann H (1938) Effect of fertiliser treatment on Transvaal highveld. South African Journal of
Science 35:246-249
41
Nutrient cycling
Chapter 2
Effects of large herbivores, fire, and N2-fixation on N
and P cycling in humid savanna
Patrick G. Cech, Harry Olde Venterink, and Peter J. Edwards
Institute of Integrative Biology, ETH Zürich, Universitätsstrasse 16, 8092, Zürich, Switzerland
43
Chapter 2
Abstract
Plant growth in tropical savannas is determined by the interaction of a seasonality in
precipitation and availabilities of nitrogen and phosphorus. Additionally, herbivores, N2fixation and fire affect short-term, long-term and relative availabilities of these nutrients.
The aim of this study was to quantify the availabilities of N and P along vegetation gradients
within a recently abandoned cattle ranch, and at reference sites in an adjacent former game
reserve. Annual input and output fluxes were estimated to assess how the activities of large
herbivores, symbiotic N2-fixation, and fires affect N and P availabilities in the long-term.
Cattle ranching has led to the spatial re-distribution of large quantities of nutrients, with
local accumulation being stronger and more persistent for phosphorus than for nitrogen.
Lawn patches intensively grazed by wild herbivores tended to have elevated availabilities
of N and P, but the nutrient balances of both nutrients were negative. In dense Acacia
stands N2-fixation enhanced N availability and caused a net annual N input. Fire was
the major cause for nutrient losses from tallgrass savanna. N inputs from the atmosphere
and symbiotic N2-fixation were not sufficient to compensate for these losses; our results
therefore call into question the common assumption that N budgets in annually burned
savanna are balanced.
Keywords: fire, herbivory, mineralization, N fixation, N:P stoichiometry, nutrient balances,
Tanzania
44
Nutrient cycling
Introduction
Savanna ecosystems are characterized by heterogeneous vegetation co-dominated by trees
and grasses. Besides water, plant growth in savannas is often limited by the availabilities
of nitrogen and/or phosphorus (Chapter 1), which are often in low supply (Medina 1987).
The availability and spatial distribution of these nutrients are affected seasonally by water
availability, and by other factors including fire, grazing by cows and/or native ungulates
and N2-fixation (Augustine 2003; Fynn et al. 2003; Ludwig et al. 2004; McNaughton et
al. 1997). The relative importance of these factors for short-term and long-term N and
P availabilities are still largely unknown, and may vary among regions; for example, the
Serengeti is likely to differ from savannas without large herds of migrating herbivores.
Fire is a natural component of savannas, but the frequency of fires has increased greatly
through human activities (Bond et al. 2005); many of these fires are accidental, but fire is
also used as a management tool, for example to attract animals and increase visibility of
wildlife in national parks (Eriksen 2007). Data from long-term fire experiments in tallgrass
prairies and savannas indicate that N mineralization may increase in annually burned
savannas (Aranibar et al. 2003b), but that recurrent fires over long periods lead to reduced
nitrogen mineralization and total soil N pools (Fynn et al. 2003; Ojima et al. 1994; Reich
et al. 2001). Data on the effect of fire on soil phosphorus availability and mineralization in
savannas are scarce, but there are some indications for increased levels of extractable P in
annually burned savannas compared to nearby protected sites (Brookman-Amissah et al.
1980; Harrington 1974; Singh et al. 1991).
The effect of herbivory on nutrient availability seems to differ spatially and among animal
species. Some domestic herbivores like cows and horses are known to redistribute nutrients
by feeding in certain areas while distributing their excreta in others, thus causing local
depletion or enrichment of nutrients (Augustine et al. 2003; Edwards and Hollis 1982;
Jewell et al. 2007). In contrast, other herbivores feed and return their excreta in the same,
intensively used areas, thereby maintaining nutrient-rich patches in otherwise nutrientpoor vegetation; the grazing lawns of hippos are well known examples of this process. For
tallgrass prairie it has been shown that grazers may have a positive net effect on nitrogen
balance because biomass consumption reduced losses through fire (Hobbs et al. 1991). The
net effect of herbivores on soil nutrient turnover is less clear: N mineralization was increased
by bison in Yellowstone National Park and by prairie dogs in the South Dakota mixed
grass prairie (Frank and Groffman 1998; Holland and Detling 1990); but it was decreased
by cattle in North Dakota rangelands and in a Kenyan semi-arid savanna (Augustine and
McNaughton 2006; Biondini et al. 1998). In Serengeti, sites with sparse animal density had
45
Chapter 2
lower N mineralization rates than sites with high numbers of resident native herbivores,
however, in a 2-years study with exclosures it could not be determined whether or not the
higher mineralization rates were a result of herbivore activity (McNaughton et al. 1997).
We are unaware of any quantitative studies in tropical savannas of the relative effects of
wild and domestic herbivores upon nutrient availability and cycling; however, it would be
useful to have such information as a basis for assessing the ecological consequences of
intensified livestock production (Augustine et al. 2003; Kauffman et al. 1995; Scholes and
Archer 1997).
Woody legumes, especially of the genus Acacia, are conspicuous elements of many savannas,
and the abundance and distribution of these trees have been found to be significantly
affected by livestock grazing (Hudak 1999; Tobler et al. 2003). Much higher amounts of
nitrogen accumulate and are mineralized beneath the trees or shrubs than in open grassland,
partly due to the woody legume’s capacity to fix atmospheric N2 (Archer 2001; BernhardReversat 1996; Geesing et al. 2000; Ludwig et al. 2004; Yelenik et al. 2004). Several authors
have suggested that the N gained by symbiotic fixation compensates for N lost due to fire
(Laclau et al. 2005; Medina and Bilbao 1991; Sanhueza and Crutzen 1998), though to our
knowledge this has never been explicitly investigated.
Most studies in tropical savannas have focussed on the effects of either fire or herbivory on
nitrogen availability. Data on the various factors influencing nitrogen availability, as well
as data on phosphorus availability in tropical savannas, are scarce despite the importance
of these nutrients as factors limiting plant growth (Chapter 1). The aim of this study was to
compare the influence of domestic and wild herbivores upon soil nutrient conditions. To
do this, we determined various measures of nitrogen and phosphorus availabilities along
vegetation gradients within a recently abandoned cattle ranch and at sites in a neighbouring
game reserve (Figure 1). We also calculated nutrient balances to compare the influence
of these herbivory factors with the effects of symbiotic N2-fixation, fire and other input
and output fluxes of N and P. Our working hypotheses are presented diagrammatically
in Figure 2B. In the former ranch area we expected the nutrient availabilities to be highest
within and around the paddocks where cattle were herded overnight, and to be lowest in
pasture areas. Since N is lost through ammonia volatilization from dung and urine as well
as from leaching (Augustine 2003; Frank and Zhang 1997; Ruess and McNaughton 1988),
we expected nutrient enrichment due to cattle excreta to be more pronounced for P than
for N. Since most Acacia species are known to nodulate, and thus potentially able to fix
atmospheric N2 symbiotically (DeFaria et al. 1989), we expected N availability to be higher
in areas of bush encroachment than elsewhere. Regarding the effects of wild ungulates,
we expected to find some N and P enrichment at heavily grazed sites, though less than for
areas used by cattle.
46
Nutrient cycling
Study area and methods
Study area
The study was conducted in the Saadani National Park on the Tanzanian coast (5° 43′ S,
38° 47′ E). The soils in this area are relatively nutrient-poor and consist of greyish fine
sand or loamy sand in the flats, and reddish loamy sand over clay on slopes and hilltops
(Klötzli 1980). Between 1954 and 2000, the northern part of the area was used as a cattle
Msangazi River
Tanzania
G1
G2
G3
former
Mkwaja Ranch
Mkwaja
Mlig
aji R
iver
GP1
TGS1
TGS2
TG 3
S
GP2
Indian Ocean
GP3
Saadani
former
Saadani Game
Reserve
N
0
3
6
km
9
iR
ive
r
m
Wa
Figure 1 Map of Saadani National Park on the Tanzanian coast. The former Mkwaja Ranch
and the former Saadani Game Reserve are indicated by the light grey and dark areas, respectively.
Filled circles indicate the location of paddocks where cattle were herded overnight, filled triangles
are villages. The three selected paddock gradients are indicated with arrows. The three grazed sites
(GP) and the three tallgrass sites (TGS) in the former game reserve are indicated with stars.
47
Chapter 2
ranch, with up to 13’000 head of cattle on ~460 km2 (Figure 1). From 1969, much of the
southern part was managed as a game reserve (Figure 1). The whole area is now grazed by
wild herbivores including warthog, waterbuck, reedbuck, buffalo, wildebeest, giraffe and
elephant. Densities and diversity of wild herbivores are much higher in the southern part of
the park compared to the former ranch area (Treydte et al. 2005). Mean annual temperature
recorded at the former ranch complex was 25° C (1973–98). Annual precipitation from
1957-98 has ranged from 610 to 1700 mm, with a mean of 1040 mm. The wet season
lasts from March until June, and there is a short rainy season from mid-October to midNovember. The driest months are January and February and August and September, and
during these periods fires are common, many of them being started deliberately by the
local people, poachers or by the park management. Rainfall, temperature and relative
humidity during the study period were recorded with tipping bucket rain gauges and data
loggers (HOBO RG3-M, HOBO H8 Pro, Onset Computer Corp., Bourne, MA, USA) at
two locations: one in the northern and one the southern part of the national park. For a
more detailed description of the study area see (Tobler et al. 2003).
The vegetation of the area is dominated mainly by bushland and grassland, but there are
areas of evergreen forest. During the operation of the ranch, cattle were kept in paddocks
with up to 1500 animals at night, and during the day were driven to pasture areas and dams
in herds of 200-400 animals. The effect of this management is reflected in the vegetation,
with five concentric vegetation zones around each paddock (Tobler et al. 2003; Figure 2):
(1) the paddock centres (PC) are dominated by the stoloniferous grass Cynodon dactylon,
which forms dense mats; (2) the margins of paddocks (PM) are characterized by the grasses
Digitaria milanjiana, Eragrostis superba and the sedge Cyperus bulbipes, reaching a height of
50-60 cm; (3) Heteropogon savanna (HS) is dominated by the grass Heteropogon contortus,
accompanied by Panicum infestum, Bothriochloa bladhii and the sedge Abildgaardia triflora, and
reaching 60 cm; (4) Acacia woodlands (AW) are dominated by Acacia zanzibarica reaching
heights up to c. 8 m – Panicum infestum, Heteropogon contortus and Abildgaardia triflora are
most abundant in the herbaceous layer; (5) tallgrass savanna (TGM) is dominated by the tall
grasses Hyperthelia dissoluta and Diheteropogon amplectens with culms up to ~2 m height.
For our study we selected three paddocks systems (Figure 1). Along each paddock gradient
we selected five locations representative of the five vegetation types; the main criteria for
selecting these locations were distance to the paddock centre and species composition,
but we also looked for sites that were on level ground, homogeneous and without termite
mounds. In the former game reserve we selected three grazed patches (sensu Archibald et
al. 2005; GP). Field observations suggest that such patches are not grazed continuously,
48
Nutrient cycling
A
Paddock
Centre
(PC)
Paddock
Margin
(PM)
Heteropogon Acacia
Woodland
Savanna
(AW)
(HS)
Tallgrass
Savanna
(TG)
Distance from Paddock Centre
B
P-availability
N-availability
TG
AW
HS
PM PC
Distance from Paddock Centre
Figure 2 Schematic cross-section of the vegetation types found along paddock gradients in the
former Mkwaja ranch area (A), and the expected patterns in relative N and P availabilities along
the vegetation gradients (B).
but are abandoned if the grass grows above a certain height; after that, they are not grazed
again until the biomass is removed by fire. The lawn sites selected were dominated by the
grasses Panicum infestum, Heteropogon contortus, Digitaria milanjiana, Bothriochloa bladhii and
the sedge Bulbostylis pilosa; the vegetation was c. 7 cm tall. As reference sites with a low
impact of native herbivores, we selected three nearby tallgrass sites dominated by the tall
grasses Hyperthelia dissoluta and Diheteropogon amplectens (TGS, Figure 1).
Measurement of nutrient pools and soil turnover rates
Aboveground herbaceous biomass was determined at all sites at the beginning of the dry
season in July 2005. Additionally, total herbaceous biomass (i.e. including root biomass)
was determined for four vegetation types (PM, AW, TGM and GP) after the short wet
season in January 2007. At each site aboveground biomass was measured by clipping four
50 × 50 cm squares per site at ground level. To measure belowground biomass, four soil
49
Chapter 2
cores (4.2 cm diameter, 20 cm depth) were extracted from each of the clipped squares and
pooled. Soil was sieved (2 mm) and root biomass collected separately from rhizomes/
shoot bases. Harvested biomass was dried until constant weight and, after Kjeldahl
digestion, analysed for total N and P content by means of a continuous flow injection
analyser (FIAStar, Foss Tecator, Höganäs, Sweden). Biomass production during 150 days
was determined by harvesting regrowth from the clipped squares in early November and
again in late December 2005 after some larger rain events. In grazed patches, squares to be
clipped were protected from grazing.
During 2005, three measures of soil N availability were conducted in the top 10 cm soil:
(1) total soil N (2) inorganic N pool, (3) net N-mineralization, as well as adsorption of
inorganic N to ion-exchange resin-bags. For soil phosphorus availability we measured the
total and inorganic P pools in the top 10 cm soil, and adsorption of inorganic P to ionexchange resin-bags. At each site we collected five pairs of soil cores in PVC tubes (2.8 cm
diameter, 10 cm length), four being arranged equidistantly (5 m) from a central one. One
of each pair of soil cores was taken immediately for extraction and drying and the other
placed back into the soil as incubating sample for measurements of net N mineralization.
Total C and N were measured on a dry combustion analyzer (CN-2000, LECO Corp., St.
Joseph, Minnesota, USA), total P was measured after Kjeldahl digestion as described for
vegetation samples. Inorganic N pool was determined as exchangeable NH4+ and NO3- by
extraction of 10 g fresh soil in 50 ml 0.2 M KCl solution. Inorganic P pool was determined
as extractable PO42- by extraction of 5 g of fresh soil with 50 ml Bray-2 extraction solution
(Bray and Kurtz 1945). Extraction was done within 12 h of collection of the soil cores.
KCl-extracts were acidified with 5% H2SO4 for conservation until analysis. Concentrations
of NO3-, NH4+ and PO42- in the extracts were measured colorimetrically using a continuous
flow injection analyzer. In situ incubation for net N mineralization measurements were done
for a 33 days interval in July 2005 and for a 19 days interval in November 2005, both at the
beginning of the dry season. Net N mineralization was calculated as the difference between
extractable inorganic N at start and at the end of the incubation interval. Finally we used
ion-exchange resin bags (Dowex 1 (1X8-100) for anions and Dowex 50W (50X8100) for
cations; Fluka, CH) to obtain an additional index of relative inorganic N and P availability
among sites. Resin bags were prepared as described by Güsewell (2002): 1.00 ± 0.01 g
of Dowex-1 and 0.76 ± 0.01 g of Dowex 50W were weighed into bags (3 cm × 4 cm) of
nylon mesh (60 μm width, Sefar AG, Heiden, CH). Resins were conditioned by overnight
immersion in a saturated KCl solution and 30 sec rinsing with deionised water. At each
site, five bags were placed carefully at 5 cm depth in 45° slant incisions made with a knife,
after which the slides were closed. After 75 days, resin bags were collected from the field,
50
Nutrient cycling
rinsed gently with de-ionized water and dried until extraction. Resin bags were extracted
by shaking each bag for 90 min with 50 ml 0.5 M HCl. The extraction solution was then
analysed colorimetrically for PO42-, NO3- and NH4+ as described above.
Additionally, total pools of C, N and P, and inorganic pools of N and P were determined
along soil profiles in five savanna vegetation types (PC, PM, TGM, TGS, GP). At each
site five soil cores to 70 cm depth were taken with a soil corer and split into five segments
(0-10, 10-20, 20-30, 30-50, 50-70 cm). For each depth, the five segments were pooled and
analysed using the methods described above. Soil bulk density was determined by drying
two large soil cores (4.2 cm diameter) per depth segment to constant weight.
Input and output fluxes of nutrient balances
(1) Wet atmospheric deposition was determined from rain samples (n = 15) collected at
regular intervals from October 2005 until November 2006. Rain samples were acidified
with 5% H2SO4 for conservation and analysed colorimetrically for PO42-, NO3- and NH4+
as described above.
(2) Symbiotic N2-fixation was estimated for Acacia trees as well as for legume herbs using
measured data on biomass production, tissue N concentrations, and the 15N natural
abundance method. We sampled the total biomass of herbaceous legumes and foliage
of re-sprouting leguminous shrubs (mainly Dichrostachys cinerea) at each site from eight
randomly placed frames of 1 × 1 m in August 2006. Samples were dried, ground, and
analysed on N concentration using a dry combustion analyzer (CN-2000, LECO Corp.,
St. Joseph, Minnesota, USA). We assumed that the biomass harvested after the long wet
season corresponded to net annual production, because for the majority of the sites average
time since the last fire was c. 12 months. For Acacia woodlands we measured diameters of all
Acacia trees at ankle height within a square of 15 × 15 m centred on the sites of soil sampling,
and used allometric equation from Cochard (2004) to estimate foliage mass after the long
wet season, which was then used as estimate for annual production. N concentration of A.
zanzibarica leaves was determined from pooled leaf samples collected from two sites at the
end of the wet season 2005; these had a mean value of 24.5 ± 0.5 mg N g-1. We calculated
% N derived from fixation following Amarger et al. (1979) using the 15N natural abundance
method:
15
Ndfa =
N ref
15
N ref
15
N leg
B
(1)
51
Chapter 2
where δ15Nleg is the 15N abundance in the N2-fixing legume species, δ15Nref is the level of δ15N
measured in the non-N2-fixing reference species growing on the same substrate, and B is
the abundance of 15N in a legume individual that obtains all its nitrogen from N2-fixation.
The use of a correct B-value is crucial for the accuracy of this method to quantify N2fixation. To avoid unrealistic values of Ndfa (> 100), B was set to the lowest detected δ15N (1.82‰ in Tephrosia villosa), as proposed by Hansen and Vinther (2001). Based on leaf δ15N
values in the two most common herbaceous legumes, Cassia mimosoides (-1.42‰ ± 0.01)
and Tephrosia villosa (-1.82‰) collected from two sites, and using leaves of four non-legume
species (Agathisanthemum bojeri, Erythrocephalum zambesiacum, Dalechampia trifoliata and
Kohautia longifolia) at the sites as references (δ15N 1.52‰ ± 0.25), we determined that these
plants derived an average of 94% of their N from symbiotic fixation. For Acacia zanzibarica
δ15N was -0.44‰ ± 0.12, yielding 58% N derived from atmosphere, using leaves of a cooccurring non-legume tree Combretum constrictum (1.23‰) as a reference.
(3) Nutrient inputs in the dung of wild ungulates were estimated in three savanna types
in the former Saadani game reserve by collecting dung in 100 m2 plots during c. 45 days
(5-14 Jan. to 22-25 Feb. 2007), after initially removing all dung from the plots. In each
plot, all fresh dung pellets were collected every two weeks, and numbers of pellets for
each herbivore species were recorded (Halsdorf et al., unpublished). Inputs of N and P
were calculated using the average dry weight of fresh dung pellets per species, and the
average N and P concentrations in fresh dung of almost all occurring herbivore species
(buffalo, giraffe, kongoni, reedbuck, warthog, waterbuck, wildebeest) as measured after
Kjeldahl digestion (Halsdorf, unpublished data). Dung concentrations of other herbivores
were assessed by using data for species with similar bodyweights and diet. Net deposition
of N in urine was calculated from the measured dung input. In the absence of data for wild
herbivores, we used values from Scholefield et al. (1991) reporting that for cows on a low
nitrogen diet 45% of excreted N was in urine. We also accounted for 3-25% urea (85% of
urinary N) to be lost through ammonia volatilization as reported for the Serengeti (Ruess
and McNaughton 1988). Due to the very low proportion of P excreted in urine (Morse et
al. 1992) this input was regarded as negligible.
(4) Nutrient losses through fire were determined experimentally. In the early dry season
2006 we conducted burning experiments using total aboveground herbaceous biomass from
tallgrass savanna, Acacia woodland, and from gaps in remnant forest. The experimental
design was fully factorial with respect to the origin of biomass and burning location, with
a replication of three sites per savanna type. Aboveground biomass was collected from all
three vegetation types around midday by clipping two adjacent 50 × 50 cm squares to c.
52
Nutrient cycling
3 cm height (one serving as pre-burning reference material). Burning experiments were
started immediately afterwards at a location within one of the vegetation types. Collected
biomass was weighed and placed on a 50 × 50 cm collecting pan and arranged upright
held by a 1.5 m tall frame in order to simulate field conditions. Samples were ignited using
a 50 cm long torch placed upwind for 30 seconds. When the burning was finished, the
residue was collected and separated into unburned green biomass, unburned dead biomass
and ash. Oven-dried samples were milled, digested (Kjeldahl method), and analysed for N
and P concentrations. Concentrations of K, Ca and Mg were determined after extraction
of the residue with 0.5 M HCl, using atomic absorption spectrometry (Hunt 1982)
(SPECTRAA 240 FS, Varian AG, Zug, Switzerland). Nutrient losses were calculated as
difference between nutrient content in pre-burning reference samples and post-burning
residues. Weather conditions during the burning of the samples (wind speed, temperature
and relative humidity) were recorded as co-variables. These did not vary significantly
between days, and there was also no effect of the location of the experimental burnings
within different vegetation types on relative nutrient losses (data not shown).
Statistical analyses
Statistical analysis was done with JMP 6.0.3 (SAS Institute, Cary, USA). Analysis of
data from paddock gradients was done with analyses of variance (type III SSQ, factors
‘vegetation type’ and ‘gradient’) and multiple comparisons between factor levels using the
Tukey-Kramer HSD test (P < 0.05). If necessary, data were transformed prior to analysis in
order to meet model assumptions. Data from tallgrass sites in the former game reserve (TGS)
were compared with data from nearby grazed patches (GP), and with data from tallgrass
sites in the former ranch (TGM) using Student’s t-test assuming unequal variances.
Results
Vegetation
Aboveground herbaceous biomass, and the pools of N and P in this biomass, were highest
in paddock centres (Table 1). There were no significant differences between the other
vegetation types along the gradient, although N pools in aboveground biomass tended
to be higher in tallgrass savanna, and P pools tended to be higher in tallgrass savanna
and paddock margins. The N pool in aboveground vegetation was significantly lower in
grazed patches than in nearby ungrazed tallgrass sites (Table 1). Aboveground herbaceous
53
Chapter 2
productivity was highest in paddock margins and tallgrass savanna (Table 1). In tallgrass
savanna aboveground herbaceous biomass contained 1.7% and 1.8% of total N and P
pools (total biomass plus top 30 cm of soil), respectively. In paddock margins and grazed
patches the corresponding values were 1.4% and 0.7% for N and 2.8% and 0.9% for P,
respectively. On average more than 50% of the nutrient pools in vegetation were stored in
belowground plant parts (Figure 3).
Relative availabilities as measured by N:P ratios in live aboveground vegetation were in
line with our expectations (see Figure 2B): thus, availability of P relative to N was high in
paddock centres and margins, while the reverse was the case in Acacia woodlands (Figure
4). There was considerable variation in N:P ratios among sites of the same vegetation type
(Figure 4). For instance, the availability of N relative to P was higher on gradients 1 and 3,
which had higher densities of Acacia trees than on gradient 2 (Figure 4).
Soil nutrient pools
Pools of carbon, nitrogen and phosphorus were higher in paddock centres than in paddock
margins, Heteropogon savanna and Acacia woodlands, and lowest in tallgrass savanna (Table
1, Appendix A). In most cases, the only significant differences were between the paddock
centre and the other sites, partly due to low statistical power. The enrichment in total soil
N and P in the paddock centres was measurable even in the deepest samples (70 cm), and
was more pronounced for P than for N (Figure 5).
Pools of inorganic N along the paddock gradients, measured in July 2005, corresponded
to the expected pattern in Figure 2B, although in most cases only the differences between
paddock centres and the other vegetation types were significant (Table 1, Appendix A).
The mean inorganic N pools in November 2005 were higher than those in July, particularly
in Heteropogon savanna. The enrichment in inorganic N-pools in the paddock centres as
compared to the other sites was most pronounced in the top 20 cm and declined with
increasing depth (Figure 5). Inorganic N pools in the Saadani area in November were
on average lower than in July, and tended to be higher in two of the three heavily grazed
patches than in nearby ungrazed tallgrass sites (Appendix A).
Pools of inorganic P along paddock gradients were rather similar in July and November,
and matched with the expected pattern (Figure 2B); again, only paddock centre sites were
significantly different from the other sites (Table 1). P-enrichment in paddock centres was
54
Nutrient cycling
Table 1 Nutrient pools in topsoil (10 cm) and aboveground vegetation along three paddock
gradients on former cattle ranch grounds, and at three grazed and ungrazed sites in the former
Saadani game reserve. Data shows means per vegetation type. PC: paddock centres, PM: paddock
margins, HS: Heteropogon savanna, AW: Acacia woodland, TGM: tallgrass savanna of the former
Mkwaja ranch, TGS: tallgrass savanna in the former Saadani game reserve, GP: grazed patches.
For vegetation types within a paddock gradient values not sharing the same letter are significantly
different (Tukey-HSD, P > 0.05). * and + indicate significant differences between TGS and GP sites
(t-test) at P < 0.05 and P < 0.10, respectively. There were no significant differences between TGM and
TGS sites (t-test, P < 0.10). Precipitation during the period of biomass production measurements
was 176 mm at paddock gradient sites and 241 mm at Saadani sites.
Paddock gradients
PM
HS
AW
PC
TGM
Saadani sites
TGS
GP
Soil
Total C (kg m-2) 3.79 ± 0.82A 0.84 ± 0.10B 1.64 ± 0.54AB 1.66 ± 0.16AB 0.72 ± 0.15B 0.89 ± 0.04 1.16 ± 0.25
Total N (g m-2)
367 ± 78A
73 ± 9B
116 ± 33B
124 ± 9B
55 ± 10B
66 ± 5
89 ± 18
Total P (g m-2)
98 ± 29A
7.8 ± 1.5B
8.6 ± 1.9B
9.2 ± 0.9B
5.3 ± 0.9B
7.3 ± 0.1
11.0 ± 2.7
Extractable N
Jul ‘05 (g m-2)
5.9 ± 2.4A 0.29 ± 0.08BC 0.19 ± 0.06C 0.41 ± 0.08B 0.21 ± 0.04BC 0.30 ± 0.07 0.44 ± 0.12
Extractable N
6.1 ± 3.4A
Nov 2005 (g m-2)
0.32 ± 0.04B 0.63 ± 0.20AB 0.59 ± 0.05AB 0.28 ± 0.09B 0.13 ± 0.04 0.30 ± 0.15
Extractable P Jul
51 ± 27A
‘05 (g m-2)
0.73 ± 0.15B 0.19 ± 0.01B 0.28 ± 0.05B 0.35 ± 0.11B 0.32 ± 0.08 0.82 ± 0.18+
Extractable P
Nov ‘05 (g m-2)
0.81 ± 0.27B 0.21 ± 0.03B 0.25 ± 0.05B 0.34 ± 0.08B 0.21 ± 0.05 0.57 ± 0.12+
63 ± 22A
Aboveground herbaceous biomass
Biomass Jul ‘05
1.44 ± 0.14A 0.56 ± 0.07B 0.45 ± 0.10B 0.35 ± 0.20B 0.98 ± 0.19AB 0.87 ± 0.09 0.16 ± 0.06*
(kg m-2)
Total N Jul ‘05 (g
m-2)
Total P Jul ‘05
(g m-2)
19 ± 1A
3.1 ± 0.2A
Production Jul291 ± 102A#
Dec ‘05 (g m-2)
#
§
2.3 ± 0.2B
1.9 ± 0.5B
2.2 ± 1.1B
3.2 ± 0.8B
2.5 ± 0.2
1.1 ± 0.2*
0.38 ± 0.11B 0.13 ± 0.04B 0.13 ± 0.07B 0.35 ± 0.13B 0.26 ± 0.06 0.13 ± 0.05
95 ± 13B#
65 ± 26B§
59 ± 13B
one of three sites burned during the measurement period (see also Appendix A)
two of three sites burned during the measurement period (see also Appendix A)
55
165 ± 12AB
225 ± 80
132 ± 9
Chapter 2
Biomass (g m-2)
3000
live aerial
dead aerial
rhizomes
roots
2500
2000
f
1500
f
1000
500
0
0 1.9 2 2.1 3 3.9 4 4.1 4.5 4.9 5 5.1 6 6.9 7 7.1 8
12
N (g m-2)
10
8
f
6
f
4
2
0
2.5
P (g m-2)
2
1.5
1
f
f
0.5
0
G1 G2 G3
G1 G2 G3
G1 G2 G3
PM
AW
TGM
1
2
3
GP
Savanna vegetation type
Figure 3 Total herbaceous biomass and biomass nutrient pools in four selected savanna types
in January 2007: paddock margins (PM), Acacia woodland (AW) and tallgrass savanna (TGM)
were located along paddock gradients, and three intensively grazed sites (GP) were located in the
former Saadani game reserve. Biomass and nutrient pools are shown for live and dead aboveground
biomass, as well as for rhizomes and roots. Sites marked with an ‘f ’ burned 1-2 months before the
measurements. Error bars indicate standard errors of total biomass (n = 4).
56
N:P ratio in biomass
Nutrient cycling
P-limitation
gradient 1
gradient 2
gradient 3
20
15
N&P-co-limitation
10
5
0
N-limitation
c
PC
c
ab
a
PM
HS
AW
b
TGM
TGS
GP
Savanna vegetation type
Figure 4 Relative availabilities of N and P along paddock gradients in a former cattle ranch
and reference sites in a former game reserve as measured by the N:P ratio in live aboveground
herbaceous biomass in July 2005. PC: paddock centres, PM: paddock margins, HS: Heteropogon
savanna, AW: Acacia woodland, TGM: tallgrass savanna of the former Mkwaja ranch, TGS: tallgrass
savanna in the former Saadani game reserve, GP: grazed patches. Data shows means per site (n
= 4, ± SE). The dashed line represents the boundary between N-limitation and NP-co-limitation
based on a study by Cech et al. (Chapter 1). The available data are not sufficient to distinguish a
critical value between NP-co-limitation and P limitation, or to determine whether there is such a
critical value all (Chapter1).
obvious down to 70 cm soil depth, and was more pronounced relative to the enrichment
observed for extractable N (Figure 5). Elevated levels of extractable P in paddock margins
and grazed patches were confined to the top 20 cm (Figure 5). Grazed patches tended to
have higher pools of extractable phosphorus than nearby tallgrass sites (Table 1).
Soil nutrient turnover rates
Rates of N mineralization varied widely, and some were even negative (Figure 6a and
6b). The trend in net N mineralization rates along paddock gradients was similar to that
observed for inorganic N pools (cf. Figure 6 and Table 1), with the highest values in the
paddock centres (cf. Figure 6 and 2B). In Acacia woodlands N mineralization tended to
be higher than in adjacent vegetation types, but only during the first incubation period
(Figure 6). It also tended to be higher in grazed areas than in nearby tallgrass sites, but the
differences were statistically not significant (Figure 6, p = 0.23 and p = 0.15 for the first and
the second incubation period, respectively).
57
Chapter 2
PC
PM
Savanna vegetation type
TGM
0-10 cm
10-20 cm
20-30 cm
30-50 cm
50-70 cm
TGS
GP
0
30
60
90
120 200 600 0
Ntot (g m-2 per 10 cm layer)
0.3 0.6 0.9 1.2 1.5
5
15
Nextr (g m-2 per 10 cm layer)
PC
PM
TGM
TGS
GP
0
5
10
15
20
25 50
150 0 0.3 0.6 0.9 1.2 1.5
Ptot (g m-2 per 10 cm layer)
2 50 100
Pextr (g m-2 per 10 cm layer)
Figure 5 Pools of total N, total P, and extractable inorganic N and P, along soil profiles in five
selected savanna types: paddock centres (PC), paddock margins (PM) and tallgrass savanna (TGM)
were located along paddock gradients; tallgrass savanna (TGS) and intensively grazed patches (GP)
were located the former Saadani game reserve. Error bars indicate standard errors (n = 3).
Figure 6 Net N mineralization rates and adsorption rates of N and P on resin along three paddock
gradients on former cattle ranch grounds, and at three tallgrass savanna and three grazed sites in
the former Saadani game reserve. Net N mineralization rates were measured in situ for a 33 days
incubation period in July/August 2005 (a), and for a 19 days incubation period in November/
December 2005 (b). Adsorption rates of inorganic N (c) and inorganic P (d) to ion-exchange resins
were determined between September and November 2005. PC: paddock centres, PM: paddock
margins, HS: Heteropogon savanna, AW: Acacia woodland, TGM: tallgrass savanna of the former
Mkwaja ranch, TGS: tallgrass savanna in the former Saadani game reserve, GP: grazed patches.
Grey symbols indicate that the site burned while resin bags were in the soil. Error bars indicate
standard errors (n = 5). Savanna types along paddock gradients not sharing the same letter are
significantly different (n = 3, Tukey-HSD, P > 0.05). Note: in panels (a) and (b) the y-axis breaks at
10 and 35, respectively, and continues with a different scale. In panels (c) and (d) data is shown on
a log scale.
58
Net N mineralization (mg N m-2 d-1)
Nutrient cycling
250
150
50
8
gradient 1
gradient 2
gradient 3
a
4
0
-4
a
b
b
ab
b
300
100
30
b
15
0
-15
Adsorbed P (mg/resin bag)
Adsorbed N (mg/resin bag)
-30
a
ab
c
bc
bc
10
c
1
0.1
a
b
b
b
b
10
d
1
0.1
0.01
0.001
a
b
b
b
PC
PM
HS
AW
b
TGM
Savanna vegetation type
59
TGS
GP
Chapter 2
Rates of inorganic N release as measured by adsorption to resin bags were generally
consistent with the patterns in N mineralization rates found in the first incubation period
(cf. Figure 6a and 6c), with values in the paddock centres being significantly higher than in
the other vegetation types (Figure 6c). The rates in grazed patches also tended to be higher
than in nearby tallgrass sites, but this difference was not significant (Figure 6c, p = 0.08).
Adsorption of inorganic P to resin bags showed a steady decline along the paddock
gradients (Figure 6d). Grazed patches showed on average higher P-adsorption than nearby
tallgrass sites, but this was statistically not significant (p = 0.28).
Nutrient input and output fluxes and balances
Mean N concentration of the rain samples was 0.67 ± 0.12 mg l-1 (0.41 ± 0.12 NO3-N
and 0.26 ± 0.05 NH4-N), and mean P concentration was 34 ± 10 μg l-1 (PO4-P). From
these concentrations and average long-term precipitation of 1040 mm yr-1, we estimated
atmospheric deposition rates at 7.0 kg N ha-1 yr-1 and 0.35 kg P ha-1 yr-1.
Annual inputs of N and P in dung were significantly higher in heavily grazed areas than
in nearby tallgrass savanna (Table 2). Values of nutrient inputs in Acacia woodlands were
intermediate with the other vegetation types.
Table 2 Nitrogen and phosphorus input rates from herbivore dung in four vegetation types in the
area of the former Saadani game reserve, as well as percentages per herbivore species. Significant
differences in N and P fluxes among habitat types are indicated with letters (ANOVA, Tukey-HSD,
P < 0.05). Data from Halsdorf et al. (unpublished).
Habitat
n
N
P
(kg ha-1 yr-1)
(kg ha-1 yr-1)
N:P
Tallgrass savanna
5
0.40 ± 0.19B
0.07± 0.04B
4.7 ± 0.2
Grazed patches
6
6.36 ± 1.81A
0.99 ± 0.32A
6.8 ± 0.6
Acacia woodland
4
1.96 ± 1.03AB
0.33 ± 0.15AB
5.3 ± 0.6
Medium height savanna
5
3.57 ± 1.42AB
0.67 ± 0.28AB
5.3 ± 0.6
60
Nutrient cycling
Table 3 Total N content in biomass of herbaceous legumes and foliage biomass of Acacia trees,
as well as calculated symbiotic N2-fixation, at the sites of Table 1. Average proportion of N derived
from the atmosphere is 94% for herbaceous and 58% for woody legumes.
PC
PM
Paddock gradients
HS
AW
TGM
Saadani sites
TGS
GP
N in herbaceous legumes (kg ha-1)
gradient/site 1
0
0.31
gradient/site 2
0
0
gradient/site 3
0
0.32
0.20
0.44
0.03
0.13
0.03
0
0.47
0.16
0.32
0.50
0.25
0.32
0.40
0.09
0.31
N in foliage of woody legumes (kg ha-1)
gradient/site 1
0
0
gradient/site 2
0
1.11
gradient/site 3
0
0
0.17
0
1.14
17.2
6.9
16.3
0
0
0
0
0
0
0
0.03
0.04
Fixed N (kg ha-1)
gradient/site 1
gradient/site 2
gradient/site 3
0
0
0
0.28
0.44
0.68
10.10
4.03
9.45
0.44
0.15
0.30
0.47
0.24
0.30
0.38
0.10
0.31
Mean total (n = 3)
0
0.29
0.63
0.30
0.41 ± 0.11 0.46 ± 0.12 7.9 ± 1.9 0.30 ± 0.08 0.34 ± 0.07 0.26 ± 0.08
Tree densities of A. zanzibarica in Acacia woodlands were 2620, 840 and 2440 per ha for the
sites on gradients 1, 2 and 3, respectively. The estimated N fixation in Acacia woodlands
was 10 kg N ha-1 yr-1 in the densest stands, but about 4 kg N ha-1 yr-1 in the least dense stand
(Table 3). The contribution from N2-fixation by herbaceous legumes was very low in all
savanna types, and did not exceed 0.45 kg N ha-1 yr-1 (Table 3).
Burning resulted in a mean loss of 71% of the N present in aboveground herbaceous
biomass, 27% of P, and between 30 and 44% of K, Ca and Mg (Table 4). The percentage
losses of N were significantly higher than those of the other nutrients (pairwise t-tests for
each vegetation type separately, p < 0.05). Percentage losses of P tended to be lower than
those of K, Ca and Mg, but differences were only significant for Acacia woodland (pairwise
t-tests Mg > P at p < 0.05) and forest gaps (pairwise t-tests Mg > P and Ca > Mg at
p < 0.05). The percentage nutrient losses were generally similar for the three savanna types;
only for K tallgrass savanna and Acacia woodlands differed (Table 4). Biomass harvested
from different vegetation types differed significantly in the proportions of dead biomass:
the lowest proportion of dead biomass was in tallgrass savanna - the vegetation with the
61
Chapter 2
highest observed fire frequency - while the highest was in forest gaps, which are protected
from fire. Relative losses of N, K, Ca and Mg tended to increase with a higher proportion
of dead biomass, but this relationship was only significant for N and Ca (N: R2 = 0.19,
p = 0.02; Ca: R2 = 0.16, p = 0.04). Based on a 71% loss of N, a 27% loss of P and data from
Table 1, a single early dry season fire event was estimated to cause losses of 133 kg N ha-1
and 8.3 kg P ha-1 in the paddock centres, and 13-22 kg N ha-1 and 0.35-1.03 kg P ha-1 in the
other vegetation types.
We calculated balances for N and P for five selected vegetation types using measured
atmospheric inputs of N and P and estimates of inputs from herbivore urine and outputs
through defoliation (Table 5). For nitrogen, balances were negative in all vegetation types
except Acacia woodlands, which tended to have a positive net N balance. For phosphorus,
balances tended to be negative in paddock centres and in grazed areas, and zero or positive
in the other vegetation types (Table 5).
Table 4 Nutrient losses through fire, as determined by combustion experiments, in percent of
total content in pre-burned aboveground herbaceous biomass. Data per site are means of nine
replicate combustions done in the early dry season in August 2006 (± SE). Values not sharing the
same letter indicate significant differences among vegetation types (Tukey-HSD, P < 0.05).
Acacia
woodland
Tallgrass
savanna
Forest gaps
1025 ± 77A
990 ± 30A
1120 ± 90A
% dead
72 ± 2B
59 ± 2C
84 ± 1A
Water content (%)
38 ± 2A
33 ± 1B
24 ± 1C
Nitrogen (%)
71 ± 2A
67 ± 3A
75 ± 3A
Phosphorus (%)
28 ± 4A
25 ± 3A
28 ± 3A
Potassium (%)
40 ± 5A
24 ± 4B
36 ± 5AB
Calcium (%)
41 ± 6A
31 ± 3A
43 ± 3A
Magnesium (%)
44 ± 5A
30 ± 2A
43 ± 3A
Characteristics of biomass prior to burning
Dry weight (g m-2)
Percentage nutrient loss through burning
62
Nutrient cycling
Table 5 Nitrogen and phosphorus balances along paddock gradients in the former Mwaja Ranch
area, and of tall grass sites and grazed patches in the former Saadani game reserve. All fluxes are
given in kg ha-1 yr-1. Site abbreviations are explained in Table 1.
Process
PC
Paddock gradients
AW
Saadani sites
TGM
TGS
GP
Nitrogen
Inputs
Atm. deposition
N2-fixation
Herbivore dung
Herbivore urine
Total input
Outputs
Fire
Herbivory
Total output
Net N balance
7.0
0
0.04
0.03
7.1
7.0
7.9
0.20
0.13-0.16
15.2
7.0
0.30
0.04
0.03
7.4
7.0
0.34
0.40
0.26-0.32
8.0-8.1
7.0
0.26
4.9-5.4
3.1-4.3
15.3-17.0
44-88
0-0.9
44-89
5.0-10.0
0-0.6
5.0-10.6
7.5-15.0
0-0.16
7.5-15.2
5.9-11.8
0-1.26
5.9-13.1
5.9-8.8
17-33
22.9-41.8
-82 to -37
4.6-10.2
-7.8 to -0.1
-5.1 to 2.2
-26.5 to -5.9
0.35
0.01
0.36
0.35
0.03
0.38
0.35
0.01
0.36
0.35
0.07
0.42
0.35
0.76-0.84
1.1-1.2
2.8-5.5
0-0.15
2.8-5.7
0.12-0.23
0-0.04
0.12-0.27
0.32-0.63
0-0.02
0.32-0.65
0.23-0.47
0-0.13
0.23-0.60
0.23-0.35
2.0-3.9
2.2-4.3
-5.3 to -2.4
0.11 to 0.26
-0.29 to 0.04
-0.18 to 0.19
-3.4 to -1.0
Phosphorus
Inputs
Atm. deposition
Herbivore dung
Total input
Outputs
Fire
Herbivory
Total output
Net P balance
Herbivore dung and urine: for the sites along the paddock gradients in the former ranch area we divided dung input rates measured
in the Saadani area by a factor 10 since it has c. 10 times lower densities of wild herbivores (Treydte et al. 2005); we assumed
PC to receive similar dung input rates like tallgrass sites (TGM); grazed patches are assumed to be grazed for 1.5-2.5 years before
being abandoned temporarily for one growing season (6 months) during which input rates are equal to those in tallgrass savanna
(TGS).
Fire: losses are given for a range of fire frequencies from one to two fires per three years (Cech et al., personal observation). Grazed
patches are assumed to burn once per two to three years after the site is abandoned by herbivores, then they lose the same amount
of nutrients as the nearby TGS sites. We calculated losses through fire as 71% of N and 27% of P (average percentage loss from
Table 4) present in aboveground biomass (Table 1).
Herbivory: based on results from Cochard (2004) we used ranges of 0-0.5% peak biomass consumed by wild herbivores for PC
and TGM and 0-3% for Acacia woodland; for TGS sites we used a range of 0-5% accounting for the 10 times higher herbivore
density. In grazed patches we measured aboveground biomass production of 132 g m-2 (Table 1), which contained 1.4 g N m-2
and 0.17 g P m-2 for a period of 5 months and 241 mm. For annual productivity we used a range of 320-550 (3.5-5.9 g N m-2
and 0.41-0.70 g P m-2), obtained by linearly extrapolating the measured production based on time (lower bracket) and based
on mean annual precipitation of 1000 mm (upper bracket). Grazed patches are assumed to be grazed for 1.5-2.5 years before
being temporarily abandoned (see above). Under intensive grazing 66% of aboveground biomass production is assumed to be
consumed (McNaughton 1985: 17-94%, mean 66%).
Net balance: maximum-minimum range calculated as minimum input minus maximum output, and maximum input minus
minimum output
63
Chapter 2
Discussion
Soil nutrient pools and turnover rates
Except for the extreme conditions in the paddock centres, total and extractable P
concentrations in soils of the Saadani region (0.04-0.08 g kg-1) were much lower than in
other savanna areas, such as Serengeti (0.6-6 g kg-1), a semi-arid savanna in Kenya (0.2 g kg-1),
and a littoral savanna in Congo (0.2-0.3 g kg-1) (Augustine 2004; Laclau 2003; Ruess and
McNaughton 1987). Total and inorganic N concentrations outside paddocks (0.4-0.8 g Ntot
kg-1 and 1-3 mg Ninorg kg-1) were in the same range as reported for humid savannas in West
Africa (0.3-0.5 g Ntot kg-1 and ≤ 2 mg Ninorg kg-1) (Abbadie 1990; de Rham 1973; Laclau
2003), but considerably lower than those in Serengeti (1-4 g Ntot kg-1 and 5-13 mg Ninorg kg-1)
(Ruess and McNaughton 1987).
Patterns of inorganic pools of N and P along our paddock gradients, generally were
consistent with our expectations (Figure 2), as well as with the relative availabilities of
these nutrients indicated by N:P ratios in plant tissue (cf. Table 1 and Figure 4). Differences
in inorganic pools of N and P between paddock centres and margins, and between those
sites and the others vegetation types were more pronounced for P than for N. Much of the
nitrogen returned in dung and urine was presumably lost through ammonia volatilization
(Augustine 2003), which is supported by significantly higher δ15N values in soils of paddock
centres compared to surrounding sites (Treydte et al. 2006). The difference between N and
P might also partly be caused by higher leaching of N ions to deeper soil layers than of
the less mobile P ions. Relatively high N and P concentrations in deeper soils layers in the
paddock centres than elsewhere support that leaching indeed took place (Figure 5). Total
and extractable nutrient pools in paddock centres were comparable to those found in 12-24
year old paddocks (called ‘bomas’) from a chronosequence of abandoned paddocks in a
semi-arid savanna in Kenya (Augustine 2003). In the same study, soil N pools in paddocks
were found to decrease more rapidly in time than soil P pools, which is consistent with our
data measured 6 years after abandonment of the ranch (cf. Table 1, Figure 5). However,
our spatial analysis shows that the enrichment in P relative to N through cattle was not
confined to the small area of the paddock centres (0.2% of the total ranch area), but
extended into the surrounding vegetation (PC + PM comprise 3.3% of the total area). In
Sahelian rangelands also, P availability was found to be increased at up to 5 km distance
from watering points (Turner 1998).
64
Nutrient cycling
Soil P release rates (adsorption to resins) gradually declined with increasing distance from
the paddock centres. Turnover rates of N (N mineralization and adsorption to resins) were
highest in the paddock centres, but for the remaining part of the paddock gradient the
expected pattern from Figure 2 was not found (Figure 6). This shows that patterns in soil
pools of inorganic P created by cattle resulted in a corresponding pattern in P availability
to plants, whereas this was only partially the case for N. Average N mineralization
rates in tallgrass savanna were estimated to range between to <0 (immobilisation) and
27 kg N ha-1 yr-1, which is similar to the range reported for other savanna ecosystems
(2-35 kg ha-1 yr-1; de Rham 1973; McNaughton et al. 1997; Scholes et al. 2003). Nitrogen
turnover rates were not significantly higher in Acacia woodlands than in adjacent tallgrass
savanna unlike what has been found for invading Acacia species in other areas (Stock et
al. 1995; Yelenik et al. 2004). We have no good explanation for this deviation from our
expectations, but we note that our N mineralization rates were measured during two early
dry season intervals, and calculations of annual rates based on this data should therefore be
treated with caution. Other measures of N availability (soil pools, N:P in plants) indicate
that N availability was indeed higher in our Acacia woodlands.
In patches grazed by wild herbivores we found large variation in total and inorganic pools
of N and P, as well as in release rates of these nutrients (Table 1, Figure 6, Appendix
A). Inorganic P pools tended to be higher in the grazed patches than in nearby tallgrass
savanna (P < 0.10). This could be because wild herbivores – like cattle – might enhance
soil P availability, but could also be because herbivores chose to feed in the more nutrientrich areas; based on these data one cannot separate between the two processes. The very
large variation in N:P ratios in grazed patches suggests that this factor does not strongly
influence food selection by wild herbivores (Figure 4).
Concerning soil nutrient pools and turnover rates, we conclude that relative availabilities of
nitrogen and phosphorus as affected by wild and domestic herbivores, N2-fixation and fire
can be reasonably well predicted. However, except for the highly enriched paddock centres,
nutrient availabilities of nutrients and mineralization rates could not fully explain patterns
in biomass and productivity, since relatively high productivity values were observed in
tallgrass savanna, which had the lowest levels of nutrient availability (Table 1). In Lamto
savanna in West Africa, Abbadie et al. (1992) concluded that grasses were independent
from soil organic matter mineralization and derived most of their nitrogen requirements
through the recycling of N stocks in dead roots before humification. This may be an
important source for nutrients in our tallgrass savanna as well.
65
Chapter 2
Nutrient input an output fluxes and balances
Based on our field measurements and experiments, as well as on published information,
we assessed annual N and P balances for the savanna types investigated in this study.
Although some of the parameter estimates were rather rough, the balances do indicate
the relative importance of the different input and output fluxes, as well as the modulating
effects of herbivores, N2-fixation and fire on long term N and P availabilities in the various
savanna types. Some studies on nitrogen cycling in savanna assume that nitrogen balances
are neutral, however without presenting evidence (Bate 1981; Laclau et al. 2005; Sanhueza
and Crutzen 1998). Below, we will evaluate whether this assumption is supported for our
study areas.
Our estimates of wet atmospheric deposition of N and P are within the range reported for
other savanna ecosystems (1.3-19 kg ha-1 yr-1 for N and 0.3-0.5 kg ha-1 yr-1 for P) (Garstang
et al. 1998; Montes and San Jose 1989; Tamatamah et al. 2005; Villecourt and Roose
1978). For the East African coast Galloway et al. (2004) estimate wet N deposition at
2.5-5 kg N ha-1 yr-1, which is lower than our estimate of 7 kg N ha-1 yr-1, nevertheless close
to natural background levels. For Northern Tanzania a rate of 0.3 kg ha-1 yr-1 is reported
for wet P deposition (Tamatamah et al. 2005). Our study area is located on the coast of
the Indian Ocean, with prevailing winds coming from the Indian Ocean. Therefore dry
deposition is likely to be minimal, as was reported for a coastal savanna in Congo (Laclau
et al. 2005).
We calculated that N2-fixation by Acacia zanzibarica trees caused an annual N input to
the ecosystem of 4-10 kg N ha-1 yr-1, depending on the density of trees. The 15N natural
abundance method used to estimate N2-fixation has some limitations in field studies,
because it assumes that fixing and reference species have the same rooting profile, the
same mycorrhiza type, the same temporal N uptake pattern and the same preference for
different N forms. Where complementary data on rooting and N transformation patterns
are lacking, an accurate estimation of N2-fixation requires a difference between fixing and
reference species of at least ±5‰ (Högberg 1997), which was not the case in this study.
Thus, the estimates of N2-fixation reported here should be interpreted with care. However,
our estimates of the proportion of N derived from the atmosphere in Acacia zanzibarica
(58%) is remarkably similar to the mean value of 55% (± 15 SE) for other African Acacia
tree species and Dichrostachys cinerea (Cramer et al. 2007). The upper limit of our N input
rates in Acacia zanzibarica woodlands is also similar those reported for planted Acacia pellita
stands in Australia (c. 12 kg N ha-1 yr-1), but much lower than those for Acacia mangium
66
Nutrient cycling
and Acacia auriculiformis plantations in Congo (120-140 kg N ha-1 yr-1; (Bernhard-Reversat
1996). In the latter study, however, N fixation was a rough estimate based on the difference
of total N pools compared with adjacent pine and eucalypt plantations.
The two herbaceous legume species derived a mean of 94% of nitrogen by symbiotic N2fixation, this value being considerably higher than those reported for herbaceous legume
species in West Africa (68-80%; Becker and Johnson 1998; Sanginga et al. 1996). However,
because of the low annual production of these species (less than 27 kg ha-1 yr-1), the N input
from herbaceous legumes was very small in all vegetation types (< 0.5 kg N ha-1 yr-1). As
demonstrated in a competition experiment, the low abundance of herbaceous legumes in
savanna is due to the competitive superiority of C4 grasses (Chapter 3). Our rough estimate
for N2-fixation by herbaceous legumes shows that this N input is probably much lower than
has been assumed for some other savannas (e.g. Laclau et al. 2005; Sanhueza and Crutzen
1998).
Input in animal excreta only affect nutrient balances substantially in heavily used areas,
such as grazed patches (Table 5). Wildebeest are the most abundant herbivore species in
Saadani National Park and occur mainly in the southern part of the park (Treydte et al.
2005). During the period when dung was sampled the herds of wildebeest were absent
from the area where the grazed patches were located (Halsdorf, personal observation).
Thus, our data may underestimate annual dung and urine input in the grazed patches as
passing of wildebeest herds may constitute an irregular but important event. Deposition
of dung by bison and elk in Yellowstone National Park ranged between 0 and 410 kg ha1
yr-1 (Frank and McNaughton 1992), but we are unaware of comparable data for wild
herbivores in savanna. Dung deposition in the grazed patches of our study were with 390
± 140 kg ha-1 yr-1 comparable to the upper end of the range at Yellowstone. The P returned
by wild herbivores in grazed patches (c. 0.8 kg P ha-1 yr-1, Table 5) was much lower than the
estimated returns by cattle in paddocks (47 kg P ha-1 yr-1) or even in the surrounding areas
(2.3 kg P ha-1 yr-1; Appendix B).
In most of our savanna types herbivore density was very low and therefore nutrient removal
in offtake is likely to have been negligible (Table 5). However, when ungulates concentrate
their feeding on small patches, the amount of nutrients removed appears to be substantial,
possibly exceeding losses through fire (Table 5). This has also been suggested for a semi-arid
savanna (Van de Vijver et al. 1999), and is partly due to the high nutrient concentrations in
regrowth, which is repeatedly grazed (Hobbs 1996).
Measuring productivity of grazed patches by repeated clipping may be less accurate than
using the moveable cage method (Augustine et al. 2003) because grazing may stimulate
67
Chapter 2
or reduce growth compared to clipping. For Serengeti annual primary productivity under
grazing measured with moveable cages was on average 6’640 kg ha-1 yr-1 (2’000-14’000 kg
ha-1yr-1) compared with 3’750 kg ha-1 yr-1 (1’000-6’000 kg ha-1yr-1) in permanent exclosures
(McNaughton 1985). Thus, we may have underestimated the yield of herbage to grazers
by 40-50%. Nevertheless the estimates can be used for a rough estimate of the amount
of nutrients removed by herbivores which, to our knowledge, has not previously been
quantified. Our results suggest that intensive grazing by wild herbivores may have a negative
effect on nutrient pools in the long-term (Table 5).
Our hypothesis that availabilities of N and P in the zone around paddock margins would
be reduced because of high rates of biomass consumption by cattle was not supported by
our results. Based on information on cattle stock on the former Mkwaja ranch we estimated
the annual net loss of phosphorus from grazing areas during the ranch period (1955-1999)
at 0.34 kg ha-1 yr-1, of which 0.22 kg ha-1 yr-1 were re-distributed to paddocks and paddock
margins and 0.12 kg ha-1 yr-1 were lost as P exported in cow sales (Appendix B). Similar
losses from grazed areas were estimated for the above mentioned cattle ranch in Kenya
(0.21-0.26 kg P ha-1 yr-1; export in animals not considered) (Augustine 2003). Considering
the low levels of inorganic P in the soil (Table 1), such rates of loss appear high, though
atmospheric inputs may have been sufficient to compensate for them (see Table 5). Thus,
there may not have been depletion of P pools during cattle ranching.
In ungrazed savanna, the largest losses of nutrients are due to fire, especially in vegetation
such as tallgrass savanna with a high aboveground biomass. Except for the paddock centres,
losses through single fire events would range between 13 to 22 kg N ha-1 and between 0.4
to 1.0 kg P ha-1 based on the losses in our burning experiment (71% of N and 27% of P lost
from aboveground biomass). Losses of N and P from other savannas through single fire
events reported in literature were rather similar with ranges of 10-23 kg ha-1 (85-97%) for
N, and 0.4-1.6 kg ha-1 (13-61%) for P (Cook 1994; Kauffman et al. 1995; Laclau et al. 2002;
Pivello and Coutinho 1992; Van de Vijver et al. 1999; Villecourt et al. 1980). However,
the accumulated losses depend also upon fire frequency (Table 5). In national parks the
vegetation is commonly burnt in the early dry season to attract wildlife and to prevent
uncontrollable late dry season fires (Eriksen 2007) and we estimated losses resulting from
typical fire frequencies of one to two every three years. We note that in our experiments
combustion was generally not complete (215 ± 53 g m-2 residues of incompletely or unburnt
biomass), which might explain the lower percentage of N lost reported in this study (Raison
et al. 1985). The fact that absolute losses of N were in line with other studies may be due
to the higher N concentration in aboveground biomass in the early dry season than in the
late dry season (Laclau et al. 2002).
68
Nutrient cycling
The list of input and output fluxes presented in Table 5 is not exhaustive and some relevant
fluxes may be missing. For comparison we present some values from other areas: Additional
N input may come from non-symbiotic biological N2-fixation in soil crusts of cyanobacteria
or by free-living bacteria in the rhizosphere of grass roots (Döbereiner and Day 1976; Eisele
et al. 1989; Isichei 1980). Estimates of this process vary widely: Isichei (1980) estimated
fixation rates of 3-9 kg N ha-1 yr-1 for cyanobacterial crusts in Nigerian savanna, whereas
Aranibar et al. (2003a) estimated maximal rates of 0.04 kg N ha-1 yr-1 in Southern Africa,
and Sanhueza and Crutzen (1998) 1-2 kg N ha-1 yr-1 for Trachypogon savannas in Venezuela.
Furthermore, N2-fixation by rhizosphere associations is estimated at 1-8 kg N ha-1 yr-1 and
10 kg N ha-1 yr-1 for savannas in Venezuela (Sanhueza and Crutzen 1998) and Ivory Coast
(Balandreau 1976), respectively. For Southern African savannas N2-fixation by rhizosphere
associations could not be demonstrated (Bate and Gunton 1982). Nitrogen can be lost from
ecosystems through denitrification, but this flux is usually considered negligible for welldrained savanna soils (Chacon et al. 1991; Le Roux and Mordelet 1995; Sanhueza and
Crutzen 1998; Serca et al. 1998). Data on leaching losses for humid savannas is scarce and
vary widely: whereas these are likely to be minimal in semi-arid savanna (Bate and Gunton
1982; Dougill et al. 1998), measured losses were 5.6 kg N ha-1 yr-1 and 0.58 kg P ha-1 yr-1
in Lamto savanna (Villecourt and Roose 1978), and 3.0 kg N ha-1 yr-1 and 0.1 kg P ha-1 yr-1
in a savanna in Congo (Laclau et al. 2005). For a Trachypogon savanna N losses through
leaching were estimated at 1.2-3.0 kg ha-1 yr-1 (Chacon et al. 1991).
In most of our savanna types N and P losses through fire could only be compensated by
atmospheric deposition if the fire frequency were once every three years or less (Table 5).
In very nutrient-rich vegetation types, like the paddock centres, the large nutrient losses
through fire cannot be compensated by inputs (Table 5). Productive tallgrass sites are
therefore likely to suffer from net N losses, and fire frequencies of less than once every
three years would be required to achieve a neutral balance. Input of N through symbiotic
fixation is only substantial in Acacia woodlands, which may have a positive N balance,
especially in dense stands with high input through fixation and small losses through fire
because of a poor herbaceous cover (Table 5). Inputs of N and P through herbivore dung
and urine appear to be significant only in intensively used areas, such as grazed patches
(Table 5). However, our results indicate that in such areas nutrient losses through biomass
consumption by herbivores may be higher than inputs by dung and urine, possibly leading
to negative balances for both N and P. Our findings demonstrate that calculation of nutrient
balances at smaller spatial scales are helpful for understanding the functioning of savanna
ecosystems in the long run, and may provide a basis for avoiding overexploitation with
cattle or long-term nutrient losses through fire by ecosystem management.
69
Chapter 2
Conclusion
Short-term availabilities of nitrogen and phosphorus were generally very low, and the
effects of herbivory, N2-fixation and fire on some nutrient availability variables could be
reasonably well predicted, especially for P. Nutrient input and output fluxes were found
to be very low, indicating that nutrient-poor savanna ecosystems may be rather sensitive
to environmental pollution or human activities. Our results demonstrate that in the longterm, the main effect of domestic herbivores on nutrient availabilities was to re-distribute
large amounts of N and P, with the effect on P being more persistent. There was also a
considerable nutrient output through sold cows. Wild herbivores may also re-distribute
nutrients, e.g. from intensively grazed patches or nutrient rich Acacia woodlands to other
vegetation types or below trees, but we could not demonstrate it. Symbiotic N2-fixation by
stands of woody legumes are an important source of N to the savanna ecosystem, in contrast
to herbaceous legumes, which are too sparse to be significant. Fire causes large, uniform
losses of nutrients and may therefore reduce heterogeneity in nutrient pools. Finally, our
results lead us to question the common assumption that the N balance in frequently burned
savanna ecosystems is neutral.
Acknowledgements
This study was financed by the Swiss National Science Foundation grant No. 2-77502-04.
We thank Thomas Kuster for his assistance in the conduct and analysis of combustion
experiments. We thank Stephanie Halsdorf and Christoph Rohrer for collecting fresh
dung of almost all herbivore species in Saadani, as well as for the analyses of N and P
concentrations in it. We thank Sabine Güsewell for statistical advice. We thank Prof. S.
L. S. Maganga, Markus Schneider-Mmary, and the authorities of Saadani National Park
for their logistical support in Tanzania, and Benjamin Donald, John Williams and Hamis
Williams for their assistance in the field.
Literature cited
Abbadie L (1990) Carbon and nitrogen mineralization and denitrification in a humid savanna of
West Africa (Lamto, Côte d’Ivoire). Acta Oecologica 11:717-728
Abbadie L, Mariotti A, Menaut JC (1992) Independence of savanna grasses from soil organic
matter for their nitrogen supply. Ecology 73:608-613
70
Nutrient cycling
Aland A, Lidfors L, Ekesbo I (2002) Diurnal distribution of dairy cow defecation and urination.
Applied Animal Behaviour Science 78:43-54
Amarger N, Mariotti A, Mariotti F, Durr JC, Bourguignon C, Lagacherie B (1979) Estimate of
symbiotically fixed nitrogen in fieldgrown soybeans using variations in 15N natural abundance.
Plant and Soil 52:269-280
Aranibar JN, Anderson IC, Ringrose S, Macko SA (2003a) Importance of nitrogen fixation in
soil crusts of southern African arid ecosystems: acetylene reduction and stable isotope studies.
Journal of Arid Environments 54:345-358
Aranibar JN, Macko SA, Anderson IC, Potgieter ALF, Sowry R, Shugart HH (2003b) Nutrient
cycling responses to fire frequency in the Kruger National Park (South Africa) as indicated by
stable isotope analysis. Isotopes in Environmental and Health Studies 39:141-158
Archer S (2001) Trees in grasslands: biogeochemical consequences of woody plant expansion.
In: Schulze E-D, Harrison SP, Heimann M, Holland EA, Lloyd J, Prentice IC & Schimel D
(eds) Global biogeochemical cycles in the climate system. Academic Press, San Diego, USA,
pp 115-137
Archibald S, Bond WJ, Stock WD, Fairbanks DHK (2005) Shaping the landscape: fire-grazer
interactions in an African savanna. Ecological Applications 15:96-109
Augustine DJ (2003) Long-term, livestock-mediated redistribution of nitrogen and phosphorus in
an East African savanna. Journal of Applied Ecology 40:137-149
Augustine DJ (2004) Temporal asynchrony in soil nutrient dynamics and plant production in a
semiarid ecosystem. Ecosystems 7:829-840
Augustine DJ, McNaughton SJ (2006) Interactive effects of ungulate herbivores, soil fertility, and
variable rainfall on ecosystem processes in a semi-arid savanna. Ecosystems 9:1242-1256
Augustine DJ, McNaughton SJ, Frank DA (2003) Feedbacks between soil nutrients and large
herbivores in a managed savanna ecosystem. Ecological Applications 13:1325-1337
Balandreau J (1976) Rhizospheric nitrogen fixation (C2H2) in Lamto savanna. Revue d’Écologie et
de Biologie du Sol 13:529-544
Bate GC (1981) Nitrogen cycling in savanna ecosystems. In: Clark FE & Rosswall T (eds) Terrestrial
nitrogen cycles, Ecol Bull 33. Royal Swedish Academy of Sciences, Stockholm, pp 463-475
Bate GC, Gunton C (1982) Nitrogen in the Burkea savanna. In: Huntley BJ & Walker BH (eds)
Ecology of Tropical Savannas. Springer, Berlin, pp 498-513
Becker M, Johnson DE (1998) Legumes as dry season fallow in upland rice-based systems of West
Africa. Biology and Fertility of Soils 27:358-367
Bernhard-Reversat F (1996) Nitrogen cycling in tree plantations grown on a poor sandy savanna
soil in Congo. Applied Soil Ecology 4:161-172
Biondini ME, Patton BD, Nyren PE (1998) Grazing intensity and ecosystem processes in a northern
mixed-grass prairie, USA. Ecological Applications 8:469-479
Blaxter K (1980) The nutrient requirement of ruminant livestock. Agricultural Research Council
(Great Britain), Slough
Bond WJ, Woodward FI, Midgley GF (2005) The global distribution of ecosystems in a world
without fire. New Phytologist 165:525-537
Bray RH, Kurtz LT (1945) Determination of total, organic and available forms of phosphorus in
soils. Soil Science 59:39-45
Brookman-Amissah J, Hall JB, Swaine MD, Attakorah JY (1980) A re-assessment of a fire
protection experiment in northeastern Ghana savanna. Journal of Applied Ecology 17:85-99
71
Chapter 2
Calegare L, Alencar MM, Packer IU, Lanna DPD (2007) Energy requirements and cow/calf
efficiency of Nellore and Continental and British Bos taurus × Nellore crosses. Journal of Animal
Science 85:2413-2422
Chacon P, Lopezhernandez ID, Lamotte M (1991) Nitrogen-cycle in a Trachypogon savanna in
central Venezuela. Revue d’Écologie et de Biologie du Sol 28:67-75
Cochard R (2004) Pattern and dynamics of secondary Acacia zanzibarica woodlands at Mkwaja
Ranch, Tanzania. PhD thesis, ETH Zurich, Zurich
Cook GD (1994) The fate of nutrients during fires in a tropical savanna. Australian Journal of
Ecology 19:359-365
Cramer MD, Chimphango SBM, Van Cauter A, Waldram MS, Bond W (2007) Grass competition
induces N2 fixation in some species of African Acacia. Journal of Ecology 95:1123-1133
de Rham P (1973) Recherches sur la minéralisation de l’azote dans les sols des savanes de Lamto
(Côte d’Ivoire). Revue d’Écologie et de Biologie du Sol 10
DeFaria SM, Lewis GP, Sprent JI, Sutherland JM (1989) Occurrence of nodulation in the
Leguminosae. New Phytologist 111:607-619
Döbereiner J, Day JM (1976) Associative symbioses in tropical grasses: characterization of
microorganisms and dinitrogen-fixing sites. In: Newton WE & Nyman CJ (eds) Proceedings
of the 1st International Symposium on Nitrogen Fixation. Vol. 2. Washington State University
Press, Pullman, Washington, pp 518-538
Dougill AJ, Heathwaite AL, Thomas DSG (1998) Soil water movement and nutrient cycling in semiarid rangeland: vegetation change and system resilience. Hydrological Processes 12:443-459
Edwards PJ, Hollis S (1982) The distribution of excreta on New Forest grassland used by cattle,
ponies and deer. Journal of Applied Ecology 19:953-964
Eisele KA, Schimel DS, Kapustka LA, Parton WJ (1989) Effects of available P and N:P ratios on
non-symbiotic dinitrogen fixation in tallgrass prairie soils. Oecologia 79:471-474
Eriksen C (2007) Why do they burn the ‘bush’? Fire, rural livelihoods, and conservation in Zambia.
Geographical Journal 173:242-256
Frank DA, Groffman PM (1998) Ungulate vs. landscape control of soil C and N processes in
grasslands of Yellowstone National Park. Ecology 79:2229-2241
Frank DA, McNaughton SJ (1992) The ecology of plants, large mammalian herbivores, and drought
in Yellostone National Park. Ecology 73:2043-2058
Frank DA, Zhang YM (1997) Ammonia volatilization from a seasonally and spatially variable
grazed grassland: Yellowstone National Park. Biogeochemistry 36:189-203
Fynn RWS, Haynes RJ, O’Connor TG (2003) Burning causes long-term changes in soil organic
matter content of a South African grassland. Soil Biology & Biochemistry 35:677-687
Galloway JN, Dentener FJ, Capone DG, Boyer EW, Howarth RW, Seitzinger SP, Asner GP,
Cleveland CC, Green PA, Holland EA, Karl DM, Michaels AF, Porter JH, Townsend AR,
Vörösmarty CJ (2004) Nitrogen cycles: past, present, and future. Biogeochemistry 70:153-226
Garstang M, Ellery WN, McCarthy TS, Scholes MC, Scholes RJ, Swap RJ, Tyson PD (1998) The
contribution of aerosol- and water-borne nutrients to the functioning of the Okavango Delta
ecosystem, Botswana. South African Journal of Science 94:223-229
Geesing D, Felker P, Bingham RL (2000) Influence of mesquite (Prosopis glandulosa) on soil
nitrogen and carbon development: Implications for global carbon sequestration. Journal of Arid
Environments 46:157-180
72
Nutrient cycling
Güsewell S (2002) Time-dependent effect of fertilization on plant biomass in floating fens. Journal
of Vegetation Science 13:795-718
Hansen JP, Vinther FP (2001) Spatial variability of symbiotic N2 fixation in grass-white clover
pastures estimated by the 15N isotope dilution method and the natural 15N abundance method.
Plant and Soil 230:257-266
Harrington GN (1974) Fire effects on a Ugandan savanna grassland. Tropical Grasslands
8:87-101
Hobbs NT (1996) Modification of ecosystems by ungulates. Journal of Wildlife Management
60:695-713
Hobbs NT, Schimel DS, Owensby CE, Ojima DS (1991) Fire and grazing in the tallgrass prairie:
contingent effects on nitrogen budgets. Ecology 72:1374-1382
Högberg P (1997) 15N natural abundance in soil-plant systems. New Phytologist:179-203
Holland EA, Detling JK (1990) Plant response to herbivory and belowground nitrogen cycling.
Ecology 71:1040-1049
Hudak AT (1999) Rangeland mismanagement in South Africa: Failure to apply ecological
knowledge. Human Ecology 27:55-78
Hunt J (1982) Dilute hydrochloric acid extraction of plant material for routine cation analysis.
Communications in Soil Science and Plant Analysis 13:49-55
Isichei AO (1980) Nitrogen fixation by blue-green algal soil crusts in Nigerian savanna. In: Rosswall
T (ed) Nitrogen cycling in West African ecosystems. Royal Swedish Academy of Sciences,
Stockholm, pp 191-198
Jewell PL, Kauferle D, Gusewell S, Berry NR, Kreuzer M, Edwards PJ (2007) Redistribution of
phosphorus by mountain pasture in cattle on a traditional the Alps. Agriculture Ecosystems &
Environment 122:377-386
Kauffman JB, Cummings DL, Ward DE, Babbitt R (1995) Fire in the Brazilian Amazon: 1.
Biomass, nutrient pools, and losses in slashed primary forests. Oecologia 104:397-408
Klötzli F (1980) Analysis of species oscillations in tropical grasslands in Tanzania due to
management and weather conditions. Phytocoenologia 8:13-33
Laclau JP (2003) Nutrient cycling in a clonal stand of Eucalyptus and an adjacent savanna
ecosystem in Congo 2. Chemical composition of soil solutions. Forest Ecology and Management
180:527-544
Laclau JP, Ranger J, Deleporte P, Nouvellon Y, Saint-Andre L, Marlet S, Bouillet JP (2005) Nutrient
cycling in a clonal stand of Eucalyptus and an adjacent savanna ecosystem in Congo 3. Inputoutput budgets and consequences for the sustainability of the plantations. Forest Ecology And
Management 210:375-391
Laclau JP, Sama-Poumba W, Nzila JD, Bouillet JP, Ranger J (2002) Biomass and nutrient dynamics
in a littoral savanna subjected to annual fires in Congo. Acta Oecologica 23:41-50
Le Roux X, Mordelet P (1995) Leaf and canopy CO2 assimilation in a West African humid savanna
during the early growing season. Journal of Tropical Ecology 11:529-545
Ludwig F, de Kroon H, Berendse F, Prins HHT (2004) The influence of savanna trees on nutrient,
water and light availability and the understorey vegetation. Plant Ecology 170:93-105
McDowell LR (1985) Calcium, phosphorus and fluorine. In: McDowell LR (ed) Nutrition of
grazing ruminants in warm climates. Academic Press, London, pp 189-212
73
Chapter 2
McNaughton SJ (1985) Ecology of a grazing ecosystem: the Serengeti. Ecological Monographs
55:259-294
McNaughton SJ, Banyikwa FF, McNaughton MM (1997) Promotion of the cycling of dietenhancing nutrients by African grazers. Science 278:1798-1800
Medina E (1987) Nutrients: requirements, conservation and cycles in the herbaceous layer. In:
Walker BH (ed) Determinants of tropical savannas. IRL Press, Oxford, pp 39-65
Medina E, Bilbao B (1991) Significance of nutrient relations and symbiosis for the competitive
interaction between grasses and legumes in tropical savannas. In: Esser G & Overdieck D (eds)
Modern ecology: basic and applied aspects. Elsevier, Amsterdam, pp 295-319
Minson DJ, McDonald CK (1987) Estimating forage intake from the growth of beef cattle. Tropical
Grasslands 21:116-122
Montes R, San Jose JJ (1989) Chemical composition and nutrient loading by precipitation in the
Trachypogon savannas of the Orinoco llanos, Venzuela. Biogeochemistry 7:241-256
Morse D, Head HH, Wilcox CJ, van Horn HH, Hissem CD, Harris B (1992) Effects of
concentration of dietary phosphorus on amount of and route of excretion. Journal of Dairy
Science 75:3039-3049
Ojima DS, Schimel DS, Parton WJ, Owensby CE (1994) Long-term and short-term effects of fire
on nitrogen cycling in tallgrass prairie. Biogeochemistry 24:67-84
Pivello VR, Coutinho LM (1992) Transfer of macro-nutrients to the atmosphere during experimental
burnings in an open cerrado (Brazilian savanna). Journal of Tropical Ecology 8:487-497
Raison RJ, Khanna PK, Woods PV (1985) Mechanisms of Element Transfer to the Atmosphere
During Vegetation Fires. Canadian Journal of Forest Research-Revue Canadienne De Recherche
Forestiere 15:132-140
Reich PB, Peterson DW, Wedin DA, Wrage K (2001) Fire and vegetation effects on productivity
and nitrogen cycling across a forest-grassland continuum. Ecology 82:1703-1719
Ruess RW, McNaughton SJ (1987) Grazing and the dynamics of nutrient and energy regulated
microbial processes in the Serengeti grasslands. Oikos 49:101-110
Ruess RW, McNaughton SJ (1988) Ammonia volatilization and the effects of large grazing
mammals on nutrient loss from East African grasslands. Oecologia 77:382-386
Sanginga N, Wirkom LE, Okogun A, Akobundu IO, Carsky RJ, Tian G (1996) Nodulation and
estimation of symbiotic nitrogen fixation by herbaceous and shrub legumes in Guinea savanna
in Nigeria. Biology and Fertility of Soils 23:441-448
Sanhueza E, Crutzen PJ (1998) Budgets of fixed nitrogen in the Orinoco savannah region: Role of
pyrodenitrification. Global Biogeochemical Cycles 12:653-666
Scholefield D, Lockyer DR, Whitehead DC, Tyson KC (1991) A model to predict transformations
and losses of nitrogen in UK pastures grazed by beef cattle. Plant and Soil 132:165-177
Scholes MC, Scholes RJ, Otter LB, Woghiren AJ (2003) Biogeochemistry: the cycling of elements.
In: du Toit JT, Rogers KH & VBiggs HC (eds) The Kruger experience: ecology and management
of savanna heterogeneity. Island Press, Washington, pp 130-148
Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annual Review of Ecology and
Systematics 28:517-544
Serca D, Delmas R, Le Roux X, Parsons DAB, Scholes MC, Abbadie L, Lensi R, Ronce O, Labroue
L (1998) Comparison of nitrogen monoxide emissions from several African tropical ecosystems
and influence of season and fire. Global Biogeochemical Cycles 12:637-651
74
Nutrient cycling
Singh RS, Srivastava SC, Raghubanshi AS, Singh JS, Singh SP (1991) Microbial C, N and P in dry
tropical savanna: effects of burning and grazing. Journal of Applied Ecology 28:869-878
Stock WD, Wienand KT, Baker AC (1995) Impacts of invading N2-fixing Acacia species on patterns
of nutrient cycling in 2 Cape ecosystems: evidence from soil incubation studies and 15N natural
abundance values. Oecologia 101:375-382
Tamatamah RA, Hecky RE, Duthie HC (2005) The atmospheric deposition of phosphorus in Lake
Victoria (East Africa). Biogeochemistry 73:325-344
Tobler MW, Cochard R, Edwards PJ (2003) The impact of cattle ranching on large-scale vegetation
patterns in a coastal savanna in Tanzania. Journal of Applied Ecology 40:430-444
Trail JCM, Sones K, Jibbo JMC, Durkin J, Light DE, Murray M (1985) Productivity of Boran
cattle maintained by chemoprophylaxis under trypanosomiasis risk. ILCA Research Reports.
ILCA, Addis Ababa, Ethiopia
Treydte AC, Bernasconi SM, Kreuzer M, Edwards PJ (2006) Diet of the common warthog
(Phacochoerus africanus) on former cattle grounds in a Tanzanian savanna. Journal of Mammalogy
87:889-898
Treydte AC, Edwards PJ, Suter W (2005) Shifts in native ungulate communities on a former cattle
ranch in Tanzania. African Journal of Ecology 43:302-311
Turner MD (1998) Long-term effects of daily grazing orbits on nutrient availability in Sahelian West
Africa: 2. Effects of a phosphorus gradient on spatial patterns of annual grassland production.
Journal of Biogeography 25:683-694
Van de Vijver CADM, Poot P, Prins HHT (1999) Causes of increased nutrient concentrations in
post-fire regrowth in an East African savanna. Plant and Soil 214:173-185
Villecourt P, Roose E (1978) Charge en azote et en éléments minéraux majeurs des eaux de pluie,
de pluviolessivage et de drainage dans la savane de Lamto (Côte d’Ivoire). Revue d’Écologie et
de Biologie du Sol 15:1-20
Villecourt P, Schmidt W, Cesar J (1980) Losses of ecosystem during bush fire (tropical savanna of
Lamto, Ivory Coast). Revue d’Écologie et de Biologie du Sol 17:7-12
Yelenik SG, Stock WD, Richardson DM (2004) Ecosystem level impacts of invasive Acacia saligna
in the South African fynbos. Restoration Ecology 12:44-51
75
Chapter 2
Appendix A
Nutrient pools in topsoil (10 cm) and aboveground vegetation, and aboveground production along
three paddock gradients on former cattle ranch grounds and at three tallgrass savanna and three
grazed sites in the former Saadani game reserve. Data shows means per site (± SE, n=5 for soil
data, n=4 for biomass related data). Production data in italics indicates that the sites had burned
during the measurement period. PC: paddock centres, PM: paddock margins, HS: Heteropogon
savanna, AW: Acacia woodland, TGM: tallgrass savanna of the former Mkwaja ranch, TGS: tallgrass
savanna in the former Saadani game reserve, GP: grazed patches. For vegetation types within the
same paddock gradient values not sharing the same letter are significantly different (Tukey-HSD,
P < 0.05).
PC
PM
Paddock gradients
HS
Total soil C (kg m-2)
gradient/site 1 2.26 ± 0.52AB 0.71 ± 0.08C 2.50 ± 0.12A
gradient/site 2 4.01 ± 0.56A 0.77 ± 0.04C 0.64 ± 0.04CD
gradient/site 3 5.08 ± 0.73A 1.02 ± 0.09C 1.78 ± 0.18B
Total soil N (g m-2)
gradient/site 1
222 ± 54A
62 ± 7C
159 ± 6AB
67 ± 4C
52 ± 2CD
gradient/site 2
394 ± 52A
A
BC
gradient/site 3
486 ± 62
90 ± 7
136 ± 24B
-2
Total soil P (g m )
4.9 ± 0.4D
11.6 ± 0.6B
gradient/site 1
45 ± 9A
gradient/site 2
104 ± 16A
8.2 ± 0.6B
5.0 ± 0.3CD
gradient/site 3
144 ±15A
10.1 ± 0.6B
9.1 ± 0.4B
-2
Extractable soil N Jul 2005 (g m )
gradient/site 1
5.8 ± 1.5A
0.26 ± 0.04B
0.29 ± 0.05B
gradient/site 2
1.7 ± 0.3A
0.16 ± 0.05B
0.08 ± 0.04B
0.44 ± 0.11B
0.20 ± 0.03B
gradient/site 3
10.1 ± 1.3A
Extractable soil N Nov 2005 (g m-2)
gradient/site 1
5.1 ± 2.9A
0.29 ± 0.04B
0.72 ± 0.13B
gradient/site 2
0.6 ± 0.1A§
0.27 ± 0.01B
0.25 ± 0.03B
gradient/site 3
12.4 ± 2.4A 0.39 ± 0.06CD 0.91 ± 0.08B
Extractable soil P Jul 2005 (g m-2)
gradient/site 1
27 ± 7A
0.44 ± 0.05B 0.20 ± 0.03C
A
gradient/site 2
21 ± 3
0.92 ± 0.20B 0.21 ± 0.01C
0.82 ± 0.18B 0.16 ± 0.01C
gradient/site 3
106 ± 3A
-2
Extractable soil P Nov 2005 (g m )
gradient/site 1
32 ± 5A
0.36 ± 0.04B
0.26 ± 0.04B
1.29 ± 0.07B 0.22 ± 0.02C
gradient/site 2
64 ± 12A§
gradient/site 3
100 ± 12A
0.79 ± 0.12B 0.15 ± 0.04C
Aboveground herbaceous biomass July 2005 (kg m-2)
gradient/site 1 1.62 ± 0.11A 0.50 ± 0.09C 0.64 ± 0.08BC
gradient/site 2 1.53 ± 0.15A 0.48 ± 0.07BC 0.32 ± 0.02C
gradient/site 3 1.17 ± 0.07A
0.69 ±0.09B
0.40 ± 0.03C
N in aboveground herbaceous biomass July 2005 (g m-2)
gradient/site 1
21 ± 2A
2.1 ± 0.3BC
2.9 ± 0.3B
A
C
2.0 ± 0.3
1.2 ± 0.1C
gradient/site 2
18 ± 2
2.7 ± 0.4BC
1.7 ± 0.1CD
gradient/site 3
18 ± 1A
P in aboveground herbaceous biomass July 2005 (g m-2)
0.17 ± 0.02C 0.21 ± 0.02C
gradient/site 1
3.4 ± 0.1A
gradient/site 2
2.7 ± 0.2A
0.3 ± 0.04B
0.07 ± 0.00D
gradient/site 3
3.2 ± 0.3A
0.53 ± 0.12B 0.11 ± 0.01CD
Herbaceous production July-Dec 2005 (g m-2)
gradient/site 1
393 ± 21A
113 ± 14BC
117 ± 13BC
41 ± 4C
gradient/site 2
burned
103 ± 15AB
gradient/site 3
189 ± 25A
70 ± 3B
37 ± 7B
§
AW
1.46 ± 0.08B
1.54 ± 0.22B
1.98 ± 0.34B
113 ± 6B
115 ± 14B
142 ± 22B
Saadani sites
TGS
GP
TGM
0.78 ± 0.03C | 0.95 ± 0.07
0.45 ± 0.03D | 0.91 ± 0.05
0.95 ± 0.08C | 0.81 ± 0.02
60 ± 3C
37 ± 2D
68 ± 6C
|
|
|
69 ± 5
72 ± 5
57 ± 3
0.99 ± 0.32
1.66 ± 0.41
0.84 ± 0.12
73 ± 19
124 ± 29
69 ± 9
9.5 ± 0.3BC
7.5 ± 0.9BC
10.6 ± 1.6B
6.2 ± 0.1CD | 7.5 ± 0.6
3.6 ± 0.2D | 7.3 ± 0.4
6.1 ± 0.3C | 7.1 ± 0.2
8.4 ± 0.7
16.4 ± 1.1
8.3 ± 0.8
0.40 ± 0.07B
0.28 ± 0.10B
0.55 ± 0.13B
0.24 ± 0.05B | 0.43 ± 0.08
0.13 ± 0.04B | 0.28 ± 0.06
0.27 ± 0.05B | 0.18 ± 0.03
0.31 ± 0.05
0.68 ± 0.13
0.34 ± 0.06
0.48 ± 0.06B
0.65 ± 0.09A
0.63 ± 0.09BC
0.45 ± 0.05B | 0.21 ± 0.08
0.14 ± 0.02C | 0.08 ± 0.02
0.26 ± 0.02D | 0.08 ± 0.01
0.11 ± 0.03
0.59 ± 0.04
0.21 ± 0.03
0.30 ± 0.03BC
0.37 ± 0.18C
0.19 ± 0.02C
0.36 ± 0.04BC | 0.26 ± 0.06
0.15 ± 0.03C | 0.22 ± 0.01
0.53 ± 0.16BC | 0.47 ± 0.03
1.05 ± 0.18
0.96 ± 0.10
0.46 ± 0.04
0.30 ± 0.02B
0.30 ± 0.04C
0.16 ± 0.03C
0.39 ± 0.03B | 0.18 ± 0.02
0.18 ± 0.01C | 0.13 ± 0.01
0.45 ± 0.14B | 0.31 ± 0.02
0.61 ± 0.07
0.76 ± 0.09
0.34 ± 0.05
0.13 ± 0.03D
0.75 ± 0.17B
0.18 ± 0.04C
1.22 ± 0.25AB | 0.83 ± 0.18
0.61 ± 0.06B | 0.73 ± 0.14
1.10 ± 0.05A | 1.04 ± 0.11
0.26 ± 0.05
0.07 ± 0.01
0.17 ± 0.00
0.9 ± 0.3C
4.3 ± 1.0B
1.3 ± 0.3D
0.04 ± 0.01D
0.27 ± 0.18BC
0.07 ± 0.02D
50 ± 12C
84 ± 23BC
43 ± 8B
4.2 ± 0.6B
1.7 ± 0.1C
3.8 ± 0.4B
| 2.8 ± 0.5
| 2.3 ± 0.3
| 2.5 ± 0.4
0.59 ± 0.13B | 0.23 ± 0.02
0.13 ± 0.01CD | 0.18 ± 0.03
0.34 ± 0.03BC | 0.37 ± 0.07
153 ± 47B
153 ± 14A
188 ± 22A
| 131 ± 22
| 160 ± 12
| 384 ± 28
1.5 ± 0.2
0.7 ± 0.1
1.1 ± 0.1
0.24 ± 0.04
0.07 ± 0.01
0.08 ± 0.01
149 ± 12
118 ± 26
130 ± 9
The PC site of gradient 2 burned in November 2005, for measurements of inorganic N pools an unburned site at c. 50 m distance had
to be selected.
76
Nutrient cycling
Appendix B
Calculation of total dry matter and phosphorus intake by cattle in the northern part of
Saadani National Park (Mkwaja) during the ranch period (1955- 1999), and re-distribution
to focal points based on P surplus in the soil observed in this study.
Table B1 Ranch area, and cow production on the former Mkwaja Ranch:
Variable
2
Comments and source
Mkwaja Ranch
Total area
Total area of paddocks
Total area of paddock margins
Total area available for grazing
46’200 ha
92 ha
1’432 ha
34’650 ha
Tobler et al. (2003)
0.2% of total area (Treydte et al. 2006b)
3.1% of total area (Treydte et al. 2006b)
75% of total area (Trail et al. 1985)
Stock and productivity
Average total stock
Average No. of cows with calves
Average No. of cows without calves
Average No. of fattening animals
Average net sales per year
Weight at weaning
Mature cow weight
Growth rate
DMI cows without calf
10’200
2’500
1’400
3’800
1’600
133 kg
286 kg
120 g d-1
5.2 kg d-1
ranch records 1955-1999
1 = cow + calf, ranch records 1955-1999
ranch records 1955-1999
ranch records 1955-1999
ranch records 1955-1999
Trail et al. (1985)
Trail et al. (1985)
calculated1
Minson and McDonald (1987): DMI for body
weight maintenance for a 290 kg cow
calculated2
Minson and McDonald (1987): DMI for an
average fattening animal in Mkwaja (210 kg) at
0.1 kg d-1 growth rate
DMI cow + calf
DMI fattening animals
1
Value
7.9 kg d-1
4.2 kg d-1
fattening time was on average 3.5 years (Trail et al. 1985)
pregnant cows require 29% more energy in the last 4 months of pregnancy and during the 8 months until weaning the energy
requirement of the cow and her calf are 63% higher than maintenance requirements of the cow (calculated from Calgare et al. 2007);
thus over a period of one year a cow and its calf require 52% more dry matter intake (DMI) than required for maintenance
Table B2 Total stock of phosphorus in the soil (0-70 cm) and excess P in focal areas of cattle
compared to tallgrass savanna areas measured in 2005 (measured soil P values are based on data
from Figure 5).
Variable
Measured variables
Total soil P (kg ha-1)
Calculated variables
Excess P1 (kg ha-1)
Total excess P (t)
P accumulation (kg ha-1 yr-1)
1
Paddocks
Paddock margins
Tallgrass savanna
2400
460
360
2040
188.5
46.6
100
143.2
2.3
0
0
-
Excess P was calculated as difference of total P stock of paddock centres and paddock margins and total P stock in tallgrass savanna
(background P stock); calculated P surplus would increase by only 6% if data measured from to a depth of 70 cm are extrapolated to
a depth of 200 cm.
77
Chapter 2
Table B3 Total and annual dry matter and phosphorus intake by cattle at Mkwaja Ranch,
re-distribution to focal points, and losses of P through sales
DMI by cattle1
P intake by cattle2
P re-distributed to focal areas
P re-distributed to grazing areas3
Loss of P through sales4
Net P loss
Total for ranch
in total
per year
693 kt
15.7 kt
831 t
18.9 t
332 t
7.5 t
322 t
7.4 t
177 t
4.0 t
177 t
4.0 t
Per area available for grazing
in total
per year
-1
20 t ha
454 kg ha-1
-1
24.0 kg ha
0.55 kg ha-1
-1
9.6 kg ha
0.22 kg ha-1
-1
9.3 kg ha
0.21 kg ha-1
-1
5.1 kg ha
0.12 kg ha-1
-1
14.7 kg ha
0.34 kg ha-1
1
calculations were done using data from yearly stock records rather than averages for 44 years showed further above
average P concentration of forage was estimated from average concentration of 0.12% P in regrowth in tallgrass savanna measured
in 2006 (Chapter 1), which compares to similarly low levels reported for forage in other tropical areas (McDowell 1985)
3
calculated as P intake minus P re-distributed to focal areas and P losses through sales
4
P content of mature cows is 0.88% of live weight (Blaxter 1980)
2
Some conclusions and discussion based on calculations with data from Tables B1-B3:
1. About 23% of all P ingested by cattle was re-distributed into paddocks (resting areas),
17% to areas around paddocks and 39% to grazing areas. The latter 39% were, however,
not likely distributed evenly, with returns around watering places probably higher than on
pasture areas.
2. Annual losses of P from the ranch through cattle sales during the ranch period
corresponded to 3.4% of total P in aboveground biomass of tallgrass savanna measured in
this study (Table 1).
3. According to our calculations only 32% of P re-distributed in dung went into paddocks
(resting areas), in which cows spent 12 hours (18h-6h). This agrees with the observation
that cows defecate less during resting time (Aland et al. 2002).
78
Legume-grass competition
Chapter 3
Why are herbaceous legumes scarce in savanna?
A grass-legume competition experiment
Patrick G. Cech, Peter J. Edwards, and Harry Olde Venterink
Institute of Integrative Biology, ETH Zürich, Universitätsstrasse 16, 8092, Zürich, Switzerland
79
Chapter 3
Abstract
Although large amounts of nitrogen are lost through fire in tropical savanna ecosystems,
N2-fixing herbaceous legumes – which might be expected to benefit from low nitrogen
conditions – are usually not abundant. To investigate possible reasons for the low proportion
of herbaceous legumes, we conducted a pot experiment in which Cassia mimosoides, a
common legume of humid African savanna, was grown with and without competition
from a C4 grass, Hyperthelia dissoluta, at different levels of water, nitrogen and phosphorus.
At high water supply, yield of both the legume and grass monocultures was significantly
increased only by the combined addition of nitrogen and phosphorus. Addition of N alone
stimulated legume growth, whereas P had no effect. For the grass, the supply of N had a
stronger effect on biomass production than did the supply of P. The legume suffered from
interspecific competition with the grass when no P was supplied, but benefited when P was
added. In the presence of the grass, N2-fixation of the legume was increased by 32% when
no nutrients were added, and by 56% when P was supplied. Even when no N was supplied
to mixtures, the N content of the grass increased and the foliar δ15N values decreased,
indicating a rapid transfer of nitrogen from the legume to the grass. Nitrogen use efficiency
of the grass, expressed as biomass production per unit of nitrogen acquired, was three times
larger than that of the legume. Water stress reduced legume growth by 17%, but this effect
was of similar magnitude in the grass species. Our results suggest that the low abundance
of herbaceous legumes in savannas is not due to the low phosphorus availability as such,
but to the greater ability of C4 grasses to compete for nutrients and their higher nutrient
use efficiency. We conclude that symbiotic N2-fixation in tropical savannas does not have
the same potential as in some other ecosystems to compensate for nitrogen lost through
volatilization.
Keywords: C4 grass, Cassia mimosoides, fire, Hyperthelia dissoluta, Hyparrhenia dissoluta,
legume, nitrogen fixation, nitrogen use efficiency, phosphorus, savanna, water
80
Legume-grass competition
Introduction
Savanna is a vegetation type characterized by the co-existence of trees and grasses (Scholes
and Archer 1997). The main factors influencing savanna ecosystems, which cover one fifth
of the world’s land surface, are strong seasonality in water availability, herbivory and fire
(Bond et al. 2005; Keeley and Rundel 2005; Sankaran et al. 2005).
Many tropical savannas have infertile soils (Medina 1987), which is often because
nutrients have been lost from the ecosystem through recurrent fire. Because the proportion
of nitrogen lost by burning is particularly high (Cook 1994; Pivello and Coutinho 1992;
Villecourt et al. 1980), it is commonly assumed that legumes and other plants with access
to atmospheric N through symbiotic N2-fixation have a competitive advantage in places
where fires are frequent (Medina and Bilbao 1991; Vitousek and Field 1999; Vitousek
and Howarth 1991). A high abundance of legumes has, indeed, been reported in fireprone Mediterranean ecosystems (Bell and Koch 1980; Kazanis and Arianoutsou 1994)
and American burned pine ecosystems (Hendricks and Boring 1999). Vitousek and Field
(1999) have proposed a simple model showing how legumes may co-exist with non-fixing
species under conditions of frequent fires. However, in tropical savannas, which are the
most frequently burnt ecosystems in the world (Bond et al. 2005), herbaceous legumes
are usually scarce, often making up less than 1% of the biomass (Ezedinma et al. 1979;
Huntley 1982; Isichei 1995; Laclau et al. 2002; Medina 1987; Menaut and Cesar 1979;
San Jose et al. 1985). A similarly low abundance of herbaceous legumes has also been
reported for prairies and semi-arid grasslands in the USA (Collins et al. 1995; Ritchie and
Tilman 1995; Woodmansee 1978). While the woody legumes of savannas, especially trees
in the genus Acacia, have been relatively well studied, much less is known about herbaceous
legumes and the factors that determine their abundance.
We can think of four possible explanations for the paucity of herbaceous legumes in
tropical savannas. First, N2-fixation and growth of leguminous plants may be limited by
the availability of another nutrient, perhaps phosphorus, as has been demonstrated for
tropical ecosystems such as rain forest (Döbereiner 1978; Medina 1982; Perreijn 2002;
Vitousek et al. 2002). Second, herbaceous legumes may be outcompeted by grasses.
Grasses usually dominate over legumes in grassland vegetation (del-Val and Crawley 2005;
Hu and Jones 2001; 2004; Norman 1966; Van Auken 1994), though the reasons for their
competitive superiority remain uncertain. There is evidence indicating that grasses are
stronger competitors for nutrients (Hu and Jones 2001; Norman 1966; Van Auken 1994;
81
Chapter 3
Viera-Vargas et al. 1995), but a higher nitrogen-use-efficiency of grasses may also play a
role (Del Pozo et al. 2000; Ludlow 1985; Tjoelker et al. 2005). Third, growth of herbaceous
legumes may be limited by lack of water (Hendricks and Boring 1999; Khadka and Tatsumi
2006; McKey 1994; Medina 1987). Fourth, legumes may be preferred by herbivores due to
the high nutritional quality of their tissues (Hulme 1996; Ritchie and Tilman 1995). These
effects, and particularly interactions among them, have been poorly studied, particularly for
non-crop species in tropical savannas. Knowledge about the constraints to symbiotic N2fixation by herbaceous legumes is, however, important for understanding vegetation and
ecosystem level properties such as species composition, nutrient turnover and resilience.
We performed a pot experiment using Cassia mimosoides, a widespread herbaceous legume
of African savannas, to investigate whether the scarcity of this species can be explained by
specific soil conditions, for example low water or phosphorus availability, or by interspecific
competition with C4 grasses. Selective herbivory, the fourth of the proposed constraints
upon legume abundance, was not considered. Specifically, the experiment was designed
to test whether growth and symbiotic N2-fixation of C. mimosoides were limited by the
supply of water, nitrogen or phosphorus, or whether growth was more constrained by the
presence of the competing grass, Hyperthelia dissoluta (one of the common tall C4 grasses in
African savannas). We hypothesized that competition is the principal ecological constraint
to legume growth, and that the grass lowers the availability of water and nutrients, thereby
reducing N2-fixation and growth by the legume.
Methods
Plants and soil
The reference system for our study was a tallgrass plant community that occurs commonly
in moist coastal savanna in Tanzania (5° 43′ S, 38° 47′ E). This community has a high
biomass of several C4 grasses including Hyperthelia dissoluta (synonym Hyparrhenia dissoluta),
while herbaceous legume biomass never accounts for more than 0.8% of the total (Chapter
2). The most common legume species is Cassia mimosoides (subfamily Caesalpinioideae).
C. mimosoides plants in the field are nodulated, and the δ15N values of their leaves indicate
that they fix a substantial proportion of their nitrogen (Chapter 2). The tallgrass communities
are burned once or twice every two years. For the experiment, seedlings of C. mimosoides
and small tillers of H. dissoluta were collected from a tallgrass site which was regrowing
after a recent fire. Topsoil (0-10 cm) was collected from a nearby unburnt patch, and sieved
before use. Total N and P contents of the soil as well as extractable N and P were very low
(Appendix A).
82
Legume-grass competition
Experimental design and treatments
The experiment was carried out at our temporary research station in Saadani National
Park. Plants were grown outside in 1-L pots (10 cm diameter; 1.4 kg soil) and placed
beneath a protective roof so that they received four hours of direct sunlight per day. Each
pot contained four plants in a replacement design: four C. mimosoides or four H. dissoluta
(monocultures), or two plants of each species (competition treatment). At planting,
C. mimosoides seedlings had three to five nodules, with any additional nodules being
removed. H. dissoluta individuals were cut to 30 cm height to reduce variation in biomass
among individuals. For the first seven days, plants were protected from direct sunlight and
watered to saturation daily in order to ensure root establishment.
The experiment consisted of a full factorial design with three species combinations (the two
monocultures and competition), two water treatments, and four nutrient treatments. Each
treatment combination was replicated six times. Half of the pots were watered every day
while the other half were watered every second or third day in order to achieve short but
frequent periods drought. From day 8 until harvest on day 45, pots assigned to the water
stress treatment were given 340 ml water in total, compared to 800 ml to pots not exposed
to water stress. The four nutrient treatments consisted of solutions of nitrogen supplied
as NH4NO3 (+N), of phosphorus supplied as Na2HPO4 (+P), of a combination of both
(+NP), and of only water (control). Fertilized pots received a total of 360 mg N and 36 mg
P, of which one sixth was added on day 8, one third on day 15, and one half on day 34; this
arrangement was intended to allow for increasing nutrient demand of the growing plants.
The fertilizer solution was applied after watering the pots to saturation. The arrangement
of the pots was randomized three times during the experiment.
Harvest and chemical analyses
After 45 days, shoots and roots of C. mimosoides and H. dissoluta in each pot were harvested
separately, and the number of nodules on C. mimosoides roots was counted. Plant material
was dried to constant weight at 70° C, weighed and ground. For a subset of samples,
δ15Ν values in shoot tissue were analysed using an elemental analyzer (NCS-2500, Carlo
Erba, Milan, Italy) coupled in continuous flow to an ion ratio mass spectrometer (Optima,
Micromass, Manchester, UK). Total N concentration in roots was measured on a dry
combustion analyzer (CN-2000, LECO Corp., St. Joseph, Minnesota, USA).
83
Chapter 3
Calculations and statistical analyses
Isotopic composition is reported as δ15N (‰) with respect to atmospheric N2 according
to (Mariotti 1984). The fraction of nitrogen in C. mimosoides derived from the atmosphere
(Ndfa) was calculated following Amarger et al. (1979):
15
Ndfa =
N ref
15
N ref
15
N leg
B
(1)
where δ15Nleg is the 15N abundance in the N2-fixing legume species, δ15Nref is the level of
δ15N measured in the non-N2-fixing reference species growing on the same substrate (in
our case H. dissoluta), and B is the abundance of 15N in a legume individual that obtains
all its nitrogen from N2-fixation. Most values reported for B are in the range of -0.2 to 2
‰ (Yoneyama 1998). Since no data were available for C. mimosoides, we used -1.7‰, a
value that has been used previously for tropical species (Sprent et al. 1996; Yoneyama et
al. 1993). It was not possible to estimate Ndfa for plants receiving nitrogen due to lack of
information on the isotopic signature of the NH4NO3 used.
We assessed the proportion of nitrogen uptake by the grass originating from the legume
(Ndf leg) using a mixing-model from Högberg (1997), adapted to our setting:
Ndf leg · ∆N · δ15Nlegume + (1- Ndf leg) · ∆N · δ15Nsoil = δ15Nstart · Nstart - δ15Nend · Nend
(2)
δ15Nstart and Nstart are the average δ15N-value and total nitrogen content of grass plants at
collection from the field, and δ15Nend and Nend are the δ15N-value and total nitrogen content of
grass plants at harvest. ∆N is the total nitrogen uptake of the grass during the experiment (=
Nend - Nstart), and δ15Nsoil is the measured isotopic signature of the soil substrate (= 6.85‰).
Statistical analysis was done using JMP 6.0.3 (SAS Institute, Cary, USA). Analyses of
variance were calculated with type III sums of squares. Data from pots in which a plant
had died early in the experiment were excluded from the analysis. If necessary, data were
transformed prior to analysis in order to meet model assumptions. For multiple comparisons
between factor levels the Tukey-Kramer HSD test was used.
Figure 1 Effect of supply of nitrogen and phosphorus and water stress on total biomass of Cassia
mimosoides and Hyperthelia dissoluta growing in monoculture (open bars) and competition (grey bars).
Average biomass of plants at start of the experiment was 0.12 g for C. mimosoides, and 0.52 g for H.
dissoluta. Bars show mean total biomass per plant (± SE; n=5 or 6). Bars within a water supply level
not connected by the same letter indicate significant differences between nutrient levels; uppercase
letters are for total biomass in monoculture, lowercase letters for total biomass in competition.
84
Legume-grass competition
Results
Biomass
Biomass production of C. mimosoides in monoculture was stimulated by the combined
addition of N and P (Figure 1, Table 1); the ANOVA with separate factors for N and P
supply showed that N supply significantly increased the biomass of C. mimosoides but that
P supply had no significant effect (Table 1). Biomass of C. mimosoides also increased when
both N and P were added in the competition pots (Figure 1), while the effect of adding P
alone was stronger than that of adding N (Table 1). Water stress significantly reduced total
High water supply
Water stress
0.8
Total biomass (g)
Cassia mimosoides
0.6
0.4
0.2
B b
B b
AB b
A a
Y x
XY x
X x
XY x
Total biomass (g)
Hyperthelia dissoluta
1.4
1.2
1
0.8
0.6
B a
B a
B a
A a
X x
X x
X x
X x
control
+P
+N
+NP
control
+P
+N
+NP
Fertilizer treatment
85
Chapter 3
Table 1 Effect of nitrogen and phosphorus supply, water and competition on total biomass,
nodule count and δ15N of Cassia mimosoides and on total biomass and δ15N of Hyperthelia dissoluta
growing in monoculture and in competition. The results of a four-way ANOVA on total biomass
including competition as a factor can be found in Appendix B.
Cassia
mimosoides
F
Test variable and source of variation
df
Total biomass (monoculture only)
N
1
P
1
Water
1
1
N×P
1
N × water
1
P × water
1
N × P × water
Total biomass (competition only)
N
1
P
1
Water
1
1
N×P
1
N × water
1
P × water
1
N × P × water
Nodules per plant without N supply
P
1
Water
1
Competition
1
1
P × water
1
P × competition
1
Water × competition
1
P × water × competition
Nodules per gram total biomass (plants without N supply)
P
1
Water
1
Competition
1
1
P × water
1
P × competition
1
Water × competition
1
P × water × competition
15
δ N at high water supply and no added N
P
1
Competition
1
1
P × competition
Hyperthelia
dissoluta
F
18.81***
2.50
11.41**
0.74
1.13
0.06
6.21*
10.12**
5.03*
0.01
2.88+
0.03
5.08*
0.85
6.37*
16.22***
12.12**
0.41
4.37*
6.75*
3.11+
3.68+
2.02
4.55*
1.84
0.06
0.01
0.37
18.44***
0.93
5.79*
0.50
1.90
0.07
0.84
-
1.32
0.60
5.55*
0.07
0.65
0.87
0.17
-
1.25
10.16**
0.64
1.05
2.62
0.10
Notes: The F ratio and significance level are given for each factor. + P < 0.10, * P < 0.05, ** P < 0.01, *** P < 0.001
86
Legume-grass competition
growth of C. mimosoides (Table 1), the mean biomass across all nutrient treatments biomass
being 17% lower than with high water supply. The relative reduction due to water stress
was positively correlated with the biomass level achieved at high water supply (Figure 1).
Biomass of H. dissoluta in monocultures only increased when N and P were applied together
(Figure 1). The ANOVA with separate factors for N and P supply showed that the effect
on grass biomass of N supply was stronger than that of P supply (Table 1). In competition
with C. mimosoides, H. dissoluta tended to produce less biomass in the +P treatments, but
this effect was not significant (Figure 1, Table 1). Overall, water stress had no significant
effect on growth of H. dissoluta, but water-stressed plants in the competition treatment were
smaller (Table 1). The fact that water had no overall significant effect on grass biomass was
due to relatively low biomass of the ‘no fertilizer’ and +N monocultures (Figure 1).
C. mimosoides tended to benefit from interspecific competition with H. dissoluta when P was
supplied, but to suffer from competition in the +N treatment as well as in unfertilized soil
with sufficient water (Figure 1; compare biomasses at competition treatments with those in
monocultures for the two species, P < 0.01).
N2-fixation and nitrogen economy
Cassia mimosoides plants produced many nodules in the control and +P treatments, but
almost none in the +N treatments (Figure 2). Compared to the unfertilized control, P
addition significantly increased the number of nodules per plant; however, the number of
nodules per gram biomass was not higher (Table 1), indicating that the increase per plant
was merely because the plants were bigger. Competition with the grass led to significantly
more nodules, both per C. mimosoides plant as well as per gram biomass (Figure 2, Table 1).
Water stress had no significant effect on nodulation (Table 1).
The values of δ15N in C. mimosoides were clearly lower than in H. dissoluta, indicating that
the legume was fixing atmospheric nitrogen (Figure 3). The proportion of the legume’s
nitrogen derived by fixation was ca. 50 % when the plants were grown in monocultures
under unfertilized conditions (Table 2; taking H. dissoluta as a reference). The presence
of the grass competitor reduced the δ15N in the legume even more (Figure 3, Table 1),
87
Chapter 3
High water supply
Water stress
Nodules per plant
40
30
20
10
0
control
+P
+N
+NP
control
+P
+N
+NP
Fertilizer treatment
Figure 2 Number of nodules in Cassia mimosoides as affected by addition of N and P, and by
competition with the grass Hyperthelia dissoluta. Open bars are for C. mimosoides plants growing
in monoculture, grey bars for plants growing in competition with H. dissoluta. Bars show mean
number of nodules per plant (± SE; n=5 or 6).
Control
+P
5
15
N (‰)
4
3
2
1
0
monoculture
competition
monoculture
competition
Figure 3 Abundance of 15N in shoot tissue of Cassia mimosoides (solid circles) and Hyperthelia
dissoluta (open circles) after 45 days of growth in monoculture and competition at high water supply,
without and with phosphorus addition. C. mimosoides and H. dissoluta individuals transplanted from
the field initially had a δ15N of -0.30 ± 0.13 ‰ and 1.96 ± 0.34 ‰, respectively. Error bars represent
standard errors (n = 5).
88
Legume-grass competition
Table 2 Effect of N and P supply on the nitrogen economy of C. mimosoides and H. dissoluta
growing in monoculture and competition. Nitrogen derived from atmosphere (Ndfa) and nitrogen
derived from legume (Ndfleg) are expressed as percent of total nitrogen content of the legume
and grass, respectively. Since nitrogen addition reduced the number of nodules almost to zero
(see Figure 2), Cassia mimosoides was assumed to derive all N from the soil in these treatments. N
uptake (mg plant-1) is calculated as total N content at harvest minus total N content at start of the
experiment, and is partitioned into the different sources of nitrogen available to the two species.
Nitrogen use efficiency (NUE) is calculated as ratio of increase in total biomass and nitrogen
uptake. Data are shown as mean ± standard error (n = 5).
no fertilizer
mono
comp
+P
mono
comp
+N
mono
+NP
mono
Cassia mimosoides
Ndfa (%)
[N] shoot (mg g-1)
[N] root (mg g-1)
N uptake soil (mg)
N uptake atmosphere (mg)
Total N content (mg)
NUE (g µg N-1)
50 ± 11
23.4 ± 0.4
17.9 ± 1.2
3.6 ± 0.7
4.0 ± 1.2
7.6 ± 1.4
48 ± 3
66 ± 2
21.9 ± 0.5
17.8 ± 1.6
1.7 ± 0.5
3.2 ± 0.9
4.9 ± 1.3
55 ± 5
41 ± 5
23.7 ± 0.4
19.8 ± 0.8
3.9 ± 0.7
2.6 ± 0.3
6.4 ± 0.6
45 ± 1
64 ± 3
24.4 ±0.3
19.5 ± 1.6
3.4 ± 0.7
5.8 ± 0.9
9.1 ± 1.4
43 ± 1
n.d.
37.0 ± 1.1
35.6 ± 2.0
16.4 ± 1.1
0
16.4 ± 1.1
25 ± 3
n.d.
33.2 ± 1.0
35.6 ± 0.7
18.6 ± 1.6
0
18.6 ± 1.6
28 ± 1
Hyperthelia dissoluta
Ndfleg (%)
[N] shoot (mg g-1)
[N] root (mg g-1)
N uptake soil (mg)
N transfer from legume (mg)
Total N uptake (mg)
NUE (g µg N-1)
0
5.2 ± 0.1
6.5 ± 0.3
2.1 ± 0.3
0
2.1 ± 0.3
135 ± 15
31 ± 7
5.2 ± 0.3
7.4 ± 0.2
3.0 ± 0.6
1.4 ± 0.4
4.3 ± 0.8
149 ± 14
0
5.5 ± 0.3
6.7 ± 0.4
2.2 ± 0.2
0
2.2 ± 0.2
134 ± 25
21 ± 6
6.6 ± 0.5
8.6 ± 0.5
2.7 ± 0.4
0.8 ± 0.3
3.5 ± 0.5
103 ± 11
0
24.2 ± 0.1
16.5 ± 1.2
14.6 ± 2.0
0
14.6 ± 2.0
18 ± 3
0
28.0 ± 0.1
16.7 ± 1.2
29.2 ± 1.4
0
29.2 ± 1.4
25 ± 2
indicating a significant increase to around 65% in the proportion of N fixed (Table 2).
Phosphorus addition did not significantly affect δ15N (Table 1).
The average δ15N of H. dissoluta was lower in the mixtures than in the monocultures.
Although the differences were not significant (Table 1), these results suggest that the grass
had access to some of the nitrogen fixed by the legume (Figure 3) and was obtaining 2131% of its nitrogen from this source (Table 2).
When no nitrogen was added, concentrations of nitrogen in plant tissue were much lower
in H. dissoluta than in C. mimosoides (Table 2). Average nitrogen use efficiency (NUE) - i.e.
biomass increment per unit of nitrogen taken up - was almost three times larger for the
grass than for the legume (Table 2).
89
Chapter 3
Discussion
In the savanna soil used in this experiment, plant growth was limited by both nitrogen and
phosphorus, though nitrogen was the more important limiting nutrient, especially for the
C4 grass. This finding is consistent with the main idea behind this study - that frequent fires
such as occur in savanna lead to nitrogen limitation (Vitousek et al. 1982). Several authors
have suggested that N-limiting conditions should confer a competitive advantage upon N2fixers, so that the net loss of N to the atmosphere due to fire might be matched by symbiotic
N2-fixation (Laclau et al. 2005; Medina and Bilbao 1991; Sanhueza and Crutzen 1998). In
this study we investigated why herbaceous legume biomass is nonetheless very low in this
ecosystem. The experiment was designed to distinguish between three possible explanations
– limitation of legumes by phosphorus, competition with grasses, and sensitivity to drought
–, which we consider briefly in turn.
Phosphorus limitation. We found that N2-fixation and growth of the legume in monoculture
were not limited by P. This finding contradicts the common view that legumes in tropical
savannas may be limited by low phosphorus availability (Bustamante et al. 2006; Döbereiner
1978; Medina 1987). Addition of P has been shown to increase N2-fixation and growth of
legumes in some tropical areas (Crews 1993; Perreijn 2002; Sanginga et al. 1995), but most
of the evidence for P limitation comes from studies with crop species, and these may have
especially high P requirements (Sprent 1999). In our study, C. mimosoides benefited most
from P supply when grown in competition with the grass, indicating that both species
were competing for phosphorus. Thus, the growth of C. mimosoides was constrained by the
reduced availability of P caused by the competing grass rather than by the availability of P
in the soil per se. In fact, other studies have shown phosphorus additions to benefit legume
species competing with C4 grasses (Anderson 1968; Brewer and Cralle 2003).
Competition with grasses. The increase in N2-fixation of C. mimosoides in response to a grass
competitor showed that both species were competing for nitrogen, and that the grass
lowered the N availability to the legume. Such a response of legumes to interspecific
competition has been found in other studies in tropical and temperate ecosystems (Cramer
et al. 2007; Ibijbijen and Ismaili 1995; Trannin et al. 2000; Viera-Vargas et al. 1995). In the
field, C. mimosoides is likely to face stronger competition than in this experiment because
grasses comprise 70% or more of total biomass in tropical savannas (Ezedinma et al. 1979;
Menaut and Cesar 1982; San Jose et al. 1985). This is confirmed by estimated N2-fixation
rates of up to 92% measured in mature C. mimosoides plants in the field used as source
90
Legume-grass competition
for this experiment (Chapter 2). The higher N content and lower δ15N of grasses growing
without N supply and in competition with Cassia mimosoides suggest that N availability to
the grass (H. dissoluta) was elevated in presence of the legume and that the grass may have
had access to N fixed by the legume. Recently, rapid transfer of N from legume to grass
has been shown for grass-legume combinations in the temperate zone (Ayres et al. 2007;
Rasmussen et al. 2007), but it remains unclear how this transfer occurs. In a humid African
savanna, dead root mineralization was shown to be the major source of N to the dominant
grasses (Abbadie et al. 1992), and it has also been suggested that arbuscular mycorrhiza
might play an important role.
Our results show not only that the C4 grass H. dissoluta was the stronger competitor for soil N,
but that it produced three times more biomass than the legume per unit nitrogen acquired.
This finding is consistent with other studies showing, first, that C4 plants generally use N
more efficiently than C3 plants (Sage and Pearcy 1987) and, second, that in this respect
savanna grasses are amongst the most efficient of all C4 plants (Le Roux and Mordelet
1995). In contrast, legume species tend to be at the lower end of the range of nitrogen use
efficiency (Del Pozo et al. 2000; Tjoelker et al. 2005). Indeed, the observation that most
legumes maintain high nitrogen concentration in their tissues led McKey (1994) to propose
that evolution of the N2-fixing symbiosis was due to the high N-demand of ancestral
legumes. According to this view, symbiotic N2-fixation is not primarily an adaptation to
N-limiting conditions.
Drought. Symbiotic N2-fixation is known to be highly sensitive to drought (Serraj et al.
1999), and nitrogen-rich leaves are likely to be a disadvantage if photosynthesis is limited
by a lack of water (McKey 1994). In our experiment, the water-stress treatment reduced
biomass production in both the grass and the legume, but the grass continued to produce
more biomass than the legume (Figure 1). This is consistent with many studies showing a
higher water use efficiency of C4 than of C3 species (McCarron and Knapp 2001; Simioni
et al. 2004; Turner et al. 1995), indicating that C4 grasses are generally better adapted to
drought. Turner (1995) suggested forbs only persist in some semi-arid regions because of
occasional years of high rainfall, and watering has been shown to increase legume biomass
in shortgrass prairie (Lauenroth and Dodd 1979).
We conclude that the competitive superiority of the C4 grass is due to both its greater
ability to capture nutrients and its higher nitrogen use efficiency. It is known that C4 grasses
can sustain high productivity despite very low N availability (Wedin and Tilman 1993),
and perform better under these conditions than C3 grasses (Ludlow 1985). Thus, the low
abundance of herbaceous legume in C4 dominated grasslands is most likely to be due
91
Chapter 3
to the competitive superiority of the C4 grasses, rather than from any particular abiotic
limiting factors. C4 grasses probably remain dominant because of a positive feedback on
low nitrogen availability by promoting fire and slowing down N cycling (Johnson and
Wedin 1997; Kauffman et al. 1995; Ojima et al. 1994; Seastedt et al. 1991). This mechanism
therefore contrasts with that proposed by (Vitousek and Field 1999), who argue that N2fixers may achieve high abundance where fire-return intervals are short. Their model
may not applicable to C4 grasslands because in these ecosystems N2-fixing plants would
only have an advantage over C4 grasses under conditions of extremely low N-availability.
Nonetheless, our data do support the prediction of the model that non-fixers may effectively
suppress N2-fixers if these are disproportionately limited by another factor such as P.
Despite their low abundance and competitiveness, the species diversity of herbaceous
legumes in tropical savannas is remarkably high (Medina 1987; Medina and Bilbao 1991).
One possible mechanism allowing herbaceous legumes to persist in C4 grasslands may be
disturbance. In tallgrass prairie, fire was found to increase the density of legume plants
compared to unburnt controls, though the biomass of legumes remained unchanged (Towne
and Knapp 1996). Fire may thus create a short-term opportunity for legumes until the C4
grasses can recover and again suppress the less efficient legumes. A further mechanism
creating temporal niches for herbaceous legumes would be the complete removal of grass
competitors, e.g. by animals digging for roots and bulbs. This may be the reason why legumes
are most common along roadsides and in fire-breaks (Cech et al., personal observation).
We could show that a symbiotic N2-fixer is at a strong competitive disadvantage compared
to a C4 grass and that this is mainly due to its high demand for N and low competitiveness
for soil nutrients. Therefore, symbiotic N2-fixation by herbaceous legumes may not be the
most important source of N to balance the ecosystem N budget, as has sometimes been
assumed (Laclau et al. 2005; Sanhueza and Crutzen 1998). Moreover, in a recent study
in a Trachypogon savanna in Venezuela, it was concluded that N2-fixation by microbial
crusts and soil-grass-rhizosphere associations seem to be more relevant than symbiotic N2fixation by legumes (Lopez-Hernandez et al. 2006). Most estimates of N losses through
fire in tropical savannas are in the range from 7-23 kg ha-1 yr-1 (Isichei 1995; Kauffman et
al. 1995; Laclau et al. 2005; Medina 1987). This contrasts with estimates of symbiotic N2fixation of 2 kg ha-1 yr-1 or less for North American prairies and grasslands (Reuss and Innis
1977; Woodmansee 1978); to our knowledge no data from measurements are available for
tropical savannas. Thus, in tropical savannas symbiotic N2-fixation probably does not have
the potential to compensate nitrogen losses through volatilization, as may be the case in
other ecosystems.
92
Legume-grass competition
Acknowledgements
This study was financed by the Swiss National Science Foundation grant No. 2-7750204. We thank Prof. S. L. S. Maganga, Markus Schneider-Mmary, and the authorities of
Saadani National Park for their logistical support in Tanzania, and Benjamin Donald,
John Williams and Hamis Williams for their assistance in the field. We also thank Stefano
Bernasconi and Nina Buchmann for providing the facilities for 15N analyses.
Literature Cited
Abbadie L, Mariotti A, Menaut JC (1992) Independence of savanna grasses from soil organic
matter for their nitrogen supply. Ecology 73:608-613
Amarger N, Mariotti A, Mariotti F, Durr JC, Bourguignon C, Lagacherie B (1979) Estimate of
symbiotically fixed nitrogen in fieldgrown soybeans using variations in 15N natural abundance.
Plant and Soil 52:269-280
Anderson GD (1968) Effects of fertilizers on botanical composition and productivity of pasture
on the sandy soils of the Tanganyika coast. East African Agricultural and Forestry Journal
34:207-216
Ayres E, Dromph KM, Cook R, Ostle N, Bardgett RD (2007) The influence of below-ground
herbivory and defoliation of a legume on nitrogen transfer to neighbouring plants. Functional
Ecology 21:256-263
Bell DT, Koch JM (1980) Post-fire succession in the northern jarrah Forest of Western Australia.
Australian Journal of Ecology 5:9-14
Bond WJ, Woodward FI, Midgley GF (2005) The global distribution of ecosystems in a world
without fire. New Phytologist 165:525-537
Brewer JS, Cralle SP (2003) Phosphorus addition reduces invasion of a longleaf pine savanna
(Southeastern USA) by a non-indigenous grass (Imperata cylindrica). Plant Ecology 167:237-245
Bustamante MMC, Medina E, Asner GP, Nardoto GB, Garcia-Montiel DC (2006) Nitrogen cycling
in tropical and temperate savannas. Biogeochemistry 79:209-237
Collins SL, Glenn SM, Gibson DJ (1995) Experimental analysis of intermediate disturbance and
initial floristic composition: decoupling cause and effect. Ecology 76:486-492
Cook GD (1994) The fate of nutrients during fires in a tropical savanna. Australian Journal of
Ecology 19:359-365
Cramer MD, Chimphango SBM, Van Cauter A, Waldram MS, Bond W (2007) Grass competition
induces N2 fixation in some species of African Acacia. Journal of Ecology 95:1123-1133
Crews TE (1993) Phosphorus regulation of nitrogen fixation in a traditional Mexican agroecosystem.
Biogeochemistry 21:141-166
Del Pozo A, Garnier E, Aronson J (2000) Contrasted nitrogen utilization in annual C3 grass
and legume crops: Physiological explorations and ecological considerations. Acta Oecologica
21:79-89
del-Val E, Crawley MJ (2005) What limits herb biomass in grasslands: competition or herbivory?
Oecologia 142:202-211
93
Chapter 3
Döbereiner J (1978) Potential for nitrogen fixation in tropical legumes and grasses. In: Döbereiner
J (ed) Limitations and potentials for biological nitrogen fixation in the tropics. Plenum Publishing
Corp., New York, pp 13-24
Ezedinma FOC, Agbim NN, Onyekwelu SSC (1979) Contribution of symbiotic nitrogen fixation
to the nitrogen economy of natural ecosystems: 1. Occurrence of herbaceous legumes in derived
savanna fallow and their nodulation in pot culture. Plant and Soil 51:503-513
Hendricks JJ, Boring LR (1999) N2-fixation by native herbaceous legumes in burned pine ecosystems
of the southeastern United States. Forest Ecology and Management 113:167-177
Högberg P (1997) 15N natural abundance in soil-plant systems. New Phytologist:179-203
Hu FD, Jones RJ (2001) Competition in pots between two tropical legumes (Stylosanthes hamata
and S. scabra) and two tropical grasses (Urochloa mosambicensis and Bothriochloa pertusa) at two
phosphorus fertiliser levels. Tropical Grasslands 35:34-42
Hu FD, Jones RJ (2004) The effect of seedbed treatment, cutting frequency and selective grass
defoliation on the production and botanical composition of experimental swards of Urochloa
mosambicensis and Bothriochloa pertusa mixed with Stylosanthes. Tropical Grasslands 38:204-216
Hulme PE (1996) Herbivores and the performance of grassland plants: A comparison of arthropod,
mollusc and rodent herbivory. Journal of Ecology 84:43-51
Huntley BJ (1982) Southern African savannas. In: Huntley BJ & Walker BH (eds) Ecology of
tropical savannas. Springer, Berlin, pp 101-119
Ibijbijen J, Ismaili M (1995) Effect of water stress on competition between Medicago truncatula and
wheat in 15N labeled soil. Arid Soil Research and Rehabilitation 9:399-408
Isichei AO (1995) Stocks of nitrogen in vegetation and soil in West African moist savannas and
potential effects of climate change and land use on these stocks. Journal of Biogeography
22:393-399
Johnson NC, Wedin DA (1997) Soil carbon, nutrients, and mycorrhizae during conversion of dry
tropical forest to grassland. Ecological Applications 7:171-182
Kauffman JB, Cummings DL, Ward DE, Babbitt R (1995) Fire in the Brazilian Amazon: 1. Biomass,
nutrient pools, and losses in slashed primary forests. Oecologia 104:397-408
Kazanis D, Arianoutsou M (1994) Vegetation structure in a post-fire successional gradient of Pinus
halepenis forest. In: Viegas DX (ed) Proceedings of the 2nd international conference on forest fire
research. Coimbra, Portugal, pp 979-994
Keeley JE, Rundel PW (2005) Fire and the Miocene expansion of C4 grasslands. Ecology Letters
8:683-690
Khadka J, Tatsumi J (2006) Difference in delta 15N signatures among plant parts of perennial
species subjected to drought stress with special reference to the contribution of symbiotic N2fixation to plant N. Plant Production Science 9:115-122
Laclau JP, Ranger J, Deleporte P, Nouvellon Y, Saint-Andre L, Marlet S, Bouillet JP (2005) Nutrient
cycling in a clonal stand of Eucalyptus and an adjacent savanna ecosystem in Congo 3. Inputoutput budgets and consequences for the sustainability of the plantations. Forest Ecology And
Management 210:375-391
Laclau JP, Sama-Poumba W, Nzila JD, Bouillet JP, Ranger J (2002) Biomass and nutrient dynamics
in a littoral savanna subjected to annual fires in Congo. Acta Oecologica 23:41-50
Lauenroth WK, Dodd JL (1979) Response of native grassland legumes to water and nitrogen
treatments. Journal of Range Management 32:292-294
Le Roux X, Mordelet P (1995) Leaf and canopy CO2 assimilation in a West African humid savanna
during the early growing season. Journal of Tropical Ecology 11:529-545
94
Legume-grass competition
Lopez-Hernandez D, Santaella S, Chacon P (2006) Contribution of nitrogen-fixing organisms to
the N budget in Trachypogon savannas. European Journal of Soil Biology 42:43-50
Ludlow MM (1985) Photosynthesis and dry matter production in C3 and C4 pasture plants, with
special emphasis on tropical C3 legumes and C4 grasses. Australian Journal of Plant Physiology
12:557-572
Mariotti A (1984) Natural 15N abundance measurements and atmospheric nitrogen standard
calibration. Nature 311:251-252
McCarron JK, Knapp AK (2001) C3 woody plant expansion in a C4 grassland: Are grasses and
shrubs functionally distinct. American Journal of Botany 88:1818-1823
McKey D (1994) Legumes and nitrogen: The evolutionary ecology of a nitrogen-demanding
lifestyle. In: Sprent JI & McKey D (eds) Advances in legume systematics 5: The nitrogen factor.
Royal Botanic Gardens, Kew, pp 211–228
Medina E (1982) Physiological ecology of neotropical savanna plants. In: Huntley BJ & Walker
BH (eds) Ecology of Tropical Savannas. Springer, Berlin, pp 308-335
Medina E (1987) Nutrients: requirements, conservation and cycles in the herbaceous layer. In:
Walker BH (ed) Determinants of tropical savannas. IRL Press, Oxford, pp 39-65
Medina E, Bilbao B (1991) Significance of nutrient relations and symbiosis for the competitive
interaction between grasses and legumes in tropical savannas. In: Esser G & Overdieck D (eds)
Modern ecology: basic and applied aspects. Elsevier, Amsterdam, pp 295-319
Menaut JC, Cesar J (1979) Structure and primary productivity of Lamto savannas, Ivory Coast.
Ecology 60:1197-1210
Menaut JC, Cesar J (1982) The structure and dynamics of a West African savanna. In: Huntley BJ
& Walker BH (eds) Ecology of Tropical Savannas. Springer, Berlin, pp 80-100
Norman MJT (1966) Katherine Research Station 1956-64: A review of published work. CSIRO,
Melbourne, Australia
Ojima DS, Schimel DS, Parton WJ, Owensby CE (1994) Long-term and short-term effects of fire
on nitrogen cycling in tallgrass prairie. Biogeochemistry 24:67-84
Perreijn K (2002) Symbiotic nitrogen fixation by leguminous trees in tropical rain forest in Guyana.
Tropenbos-Guyana Series 11, Tropenbos, Georgetown, Guyana
Pivello VR, Coutinho LM (1992) Transfer of macro-nutrients to the atmosphere during experimental
burnings in an open cerrado (Brazilian savanna). Journal of Tropical Ecology 8:487-497
Rasmussen J, Eriksen J, Jensen ES, Esbensen KH, Hogh-Jensen H (2007) In situ carbon and
nitrogen dynamics in ryegrass-clover mixtures: Transfers, deposition and leaching. Soil Biology
and Biochemistry 39:804-815
Reuss JO, Innis GS (1977) Grassland nitrogen flow simulation model. Ecology 58:379-388
Ritchie ME, Tilman D (1995) Responses of legumes to herbivores and nutrients during succession
on a nitrogen-poor soil. Ecology 76:2648-2655
Sage RF, Pearcy RW (1987) The nitrogen use efficiency of C3 and C4 plants. 2. Leaf nitrogen effects
on the gas-exchange characteristics of Chenopodium album (L.) and Amaranthus retroflexus (L.).
Plant Physiology 84:959-963
San Jose JJ, Montes R, Garcia-Miragaya J, Orihuela BE (1985) Bio-production of Trachypogon
savannas in a latitudinal cross-section of the Orinoco Llanos, Venezuela. Acta Oecologica
6:25-43
Sanginga N, Danso SKA, Zapata F, Bowen GD (1995) Phosphorus requirements and nitrogen
accumulation by N2-fixing and non-N2-fixing leguminous trees growing in low P soils. Biology
and Fertility of Soils 20:205-211
95
Chapter 3
Sanhueza E, Crutzen PJ (1998) Budgets of fixed nitrogen in the Orinoco savannah region: Role of
pyrodenitrification. Global Biogeochemical Cycles 12:653-666
Sankaran M, Hanan NP, Scholes RJ, Ratnam J, Augustine DJ, Cade BS, Gignoux J, Higgins SI,
Le Roux X, Ludwig F, Ardo J, Banyikwa F, Bronn A, Bucini G, Caylor KK, Coughenour MB,
Diouf A, Ekaya W, Feral CJ, February EC, Frost PGH, Hiernaux P, Hrabar H, Metzger KL,
Prins HHT, Ringrose S, Sea W, Tews J, Worden J, Zambatis N (2005) Determinants of woody
cover in African savannas. Nature 438:846-849
Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annual Review of Ecology and
Systematics 28:517-544
Seastedt TR, Briggs JM, Gibson DJ (1991) Controls of nitrogen limitation in tallgrass prairie.
Oecologia 87:72-79
Serraj R, Sinclair TR, Purcell LC (1999) Symbiotic N2 fixation response to drought. Journal of
Experimental Botany 50:143-155
Simioni G, Le Roux X, Gignoux J, Walcroft AS (2004) Leaf gas exchange characteristics and waterand nitrogen-use efficiencies of dominant grass and tree species in a West African savanna. Plant
Ecology 173:233-246
Sprent J, Geoghegan IE, Whitty W, James EK (1996) Natural abundance of 15N and 13C in
nodulated legumes and other plants in the cerrado and neighbouring regions of Brazil. Oecologia
105:440-446
Sprent JI (1999) Nitrogen fixation and growth of non-crop legume species in diverse environments.
Perspectives in Plant Ecology, Evolution and Systematics 2:149-162
Tjoelker MG, Craine JM, Wedin D, Reich PB, Tilman D (2005) Linking leaf and root trait
syndromes among 39 grassland and savannah species. New Phytologist 167:493-508
Towne EG, Knapp AK (1996) Biomass and density responses in tallgrass prairie legumes to annual
fire and topographic position. American Journal of Botany 83:175-179
Trannin WS, Urquiaga S, Guerra G, Ibijbijen J, Cadisch G (2000) Interspecies competition and N
transfer in a tropical grass-legume mixture. Biology and Fertility of Soils 32:441-448
Turner CL, Kneisler JR, Knapp AK (1995) Comparative gas-exchange and nitrogen responses
of the dominant C4 grass Andropogon gerardii and 5 C3 forbs to fire and topographic position in
tallgrass prairie during a wet year. International Journal of Plant Sciences 156:216-226
Van Auken OW (1994) Changes in competition between a C4 grass and a woody legume with
differential herbivory. The Southwestern Naturalist 39:114-121
Viera-Vargas MS, Souto CM, Urquiaga S, Boddey RM (1995) Quantification of the contribution of
N2 fixation to tropical forage legumes and transfer to associated grass. Soil Biology & Biochemistry
27:1193-1200
Villecourt P, Schmidt W, Cesar J (1980) Losses of ecosystem during bush fire (tropical savanna of
Lamto, Ivory Coast). Revue d’Ecologie et de Biologie du Sol 17:7-12
Vitousek PM, Cassman K, Cleveland C, Crews T, Field CB, Grimm NB, Howarth RW, Marino R,
Martinelli L, Rastetter EB, Sprent JI (2002) Towards an ecological understanding of biological
nitrogen fixation. Biogeochemistry 57:1-45
Vitousek PM, Field CB (1999) Ecosystem constraints to symbiotic nitrogen fixers: a simple model
and its implications. Biogeochemistry 46:179-202
Vitousek PM, Gosz JR, Grier CC, Melillo JM, Reiners WA (1982) A comparative analysis of
potential nitrification and nitrate mobility in forest ecosystems. Ecological Monographs
52:155-177
96
Legume-grass competition
Vitousek PM, Howarth RW (1991) Nitrogen limitation on land and in the sea: How can it occur?
Biogeochemistry 13:87-115
Wedin D, Tilman D (1993) Competition among grasses along a nitrogen gradient: Initial conditions
and mechanisms of competition. Ecological Monographs 63:199-229
Woodmansee RG (1978) Additions and losses of nitrogen in grassland ecosystems. Bioscience
28:448-453
Yoneyama T (1998) Variations in fractionation of carbon and nitrogen isotopes in higher plants:
N metabolism and partitioning in phloem and xylem. In: Griffiths H (ed) Stable Isotopes. BIOS
Scientific Publishers, Oxford, UK, pp 99-109
Yoneyama T, Muraoka T, Murakami T, Boonkerd N (1993) Natural abundance of 15N in tropical
plants with emphasis on tree legumes. Plant and Soil 153:295-304
97
Chapter 3
Appendix A
Characteristics of the soil used as substrate in the pot experiment.
A
B
Soil characteristics
mean ± SE
Total carbon (mg g-1)
Total nitrogen (mg g-1)
Total phosphorus (µg g-1)
Extractable nitrate-N (µg g-1)A
Extractable ammonia-N (µg g-1)A
Extractable phosphate-P (µg kg-1)B
pH1
5.47 ± 0.08
0.34 ± 0.01
4.38 ± 0.04
1.16 ± 0.22
3.56 ± 0.18
2.82 ± 0.24
5.86 ± 0.02
in 0.2 M KCl
Bray-II
Appendix B
Effect of nitrogen and phosphorus supply, water and competition on total biomass of Cassia
mimosoides and Hyperthelia dissoluta growing in monoculture and in competition. All possible
interactions are included.
Cassia
mimosoides
Hyperthelia
dissoluta
Source of variation
df
F
F
N
P
Water
Competition
N×P
N × water
N × competition
P × water
P × competition
Water × competition
N × P × water
N × P × competition
N × water × competition
P × water × competition
N × P × water × competition
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
22.48***
17.19***
23.72***
0.01
1.10
5.32*
0.67
4.72*
4.49*
0.29
8.80**
0.00
0.89
3.49+
0.05
11.47**
0.00
3.08+
1.71
4.37*
0.08
0.01
1.69
5.98*
3.55+
1.06
0.05
0.01
1.39
0.00
Notes: The F ratio and significance level are given for each factor. + P < 0.10, * P < 0.05, ** P < 0.01, *** P < 0.0
98
Grazing in humid savanna
Chapter 4
Effects of clipping and fertilization on productivity
and species composition of abandoned cattle grounds
in humid savanna, Tanzania
Patrick G. Cech, Harry Olde Venterink, and Peter J. Edwards
Institute of Integrative Biology, ETH Zürich, Universitätsstrasse 16, 8092, Zürich, Switzerland
99
Chapter 4
Abstract
During the past century, numerous ranches were established in African savannas to produce
cattle intensively, often causing the displacement of native ungulate herbivores. Many of
these enterprises failed, but the abandoned cattle grounds were not always re-colonized
by wild herbivores. In this study we assessed the suitability of three different savanna
vegetations on an abandoned cattle ranch for native grazers, investigating whether clipping
of the vegetation with or without fertilization (to simulate effects of grazing animals), has a
rapid positive feedback on the quality of grazed areas. Fertilization and clipping increased
biomass production, but the effect of fertilization was much stronger. More frequent clipping
did not lead to a further increase in biomass production. Fertilizer application promoted
some plant species preferred by domestic and wild herbivores, whereas the plant species
that responded negatively were less favoured species. Stoloniferous species benefited from
fertilization, whereas other growth forms mainly showed a negative response. Our results
suggest that herbivores introduced into abandoned savanna vegetation can potentially
facilitate higher biomass production and increase the abundance of preferred fodder species;
however, they only do this if the nutrients in their excreta are returned to the grazed areas.
Savanna restoration through facilitation may be possible, but its success may be limited by
a shortage of herbivores to maintain the facilitated areas.
Keywords
grazing, herbivory, paddocks, positive feedback, selective feeder, ungulates
100
Grazing in humid savanna
Introduction
South and East African savannas are characterized by their high abundance and diversity of
wild herbivores (du Toit and Cumming 1999; McNaughton 1985). During the past century
land-use for livestock production in the savanna biome has been intensified (Augustine
et al. 2003; Hudak 1999; Scholes and Archer 1997), causing the displacement of native
ungulate herbivores (du Toit and Cumming 1999). Intensive cattle grazing often leads to a
decline in nutrient availability and forage quality (Boddey 2004), or to the replacement of
palatable grass species by unpalatable grasses, forbs or shrubs (Anderson and Briske 1995;
Brown and Stuth 1993; Kelly and Walker 1976; Perkins and Thomas 1993). Cattle ranching
also cause large re-distributions of nutrients to focal points such as watering points and
paddocks where cattle are herded overnight (Augustine 2003; Turner 1998).
Due to bush encroachment and pasture deterioration, many intensive ranching enterprises
in tropical Africa have failed (Okigbo 1985; Tobler et al. 2003). Although abandoned
cattle grounds could potentially become new habitats for displaced wild herbivores, such
areas often only support small numbers of wild ungulates (Treydte et al. 2005). This can
be explained by lack of potential colonizers from surrounding herbivore populations, by
habitat impoverishment caused by ranching, or by the fact that a savanna largely devoid
of herbivores is a difficult environment for re-colonization by grazers because most of the
nutrients are diluted in large amounts of biomass. For effective restoration management it
is important to understand how these factors limit re-colonization, but there have been few
studies to investigate this topic.
By enhancing plant biomass production, grazers may function as ecological engineers
(Jones et al. 1994). Grazing optimization theory suggests that net primary production is
maximized at an intermediate grazing intensity (Hilbert et al. 1981; McNaughton 1979b),
and this conclusion is supported by model simulations (de Mazancourt et al. 1998;
1999; Hilbert et al. 1981; Holdo et al. 2007). For savannas, experimental evidence for
overcompensatory growth at moderate grazing frequencies under field conditions has been
reported only for Serengeti (McNaughton 1983a). For a humid savanna of West Africa,
grazing optimization was not supported, at least in terms of dry matter yield, though
nitrogen yield to grazers was increased at moderate clipping frequencies when fertilizer
was added (Leriche et al. 2003).
Ungulate herbivores can also alter the species composition and structure of grassland
vegetation (Belsky 1986; McNaughton 1983b; Thornton 1971), through which they may
101
Chapter 4
affect quality of their forage. Defoliation by grazers may alter the relative advantages of
different species in competition for light, e.g. between tall and prostrate grasses (Belsky
1986; McNaughton 1979a), or differentially affect recruitment in species with contrasting
life-history traits (O’Connor 1994). Increased nutrient availability through herbivore excreta
is also likely to affect competition for nutrients among plant species (Wedin and Tilman
1993). Whether the net effect of grazing on fodder quality is positive or negative depends
on grazing intensity, environmental limits on primary production, and the response of
plants to defoliation (see review by Huntly 1991).
Whether or not the feedback of grazing on patch quality is beneficial for grazers also depends
on grazer type: for bulk feeders, average quantity and quality of forage are important,
while for selective herbivores habitat quality is mainly determined by the abundance of
preferred high quality species (East 1984). Therefore, re-colonization by native herbivores
of a nutrient-poor humid savanna, where selective herbivores are the dominant ungulates,
may depend less on the yield of grazing patches than on the ability of grazers to promote
nutritionally valuable plant species.
Wild herbivores have been observed to abandon intensively grazed patches because they
were attracted to recently burned areas (Archibald et al. 2005). Abandoned patches could
only be grazed again if the accumulated biomass of low nutrient concentrations was
removed, e.g. by fire. Such an interaction of grazing and fire is most likely to occur at high
productivity and low herbivore density, and causes a more homogeneous distribution of
grazing pressure across the ecosystem. Any feedback mechanism of grazing on quality of
temporarily grazed patches would therefore need to be rapid if grazers are to benefit from
them. However, most studies on the responses of grass communities to grazing in savanna
ecosystem have focussed on long-term changes at high herbivore densities.
The aim of this study was to investigate the suitability for native grazers of three vegetation
types on an abandoned cattle ranch, by investigating whether defoliation with or without
fertilization has a rapid positive feedback on patch quality. We conducted a factorial
clipping × fertilization experiment with three levels of clipping frequency and two levels
of fertilizer. Our objective was to examine effects of these treatments on differences in
patch quality by measuring dry matter production as well as the relative abundance of
grass species. We made the following predictions: (i) clipping at low frequency increases
dry matter production and promotes grazing-tolerant species, (ii) fertilization increases dry
matter production and the abundance of palatable grass species preferred by herbivores,
because low nutrient levels promote the dominance of unpalatable species with chemical
defences against defoliation (Chapin and McNaughton 1989), and (iii) the vegetation
102
Grazing in humid savanna
types close to focal points show the strongest responses in terms of yield, whereas tallgrass
savanna vegetation responds more in terms of species composition. The latter expectation
is based on the assumption that focal areas are enriched in nutrients through high input of
excreta in the past and species composition still reflects the high grazing pressure by cattle,
whereas tallgrass savanna is dominated by fire-tolerant species adapted to very low nutrient
levels.
Materials and Methods
Study area
The study area is located in Saadani National Park on the Tanzanian coast (5° 43′ S,
38° 47′ E). The relatively nutrient-poor soils consist of greyish fine sand or loamy sand
in the flats and reddish loamy sand over clay on slopes and hilltops (Klötzli 1980). The
northern part of the national park was operated as a cattle ranch from 1954–2000 with up
to 13’000 head of cattle on ~460 km2. During the ranch period, cattle were led to pasture
areas and dams in groups of 200-400 animals by herdsmen during the daytime, and kept
in paddocks with up to 1500 animals overnight. The southern part of the national park
was a game reserve from 1969 until it became national park in 2002. Both parts of the
national park are grazed by wild herbivores such as warthog, waterbuck, reedbuck, buffalo,
wildebeest, giraffe and elephant, but densities and diversity of wild herbivores are much
higher in the south, and no increase in ungulate abundance was observed in the former
ranch area in the first three years following abandonment (Treydte et al. 2005).
We selected three savanna vegetation types in the former ranch area, ordered by their
distance from the paddocks: (1) the margins of paddocks (PM) are characterised by the
grasses Digitaria milanjiana, Eragrostis superba and the sedge Cyperus bulbipes, the height
of this cover usually not exceeding 50-60 cm; (2) in Heteropogon savanna (HS) the grass
Heteropogon contortus is most abundant and accompanied by Panicum infestum, Bothriochloa
bladhii and the sedge Abildgaardia triflora; vegetation is < 60 cm high; (3) tallgrass savanna
(TG) is dominated by the tall grasses Diheteropogon amplectens and Hyperthelia dissoluta with
culms < 2 m. The rationale for this selection was the expectation that these three savanna
grassland types corresponded to a gradient of decreasing grazing pressure during the
operation of the ranch (PM being the most grazed and TG the least).
Mean annual temperature recorded at the former ranch complex is 25° C (1973–98).
Annual precipitation from 1957-98 ranged from 610 to 1700 mm, with a mean of 1040
103
Chapter 4
mm. The wet season lasts from March until June, and there is a short rainy season from
mid-October to mid-November. The driest months are January and February and August
and September, and during these periods fires are common, many of them being started
deliberately by the local people, poachers, or the park management. For a more detailed
description of the study area see Tobler et al. (2003) and Chapter 2.
Experimental design and treatments
In June 2005 (at the end of the long wet season), we set up exclosures with fences 1.6 m high
to exclude small and large mammalian herbivores. There were four replicate exclosures in
each of the three vegetation types along one paddock vegetation gradient. A factorial splitplot design with two levels of fertilization (with fertilizer or without) and three levels of
clipping frequency (zero, once, and thrice) was applied in each exclosure. The two halves
of an exclosure assigned to a fertilizer treatment were separated by a 2 m buffer zone, and
contained three 2 × 2 m plots each. Fertilizer was applied at weeks 0 and 8 by scattering
granules of NPK fertilizer, resulting in a cumulative input of 40 g N m-2, 10 g P m-2, and
20 g K m-2; this N input is equivalent to the local enrichment due to a single urination
(Hamilton et al. 1998). The clipping treatment involved removing all aboveground biomass
(including litter) to a height of c. 5 cm. For the treatment without clipping (‘zero’), the
vegetation was only clipped at the final harvest at week 52; for the low frequency clipping
treatment (‘once’), it was clipped at week 0 and at week 52; for the high frequency treatment,
it was clipped three times (at weeks 0, 8, 20, and 52).
Biomass and soil measurements
Harvested biomass from the central 1 × 1 m square of each plot was sorted into individual
grass species, herbs and Cyperaceae, separated into dead and living, and dried to constant
weight at 70°C. The biomass outside the 1 × 1 m plot was also clipped according the
treatment schedule.
Habitat descriptors (vegetation and soil characteristics) at the start of the experiment are
shown in Table 1. Total standing biomass was on average highest in HS and lowest in PM,
although differences were not significant. Stoloniferous species tended to be most abundant
in the paddock margin, but there was large variation among exclosures. Caespitose species
were most abundant in tallgrass savanna, and tussock-forming species tended to be most
dominant in Heteropogon savanna. Total soil N and P was highest in the paddock margin
104
Grazing in humid savanna
Table 1 Habitat descriptors of the studied savanna grassland types at start of the experiment.
Topsoil (0-10 cm) nutrient contents were determined from five soil cores extracted at a representative
location central to the four exclosures (data from Chapter 2). Vegetation characteristics were
obtained at the first clipping (week 0) and data are shown as means of four exclosures (± SE).
Values not sharing the same letter are significantly different (Tukey-HSD, P < 0.05).
Soil characteristics
Total soil C (kg m-2)
Total soil N (g m-2)
Total soil P (g m-2)
Extractable soil N (g m-2)
Extractable soil P (g m-2)
Tallgrass
savanna
0.77 ± 0.04A
67 ± 4A
8.2 ± 0.6A
0.16 ± 0.05A
0.92 ± 0.20A
0.64 ± 0.04A
52 ± 2B
5.0 ± 0.3B
0.08 ± 0.04A
0.21 ± 0.01B
0.45 ± 0.0B
37 ± 2C
3.6 ± 0.2C
0.13 ± 0.04A
0.15 ± 0.03 B
287 ± 39A
0.76 ± 0.02A
481 ± 87A
0.68 ± 0.03AB
349 ± 11A
0.63 ± 0.02B
0.1 ± 0.1B
76.4 ± 14.7AB
23.4 ± 14.6A
1.6 ± 1.5B
93.7 ± 4.6A
4.7 ± 4.6A
51.6 ± 6.4A
46.7 ± 6.1B
1.7 ± 1.3A
300
250
30
2005
2006
28
200
26
150
100
24
50
0
J J A S O N D J F M A M J
Temperature (°C)
Precipitation (mm)
Vegetation characteristics
Aboveground biomass (g m-2)
Proportion of live biomass
Composition by functional groups
(% of total biomass)
Caespitose species
Tussock-forming species
Stoloniferous species
Paddock
margin
grassland type
Heteropogon
savanna
22
Figure 1 Seasonal course of monthly precipitation and monthly mean temperature during the
experiment.
105
Chapter 4
and lowest in the tallgrass savanna. Extractable P was higher in PM compared with the HS
and TG, whereas extractable N did not differ significantly among these savanna vegetation
types.
Because treatment effects on vegetation composition could affect soil moisture conditions
through differences in evapotranspiration, soil moisture conditions were also monitored.
Using a ML2x ThetaProbe (Delta-T Devices, Cambridge, UK) we made four random
measurements in the top 5 cm in every plot at weeks 0, 4, and 9. Rainfall during the
study period was recorded with a tipping bucket rain gauge and (HOBO RG3-M, Onset
Computer Corp., Bourne, MA, USA). In 2005, the short rains in October/November failed
and caused an extended dry season lasting until December (Figure 1).
Calculations and statistical analyses
For clipped plots, production was calculated as the sum of all regrowth harvested after first
clipping at week 0. For control plots, production was calculated as the difference of biomass
harvested at week 52 and the average biomass harvested from clipped plots at week 0. All
plant species were assigned to one of three functional types according to their growth form:
(c) caespitose without stolons, (t) tussock forming without stolons, and (s) stoloniferous. We
calculated the relative abundance of species and functional types (expressed as percentage
of total biomass), as well as changes in relative abundance between week 52 and week 0. For
unclipped plots, relative abundances of the species and functional groups were calculated
from the means of the clipped plots.
Statistical analysis was done with JMP 6.0.3 (SAS Institute, Cary, USA). Because initial
biomass and relative abundances of the unclipped control plots was calculated from data
of the clipped plots, production and change in relative abundances for unclipped control
plots were not independent from data for clipped plots. We therefore analysed data in
three steps. In the first step (analysis 1) we computed ANOVAs on data of unclipped plots
with vegetation type and fertilizer as factors. In the second step (analysis 2) we computed
ANOVAs on data of clipped plots with vegetation type, fertilizer, and clipping level (clipped
once and clipped thrice) as factors. In a third step (analysis 3) we computed ANOVAs for
the complete dataset with vegetation type, fertilizer, and clipping level (no clipping, clipped
once and clipped thrice) as factors. The first two analyses comply with the condition of
statistical independence, but do not allow any conclusion about the effect of clipping
versus no clipping. Conclusions about the effect of clipping from the third analysis must be
interpreted with care.
106
Grazing in humid savanna
Three out of four exclosures in the Heteropogon savanna accidentally burned at week 7.
As a consequence biomass from unclipped plots was completely burned, whereas clipped
plots were not affected (few regrowth that did not burn). We therefore excluded data of the
Heteropogon savanna type from all statistical analyses that included unclipped plots (analyses
1 and 3). For analysis 2 we computed additional ANOVAs including Heteropogon savanna
data. For statistical analyses on changes in relative abundance, species or functional groups
with zero values at both week 0 and week 52 were excluded (i.e. species that did not increase
or decrease due to a treatment because they were not there at the start of the experiment).
All data were transformed prior to analysis to meet model assumptions.
Results
Aboveground biomass production
Fertilizer addition clearly enhanced biomass production in all treatments and vegetation
types (Figure 2, Table 2). Although clipping had a smaller effect on productivity than
fertilization, it also significantly enhanced biomass production (Figure 2, Table 2). Clipping
frequency had no significant effect on biomass production (Figure 2, Table 2). The interaction
between clipping frequency and fertilizer treatment was significant for tallgrass sites only.
Without fertilizer addition, production was highest in the paddock margin (Figure 2, Table
2). Highest productivities were measured in fertilized paddock margin plots clipped once
and in fertilized tallgrass plots clipped three times (Figure 2).
To evaluate whether plant growth during the dry season was still limited by nutrients and
not only by water, we calculated dry season production. This was only possible for the
most frequently clipped plots. Fertilizer increased dry season production (P < 0.0001), and
this effect was largest in tallgrass savanna (grassland type × fertilizer interaction P = 0.046)
(Figure 3). Dry season production was significantly lower in Heteropogon savanna (HS) than
in the other two vegetation types (Figure 3, P < 0.05, Tukey-HSD).
107
Chapter 4
1600
PM
1200
800
Aboveground biomass production (g m-2 yr-1)
400
0
ctrl
1x
3x
ctrl‡
1x
3x
ctrl
1x
3x
1600
HS
1200
800
400
0
1600
TG
1200
800
400
0
Clipping treatment
Figure 2 Aboveground biomass production in three savanna vegetation types as a function of
clipping frequency with (open symbols) and without fertilizer (filled symbols) addition. Data show
average production (week 0 until week 52) of four exclosures (± SE). PM: paddock margin, HS:
Heteropogon savanna, TG: tallgrass savanna. Since unclipped plots of three exclosures in HS burned
accidentally in Aug 2005, data indicated by ‡ show production of a single unburned exclosure only
(no SE).
108
Grazing in humid savanna
Table 2 Effect of fertilizer and clipping on aboveground biomass production. Results from
ANOVAs are shown for three types of datasets: analysis 1 comprises data from unclipped plots
only (with or without fertilizer), analysis 2 comprises data from clipped plots only (clipping once,
clipping thrice, fertilizer), analysis 3 includes all treatments (see methods). -HS/+HS indicates
that data from Heteropogon savanna exclosures was excluded/included in the analysis. Significance
levels are given for each effect/interaction: + P < 0.10, * P < 0.05, ** P < 0.01, *** P < 0.001.
Interactions are only listed if significant. The effects tested were V: vegetation type, F: fertilizer, CX:
clipping 1× vs. clipping 3×, C: no clipping vs. clipping 1× vs. clipping 3×
Analysis 1
-HS
V*
F*
(g m-2 20wk-1)
Biomass production
Test variable
Biomass production
Analysis 2
-HS
V*
F***
CX
V×F× CX*
+HS
V*
F***
CX
V×F× CX+
Analysis 3
-HS
V*
F***
C***
600
400
200
0
PM
HS
TG
Savanna grassland type
Figure 3 Aboveground biomass production of the most frequently clipped plots during the
dry season (Jun-Dec 2005). Data show average cumulative regrowth (week 0 until week 20) of
four exclosures (± SE). Open symbols are for plots without fertilizer, closed symbols are for plots
receiving fertilizer.
109
Chapter 4
Species composition and functional types
Fertilizer addition significantly reduced the abundance of Andropogon schirensis in the
unclipped plots (Table 3, Table 4). Diheteropogon amplectens also tended to suffer from
fertilizer addition while Digitaria milanjiana benefited (Table 3, Table 4). In clipped plots,
relative amounts of A. schirensis and D. amplectens were significantly reduced by fertilizer
addition, whereas D. milanjiana increased significantly. Panicum infestum tended to benefit
from fertilizer in clipped plots, and this effect was significant when data from Heteropogon
savanna was included. Heteropogon contortus suffered from fertilizer addition in clipped
plots (Table 1, Table 4). In tallgrass savanna, relative abundance of Cyperaceae decreased
with fertilizer addition (P = 0.002) and increased with clipping (P = 0.0361), but in the two
other vegetation types there were no responses to fertilization or clipping (Table 3, Table
4). Clipping had no effect on relative abundance of any other species (Table 4).
The relative abundance of caespitose species tended to decrease with fertilizer addition,
which was largely due to the large decrease of D. amplectens in tallgrass savanna (Figure 4,
Table 4, Table 1). The relative abundance of tussock-forming species tended to decrease in
response to fertilizer addition, which was largely caused by the response of A. schirensis and
H. contortus (Table 1, Table 4). The relative abundance of stoloniferous species increased
with fertilizer treatment and showed the strongest response in tallgrass savanna (Figure 4,
Table 4). Clipping had no effect on changes in relative abundance of any functional type
(Table 4).
Discussion
Aboveground biomass production
Adding fertilizer increased aboveground biomass production in all vegetation types,
showing that growth was limited by nutrient supply (Figure 2). Even in the dry season,
production was stimulated by nutrient addition, with the response being strongest in
tallgrass exclosures, perhaps because of their higher water availability (Figure 3, Figure
5). Clipping also increased biomass production, but less than fertilizer addition, and there
was no difference between the two clipping frequencies applied. This is in line with other
studies in African grasslands reporting no further increase or a decline in dry matter yield
with increased cutting frequency (Brockington 1960; Evans and Mitchell 1962). The main
effect of clipping on productivity was probably to remove excess aboveground biomass and
thus reduce light limitation; such limitation is known to occur in grass canopies even when
110
Grazing in humid savanna
Table 3 Mean relative abundance of species (expressed as percent of total aboveground biomass)
as a function of fertilizer and clipping treatment. Data are shown for the initial conditions at start
of the experiment in June 2005, and at final harvest in July 2005 (n = 4, ± SE). Only species
occurring at ≥ 1% at either start or end of the study period are listed. Preference of plant species
by zebu cattle as reported by Kozak (1983) is indicated by superscripts: +++ high, ++ medium, and
+
low preference. For each species the assigned functional type is indicated: caespitose (c), tussockforming (t), and stoloniferous (s).
Jul 2006
Jun
2005
fct.
type
‡
no fertilizer
ctrl
1x
+fertilizer
3x
ctrl
1x
3x
Paddock margin
Eragrostis superba++
Digitaria milanjiana+++
Cyperaceae
Panicum infestum+++
Dactyloctenium geminatum
t
s
t
t
s
60 ± 20 68 ± 19 65 ± 20
63± 17
57 ± 25 53 ± 27 59 ± 25
23 ± 15 25 ± 18 31 ± 18 19 ± 11 34 ± 24 27 ± 24 38 ± 23
8.7 ± 1.6 5.7 ± 1.6 3.4 ± 1.5 8.7 ± 1.0 1.5 ± 1.2 3.9 ± 3.0 2.7 ± 1.5
8.2 ± 5.0 0.3± 0.3 0.4 ± 0.4 8.4 ± 8.4 0.8 ± 0.8 16 ± 16 0.1 ± 0.1
0
0
0
0
6.3 ± 6.3
0
0
Heteropogon savanna‡
Heteropogon contortus++
Panicum infestum+++
Digitaria milanjiana+++
Cyperaceae
Bothriochloa bladhii
Sporobolus pyramidalis
Andropogon schirensis
Herbs+
t
t
s
t
c
t
t
c
80 ± 8
8.7 ± 2.6
4.7 ± 4.6
3.0 ± 1.2
1.3 ± 1.3
1.2 ± 1.2
1.1 ± 1.1
0.3 ± 0.3
67 ± 8
8.4 ± 6.4
2.1 ± 2.0
4.8 ± 2.1
2.0 ± 2.0
12 ± 9.3
2.9 ± 2.9
0.9 ± 0.9
74 ± 14
12 ± 6.6
2.0 ± 1.7
4.4 ± 2.4
0.2 ± 0.2
1.3 ± 1.3
5.7 ± 5.7
0.2 ± 0.2
85 ± 6
45 ± 23
4.9 ± 3.7 32 ± 20
4.1 ± 4.1 3.3 ± 2.9
2.7 ± 1.6 8.2 ± 6.8
1.6 ± 1.6
0
0.9 ± 0.9 11 ± 11
0
0
1.2 ± 1.0 0.8 ± 0.7
55 ± 16
29 ± 13
9.2 ± 7.5
2.9 ± 1.4
0
2.9 ± 2.9
0.3 ± 0.3
1.2 ± 1.2
Tallgrass savanna
Diheteropogon amplectens++
Andropogon schirensis
Hyperthelia dissoluta+
Panicum infestum+++
Cyperaceae
Heteropogon contortus++
Digitaria milanjiana+++
Herbs+
c
t
t
t
t
t
s
c
50 ± 6
20 ± 4
15 ± 8
5.8 ± 5.3
3.4 ± 0.6
3.3 ± 3.3
1.7 ± 1.3
1.4 ± 0.8
54 ± 6
24 ± 8
5.1 ± 3.0
3.3 ± 3.2
2.9 ± 1.5
1.1 ± 1.0
6.4 ± 3.0
2.2 ± 2.2
57 ± 10
17 ± 5
10 ± 7
5.4 ± 3.6
5.1 ± 1.6
0.6 ± 0.5
2.5 ± 0.7
2.8 ± 1.3
31 ± 3
22 ± 9
30 ± 12 2.6 ± 1.6
15 ± 9
16 ± 11
3.6 ± 3.3 20 ± 12
7.3 ± 2.2 0.1 ± 0.1
1.5 ± 1.1
0
11 ± 5
32 ± 2
0.8 ± 0.5
0
9.7 ± 5.5 12 ± 8
0.4 ± 0.3 3.7 ± 1.2
15 ± 14
15 ± 9
0.9 ± 0.9 22 ± 22
2.2 ± 1.7 2.0 ± 1.1
0.1 ± 0.1
0
71 ± 19 45 ± 12
0
0.3 ± 0.3
Unclipped plots of three exclosures burned in Aug 2005, therefore data in italics do not represent the unclipped control.
111
50 ±16
23 ± 15
8.9 ± 5.2
7.0 ± 6.1
0
10 ± 10
0.2 ± 0.2
0.2 ± 0.1
Chapter 4
Table 4 Effect of fertilizer and clipping frequency on change in relative abundance of species
and functional types. Results from ANOVAs are shown for three types of datasets: analysis 1
comprises data from unclipped plots only (with and without fertilizer), analysis 2 comprises data
from clipped plots (clipping once, clipping thrice, fertilizer), analysis 3 includes all treatments (see
methods). For species data from the vegetation type was pooled due to the low number of replicates,
thus vegetation type could be included as additional factor only in the analysis for functional types.
-HS/+HS indicates that data from Heteropogon savanna exclosures was excluded/included in the
analysis. Significance levels are given for each effect/interaction: + P < 0.10, * P < 0.05, ** P < 0.01,
*** P < 0.001. Interactions are only listed if significant. Up and down arrows indicate increase
and decrease in response to fertilizer addition. ‘n.s.’ indicates that no effect was significant, and
‘-‘indicates that there were too few replicates for the particular analysis. The effects tested were
V: vegetation type, F: fertilizer, CX: clipping 1× vs. clipping 3×, C: no clipping vs. clipping 1× vs.
clipping 3×
Analysis 1
-HS
Test variable
-HS
Analysis 2
+HS
Analysis 3
-HS
Change in relative abundance by species
Andropogon schirensis
↓F*
↓F***
CX
↓F**
CX
↓F***
C
Cyperaceae
n.s.
n.s.
n.s.
n.s.
Digitaria milanjiana
+
↑F
↑F**
CX
↑F***
CX
↑F***
C
Diheteropogon amplectens
↓F+
↓F**
CX
↓F**
CX
↓F***
C
Eragrostis superba
n.s.
n.s.
n.s.
n.s.
Herbs
n.s.
-
n.s.
n.s.
Heteropogon contortus
-
-
↓F***
CX
-
Hyperthelia dissoluta
-
n.s.
n.s.
n.s.
n.s.
↑F
CX
↑F***
CX
n.s.
Panicum infestum
+
Change in relative abundance by functional type
Caespitose
n.s.
V*
↓F+
CX
V×F+
V***
↓F+
CX
V×F**
V**
↓F*
C
V×F*
Tussock-forming
n.s.
n.s.
V
↓F*
CX
V
↓F+
C
Stoloniferous
V
↑F*
V**
↑F**
CX
V**
↓F***
CX
V+
↑F***
C
112
Grazing in humid savanna
Tussock-forming
Stoloniferous
60
60
40
40
20
20
0
0
-20
-20
-40
-40
Paddock margin
Caespitose
10
5
-5
-60
ctrl
1x
3x
10
-60
ctrl
1x
3x
30
30
20
20
10
10
0
0
-10
-10
-20
-20
ctrl
1x
3x
ctrl‡
1x
3x
ctrl
1x
3x
Heteropogon savanna
-10
5
0
-5
-10
-30
ctrl‡
1x
-30
ctrl‡
3x
1x
3x
60
80
80
Tallgrass savanna
Change in relative abundance (% of total biomass)
0
40
40
40
20
0
0
0
-20
-40
-40
-40
-80
-80
-60
ctrl
1x
3x
Clipping treatment
ctrl
1x
3x
Clipping treatment
Clipping treatment
Figure 4 Change in relative abundance of functional types in three savanna vegetation types
as a function of clipping and fertilizer treatment. Data show means of four exclosures (± SE).
Open symbols are for plots without fertilizer, closed symbols are for plots receiving fertilizer. Since
unclipped plots of three exclosures in Heteropogon savanna burned accidentally in Aug 2005, data
indicated ‡ by show change in relative abundance of a single unburned exclosure only (no SE).
nutrients are limiting growth (Schimel et al. 1991). Under natural conditions, herbivores
may also be responsible for removing excess biomass, leading to an increase in productivity;
megaherbivores are often the first to feed in such areas, because of the lower nutritional
quality of accumulated biomass, and these facilitate grazing by mesoherbivores (Verweij
et al. 2006). In addition, fire can also increase productivity for the same reason (San Jose
and Medina 1975).
113
Chapter 4
Fertilizer addition also increased growth in the clipped plots, with the highest production
being achieved by a combination of clipping and fertilizing (Figure 2). This is in line
with findings that grasses require a critical N concentration to be able to compensate for
leaf removal (Hamilton 1998). Thus, herbivores may enhance production if they return
nutrients to the grazed areas. Ungulate herbivores are known to differ with respect to the
spatial distribution of excreta to the area they use (Edwards and Hollis 1982; Hutchings et
al. 2001; Jewell et al. 2007). Therefore, the increase in productivity in grazed patches will
depend on the herbivore species using them. However, areas intensively grazed by wild
herbivores may also suffer from a long-term loss of nutrients if the nutrients returned in
excreta are insufficient to offset losses by grazing (Chapter 2).
Grazing may also have an indirect effect upon productivity by altering soil moisture
conditions, though this process has usually been neglected in previous studies. Defoliation
exposes the soil surface, thereby increasing evaporation, and increased productivity through
nutrient return by grazers may increase transpiration. In our experiment, the fertilized
plots had a higher soil moisture than unfertilized plots in the short term (4 weeks), but
this pattern was opposite in the longer term (9 weeks) (Figure 5). The short-term effect
may be explained by a lower soil evaporation through increased biomass and shade in the
fertilized plots, while in the longer-term increased transpiration in these plots results in
lower soil moisture. Furthermore, clipping reduced soil moisture, but only in plots clipped
once (Figure 5). This is comparable to the negative effect of fire on soil moisture reported
for Venezuelan savanna, which was largely attributed to the exposure of soil to evaporation
(San Jose and Medina 1975). Soil moisture between plots that had been clipped at week
0 and 8 was not lower than in unclipped plots at week 9 (Figure 5). This was probably
because of heavy rain that fell two days after the second clipping (week 8), and the very
low interception due to defoliation. Eventually, increased evaporation from the soil surface
would have reduced soil moisture, as observed in the ‘clipped once’ plots. We conclude that
grazing leads to reduced soil moisture, which in turn can reduce dry-season productivity,
especially on soils with a low water-holding capacity.
Production in fertilized plots was not significantly different between the two clipping
frequencies, but the pattern varied among vegetation types (Figure 2, Table 2). The increase
in production by more frequent clipping in fertilized tallgrass plots probably reflects the
change in vegetation composition, with a large increase in Digitaria milanjiana (Table 3, see
below).
114
Grazing in humid savanna
0.15
Week 9
Paddock margin
Week 4
0.1
ctrl
0.15
Week 4
clipped clipped
W0 W0+W8
Week 4:
V***, F***, C+ , V F**
0.1
Week 9:
V**, F***, C**, V F+
0.05
0
Heteropogon savanna
0
0.15
Week 9
Week 4
Tallgrass savanna
Volumetric soil water content (m3 m-3)
0.05
0.1
0.05
0
ctrl
clipped
W0
ctrl
clipped clipped
W0 W0+W8
Clipping treatment
Figure 4 Topsoil moisture (0-5 cm) as affected by clipping and fertilization in three savanna
vegetation types. Measurements were done at week 4 and 9. For Heteropogon savanna exclosures no
data is available at week 9 since the soil moisture probe could not be inserted into the soil. Clipping
treatments indicate the weeks (0 and/or 8 weeks) after the start in which the plots were clipped.
Significant effects and interactions from three-way ANOVAs with vegetation type (V), fertilizer (F),
and clipping (C) are shown in the figure; significance levels are + P < 0.10, * P < 0.05, ** P < 0.01,
*** P < 0.001.
115
Chapter 4
Changes in species composition
The very rapid changes in species abundance observed in this study were mainly caused
by fertilizer addition, with cutting having little effect (Table 4). Repeated cutting has been
reported to alter species abundance in natural grasslands in Zambia, but these effects were
only observed after the first 2 to 3 years (van Rensburg 1968). Most data on effects of grazers
on species composition in savanna ecosystems come from studies on long-term exclosure
of herbivores. These studies often report marked changes in species composition due to
herbivore exclusion (Anderson et al. 2007; Belsky 1986; McNaughton 1983b; Thornton
1971), hence separate effects of defoliation and nutrient return in excreta could not be
assessed.
Panicum infestum and Digitaria milanjiana showed a positive response to fertilizer (Table
3, Table 4). Both species were among the most preferred fodder species by cattle (Kozak
1983), and P. infestum was the most frequently found species in the dung of wild herbivores
in the study area (Halsdorf et al., unpublished data). P. infestum and D. milanjiana are often
found in patches grazed by wild herbivores in the former Saadani Game Reserve, and
D. milanjiana is also one of the dominant species in paddock margins (Chapter 2), indicating
that these two species are grazing-tolerant. Diheteropogon amplectens and Andropogon schirensis
showed the largest decrease in relative abundance in response to fertilizer; these species
were not found in wild herbivore dung (Halsdorf et al., unpublished data), and were also
not preferred fodder species by cattle (Kozak 1983). Our results thus indicate that shortterm facilitation of preferred fodder species is possible when herbivores return a significant
amount of nutrients to the patches they graze.
Stoloniferous species (Digitaria milanjiana and Dactyloctenium geminatum) benefited from
fertilization, both with and without clipping. This is consistent with the observation that
highly nutrient-enriched paddocks in our studied region are also dominated by stoloniferous
plants, although the species – Cynodon dactylon and Paspalum dilatatum – are different (Cech
et al., personal observation). Other studies in tropical grasslands have also reported an
increased abundance of stoloniferous species (mainly Cynodon spp. and Digitaria spp.) in
response to combined cutting and fertilizing (Brockington 1961; McIvor et al. 2005) or
under heavy grazing (Hendy 1975). Moreover, the abundance of stoloniferous or matforming species was found to decrease when herbivores were excluded or removed (Belsky
1986; Thornton 1971). Another study, however, showed that repeated cutting without
fertilization increased the abundance of stoloniferous species (van Rensburg 1968),
indicating that stoloniferous species can be stimulated by both defoliation and increased
116
Grazing in humid savanna
nutrient availability; however, we did not find this effect of defoliation in our short-term
experiment. In our study, Digitaria milanjiana was present in all three savanna vegetation
types, but comprised only a small fraction of total biomass in tallgrass and Heteropogon
savanna. The rapid response upon nutrient addition suggests that this species exhibits the
‘gearing-down’ strategy proposed by Grubb (1998); thus, it remains present in the sward
with a low biomass under nutrient-poor conditions, but can respond quickly when deposited
excreta cause a sudden increase in the available nutrients. Stoloniferous species have been
found to be more prevalent in grasslands with a long evolutionary history of grazing (Mack
and Thompson 1982), perhaps because these species are better able than tussock species to
exploit localized pulses of nutrients.
The decrease in the relative abundance of caespitose species upon fertilization was largely
due to the negative response of the tall grass Diheteropogon amplectens. Thus, a general
conclusion about the response of this growth form cannot be made. The relative abundance
of tussock-forming species decreased on average in response to fertilizer treatment (Table 4),
but there were also some tussock-forming species with a neutral response (Eragrostis superba)
or a positive one (Panicum infestum) (Table 3, Table 4). In Australian grasslands, the
decrease of tussock species in a cutting and fertilization experiment was largely attributed
to repeated defoliation (McIvor et al. 2005). Our data are in line with the results of a
recent meta-analysis demonstrating that grazing generally favours prostrate over erect and
stoloniferous over tussock-forming species (Diaz et al. 2007). However, our results show
that the response of a single tussock-forming species upon grazing may not be predicted
solely based on its growth form.
Differences among vegetation types
Tallgrass savanna, which makes up much of the study area, was the vegetation type that was
most responsive to fertilization and clipping. This vegetation contained the palatable species
Panicum infestum that tends to increase in areas where excreta are deposited. Another species
present in tallgrass savanna, Digitaria milanjiana, was favoured by zebu cattle (Table 3) but
apparently not by wild herbivores (Halsdorf, unpublished data); however, it may be that
this species would also be eaten by wild herbivores if present in sufficient amounts. Thus,
herbivores may turn tallgrass savanna into better-quality grazing areas through intensive
grazing and nutrient return. Such a replacement of tallgrass species by short species had
indeed be observed during operation of the cattle ranch (Klötzli 1980). In the absence
of nutrient addition, paddock margin vegetation was the most productive vegetation
117
Chapter 4
type, and it also showed the smallest changes in species composition when fertilized; this
was probably because the species that increased in the other savanna types were already
present in relatively high abundance at the start of the experiment. The high abundance
of Eragrostis superba, which is also frequently found in dung of wild herbivores (Halsdorf
et al., unpublished data), makes paddock margins a valuable habitat, but the small area
(3.1% of the former Mkwaja ranch) is likely to limit its potential to sustain large numbers
of wild herbivores. Heteropogon savanna may yield a significant amount of preferred fodder
species, since wild herbivores were also found to consume Heteropogon contortus (Halsdorf
et al., unpublished data). However, as shown by our results, biomass production may be
significant only during the wet season.
Conclusions and management implications
Our results indicate that herbivores can increase the abundance of preferred species in
poor savanna grassland – and thus improve grazing quality – providing the nutrients they
remove are returned to the grazed areas. This effect on species composition adds a new
insight to the facilitative effect of grazing through maintaining lawns, which was thought
to be mainly the effect of nutrient-rich and palatable regrowth (Leriche et al. 2003; Verweij
et al. 2006). Furthermore, the positive effect of grazing on the quality of vegetation also
depends on the availability of water and nutrients (Hamilton et al. 1998; Hobbs 1996).
Before herbivores can re-colonise a nutrient-poor, defaunated savanna, it is necessary to
find some way of removing excess biomass and/or adding nutrients. There are several
possible ways to achieve this, but all of them may present problems. One is to mow or burn
the overgrown savanna; but mowing is likely to be costly while fire may cause nutrient
losses in the long-term (Chapter 2, Chapter 1). In some cases, it may be possible to redistribute nutrient-rich soil from paddocks to depleted areas of tallgrass savanna; however,
paddock soils are strongly enriched in P relative to N (Chapter 2) making them a poor
source of N. Sowing native herbaceous legumes, as done in agriculture, may increase N
input by symbiotic N2-fixation; however, the competitive superiority of C4 grasses over
herbaceous legumes will limit the success of such a measure, and sown legumes are likely
to persist only at very low abundance in C4 dominated grasslands (Chapter 3). Whatever
methods are used, however, the density of grazing animals must be high enough to ensure
that lawn areas of high productivity and nutritional quality are maintained. Ultimately, a
shortage of grazing animals may be the factor that prevents the successful re-establishment
of a grazing system.
118
Grazing in humid savanna
Acknowledgements
This study was financed by the Swiss National Science Foundation grant No. 2-77502-04.
We thank Sabine Güsewell and Dieter Ramseier for statistical advice. We thank Prof. S.
L. S. Maganga, Markus Schneider-Mmary, and the authorities of Saadani National Park
for their logistical support in Tanzania, and Benjamin Donald, John Williams and Hamis
Williams for their assistance in the field.
Literature cited
Anderson TM, Ritchie ME, McNaughton SJ (2007) Rainfall and soils modify plant community
response to grazing in Serengeti National Park. Ecology 88:1191-1201
Anderson VJ, Briske DD (1995) Herbivore-induced species replacement in grasslands: is it driven
by herbivory tolerance or avoidance? Ecological Applications 5:1014-1024
Archibald S, Bond WJ, Stock WD, Fairbanks DHK (2005) Shaping the landscape: fire-grazer
interactions in an African savanna. Ecological Applications 15:96-109
Augustine DJ (2003) Long-term, livestock-mediated redistribution of nitrogen and phosphorus in
an East African savanna. Journal of Applied Ecology 40:137-149
Augustine DJ, McNaughton SJ, Frank DA (2003) Feedbacks between soil nutrients and large
herbivores in a managed savanna ecosystem. Ecological Applications 13:1325-1337
Belsky AJ (1986) Population and community processes in a mosaic grassland in the Serengeti,
Tanzania. Journal of Ecology 74:841-856
Brockington NR (1960) Studies of the growth of a Hyparrhenia-dominant grassland in northern
Rhodesia. I. Growth and reaction to cutting. Journal of the British Grassland Society
15:323-338
Brockington NR (1961) Studies of the growth of a Hyparrhenia-dominant grassland in Northern
Rhodesia. II. Fertilizer responses. III. The effect of fire. Journal of the British Grassland Society
16:54-65
Brown JR, Stuth JW (1993) How herbivory affects grazing tolerant and sensitive grasses in a central
Texas grassland: integrating plant responses across hierarchical levels. Oikos 67:291-298
Chapin FS, McNaughton SJ (1989) Lack of compensatory growth under phosphorus deficiency in
grazing-adapted grasses from the Serengeti Plains. Oecologia 79:551-557
de Mazancourt C, Loreau M, Abbadie L (1998) Grazing optimization and nutrient cycling: when
do herbivores enhance plant production? Ecology 79:2242-2252
de Mazancourt C, Loreau M, Abbadie L (1999) Grazing optimization and nutrient cycling: potential
impact of large herbivores in a savanna system. Ecological Applications 9:784-797
Diaz S, Lavorel S, McIntyre S, Falczuk V, Casanoves F, Milchunas DG, Skarpe C, Rusch G,
Sternberg M, Noy-Meir I, Landsberg J, Zhang W, Clark H, Campell BD (2007) Plant trait
responses to grazing - a global synthesis. Global Change Biology 13:313-341
du Toit JT, Cumming DHM (1999) Functional significance of ungulate diversity in African savannas
and the ecological implications of the spread of pastoralism. Biodiversity and Conservation
8:1643-1661
119
Chapter 4
East R (1984) Rainfall, soil nutrients and biomass of large African savanna mammals. African
Journal of Ecology 22:245-270
Edwards PJ, Hollis S (1982) The distribution of excreta on New Forest grassland used by cattle,
ponies and deer. Journal of Applied Ecology 19:953-964
Evans AC, Mitchell HW (1962) Soil fertility studies in Tanganyika. I. Improvement to crop and
grass production on a leached sandy soil in Bukoba. East African Agricultural and Forestry
Journal 27:189-196
Grubb PJ (1998) A reassessment of the strategies od plants which cope with shortages of resources.
Perspectives in Plant Ecology, Evolution and Systematics 1:3-31
Hamilton EW, Giovannini MS, Moses SA, Coleman JS, McNaughton SJ (1998) Biomass
and mineral element responses of a Serengeti short-grass to nitrogen supply and defoliation:
compensation requires a critical [N]. Oecologia 116:407-418
Hendy K (1975) Review of natural pastures and their management problems on the north coast of
Tanzania. East African Agricultural and Forestry Journal 41:52-57
Hilbert DW, Swift DM, Detling JK, Dyer MI (1981) Relative growth rates and the grazing
optimization hypothesis. Oecologia 51:14-18
Hobbs NT (1996) Modification of ecosystems by ungulates. Journal of Wildlife Management
60:695-713
Holdo RM, Holt RD, Coughenour MB, Ritchie ME (2007) Plant productivity and soil nitrogen as
a function of grazing, migration and fire in an African savanna. Journal of Ecology 95:115-128
Hudak AT (1999) Rangeland mismanagement in South Africa: Failure to apply ecological
knowledge. Human Ecology 27:55-78
Huntly N (1991) Herbivores and the dynamics of communities and ecosystems. Annual Review of
Ecology and Systematics 22:477-503
Hutchings MR, Gordon IJ, Kyriazakis I, Jackson F (2001) Sheep avoidance of faeces-contaminated
patches leads to a trade-off between intake rate of forage and parasitism in subsequent foraging
decisions. Animal Behaviour 62:955-964
Jewell PL, Kauferle D, Gusewell S, Berry NR, Kreuzer M, Edwards PJ (2007) Redistribution of
phosphorus by mountain pasture in cattle on a traditional the Alps. Agriculture Ecosystems &
Environment 122:377-386
Jones CG, Lawton JH, Shachak M (1994) Organisms as ecosystem engineers. Oikos 69:373-386
Kelly RD, Walker BH (1976) The effects of different forms of land use on the ecology of a semiarid region in south-eastern Rhodesia. Journal of Ecology 64:553-576
Klötzli F (1980) Analysis of species oscillations in tropical grasslands in Tanzania due to
management and weather conditions. Phytocoenologia 8:13-33
Kozak AJ (1983) Der Nährwert einer tropischen Naturweide in Tansania. PhD thesis, Swiss Federal
Institute of Technology (ETH), Zurich
Leriche H, Le Roux X, Desnoyers F, Benest D, Simioni G, Abbadie L (2003) Grass response to
clipping in an African savanna: testing the grazing optimization theory. Ecological Applications
13:1346-1354
Mack RN, Thompson JN (1982) Evolution in steppe with few large, hooved mammals. The
American Naturalist 119:757-773
McIvor JG, McIntyre S, Saeli I, Hodgkinson JJ (2005) Patch dynamics in grazed subtropical native
pastures in south-east Queensland. Austral Ecology 30:445-464
120
Grazing in humid savanna
McNaughton SJ (1979a) Grassland-herbivore dynamics. In: Sinclair ARE & Norton-Griffiths M
(eds) Serengeti: dynamics of an ecosystem. University of Chicago Press, Chicago, pp 46-81
McNaughton SJ (1979b) Grazing as an optimization process: grass-ungulate relationships in the
Serengeti. The American Naturalist 113:691-703
McNaughton SJ (1983a) Compensatory plant growth as a response to herbivory. Oikos
40:329-336
McNaughton SJ (1983b) Serengeti grassland ecology: the role of composite environmental factors
and contingency in community organization. Ecological Monographs 53:291-320
McNaughton SJ (1985) Ecology of a grazing ecosystem: the Serengeti. Ecological Monographs
55:259-294
O’Connor TG (1994) Composition and population responses of an African savanna grassland to
rainfall and grazing. Journal of Applied Ecology 31:155-171
Okigbo BN (1985) Land use and production potentials of African savanna. In: Tothill JC & Mott
JJ (eds) Ecology and management of the world’s savannas. Australian Academy of Science,
Canberra, Australia, pp 95–113
Perkins JS, Thomas DSG (1993) Environmental responses and sensitivity to permanent cattle
ranching, semi-arid western central Botswana. In: Thomas DSG & Allison RJ (eds) Landscape
Sensitivity. John Wiley and Sons, Chichester, UK, pp 273–286
San Jose JJ, Medina E (1975) Effect of fire on organic matter production and water balance in
a tropical savanna. In: Golley FB & Medina E (eds) Tropical ecological systems. Ecological
Studies 11. Springer, Berlin, pp 251-264
Schimel D, Kittel TGF, Knapp AK, Seastedt TR, Parton WJ, Brown VB (1991) Physiological
interactions along resource gradients in a tallgrass prairie. Ecology 72:672-684
Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annual Review of Ecology and
Systematics 28:517-544
Thornton DD (1971) The effect of complete removal of Hippopotamus on grassland in the Queen
Elizabeth National Park, Uganda. East African Wildlife Journal 9:47-55
Tobler MW, Cochard R, Edwards PJ (2003) The impact of cattle ranching on large-scale vegetation
patterns in a coastal savanna in Tanzania. Journal of Applied Ecology 40:430-444
Treydte AC, Edwards PJ, Suter W (2005) Shifts in native ungulate communities on a former cattle
ranch in Tanzania. African Journal of Ecology 43:302-311
Turner MD (1998) Long-term effects of daily grazing orbits on nutrient availability in Sahelian
West Africa: I. Gradients in the chemical composition of rangeland soils and vegetation. Journal
of Biogeography 25:669-682
van Rensburg HJ (1968) Growth and seasonal composition of natural grassland in Zambia. Journal
of the British Grassland Society 23:51-52
Verweij RJT, Verrelst J, Loth PE, Heitkonig IMA, Brunsting AMH (2006) Grazing lawns contribute
to the subsistence of mesoherbivores on dystrophic savannas. Oikos 114:108-116
Wedin D, Tilman D (1993) Competition among grasses along a nitrogen gradient: initial conditions
and mechanisms of competition. Ecological Monographs 63:199-229
121
Acknowledgements
I am very grateful to Peter Edwards and Harry Olde Venterink for supervising my thesis,
and to Michael Scherer-Lorenzen for accepting being co-examiner.
Many thanks to Peter, who again took the risk of running a PhD project in Tanzania,
where broken cars, muddy roads, wildfires and wildlife with little respect for nice fences
constantly threaten the collection of data and the course of experiments. It was great that
Peter and Harry took the trouble to come the long way to Mkwaja and Saadani – our
inspiring discussions in the field changed our view about the study area quite dramatically.
Special thanks to Harry for the thorough testing of my homemade PVC soil incubation
tubes and for the invention of a simple but effective shaker for soil extraction bottles.
Thanks to Stephanie Halsdorf to keeping the project in Tanzania alive, her help and
company during field research in Mkwaja and her catching enthusiasm about warthogs.
Many thanks also to Roland Cochard for having put up an amazing infrastructure in our
field research station at Mkwaja Ranch, and for coming to Tanzania to share his knowledge
about the vegetation of the region.
Having Thomas Kuster as a Master student with me for one field season was a great pleasure.
Thanks to him for his help, and especially for reminding me to drink and rest during field
work! I also appreciated the company of Bettina Gutbrodt and Christoph Rohrer, Master
students working together with Stephanie.
I am very appreciative of Prof. S. Maganga of the Sokoine University of Agriculture in
Morogoro for his logistical support in Tanzania. I am also indebted to the authorities and
rangers of Saadani National Park providing me with a first class infrastructure (power,
running water and internet) for research. Thanks also to all those maintaining the roads,
which is a Sisyphean task.
Many thanks to Benjamin for his help with my field work and his fantastic dishes – he made
Mkwaja a home for me. Thanks also to John for assisting me in field work and locating
all plots or sampling locations with a higher accuracy than any GPS on the market. I shall
also mention Hamis, Jacobo, Elias and Peter for the hours they spent clipping the savanna
together with me.
123
Thanks to the management and the staff of the Saadani Safari Lodge for their hospitality
and help with broken cars. Special thanks to Coenraad Bantjes for taking care of the Saadani
rain gauge and for being a friend.
I am very much indebted to Markus “Kusä” Schneider and his wife Tunu. Not only because
of their invaluable logistical support, without which much of this research would not have
been possible, but also for welcoming me into their home. Thanks also to Seku Msolomi
for his help and precious advice, and his family for their hospitality. So to Sele, Mudi,
Pascal, Karimu for helping in the Dar es Salaam jungle and with bargaining in Kariakoo.
Thanks to Rhoda her for reliable car repair service and for her long-distance breakdown
service when I was stranded with broken bearings in Msata.
Thanks to Nissan for building the Nissan Patrol and to Thuraya for their satellite phones.
I am very grateful to all the people from the Institute of Integrative Biology for many
inspiring ideas and a great working environment. I would like to thank my office mates
Myriam, Stephanie and Jake for their patience with my cough and my loud voice. Thanks
to Sabine Güsewell and Dieter Ramseier for their statistical expertise. Special thanks to
Rose Trachsler for teaching me all the analytical methods and her for patience with my
difficult samples.
This project was funded by the Swiss National Science Foundation.
Finally, my biggest gratitude goes to my wife Patrizia, who had to go without me for a long
time and who supported me most throughout my entire PhD. When fires had consumed
my plots or when I was trapped on muddy roads she was always at the other end of the
(satellite phone) line or next to me to cheer me up and keep me going. Thanks also to my
daughter Luisa, born in June 2007, for being a sound sleeper and making me forget about
my thesis with her smile. From now on I shall put up fences and fight fires to protect what
I love and value most: my little family.
124
Curriculum Vitae
Name
Patrick Georges Cech
Date of birth
28th April 1976
Nationality
Swiss
Education
1996 – 2002
Diploma (MSc) in organismic biology at the University of Basel,
Switzerland
Thesis: Effects of elevated CO2 on water relations of mature
deciduous forests
1987 – 1995
Grammar school,
Gymnasium Basel
Mathematisch-Naturwissenschaftliches
Work experience
2005 – 2008
Scientific assistant/PhD student, Institute of Integrative Biology,
ETH Zürich
2003 – 2005
Clinical Study Data Management, Hoffmann La-Roche AG,
Basel, Switzerland
2000 – 2002
Scientific assistant, Botanical Institute, University of Basel,
Switzerland
2000 – 2001
Clinical Study Data Management, Hoffmann La-Roche AG,
Basel, Switzerland
1997 – 1999
Research assistant, Botanical Institute, University of Basel,
Switzerland.
Project: Changes in land use and urban development in the
region of Basel
125
Publications
Cech P, Pepin S, Körner C (2003) Elevated CO2 reduces sap flux in mature deciduous
forest trees. Oecologia 137:258-268
126