Journal of Applied Ecology 2011, 48, 1145–1154
doi: 10.1111/j.1365-2664.2011.02011.x
A degradation threshold for irreversible loss of soil
productivity: a long-term case study in China
Yang Gao1,2, Binglin Zhong3, Hui Yue3, Bin Wu2 and Shixiong Cao2*
1
Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural
Resources Research, Chinese Academy of Sciences, Beijing 100101, China; 2Key Laboratory of Soil and Water
Conservation & Desertification Combat, Beijing Forestry University, No. 35, Qinhuadong Road, Haidian District,
Beijing 100083, China; and 3Water and Soil Conservation Bureau of Changting County, Fujian, 366300, China
Summary
1. During the past three decades, conservation and restoration biologists have increasingly recognized that ecological communities are likely to exhibit threshold changes in structure. However,
because long-term monitoring data are generally lacking, little is known about the consequences of
such ecological thresholds for the processes of ecosystem degradation and recovery.
2. To identify whether a degradation threshold exists that defines the boundary between the possibility of natural recovery and the need for artificial restoration of an ecosystem and to use this
knowledge to support the development of a suitable strategy for environmental restoration, we performed long-term monitoring of vegetation recovery in China’s Changting County since 1984.
3. A major problem was identified, which we refer to as the ‘irreversible loss of soil services’; when
vegetation cover decreases below a degradation threshold, this leads to sustained degeneration of
the vegetation community, erosion of the surface soil and declining soil fertility. These changes represent a severe and long-lasting disturbance that will prevent ecosystem recovery in the absence of
comprehensive artificial restoration measures.
4. Synthesis and applications. We identified a degradation threshold at about 20% vegetation cover
suggesting that for some sites, vegetation cover can serve as a simple proxy for more sophisticated
approaches to identifying thresholds; restoration must start with the restoration of soil fertility and
continue by facilitating vegetation development. Our results support the concept of ecological
thresholds (specifically, for soil services in a warm and wet region) and provide a model to inform
restoration strategies for other degraded ecosystems.
Key-words: degradation threshold, ecological restoration, landscape disturbance, natural
recovery, species richness
Introduction
Ecological communities are most likely to exhibit threshold
changes in structure when perturbations cause large changes in
(i) limiting soil or other resources, (ii) dominant or keystone
species and (iii) attributes of the disturbance regime that influence the recruitment of organisms by a community (e.g. Lamb,
Erskine & Parrotta 2005; Srinivasan et al. 2008). There are several types of ecological threshold, and their potential uses in
ecosystem management differ accordingly (Bestelmeyer 2006).
Restoration ecology could be more effective if environmental
managers better understood which type of ecological threshold
is most relevant for a given site and their potential roles in ecosystem management. Unfortunately, although the ecological
*Correspondence author. E-mail: [email protected]
literature contains many conceptual models of thresholds and
discussions of ecosystems in which multiple states are possible,
there is little guidance about which model is most appropriate
for a given situation (Martin & Kirkman 2009). The concept of
irreversible soil degradation because of vegetation removal
and overgrazing and its link to vegetation cover has been well
established in semi-arid regions (e.g. van de Koppel, Rietkerk
& Weissing 1997; Rietkerk & van de Koppel 1997), but there
has been no study in wetter regions, particularly where
harvesting of trees and vegetation is the primary cause of the
degradation. Thus, knowledge of how these factors operate in
warm and wet regions is required before we can understand
the ecological thresholds that determine whether a human-disturbed ecosystem will recover naturally or whether artificial
restoration will be required. In the latter case, identification of
a suitable strategy for environmental restoration could be
based on the nature of the threshold.
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society
1146 Y. Gao et al.
Changes in the environmental variables that determine the
distribution and abundance of vegetation, which differ among
sites with different ecological histories, must also be understood. Austin (1980, 2002) grouped these environmental variables into three main types: (i) gradients in resources that are
consumed by plants (e.g. CO2, water, light and nutrients); (ii)
gradients in resources that are not directly consumed by plants
but that have direct physiological influences on growth (e.g.
wind, air and soil temperatures, pH) and (iii) indirect gradients
that have no direct physiological influence on plant growth but
that are correlated with species distribution because of their
correlation with variables such as temperature, soil moisture
and precipitation (e.g. slope aspect, elevation, longitude and
latitude, distance from the coast, relative landscape position).
To facilitate the recovery of degraded or damaged ecosystems,
it is necessary to understand the state of the original ecosystem
and identify what factor or factors have altered that state
(Jackson & Hobbs 2009). When multiple factors affect an ecosystem’s state, they may act concurrently (i.e. at the same time)
or in series (i.e. one after the other), and these patterns will have
different effects on species survival. Where factors act in series,
their order can be very important; for example, a species may
be able to cope with events A, B and then C, but unable to cope
with events A, C and then B.
For these reasons, one of the greatest challenges in environmental biology is to predict the effects of human activity on the
complex webs of interactions among species and systems
(Berlow et al. 2009). A common example is when vegetation
degradation leads to a loss of topsoil and a reduction in soil fertility, thereby impeding recolonization of a site by many of the
original species (Lamb, Erskine & Parrotta 2005; Ludwig,
Wilcox & Breshears 2005). If the site degradation becomes sufficiently severe, a threshold is crossed; beyond that threshold,
the ecosystem cannot recover to its original state without
human intervention or the passage of years or even decades
without further disturbance. Human impacts can widen the
range of habitats in which such threshold dynamics occur, and
these impacts can shift communities into new states that are
difficult to reverse (Suding & Hobbs 2008). Therefore, the concept of a degradation threshold can guide the development of
targets for sustaining landscape functions and developing
management regimes for variables such as habitat loss, habitat
fragmentation, connectivity changes and soil erosion (Huggett
2005).
In the present paper, we discuss key issues associated with
the degradation threshold concept, and the management implications, using a case study of vegetation loss and soil erosion in
China. Our goal is to show how an improved understanding of
thresholds can guide the restoration of a degraded ecosystem.
The high intensity of monsoonal rainfall in our study area
occurs at the end of a fairly long dry period, which means that
herbaceous vegetation cover has decreased and the risk of
severe soil erosion is higher than at other times of year. Soil
erosion is a particular problem in China because of the country’s huge population and rapid economic development, which
have led to unsustainable agricultural land uses, combined
with adverse climatic changes that have led to the expansion of
deserts in many parts of arid and semi-arid China (Cao, Chen
& Yu 2009a). Large-scale restoration projects have been implemented in an effort to combat these problems, but for ecological restoration to succeed, we must recognize the potentially
severe consequences caused by crossing a degradation threshold, their effect on afforestation and other forms of artificial
vegetation restoration, and the resulting consequences for traditional restoration activities.
Recent theoretical advances have emphasized that the existence of thresholds and the alternative stable states that result
when thresholds are exceeded are key factors that influence the
outcomes of ecological restoration efforts. Abiotic thresholds,
such as those associated with changes in soil or water conditions, are widely recognized as crucial factors in efforts to
restore degraded ecosystems (Norton 2009), because these factors partly determine which species can survive at a given site.
Although many studies have examined the concept of ecological thresholds in ecological restoration, the complexity of coupled systems is not well understood (Lamb, Erskine & Parrotta
2005; Sasaki et al. 2008; Jackson & Hobbs 2009; Norton
2009), and most of the previous work has been theoretical
rather than observation based (Radford, Bennett & Cheers
2005; Liu et al. 2007; Sasaki et al. 2008). Moreover, the lack of
long-term field research means that there are insufficient data
to confirm the various theoretical models of ecological thresholds (Andersen et al. 2009). In addition, most efforts to overcome ecosystem degradation have involved interventions such
as tree planting (McVicar et al. 2007, 2010) or protecting sites
from impacts such as grazing so that they can undergo natural
recovery (Suding & Hobbs 2008).
In the management of degraded land, thresholds reflect
changes in vegetation and soils that are expensive or impossible
to reverse. Uncritical use of thresholds may lead to the abandonment of management efforts in areas that would otherwise
benefit from an intervention (Bestelmeyer 2006). Because of
the paucity of field studies, we lack a sophisticated understanding of the ecological degradation threshold that determines
whether a human-disturbed ecosystem has the potential for
natural recovery if it is protected against further disturbance or
whether artificial restoration will be necessary. To provide
empirical evidence for the existence of an ecological ‘degradation threshold’ and to demonstrate how this knowledge can be
used to guide successful ecological restoration, we have carried
out long-term monitoring of vegetation recovery in a warm,
wet region of China’s Changting County since 1984. We found
a severe problem in this area, which we refer to as the ‘irreversible loss of soil services’: when vegetation cover decreases
below an ecological degradation threshold, leading to sustained degeneration of the vegetation community, erosion of
the surface soil and declining soil fertility occur. These changes
represent a severe and long-term disturbance of the vegetation,
the soil and the landscape.
Materials and methods
Our research area is located in a warm and wet part of China’s
Changting County that covers 309 720 ha in western Fujian
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
Degradation threshold of soil productivity 1147
Province (2518¢40¢¢N to 2602¢05¢¢N, 11600¢45¢¢E to 11639¢20¢¢E).
Based on meteorological data collected at a soil and water conservation monitoring station established in 1940 in the study area, the
mean annual precipitation is 1721 mm year)1, the average annual
potential evapotranspiration is 892 mm year)1, and the mean annual
temperature is 18Æ3 C, with a minimum temperature of )7Æ9 C and
a maximum temperature of 39Æ4 C based on historical climate
records from 1952 to 2004 (Yang, Zhong & Xie 2005). Planting of
crops is the primary agricultural activity, and there is little or no livestock grazing other than to feed domestic animals. However, because
a half-century of China’s Grain First Strategy led to widespread
replacement of forests with agricultural land, much of the agriculture
occurred on unsuitable land, leading to severe erosion and long-term
site degradation in many areas. These conditions have contributed
to the inability of the degraded sites to recover naturally. Moreover,
the region’s monsoon climate means that there is a relatively dry season, when vegetation cover decreases, followed by a season with
heavy rainfall. The combination of degraded land, a lack of vegetation cover and high inputs of rain have increased the frequency and
scale of water erosion of the soil and the severity of floods, leading to
further degradation of the county’s forests and landscape. Another
problem is that the region’s poor farmers have little money to buy
coal for cooking, leading to harvesting of trees and woody vegetation for use as cooking fuel; this unsustainable exploitation of the
remaining vegetation is a primary contributor to the environmental
damage (Cao et al. 2009b).
To examine the ecosystem’s capacity for natural recovery, we randomly selected 30 representative hilly plots, each 400 m2 (20 · 20 m)
in size, in four towns (Cewu, Hetian, Sanzhou and Zhuotian). The
plots were established at similar topographic positions, at mid-slope
positions on slopes ranging from 20% to 30%. Before 1984, none of
the plots was formally managed, and local residents harvested trees
for use as firewood or construction materials. Based on discussions
with local government officials, there appears to have been only
minor differences among study sites in the historical or cultural factors that might have affected the locations of the most severely
degraded plots. We grouped these plots into six groups (with five plots
per group) based on the original vegetation cover (both annual and
perennial species) in 1984: (i) £20%, (ii) 20Æ01–25%, (iii) 25Æ01–30%,
(iv) 30Æ01–35%, (v) 35Æ01–40% and (vi) >40%. To restore the ecological environment, tree harvesting and grazing of domestic livestock
were prohibited throughout Changting County starting in 1984. To
ensure that this prohibition would be effective, the purpose of the
plots was explained to local residents, and incentives were provided,
including fuel and support for the purchase of livestock fodder,
thereby making it unnecessary to collect firewood or to graze domestic animals in the study plots (Cao et al. 2009b).
To better understand the specific reasons (mechanisms) behind the
existence of any ecological thresholds at these sites, to propose a
restoration approach in the plots where threshold dynamics were
occurring and to determine whether it was possible to develop a new,
rapid way to perform ecological restoration, we randomly selected
an additional 28 representative plots (for a combined total of 58
plots), each 400 m2 (20 · 20 m) in size, for a new study that was
conducted from 1999 to 2009. In each of these 28 plots, the initial
cover of annual vegetation was <30%; in 14 of the plots, we allowed
natural recovery to occur, but in 14 plots, we performed artificial
restoration using local species starting in 1999. In each of the two
groups of 14 plots, seven plots had initial vegetation cover of <20%,
and the remaining seven plots had initial vegetation cover
>20% (actually 20Æ1% or higher, but initial cover values ranging
from 20% to 30%). Artificial restoration was performed by planting
600 trees (Schima superba, Morella rubra, Liquidambar formosana or
Castanopsis fissa) and 2400 shrubs (Lespedeza davidii) per hectare. In
addition, Paspalum wettsteinii was seeded at a rate of 105 kg ha)1, or
Paspalum notatum was seeded at a rate of 45 kg ha)1, with
900 kg ha)1 of oil cake fertilizer (the organic matter that remains
after extraction of the edible oils from rapeseed) added to improve
the soil’s organic matter content.
To assess the amount of vegetation cover at the study site, we
measured both the crown area of the trees and coverage of the
ground by understorey vegetation (only green plants, therefore
photosynthetically active vegetation). Using a steel tape, we measured the crowns of 20 randomly selected trees in each plot each year
during the middle of the growing season (between the last 10 days of
June and the end of August) to determine crown area, which we used
to represent the mean crown cover per tree. We measured the maximum and minimum crown radii and modelled the crown as an
ellipse, with these radii representing the semi-major and semi-minor
axes, and calculated the mean canopy area for each species using
geometric mean values to account for extreme values. Total tree canopy cover (the proportion of the total site area accounted for by a
vertical projection of the elliptical crowns of the trees, including the
leaves plus the stems and branches) was calculated by multiplying
the mean crown area in a given year by the number of trees that were
present in that year, then dividing this total by the total area planted
with that species. Where canopies overlapped, we carefully determined the extent of the overlap and calculated its area; we then
divided this area equally between the two trees to avoid doublecounting.
In each portion of the plot that was not covered by tree or shrub
canopies or where grass was growing below trees, we performed lineintersect sampling using two 10-m transects at right angles to each
other to survey herbaceous vegetation (ground cover). We calculated
the net vegetation cover for the grass and woody cover by averaging
the two cover values (i.e. for woody and herbaceous vegetation). We
identified the vegetation cover every year at the same time (between
the last 10 days of June and the end of August). Total vegetation
cover (the combined cover of trees and herbaceous vegetation such as
grasses, forbs and herbs) for a given area was calculated by multiplying the mean cover value for a given vegetation type (woody vs. nonwoody vegetation) by the proportion of the total plot area occupied
by that type of vegetation. To describe the plant species richness in
the study plots, we collected samples of all plant species annually in
each plot in August. The samples were brought to Fujian Normal
University for identification if we could not confirm their identity in
the field.
To monitor soil erosion at each of the 30 selected representative
natural restoration plots, a sand sedimentation pond (a run-off pond)
was established at each plot. We selected 20-m-long by 5-m-wide
observation sections along the slopes in each test plot and constructed a stone and concrete sand sedimentation pond with a 15-m3
capacity at the bottom of the slope. We calculated the total water
input by multiplying the amount of rainfall (mm) by the surface area
of the observation section (100 m2); we then estimated the volume of
water collected by the sedimentation pond (i.e. the run-off) and
expressed this volume as a proportion (5%) of the total input of
water. In addition, all the soil was removed from the bottom of each
pond within 24 h after the rain, and three random samples of this soil
were dried for 12 h at 105 C and were weighed to determine the
water content of the sediments. This was then used to determine the
total oven-dry quantity of soil eroded by the rain event. We used
these data to calculate the erosion modulus (kt soil per km2 per rain
event).
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
1148 Y. Gao et al.
Results
Vegetation cover differed significantly among the initial cover
classes (anova). LSD tests revealed that the sites with an initial
cover of <20% had significantly (P < 0Æ05) lower initial vegetation cover than all other sites throughout the study period.
By 1989, sites with an initial vegetation cover >30% had significantly higher vegetation cover than sites with an initial vegetation cover between 20% and 30% (P < 0Æ05); by the end of
the study period, only the site with an initial vegetation cover
of 20–25% had significantly lower vegetation cover than the
other sites (P < 0Æ05), and vegetation cover was increasing so
rapidly in these plots that we expect it to catch up with the
other sites within a few years (Fig. 1). At a vegetation cover of
<20%, erosion of the surface soil accelerated greatly during
rain events, and the resulting loss of fertile topsoil resulted in a
decrease in soil nutrient contents (Fig. 2) that led to a sustained
loss of vegetation cover and very slow recovery of the vegetation community in the absence of human intervention (Fig. 3).
These changes represent a severe and long-term disturbance of
both the vegetation and the soil. When the vegetation cover
was higher than 20% in the natural restoration plots, the vegetation cover and the soil characteristics in the study plots were
both able to recover significantly (P < 0Æ001) in the absence of
human disturbance and without requiring artificial restoration,
although the recovery was faster at higher initial vegetation
cover values. The differences compared with the <20% vegetation cover were significant for all other vegetation covers
throughout the study period (P < 0Æ05).
In the absence of an improvement in vegetation cover (initial
cover <20%), the soil underwent continuing degradation,
110
100
90
80
Vegetation cover (%)
We also sampled the uppermost 30 cm of the soil, where the majority of the seedling and grass roots would be found, using an auger at
three randomly selected locations at 5-year intervals, in October of
1984, 1989, 1994, 1999, 2004 and 2009, to measure the soil nutrient
contents. All other plot parameters were measured in the same years.
Study parameters were measured annually in each plot, but for simplicity, we have reported data only at 5-year intervals, corresponding
to the interval used for the soil nutrient measurements. The uppermost 5 cm of the soil was also sampled on these dates using the same
auger at three randomly selected locations in each plot, and the samples were passed through a series of sieves to determine the content of
sand and coarser materials (>1 mm). Soil organic matter content was
determined by means of oxidation with potassium dichromate in a
heated oil bath. Total nitrogen was measured by means of alkali distillation. Total phosphorus was measured by means of atomic
absorption spectrophotometry (with a Varian spectrophotometer;
Varian Inc., Palo Alto, CA, USA). Total potassium was determined
by digestion with hydrofluoric acid and perchloric acid.
The data from the 30 natural recovery plots in the initial year of the
study and the values collected every 5 years thereafter are expressed
as means ± SD, as are the results for the additional 28 plots that
were monitored starting in 1999. We performed repeated-measures
anova to identify whether significant differences among treatments
existed, and when they did, we used the least-significant-difference
(LSD) test to determine which specific combinations of values were
significantly different. All tests were performed using version 12.0 of
the spss software (SPSS Inc., Chicago, IL, USA).
70
60
50
40
30
20
10
0
1984
1989
1994
1999
2004
2009
Year
≤20%
20·01–25%
25·01–30%
30·01–35%
35·01–40%
>40%
Fig. 1. Change in vegetation cover since 1984 in the plots that underwent natural restoration. Symbols are the mean values, and the bars
represent the standard deviation (n = 5).
indicated by ongoing erosion of the surface soil and depletion
of soil fertility (Figs 2 and 3), although the number of plant
species increased over time even in the most severely degraded
plots (Fig. 3). As a result, soil erosion and run-off in the plots
with an initial vegetation cover of <20% increased throughout the study period, leading to a continuing loss of fine particles and enrichment of sand and coarser particles (>1 mm) in
the exposed surface soil (Fig. 3). In contrast, when the initial
vegetation cover was >20%, soil nutrient properties improved
continuously (Fig. 2), and soil erosion and run-off decreased
continuously (Fig. 3), leading to a gradual decrease in the proportion of coarse materials in the surface soil (Fig. 3). All these
changes were statistically significant (P < 0Æ05).
Our 25 years of monitoring data also revealed strong and
significant relationships between vegetation cover and the soil
characteristics after human disturbance of the sites was prohibited (Fig. 4). The increase in vegetation cover (y) was significantly positively correlated with the following independent
variables (x): soil organic matter (y = 0Æ0958x + 1Æ373; R =
0Æ948, P < 0Æ001), total N (y = 0Æ0056x + 0Æ0811; R =
0Æ940, P < 0Æ001), total K (y = 0Æ1525x + 3Æ379; R = 0Æ946,
P < 0Æ001), total P (y = 0Æ0029x + 0Æ0992; R = 0Æ929,
P < 0Æ001) and the number of plant species (y =
0Æ3338x + 0Æ4621; R = 0Æ890, P < 0Æ001). It was significantly negatively correlated with soil erosion (y =
)0Æ0832x + 8Æ3222; R = )0Æ837, P < 0Æ001), with run-off
(y = )0Æ2612x + 54Æ761; R = )0Æ921, P < 0Æ001), and with
the proportion of sand and coarser material (>1 mm) in the
surface soil (y = )0Æ1144x + 47Æ311; R = )0Æ823, P <
0Æ001).
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
Degradation threshold of soil productivity 1149
0·8
12
0·7
10
0·6
Total N (g kg–1)
Organic matter (g kg–1)
14
8
6
0·5
0·4
0·3
4
0·2
2
0·1
0
1984
1989
1994
1999
2004
0
1984
2009
1989
1994
Year
1999
2004
2009
Year
≤20%
20·01–25%
25·01–30%
30·01–35%
35·01–40%
>40%
≤20%
30·01–35%
20·01–25%
25·01–30%
35·01–40%
>40%
24
0·5
0·45
20
0·4
16
Total K (g kg–1)
Total P (g kg–1)
0·35
0·3
0·25
0·2
12
8
0·15
0·1
4
0·05
0
1984
1989
1994
1999
2004
2009
0
1984
1989
Year
1994
1999
2004
2009
Year
≤20%
20·01–25%
25·01–30%
≤20%
20·01–25%
25·01–30%
30·01–35%
35·01–40%
>40%
30·01–35%
35·01–40%
>40%
Fig. 2. Change in soil nutrient levels since 1984 in the plots that underwent natural restoration. Symbols are the mean values, and the bars represent the standard deviation (n = 5).
In contrast to the results for the 30 natural restoration plots,
artificial restoration was able to reverse the ecosystem degradation even when the original vegetation cover was <20%; the
recovery became significantly faster than natural recovery
within 5 years for all parameters (P < 0Æ05; Table 1). The
results also indicate that ecosystem properties, including the
vegetation cover and soil characteristics, recovered significantly (P < 0Æ05) faster in plots with an initial vegetation
cover >20% as a result of the artificial restoration, and that all
properties were significantly better at the sites with artificial
restoration (P < 0Æ05; Table 2). In the 14 plots with vegetation cover <20% and where only natural restoration
occurred, the results were similar to those in the 30 natural restoration plots: little change or continued degradation.
Discussion
Traditional ecosystem restoration efforts have focused on reestablishing historical disturbance regimes or abiotic conditions and have relied on successional processes to assist the
recovery of biotic communities. However, strong feedback
between biotic factors and the physical environment can alter
the efficacy of these succession-based management efforts
(Suding, Gross & Houseman 2004). In our study, an initial
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
1150 Y. Gao et al.
55
45
40
50
Coarse particles (>1 mm) (%)
Plant species (no·)
35
30
25
20
15
45
40
35
10
30
5
0
1984
1989
1994
1999
2004
25
1984
2009
1989
1994
≤20%
30·01–35%
1999
2004
2009
Year
Year
20·01–25%
35·01–40%
≤20%
30·01–35%
25·01–30%
>40%
10
20·01–25%
35·01–40%
25·01–30%
>40%
55
9
50
7
45
6
Runoff (%)
Erosion modulus (kt km–2)
8
5
4
40
35
3
2
30
1
0
1984
1989
1994
1999
2004
2009
25
1984
1989
1994
1999
2004
2009
Year
Year
≤20%
20·01–25%
25·01–30%
≤20%
20·01–25%
25·01–30%
30·01–35%
35·01–40%
>40%
30·01–35%
35·01–40%
>40%
Fig. 3. Change in species richness, soil texture, soil erosion and water conservation since 1984 in the plots that underwent natural restoration.
Symbols are the mean values, and the bars represent the standard deviation (n = 5).
vegetation cover of about 20% represented such a threshold:
at a vegetation cover below 20%, the vegetation cover continued to decline slowly throughout the 25-year study period
despite a lack of human disturbance, whereas at higher initial
vegetation covers, natural recovery led to complete restoration
of the ecosystem (to 100% vegetation cover), even in the
absence of artificial restoration by planting and seeding combined with soil remediation; recovery was slower at an initial
vegetation cover of 20Æ01–25%, but vegetation cover in these
plots nonetheless reached 81% by the end of the 25 years and
showed signs of exponentially approaching 100% within only
a few more years (Fig. 1). Many soil services are closely associated with the degree of ecosystem resilience – the amount of
change a system can undergo whilst retaining the same struc-
ture, functions and feedbacks (Suding & Hobbs 2008; McVicar
et al. 2010). If this resilience declines, the ecosystem services
can generally be expected to decline (Myers 1996). Degraded
systems resist traditional restoration efforts owing to constraints such as changes in landscape connectivity and structure, species loss, changes in the dominant species, interactions
among different trophic levels, and simultaneous changes in
soil and other biogeochemical processes (Sasaki et al. 2008).
Exceeding a disturbance threshold leads to a loss of ecosystem
functions, and it may become impossible for these functions to
recover naturally, even after periods as long as 25 years. When
the vegetation cover was >20%, the vegetation cover in the
study plots was able to recover naturally in the absence of
human disturbance (Fig. 1) because the soil fertility and other
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
Degradation threshold of soil productivity 1151
35
30
Parameter value
25
20
15
10
5
0
0
10
20
30
40
50
60
70
80
90
100
Vegetation cover (%)
Organic matter (g kg–1)
Plant species (No.)
Total K (g kg–1)
Erosion modulus (kt km–2)
55
0·8
0·7
0·6
45
0·5
40
0·4
0·3
35
Total N and P (g kg–1)
Runoff, sand and coarser (%)
50
0·2
30
0·1
25
0
10
20
30
40
50
60
70
80
90
0
100
Vegetation cover (%)
Sand and coarser (%)
Runoff (%)
Total N (g kg–1)
Total P (g kg–1)
Fig. 4. Relationships between vegetation cover and various soil and
other parameters at the study plots. Data represent the results of restoration by natural forces after human disturbance was prohibited in
1984.
soil parameters also improved gradually (Figs 2 and 3),
thereby increasing the survival and supporting the restoration
of natural vegetation. This demonstrates that the reconnection
of feedback cycles between the vegetation and disturbance
dynamics was a necessary initial goal to allow the study sites to
cross the threshold between alternative stable states (Martin &
Kirkman 2009), thereby moving from irreversible degradation
to recovery. It is important to note that the specific threshold
of 20% vegetation cover that we identified in the present study
is applicable only to our study sites; different ecological proxies
for a threshold and different values of the threshold for those
parameters will exist at sites with different disturbance histories, soils, climates and vegetation communities.
Local species extinction can occur if a population falls below
a recovery threshold or if the species cannot adapt sufficiently
rapidly to environmental changes (Schlaepfer, Runge & Sher-
man 2002). Theory suggests that a disproportionate loss of species occurs when vegetation cover decreases to between 10%
and 30% (Lindenmayer, Fischer & Cunningham 2005; Radford, Bennett & Cheers 2005). Hence, some degraded systems
will shift to a new state that cannot recover to the original conditions solely through natural recovery processes, even when
further human disturbance is prevented. Interestingly, our
study suggests that even when the vegetation cover was<20%,
the number of species continued to increase throughout the
study period (Fig. 3) as a result of colonization by seeds from
adjacent areas or germination of seeds in the soil seed bank,
but other ecosystem services failed to improve (Figs 2 and 3).
This suggests that different ecological thresholds may exist for
different parameters and different species. Our results suggest
that the degraded sites with the lowest levels of vegetation
cover have shifted to a new state that cannot recover naturally
to their original conditions, even when further human disturbance was prevented. This has been previously reported for
semi-arid regions (e.g. van de Koppel, Rietkerk & Weissing
1997; Rietkerk & van de Koppel 1997), and the present study
confirms that these results are also valid for the warm, wet area
of China. Because of the prominence of the soil disturbance at
our study sites, we named this phenomenon ‘irreversible loss of
soil services’. When vegetation cover decreased below the vegetation cover threshold of 20%, the soil underwent continuing
degradation, accompanied by ongoing erosion of the surface
soil and depletion of soil fertility (Figs 2 and 3). Hence, successful restoration often requires bold and innovative management to disrupt the feedbacks that lead to long-term, sustained
degradation and to mitigate the constraints imposed by abiotic
conditions in the degraded system (McVicar et al. 2010). At
our sites, this involved implementing aggressive artificial restoration measures including planting of trees, shrubs and grasses,
combined with the addition of organic matter to restore soil
fertility.
Identifying threshold behaviour is difficult in terrestrial ecosystems because the main components of the systems change
slowly before the threshold is reached (Sasaki et al. 2008). In
addition, some changes are short-term changes caused by climate variability, which affects the life cycle (e.g. recruitment
strategies) of individual plant species. Therefore, ecologists
have been eager to develop complex predictive tools and a
broader conceptual framework capable of identifying thresholds before they are reached, thereby helping managers to prevent irreversible degradation from occurring or guiding the
restoration of degraded ecosystems (Garten & Ashwood 2004;
Suding, Gross & Houseman 2004). Unfortunately, these
frameworks have been too complex for many managers and
local residents, who can seldom use these tools to predict when
thresholds will be reached (Chapin et al. 2006). Rapid environmental change renders this task even more daunting, so a
major challenge for ecologists will be to develop effective
means of quickly assessing the status of, and prognosis for,
ecosystems that are undergoing various alterations (Jackson &
Hobbs 2009). Our results suggest that for sites similar to those
in our study, vegetation cover may be a useful proxy for more
complex threshold calculations.
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
1152 Y. Gao et al.
Table 1. Change of vegetation cover and soil parameters in 14 plots where the original vegetation cover was <20% under natural recovery and
artificial restoration since 1999
First year
Vegetation cover (%)
Organic matter (g kg)1)
Total N (g kg)1)
Total P (g kg)1)
Total K (g kg)1)
Erosion modulus (kt km)2)
Run-off (%)
Sand and coarser (>1Æ0 mm) (%)
5 years after
10 years after
Natural
recovery
Artificial
restoration
Natural
recovery
Artificial
restoration
Natural
recovery
Artificial
restoration
22Æ8
3Æ5
0Æ2
0Æ18
6Æ7
6Æ5
49Æ7
45Æ3
20Æ0
2Æ2
0Æ1
0Æ15
4Æ3
7Æ3
52Æ0
47Æ8
25Æ4
4Æ4
0Æ1
0Æ10
7Æ8
5Æ9
47Æ3
45Æ1
88Æ0
6Æ8
0Æ6
0Æ32
14Æ9
0Æ4
31Æ0
35Æ6
31Æ7
5Æ2
0Æ3
0Æ20
8Æ9
5Æ3
45Æ2
44Æ9
96Æ0
10Æ3
0Æ7
0Æ47
18Æ5
0Æ4
28Æ9
35Æ6
±
±
±
±
±
±
±
±
3Æ6ab
0Æ5a
0Æ07a
0Æ03a
1Æ7a
1Æ7a
2Æ3ab
2Æ7a
±
±
±
±
±
±
±
±
5Æ3a
0Æ3b
0Æ03b
0Æ01b
0Æ3b
1Æ3b
1Æ5a
2Æ1b
±
±
±
±
±
±
±
±
6Æ3b
1Æ3c
0Æ10ad
0Æ05a
2Æ6ac
2Æ2a
4Æ2b
3Æ2a
±
±
±
±
±
±
±
±
2Æ5c
0Æ5d
0Æ04c
0Æ02d
0Æ48d
0Æ02c
0Æ6d
0Æ5c
±
±
±
±
±
±
±
±
14Æ3d
1Æ0e
0Æ13d
0Æ06a
3Æ4c
2Æ7a
6Æ0c
3Æ7a
±
±
±
±
±
±
±
±
1Æ3e
0Æ5f
0Æ02e
0Æ03e
0Æ7e
0Æ03ac
0Æ8d
0Æ5c
Natural recovery means no human intervention and artificial restoration comprised planting and soil remediation. Values of a parameter
followed by different letters differ significantly between the artificial restoration and natural recovery treatments (P < 0Æ05). Values represent means and standard deviations (n = 7). Run-off represents the estimated proportion of rainfall that was recovered in sedimentation ponds at the base of the sample plots (see the Materials and methods section for details).
Table 2. Change of vegetation cover and soil parameters in 14 plots where the original vegetation cover was >20% under natural recovery and
artificial restoration since 1999
First year
Vegetation cover (%)
Organic matter (g kg)1)
Total N (g kg)1)
Total P (g kg)1)
Total K (g kg)1)
Erosion modulus (kt km)2)
Run-off (%)
Sand and coarser (>1Æ0 mm) (%)
5 years after
10 years after
Natural
recovery
Artificial
restoration
Natural
recovery
Artificial
restoration
Natural
recovery
Artificial
restoration
25Æ0
3Æ9
0Æ26
0Æ20
8Æ0
5Æ37
48Æ2
43Æ9
25Æ4
2Æ5
0Æ18
0Æ16
6Æ7
5Æ98
50Æ8
46Æ2
29Æ2
5Æ3
0Æ31
0Æ22
9Æ6
4Æ39
44Æ4
43Æ3
89Æ8
7Æ1
0Æ60
0Æ33
14Æ9
0Æ46
30Æ7
35Æ4
39Æ4
6Æ6
0Æ36
0Æ24
11Æ3
3Æ41
41Æ0
42Æ6
96Æ6
10Æ7
0Æ69
0Æ49
18Æ9
0Æ39
28Æ6
34Æ4
±
±
±
±
±
±
±
±
2Æ8ab
0Æ5a
0Æ07a
0Æ04a
1Æ3a
1Æ21a
1Æ9ab
1Æ8a
±
±
±
±
±
±
±
±
4Æ2a
0Æ3b
0Æ06b
0Æ02b
1Æ0b
1Æ17b
2Æ1a
2Æ7b
±
±
±
±
±
±
±
±
3Æ7b
1Æ1c
0Æ1ad
0Æ05a
2Æ4ac
1Æ05a
3Æ4b
2Æ8a
±
±
±
±
±
±
±
±
4Æ5c
0Æ8d
0Æ07c
0Æ03c
1Æ7d
0Æ05c
1Æ2d
1Æ2c
±
±
±
±
±
±
±
±
8Æ9d
1Æ1e
0Æ09d
0Æ06a
2Æ1c
1Æ53a
5Æ6c
3Æ2a
±
±
±
±
±
±
±
±
2Æ4e
0Æ8f
0Æ11e
0Æ08d
2Æ6e
0Æ05ac
1Æ9d
1Æ3c
Natural recovery means no human intervention and artificial restoration comprised planting and soil remediation. Values of a parameter
followed by different letters differ significantly between the artificial restoration and natural recovery treatments (P < 0Æ05). Values represent means and standard deviations (n = 7). Run-off represents the estimated proportion of rainfall that was recovered in sedimentation ponds at the base of the sample plots (see the Materials and methods section for details).
In addition, most degraded landscapes are a mosaic of land
uses that may include patches of intact natural vegetation and
productive agricultural lands as well as degraded lands. It is
rarely possible to revegetate the whole landscape, especially if
it includes many small areas of human land use such as farms
(Garten & Ashwood 2004). Hence, a simple predictive tool or
index that would let managers and other stakeholders predict
when an ecosystem is nearing a threshold condition is urgently
needed to facilitate restoration of such ecosystems by allowing
interventions to occur before the threshold is reached. In our
study area, we identified the threshold based on the observed
difference in trends for plots with different initial levels of vegetation cover (i.e. we identified the level of vegetation cover at
which the trend changed from deterioration to improvement).
This analysis is important in China because by the early 1990s,
3Æ67 million km2 (about 38% of the land area) was experiencing the kind of soil erosion and vegetation loss described in the
present study; this comprised 1Æ79 million km2 of water erosion and 1Æ88 million km2 of wind erosion (Anonymous 1993).
In our study area, 238 km2 has been affected by water erosion
and vegetation loss, accounting for 8% of Changting County’s
total land area; 16% of this land has a vegetation cover of
<20% and will therefore require artificial restoration instead
of only being protected in the hope that natural recovery will
occur (Cao et al. 2009b).
Our results should only be generalized to other regions with
great care. Our study was conducted in a specific warm and
rainy region of China, with its own unique history and vegetation and environmental conditions; therefore, researchers in
other regions must study the unique vegetation recovery processes in their regions to identify the level of vegetation cover
that represents a degradation threshold for their ecosystems.
That is, even when a simple index such as vegetation cover can
be identified, a more complex analysis such as that of Sasaki
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
Degradation threshold of soil productivity 1153
et al. (2008) or long-term research such as that in the present
study will be required to identify the actual threshold value for
that index.
Ecosystem restoration sometimes fails because ecological
interactions are more complex or human intervention is more
difficult than anticipated (Byers et al. 2006); factors other than
human disturbance, such as climate variability (e.g. a drought
shortly after planting of vegetation), can result in failure of a
strategy that would succeed under better conditions. Some
degraded ecosystems can only be sustained through ongoing
management, but many conservation efforts preclude such
interventions (Liu et al. 2007). Although ecologists can recognize many of the species changes that are likely to precipitate
threshold changes in community composition, biotic interactions can be unexpected, and because responses often depend
strongly on local conditions, they cannot be broadly generalized (Chapin et al. 2006). For example, complex ecosystems
with multiple interacting species may have a variety of thresholds (Garten & Ashwood 2004; Hunt et al. 2008). Desertification is another example and has been shown to result from
strong biogeomorphic feedbacks that operate across several
spatial scales (Suding & Hobbs 2008). When overgrazing of
arid grasslands reduces vegetation cover, water infiltration
decreases, further limiting plant growth and leading to persistent desertification (Byers et al. 2006). Such spatial discontinuities, called ecotones, can be detected using multivariate data
ordered in one dimension through comparisons of measures of
dissimilarity computed between the systems on either side of
the discontinuity (Andersen et al. 2009).
The ‘irreversible loss of soil services’ described in our study,
whether at national, regional or local scales, will have a variety
of thresholds, and it will be necessary to calibrate this index for
different regions before it becomes a useful management tool.
However, as our results show, it is possible to identify useful
proxies for thresholds and use them to guide subsequent management of degrading sites.
Acknowledgements
This work was supported by the Commonweal Project of the State Forestry
Administration of China (200804008) and the National Natural Science Foundation of China (41001298 and 40901098). We thank Geoffrey Hart (Montréal,
Canada) for his help in writing this paper. We also thank Jianguo Liu (Rachel
Carson Chair in Sustainability & University Distinguished Professorship,
Michigan State University), David Shankman (Department of Geography,
University of Alabama) and the journal’s editors and anonymous reviewers for
their comments on an earlier version of this manuscript.
References
Andersen, T., Carstensen, J., Hernandez-Garcia, E. & Duarte, C.M. (2009)
Ecological thresholds and regime shifts: approaches to identification. Trends
in Ecology and Evolution, 24, 49–57.
Anonymous (1993) The latest result of using remote sensing technology to
survey the area of soil loss in China released by the Ministry of Water
Resources. Soil and Water Conservation in China, 132, 1 (In Chinese).
Austin, M.P. (1980) Searching for a model for use in vegetation analysis. Vegetatio, 42, 11–21.
Austin, M.P. (2002) Spatial prediction of species distribution: an interface
between ecological theory and statistical modelling. Ecological Modelling,
157, 101–118.
Berlow, E.L., Dunne, J.A., Martinez, N.D., Stark, P.B., Williams, R.J. &
Brose, U. (2009) Simple prediction of interaction strengths in complex food
webs. PNAS, 106, 187–191.
Bestelmeyer, B.T. (2006) Threshold concepts and their use in rangeland management and restoration: the good, the bad, and the insidious. Restoration
Ecology, 14, 325–329.
Byers, J.E., Cuddington, K., Jones, C.G., Talley, T.S., Hastings, A.,
Lambrinos, J.G., Crooks, J.A. & Wilson, W.G. (2006) Using ecosystem
engineers to restore ecological systems. Trends in Ecology and Evolution, 21,
493–500.
Cao, S., Chen, L. & Yu, X. (2009a) Impact of China’s Grain for Green Project
on the landscape of vulnerable arid and semiarid agricultural regions: a
case study in northern Shaanxi Province. Journal of Applied Ecology, 46,
536–543.
Cao, S., Zhong, B., Yue, H., Zeng, H. & Zeng, J. (2009b) Development
and testing of a sustainable environmental restoration policy on eradicating the poverty trap in China’s Changting County. PNAS, 106, 10712–
10716.
Chapin Jr, F.S., Robards, M.D., Huntington, H.P., Johnstone, J.F., Trainor,
S.F., Kofinas, G.P., Ruess, R.W., Fresco, N., Natcher, D.C. & Naylor, R.L.
(2006) Directional changes in ecological communities and social-ecological
systems: a framework for prediction based on Alaskan examples. The American Naturalist, 168, S36–S49.
Garten Jr, C.T. & Ashwood, T.L. (2004) Modeling soil quality thresholds to
ecosystem recovery at Fort Benning, GA, USA. Ecological Engineering, 23,
351–369.
Huggett, A.J. (2005) The concept and utility of ‘‘ecological thresholds’’ in biodiversity conservation. Biological Conservation, 124, 301–310.
Hunt, D.E., David, L.A., Gevers, D., Preheim, S.P., Alm, E.J. & Polz, M.F.
(2008) Resource partitioning and sympatric differentiation among closely
related bacterioplankton. Science, 320, 1081–1085.
Jackson, S.T. & Hobbs, R.J. (2009) Ecological restoration in the light of ecological history. Science, 325, 567–568.
van de Koppel, J., Rietkerk, M. & Weissing, F.J. (1997) Catastrophic vegetation shifts and soil degradation in terrestrial grazing systems. Trends in Ecology and Evolution, 12, 352–356.
Lamb, D., Erskine, P.D. & Parrotta, J.A. (2005) Restoration of degraded tropical forest landscapes. Science, 310, 1628–1632.
Lindenmayer, D.B., Fischer, J. & Cunningham, R.B. (2005) Native vegetation
cover thresholds associated with species responses. Biological Conservation,
124, 311–316.
Liu, J., Dietz, T., Carpenter, S.R., Alberti, M., Folke, C., Moran, E., Pell,
A.N., Deadman, P., Kratz, T., Lubchenco, J., Ostrom, E., Ouyang, Z.,
Provencher, W., Redman, C.L., Schneider, S.H. & Taylor, W.W. (2007)
Complexity of coupled human and natural systems. Science, 317, 1513–
1516.
Ludwig, J.A., Wilcox, B.P. & Breshears, D.D. (2005) Vegetation patches and
runoff-erosion as interacting ecohydrological processes in semiarid landscapes. Ecology, 86(2), 288–297.
Martin, K.L. & Kirkman, L.K. (2009) Management of ecological thresholds to
re-establish disturbance-maintained herbaceous wetlands of the south-eastern USA. Journal of Applied Ecology, 46, 906–914.
McVicar, T.R., Li, L.T., Van Niel, T.G., Zhang, L., Li, R, Yang, Q.K., Zhang,
X.P., Mu, X.M., Wen, Z.M., Liu, W.Z., Zhao, Y.A., Liu, Z.H. & Gao, P.
(2007) Developing a decision support tool for China’s re-vegetation
program: simulating regional impacts of afforestation on average annual
streamflow in the Loess Plateau. Forest Ecology and Management, 251,
65–81.
McVicar, T.R., Van Niel, T.G., Li, L.T., Wen, Z.M., Yang, Q.K., Li, R. & Jiao,
F. (2010) Parsimoniously modelling perennial vegetation suitability and
identifying priority areas to support China’s re-vegetation program in the
Loess Plateau: matching model complexity to data availability. Forest Ecology and Management, 259, 1277–1290.
Myers, N. (1996) Environmental services of biodiversity. PNAS, 93, 2764–
2769.
Norton, D.A. (2009) Species invasions and the limits to restoration: learning
from the New Zealand experience. Science, 325, 569–571.
Radford, J.Q., Bennett, A.F. & Cheers, G.J. (2005) Landscape-level thresholds
of habitat cover for woodland-dependent birds. Biological Conservation,
124, 317–337.
Rietkerk, M. & van de Koppel, J. (1997) Alternate stable states and threshold
effects in semi-arid grazing systems. Oikos, 79, 69–76.
Sasaki, T., Okayasu, T., Jamsran, U. & Takeuchi, K. (2008) Threshold changes
in vegetation along a grazing gradient in Mongolian rangelands. Journal of
Ecology, 96, 145–154.
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154
1154 Y. Gao et al.
Schlaepfer, M.A., Runge, M.C. & Sherman, P.W. (2002) Ecological and evolutionary traps. Trends in Ecology and Evolution, 17, 474–480.
Srinivasan, U.T., Carey, S.P., Hallstein, E., Higgins, P.A.T., Kerr, A.C., Koteen, L.E., Smith, A.B., Watson, R., Harte, J. & Norgaard, R.B. (2008) The
debt of nations and the distribution of ecological impacts from human activities. PNAS, 105, 1768–1773.
Suding, K.N., Gross, K.L. & Houseman, G.R. (2004) Alternative states and
positive feedbacks in restoration ecology. Trends in Ecology and Evolution,
19, 46–53.
Suding, K.N. & Hobbs, R.J. (2008) Threshold models in restoration and conservation: a developing framework. Trends in Ecology and Evolution, 24,
271–279.
Yang, X., Zhong, B. & Xie, X.(2005) Soil Erosion and Conservation in Red-Soil
Hill Area. China Agricultural Press, Beijing. (In Chinese).
Received 6 February 2011; accepted 5 May 2011
Handling Editor: Jan Leps
2011 The Authors. Journal of Applied Ecology 2011 British Ecological Society, Journal of Applied Ecology, 48, 1145–1154