Agriculture, Ecosystems and Environment 117 (2006) 205–217
www.elsevier.com/locate/agee
Determinants of lowland use close to urban markets along
an agro-ecological gradient in West Africa
Olaf Erenstein a,1,*, Andreas Oswald b,1, Moussa Mahaman c
a
International Maize and Wheat Improvement Centre (CIMMYT), NASC Complex, Pusa Campus, New Delhi 110012, India
b
International Potato Center (CIP), Apartado 1558, Lima 12, Peru
c
Africa Rice Center (WARDA), 01 B.P. 2031, Cotonou, Benin
Received 25 June 2004; received in revised form 10 February 2006; accepted 30 March 2006
Available online 15 May 2006
Abstract
Lowland development efforts in West Africa have a mixed record. The paper argues that this is due to the neglect of market opportunity as a
driving force for lowland use and the agro-ecological gradient as an important modifier. The gradient is linked to three modifiers of lowland
use: the relative value of lowland cropping with respect to other livelihood strategies; the biophysical productivity of lowland cropping; and
the access rights to lowlands. The paper applies a regression-based decomposition framework to analyze the factors affecting lowland use in
West Africa. It uses community-level data from 1014 geo-referenced lowland units around four urban centers along an agro-ecological
gradient in Côte d’Ivoire and Mali. Tobit models are used to explain the extent of lowland non-use in terms of seasonal fallow, its diversity in
terms of rice and other crop cultivation and its land use intensity in terms of double cropping. A multinomial logit model is used to explain
lowland cropping systems. Results highlight a positive link between closeness to urban markets and the extent, diversity and intensity of
lowland use. The agro-ecological gradient in West Africa has a pronounced influence on agricultural lowland use, which tends to be more
widespread, diverse and intensive proceeding towards the drier ecologies. Lowland development did increase the extent and intensity of
lowland use, but by favoring the cultivation of rice, had a negative impact on the diversity of lowland use. Lowland use is also associated with
migrants, particularly in the more humid ecologies. Research, policy and development implications are explored. An important lesson for
scaling out is that it is not proximity to urban markets per se, but the associated market access which is the key driver for more significant
lowland use.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Urban–rural linkages; Market access; Agro-ecological gradient; West Africa; Lowland use; Peri-urban agriculture
1. Introduction
Addressing the challenge of sustainable agricultural
development in Sub-Saharan Africa will be helped by better
understanding intensification processes. Factors driving
change in agricultural land use have previously been related
to land scarcity induced by population-growth (Boserup,
1965; Ruthenberg, 1976; Andriesse and Fresco, 1991; Turner
et al., 1993) and policy and market opportunity (Pingali et al.,
1987; Lele and Stone, 1989; Izac et al., 1991; Smith, 1992;
* Corresponding author.
E-mail address: [email protected] (O. Erenstein).
1
The work was undertaken when these two authors were at WARDA.
0167-8809/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.agee.2006.03.033
Goldman, 1993; Manyong et al., 1996; Lavigne-Delville and
Boucher, 1998). Specifically, market opportunity is associated with increasing urbanization, development of transportation infrastructure, changing food consumption patterns
and general economic liberalization. In West Africa such
trends provide increasing opportunities for intensification of
agricultural systems, particularly around towns and secondary
urban centers (Club du Sahel, 2000). Simultaneously, there
has been an explosion of interest among researchers in the
development potential of farming in and around Africa’s
urban areas, or what is now widely referred to as ‘urban and
peri-urban agriculture’ (Egziabher et al., 1994; Smit et al.,
1996; Moustier and Mbaye, 1998; Ellis and Sumberg, 1998;
Bruinsma, 2001; Drechsel and Kunze, 2001; FAO, 2002).
206
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
There is a long-standing interest in greater and more
effective use of lowlands for crop production in West Africa
(Adams, 1993; Andriesse and Fresco, 1994; Kolawole et al.,
1994; Ahmadi and Teme, 1998; Hirose and Wakatsuki,
2002). Lowlands are defined here as areas that are subject to
some degree of annual flooding and comprise wetlands,
swamps, inland valleys and flood plains. Lowlands and
hydromorphic fringes are estimated to occupy more than 22
million ha of land in West Africa (Andriesse and Fresco,
1994). Much of the interest in lowland development in West
Africa has focused on the potential for increased rice
production (Pearson et al., 1981; Richards, 1986; Andriesse
and Fresco, 1991; Raunet, 1993; Windmeijer and Andriesse,
1993). Improving rice production and productivity are
priorities throughout the region and it is imperative to gain
full value from investments in lowland improvement and
irrigation development. Nevertheless, lowland development
efforts in West Africa have a mixed record of success.
Lowlands are biophysically complex and heterogeneous
(Windmeijer and Andriesse, 1993; Andriesse and Fresco,
1994) and lowland research and development efforts have
focused on biophysical constraints and corresponding
technological needs. Socio-economic factors have been
largely ignored, or acknowledged only in passing. Proximity
to urban markets, or linkage to urban markets by efficient
transportation networks, offers particular promise for more
significant agricultural use of lowlands.
A more significant use of lowlands for crops can take
various forms: extending the area under cultivation,
increasing the frequency of cropping per unit area and/or
increasing the value of agricultural production per unit area
by changing lowland use. In the end, whether farmers
extend, intensify and/or diversify agricultural lowland use is
likely a reflection of both biophysical and socio-economic
factors (Izac et al., 1991; Lavigne-Delville and Boucher,
1998; Erenstein et al., 2006). Yet little is known of the
relative contribution of these factors to the choice of
agricultural lowland use strategies in West Africa.
It is in this light that further analysis of the factors
affecting lowland use close to towns in West Africa is
warranted. In this paper we analyze the dimensions of
lowland use in the proximity of four urban centers along an
agro-ecological gradient in West Africa and present
empirical models to test the contribution of the driving
and modifying factors. Finally we discuss some research,
policy and development implications of the analysis.
2. Research sites, data and methods
2.1. Site characteristics
Four West African urban centers were selected along an
agro-ecological gradient ranging from the humid forest in the
South to the Guinea savanna zone in the North (Table 1). Each
site has a large urban settlement surrounded by ‘‘satellite’’
villages. Each town is located on a major transport axis and
functions as a market hub integrated in regional trade
networks, thereby exhibiting strong urban–rural linkages
(Bah et al., 2003). Across all sites, farming is the predominant
economic activity in the villages. All sites have an abundance
of lowlands in and around town. The agro-ecology is a major
discriminatory factor amongst sites, particularly in terms of
the duration of the growing season, precipitation and any
tradition of lowland use. In addition, the sites also vary in
terms of town size, the structure and organization of
agricultural systems and markets, the origin of lowland users
and the local policy environment. Their characteristics and
regional importance has been variously documented (Alla,
1991; Coulibaly et al., 1998; ICEF et al., 1999; Club du Sahel,
2000; ICEF and ENSEA, 2002; Erenstein et al., 2004).
Each urban center serves as a hub for regional agricultural
development efforts. Around each site various lowlands have
been developed through outside intervention; for example in
Côte d’Ivoire between 1970 and 1975 by Soderiz (a rice
development parastatal, now defunct) and in Mali in the
Table 1
Selected characteristics of research sites
Urban center
Daloa (Côte d’Ivoire)
Bouake (Côte d’Ivoire)
Korhogo (Côte d’Ivoire)
Sikasso (Mali)
Agro-ecological zone
Forest
Rainfall (mm year1)
Humid period
Location
Urban population (estimate, 2000)
Rural population density
(inhabitants km2)
1300
March–November
68530 N, 68270 W
260000
39
Transition
forest-savannah
1000
April–October
78410 N, 58020 W
550000
30
Southern Guinea
(SG) savannah
1100
May–September
98270 N, 58390 W
190000
44
Northern Guinea
(NG) savannah
1100
June–September
118190 N, 58400 W
130000
35
84
345
4.11 a
[13]
122
226
1.85 b
[6]
144
292
2.03 b
[14]
150
151
1.01 c
[2]
Lowland inventory data:
Number of villages (n = 500)
Number of lowland units (n = 1014)
Average number of lowland
units per village [max.]a
Source: Erenstein et al. (2004).
a
Data followed by different letters differ significantly—Duncan (0.10), within row comparison.
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
mid-1990s by the CMDT (the former parastatal cotton
board, now privatized, also responsible for rural development and extension).
The rural population densities in the four sites are
relatively low, ranging from 30 to 44 inhabitants per km2
(Table 1). However, all sites have seen a rapid population
growth over the second half of the 20th century. This was
particularly acute in the forest and transition zone sites of
Côte d’Ivoire, which attracted significant numbers of
migrants from drier areas in response to labor opportunities.
Each site also witnessed an increasing urbanization (Kalasa,
1994; Erenstein et al., 2004).
2.2. Data and limitations of the study
For each site a geo-referenced inventory was made of
lowlands and their use within an approximate 30 km radius
around each urban center. The radius was expected to
capture a gradient of proximity to the urban center, while
still keeping fieldwork manageable. The fieldwork was
implemented in two batches (Bouake and Korhogo in 1998–
1999 and Sikasso and Daloa in 2001) as it could be
completed only after mobilizing the necessary resources.
The two time periods are similar enough in climatic and
socio-economic conditions to allow for joint analysis. The
data gathering tool included the passing of a one-page
questionnaire to make an inventory of selected village and
lowland characteristics. The questionnaire was administered
to one or more key informants for each lowland unit. The
corresponding lowland area was geo-referenced with a
global positioning system (GPS) device. Other georeferenced information was digitized from topographical
maps and complemented as needed during the fieldwork.
The data collection unit for this study was the lowland–
village combination: i.e. a specific lowland area with the
corresponding village that was entitled to its use. A specific
lowland area is thereby defined as an adjoining geographically
delimited lowland area, comprising either an entire lowland or
a part thereof. ‘Village’ is used here as synonymous with the
habitation center of entitled lowland users and comprises both
villages and neighborhoods within urban centers. It thereby
includes villages with use-rights over lowlands even if they
opt not to use the lowland for crop cultivation purposes, but
excludes villages without any lowland entitlements. Traditional land tenure systems prevail across the sites with a key
role for the village land chief. The village thus was a relevant
entry point and aggregation level. However, this aggregation
level does not imply that villagers are a coherent homogeneous group with equitable rights or that they act in a
coordinated way. One large inland valley can have several
entitled villages, and one village can have use rights over
several different lowlands, each combination would then
represent a different lowland unit.
Over all sites a total of 500 villages with lowland
entitlements were identified with 1014 lowland units. There
is a marked gradient in terms of the number of lowland units
207
per village, with an average of 4.1 in the forest zone and 1.0
in Northern Guinea savannah (Table 1). This gradient is the
net result of an increasing number of entitled villages and a
decreasing number of lowlands per site. Field and
cartographical observations support these observed trends:
lowlands become wider and population hamlets more spread
out as one proceeds from the forest zone to the Northern
Guinea (NG) savannah.
Some limitations of the data set should be noted. The
short rapid survey instrument allowed for the compilation of
a large baseline but with relatively limited depth. Focus
thereby was on aggregate indicators of villages, lowlands,
crop use and users, including factors that were both
exogenous and endogenous to lowland use. Subsequent
data gathering efforts were to provide more disaggregated
indicators including differentiation among households/
social groups, system linkages with upland, livestock and
markets and relative economics of livelihood strategies.
However, these efforts were discontinued at an early stage
when civil strife erupted in September 2002 in Côte d’Ivoire.
There are also some data gaps, particularly in some of the
quantitative variables for the sites covered in the first batch,
e.g. lowland size. Compared to non-used lowlands, used
lowlands are less likely to be underreported and are more
likely to be subdivided into different lowland units.
Lowland-unit-based estimates are therefore an imperfect
proxy for lowland use intensity in absolute terms, but this is
less of an issue when comparing land uses in relative terms.
Another limitation of the data set is the inability to
control for all modifying variables. Having four urban sites
along the agro-ecological gradient implies that certain
variables are entangled, particularly as available data do not
allow us to disaggregate certain variables per site like
rainfall and population density.
2.3. Characteristics of lowland use in study sites
Eighty-four percent of the surveyed lowland units around
the four urban centers are used for crop cultivation. Rice is
the predominant crop reported in 75% of the lowland units.
Other crops were reported in 44% of the lowland units. Land
use intensity in lowlands is limited. On average only about
half of the lowland-unit area is used for crop cultivation. For
simplicity, we only consider two cropping seasons: the rainy
season and the off-season. Cumulative over two seasons, the
annual area share averages 57% for rice, 36% for other crops
and 107% for fallow. Merely 37% of the lowland-unit area is
double cropped.
Rice-only systems are the main cropping system across
sites, reported in 42% of the surveyed lowland units. This
comprises 31% with rice cultivation during the rainy season
followed by an off-season fallow, and 11% with rice double
cropping. Mixed systems were reported in 27% of the
lowland units and imply rainy season rice cropping followed
by an off-season non-rice crop. Twenty-one percent of the
lowland units can be considered unused for cropping, i.e.
208
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
Table 2
Selected lowland use characteristics of research sites
Forest
(Daloa)
Transition
(Bouake)
Annual area share with (cumulative over two seasons, n = 929, max = 200%)a
Seasonal fallow
164% d
113% c
Rice crop
35% a
51% b
Other crop
1% a
36% b
200%
Lowland share with double cropping (n = 929, max = 100%)
Lowland cropping system (n = 988)b
Fallow
Rice only
Mixed
Non-rice only
7% a
200%
Southern Guinea
savannah (Korhogo)
47% a
99% c
54% c
200%
Northern Guinea
savannah (Sikasso)
86% b
33% a
81% d
200%
Across all
sites
107% (0.00)
57% (0.00)
36% (0.00)
200%
40% b
68% c
41% b
37% (0.00)
24%
74%
1%
1%
42%
24%
26%
9%
8%
36%
55%
2%
5%
8%
35%
52%
21%
42%
27%
11%
100%
101%
101%
100%
101%
Source: Erenstein et al. (2004).
a
Data followed by different letters differ significantly—Duncan (0.10), within row comparison. Figure in brackets in last column is the probability.
b
Chi-square of site effect highly significant (0.00).
fallow. Non-rice systems were reported in 11% of the
lowland units, comprising 6% single cropped and 5% double
cropped.
There is significant variation in lowland use across the
four research sites (Table 2), which is partly a reflection of
the agro-ecological gradient. There is also a significant
spatial variation of lowland use around each urban center
(Fig. 1). The empirical models aim to explain observed
differences in lowland use based on the various lowland and
village characteristics inventoried.
2.4. Empirical models
Von Thünen provided a basic model of agricultural land
use with increasing proximity to an urban center (Katzman,
1974; Nelson, 2002). The original model was based on the
Fig. 1. Distribution of predominant cropping system by lowland unit by site (30 km radius).
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
value of two agricultural products and the corresponding
transport costs to one central market. In much the same way,
we distinguish rice, other crops and fallow and use travel
time to the urban market as proxy for transport cost. The
current study distinguishes two cropping seasons which adds
a temporal dimension of land use to the purely spatial
dimension of the original model. Compounding the
seasonality, the two crop types differ in terms of their
appropriateness for each season, whereby rice is the main
lowland option in the rainy season because of flooding, other
crops in the off-season. We seek to explain five indicators of
lowland use across the four research sites:
1. Fallow: extent to which lowlands are not used for
cropping.
2. Rice cultivation: extent to which lowlands are used for
rice.
3. Other crop cultivation: extent to which lowlands are used
for non-rice crops.
4. Double cropping: extent to which lowlands are used in
both rain season and off-season.
5. Cropping system: main cropping system observed in
lowland.
For each indicator of lowland use we construct a separate
model. For the first three indicators we use Tobit models to
estimate the annual area share devoted to fallow, rice and
other crops, respectively. The dependent variables for these
models are censored between 0 and 200%, reflecting the
cumulative area share over two seasons devoted to the
respective use (Table 3). The first three models are
complementary as the sum of the three dependent variables
for each lowland unit adds up to 200%, the maximum
cropping intensity over the rain and off-season combined.
The fallow indicator reflects the extent of lowland non-use
whereas the other crops are an indicator for the agrobiodiversity of the lowland use system.
209
For the fourth indicator we use a Tobit model to estimate
the lowland share which is double cropped. The dependent
variable for this model is censored between 0 and 100%
(Table 3). The double cropping indicator reflects the lowland
use intensity.
For the fifth indicator we present one multinomial
logit model. The dependent variable for the cropping
systems model can assume three mutually exclusive
cropping systems: rice only (single or double cropped),
mixed (rice and non-rice) and non-rice (single or double
cropped) (Table 3). The cropping systems had to be
grouped as certain systems are infrequent and do not
occur in all sites. The cropping system model thus cannot
capture the intensity and diversity dimensions simultaneously. Qualitative dependent variable models have been
variously applied to land-use analysis in other settings
(Gobin et al., 2002; Staal et al., 2002; Munroe et al.,
2002).
The descriptive statistics of the independent variables
included in the empirical models are given in Table 3. The
independent variables included in the models cover a range
of relatively fixed and exogenous characteristics of lowlands
and villages that are expected to be associated with lowland
use decisions. To facilitate comparison, the models use the
same explanatory variables. The first four (Tobit) models are
based on all lowlands with valid data. The fifth (multinomial
logit) model contrasts cropping systems given that the
lowland is used.
2.5. Proximity to urban market
Proximity to urban centers reduces transport costs and
increases the exposure to market opportunities, thereby
providing the basic drive to agricultural land use change.
Proximity is central to the notion of peri-urban agriculture,
which has been defined as ‘characterized by strong urban
influences, including increased possibilities for marketing
Table 3
Descriptive statistics for variables used in empirical models
Model
Description
Mean
S.D.
Minimum
Maximum
Cases
Independent
1–5
1–5
1–5
1–5
1–5
1–5
1–5
1–5
1–5
variables
Travel time of lowland to urban market (rainy season, h)
Square of travel time of lowland to urban market (rainy season, h2)
Transition zone (1: yes; 0: no)
Southern Guinea savannah zone (1: yes; 0: no)
Northern Guinea savannah zone (1: yes; 0: no)
Lowland formal development (1: yes; 0: no)
Distance of lowland to village (km)
Square of distance of lowland to village (km2)
Migrant village (1: yes; 0: no)
0.92
1.35
0.22
0.29
0.15
0.17
1.61
5.07
0.11
0.71
2.08
0.42
0.45
0.36
0.38
1.58
10.30
0.31
0.01
0
0
0
0
0
0
0
0
4.2
17.8
1
1
1
1
9.7
94.0
1
1012
1012
1012
1012
1012
1012
1012
1012
969
Dependent variables
1
Annual area share with fallow (cumulative over two seasons)
2
Annual area share with rice (cumulative over two seasons)
3
Annual area share with other crops (cumulative over two seasons)
4
Lowland share with double cropping
5
Lowland cropping system (0: rice only; 1: mixed; 2: non-rice only)
1.07
0.57
0.36
0.37
0.60
0.79
0.56
0.50
0.44
0.71
2
2
2
1
2
929
929
929
929
785
0
0
0
0
0
210
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
of farm produce, provision of input and services, and nonfarm employment, but exacerbated competition for land,
inequities in its distribution and risks from pollution’
(Adam, 2001). The latter constraints have to be put into
context with limited agricultural development opportunities
in remote rural areas in West Africa (Binswanger and
McIntire, 1987). Indeed, the availability of road infrastructure and proximity to markets are increasingly seen as
crucial for agricultural and economic development (Christiaensen et al., 2003; Tiffen, 2003; Bah et al., 2003).
Therefore, we project primarily positive effects on
agricultural lowland use in the hinterland of urban centers
in West Africa.
The extent, intensity and diversity of lowland use are all
expected to be positively associated with proximity to
urban markets. Non-rice lowland crops often comprise
vegetables, roots and tubers and maize and are often of
higher value and more perishable than rice. In addition,
lowland rice production reflects a dual objective of market
and home consumption and tends to rely on extensive
cultivation practices, whereas lowland production of other
crops is rather market-oriented and more intensive.
Proximity of urban markets will also increase opportunity
costs of resources and provide livelihood diversification
options out of agriculture. These may well displace nonmarket-oriented, low-input lowland rice cultivation further
away from the urban market. With increasing proximity to
urban markets, lowland cultivation is expected to increase
and production systems to become more intensive and
move through a continuum from fallow or rice-only, to
mixed, and finally to non-rice (Erenstein et al., 2006). This
implies that the effect of proximity to urban markets on rice
cultivation is unlikely to be linear but subject to a
maximum.
The travel time to the central market is a variable that
captures the temporal distance of the lowland to the urban
market. It is used here as the main indicator of proximity to
urban markets. Travel time was estimated for the rainy
season based on the digitized road network and taking into
account the distance from the lowland to the nearest road,
road quality and assumed maximum travel speed
(Erenstein et al., 2004). Temporal distance is expected
to be a better proxy of lowland market access and related
transaction costs than spatial distance measures (e.g. radial
distance or travel distance), particularly in view of the low
density and quality of transportation infrastructure in West
Africa (Manyong et al., 1996; Club du Sahel, 2000). In the
research sites only three to six surfaced roads radiate out of
each urban center, which implies most lowlands depend
primarily on un-surfaced roads for market access. Travel
time in this context is the minimum temporal distance as
most people do not have own vehicles and ‘‘reliability’’ of
transport (waiting time, cost, etc.) may increase actual
travel time considerably. Travel time to the central market
is also included in its squared form to capture the nonlinear effects.
2.6. Site/zone
The four research sites are located along an agroecological gradient, with length of growing season
decreasing as one proceeds south to north from forest zone
to NG savannah. This gradient is also associated with
historical, economic and social factors and finds its
expression in the intensity of upland and lowland farming
activities. The gradient is indeed linked to three modifiers of
lowland use.
A first modifier is the relative value of lowland cultivation
in comparison to other livelihood strategies, such as
cropping the surrounding uplands. In the forest zone of
West Africa, the extended growing season allows for
prolonged cultivation of uplands, including perennial cash
crops. Historically, farmers around the Daloa site in the
forest zone focused their attention on upland cultivation,
including cash crops such as cocoa and coffee, and food
crops including upland rice, maize, yams and cassava. In the
savannah zone, the growing season for upland cultivation is
shorter, and lowlands are almost the only option to grow
crops outside irrigated areas during the dry season (Izac
et al., 1991). Around the Korhogo and Sikasso sites in the
savannah zone farmers have long cultivated crops like
cotton, maize and sorghum in uplands and rice, (sweet)
potato and vegetables in lowlands, with a seasonal shift
between the two types of land. During the growing season,
farmers devote more attention to upland cultivation
(Lavigne-Delville, 1998). In areas like the forest zone
where both upland and lowland rice can be cultivated,
farmers prefer upland rice. Lowland cultivation is considered more onerous and labor-demanding than upland
cultivation (Spencer and Byerlee, 1976; Richards, 1986).
Lowlands are associated with considerable investments in
water control structures and also human health hazards (Izac
et al., 1991; Windmeijer and Andriesse, 1993; McMillan
et al., 1998; Henry et al., 2003). Consequently, lowlands
remain largely unused compared to uplands despite potential
nutrient and moisture benefits (Pingali et al., 1987).
A second modifier is the productivity of lowland cropping
per se, both in terms of yield and resources needed. Rice
yields in West African inland valleys are reported to range
from <1 to >7 tonne ha1 (Becker and Johnson, 1999).
Lowlands are diverse in their bio-physical characteristics, be
it lowland form and water regime, micro-topography, soil
type or biophysical constraints (Raunet, 1985; Windmeijer
and Andriesse, 1993). Inland valleys in the forest zone are
relatively narrow and experience prolonged water excess.
These excess moisture conditions imply, even in the offseason, the need for ridges, raised beds or mounds to grow
crops under aerobic conditions in the forest zone, whereas in
the NG savannah crops can be grown without these
structures. In the savannah, inland valleys are wider and
prone to both flash floods and drought stress during the rainy
season (Raunet, 1985; Andriesse and Fresco, 1994). The
prospects for un-irrigated off-season cultivation diminish as
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
one proceeds northwards along the gradient. In the NG
savannah off-season lowland cultivation implies laborintensive hand watering from shallow wells. In the transition
and Southern Guinea (SG) savannah, off-season lowland
cultivation is mainly done in the run-up to the rainy season,
using pre-season showers before significant lowland flooding sets in.
A third modifier is that of access rights to lowlands.
Traditions and power relations imply different cropping
opportunities based on wealth, age, gender, ethnicity or
origin (Diallo, 1998; De Groote et al., 1998). Low
productivity of undeveloped lowlands implies they are
often used for subsistence agriculture only and left to women
while men manage cash-crops in the adjacent uplands.
Lowlands may have other uses such as fishing, hunting,
gathering, grazing, watering and passage that constrain their
potential for cropping (Saidu, 1994; Lavigne-Delville,
1998). Livestock becomes a more important component
of rural livelihoods as one proceeds from forest to savannah
sites and trypanosomiasis becomes less of a threat. In the
savannah zone, pastoralists’ demands for lowland access can
impose seasonal or year-round constraints on crop production, and generate crop-livestock conflicts (Shikano, 2002).
Lowlands are also often associated with the spiritual world
in West Africa, be it sacred forests, cult sites or evil spirits
(Terrin, 1995; Lavigne-Delville, 1998). The foregoing
factors contribute to a lowland avoidance strategy in the
forest zone and an increasing tradition of lowland use
proceeding northwards (Becker and Diallo, 1996). Corresponding use rights have long been established and thereby
inherently modify lowland use dynamics.
Two caveats merit mentioning here. Due to data
limitations we cannot disaggregate the relative contribution
of each of these modifiers per site. Site-specific differences
other than those related to the aforementioned gradient and
modifiers also exist and these are also captured by the site
effect, e.g. town size and the corresponding market pull.
The extent, intensity and diversity of lowland use are
expected to increase proceeding northwards as lowlands in
drier areas have a higher use value for agriculture and offer a
wide range of uses for crop production. The data set
comprises four sites in distinct agro-ecological zones. We
normalize on the forest zone site, and included three dummy
variables to capture the distinctness of the three other sites.
2.7. Lowland and user characteristics
The dataset comprises a limited number of additional
characteristics of lowlands and entitled villages that might
help explain lowland use decisions. The dummy lowland
development implies that there are structures to control
water at the lowland level. This is key in making crop
production more effective and less flood and/or drought
prone (Becker and Johnson, 1999, 2001) and opens the way
to intensification of rice production. Lowland development
usually was done with outside assistance to develop small
211
surface irrigation systems based on a stream diversion or by
a dam and storage reservoir (Humphreys, 1981; Randolph
et al., 2000). Rice is closely associated with surface
irrigation systems, and water control is a conditio sine qua
non for rice double cropping in West Africa. Non-rice crops
often rely on low levels of irrigation technology. Maintenance of lowland structures is often deficient across sites
but still lowland development is expected to be positively
associated with the extent and intensity of lowland use and
rice cultivation.
The distance of the lowland to the village indicates the
remoteness of lowland fields from the homestead. It is
included as a proxy for ease of lowland use and as a spatial
distance measure (geometric distance, ‘as the crow flies’)
lacking better estimates across all sites. It is expected to be
negatively associated with the extent and intensity of
lowland use. This distance measure is also included in its
squared form to capture non-linear effects.
Across the study sites traditional land tenure systems still
prevail. Migrants can thereby access land through the land
chief, albeit on more stringent conditions than indigenous
groups. Due to the traditional preference for upland
cultivation, lowland cultivation is often done by migrants
in the forest and transition zone (Sakurai, 2002; Tachibana
et al., 2002). Like most cash-crop areas in West Africa,
Daloa received many migrants from drier areas to the north
who were attracted by the labor opportunities in the cocoa
and coffee plantations. These economic migrants eventually
diversified their livelihoods through crop production. Many
of them had experience in lowland cultivation and as
lowlands had remained largely unused by the landowners,
they developed these lands to cultivate rice and other crops.
The dummy migrant village is expected to be positively
associated with the extent and intensity of lowland use.
3. Results
The results of the four Tobit models are presented in
Table 4. The four models have reasonable explanatory power
with an adjusted R-squared in-between 0.4 and 0.5. The
results of the cropping system model are presented in
Table 5. The model predicts 74% of the cases correctly.
Several of the explanatory variables are statistically
significant in explaining lowland use in each model.
Significant variables also have the expected algebraic signs.
Travel time to the urban market proved significant in all
models except for rice. Only in the case of fallow was the
sign positive: increasing travel time decreases the extent of
lowland use. Increasing travel time adversely affects the
extent of cultivation of other crops and double cropping in
lowlands and reduces the likelihood of finding mixed and
non-rice systems. The quadratic travel time term was
significant for the other cropped and double-cropped model.
This implies the cultivation of other crops and double
cropping decrease non-linearly with travel time to a
212
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
Table 4
Factors affecting lowland use (four Tobit models, normalized on forest zone)
Independent variable
Model 1:
fallow
Model 2: rice
cultivation
Model 3: other
crop cultivation
Model 4: double
cropping (any crop)
Constant
1.61 (0.10) ***
0.19 (0.062)***
0.81 (0.13)***
0.42 (0.13)***
0.31 (0.13) **
0.048 (0.042)
0.051 (0.079)
0.0078 (0.026)
0.47 (0.12)***
0.067 (0.039)*
0.68 (0.15)***
0.11 (0.049)**
0.29 (0.11)***
1.70 (0.084)***
1.05 (0.095)***
0.056 (0.065)
0.75 (0.047)***
0.042 (0.059)
1.04 (0.12)***
1.64 (0.11)***
2.09 (0.12)***
0.71 (0.13)***
1.84 (0.12)***
1.41 (0.13)***
Lowland and user characteristics
Lowland development (dummy)
Distance lowland–village (km)
Square of distance lowland–village (km2)
Migrant village (dummy)
0.70 (0.088)***
0.13 (0.055)**
0.016 (0.0078)**
0.51 (0.098)***
0.73 (0.051)***
0.051 (0.033)
0.0068 (0.0047)
0.34 (0.059)***
0.19 (0.080)**
0.058 (0.062)
0.0013 (0.011)
0.33 (0.11)***
0.76 (0.098)***
0.13 (0.067)**
0.017 (0.010)*
0.41 (0.12)***
Model parameters
Adj. R-squared
Significance level
Parameters/degrees of freedom
Valid cases
Basis
Censoring
0.49
0.000
10/889
899
Annual area
0–200%
0.46
0.000
10/889
899
Annual area
0–200%
0.39
0.000
10/889
899
Annual area
0–200%
0.43
0.000
10/889
899
Lowland area
0–100%
Proximity to urban market
Travel time lowland–town (h)
Square of travel time lowland–town (h2)
Site/zone
Transition (dummy)
SG savannah (dummy)
NG savannah (dummy)
Standard errors
*
Significant
**
Significant
***
Significant
are in parenthesis.
at 10%.
at 5%.
at 1%.
Table 5
Factors affecting cropping systems in used lowlands (multinomial logit model, normalized on forest zone and on rice only systems)
Independent variable
Model 5: cropping system (used lowlands)
Mixed
Non-rice
***
3.73 (0.74)***
Constant
3.61 (0.71)
Proximity to urban market
Travel time lowland–town (h)
Square of travel time lowland–town (h2)
1.25 (0.48)***
0.13 (0.16)
1.07 (0.61)*
0.18 (0.18)
Site/zone dummies
Transition (dummy)
SG savannah (dummy)
NG savannah (dummy)
4.75 (0.68)***
5.48 (0.66)***
6.71 (0.73)***
3.74 (0.67)***
1.29 (0.83)
6.83 (0.69)***
Lowland and user characteristics
Lowland development (dummy)
Distance lowland–village (km)
Square of distance lowland–village (km2)
Migrant village (dummy)
1.17 (0.25)***
0.054 (0.23)
0.0047 (0.039)
0.67 (0.56)
2.80 (0.55)***
0.32 (0.35)
0.055 (0.062)
0.50 (0.60)
Model parameters
Cases predicted correct
Log-likelihood
Chi-squared
Degrees of freedom
Significance level
Data set
Valid cases
74%
410
629
18
0.000
Used lowlands
751
Standard errors are in parenthesis.
*
Significant at 10%.
***
Significant at 1%.
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
minimum. These results are in line with expectations, i.e.
increasing travel time to urban markets significantly reduces
the extent, intensity and diversity of lowland use.
Having been normalized on the forest zone, the sitedummies show that all other sites have lower rates of
seasonal fallow and higher rates of other crops and double
cropping. The magnitude of the coefficients reiterates the
importance of the south–north gradient, with fallow
declining to a minimum in the SG savannah site, double
cropping increasing to a maximum in the same and
cultivation of other crops increasing continuously. The
cropping system model conveys similar results, with the
likelihood of finding mixed systems being markedly higher
outside the forest zone and increasing along the gradient.
The site-dummy in the rice model is only significant and
positive for the SG savannah site. Conversely, the likelihood
of finding non-rice systems is markedly higher in the
transition and particularly the NG savannah zone. All
models thus highlight the limited extent, intensity and
diversity of lowland use in the forest zone.
Lowland development plays a significant role in all
models. The models confirm the positive association of
lowland development with the extent and intensity of
lowland use, being negatively associated with fallow and
positively with double cropping cycles. Lowland development also has a clear negative effect on agro-biodiversity,
reducing the number of crops cultivated in a lowland in favor
of rice, being positively associated with rice and negatively
associated with other crops and mixed and non-rice cropping
systems. This also indicates that water control can make rice
a competitive crop, even in proximity of urban centers, and
that certain farmers prefer rice cultivation to alternative
lowland uses.
Migrant villages clearly enhance the extent and intensity
of lowland use, being negatively associated with fallow area
and positively with rice, other crops and double cropping.
This underscores the important role migrants play in lowland
cultivation and the associated low interest of the indigenous
population in the forest zone, where most migrant villages
are found. The distance of the lowland to the village and its
square only proved significant in the fallow and double
cropped models. This implies that both the extent and
intensity of lowland cultivation decrease non-linearly with
increasing distance to the homestead. There is however no
significant effect on the choice of crop.
4. Discussion and implications
The contrasting research sites illustrate the variety of
lowland use systems in West Africa, whereas the empirical
models help to explain a significant part of the observed
differences. No single lowland use indicator and corresponding model fully captures all dimensions of lowland
use. The fallow indicator only reflects the extent of lowland
non-use, double cropping the intensity and cultivation of
213
other crops the diversity. The Tobit models help explain the
degree of lowland use, but only for one indicator at the time.
The multinomial logit model has the advantage that the
diversity and intensity dimensions can be potentially
incorporated in one model. This implies that each of the
five models added some unique explanatory power. Taken
together, the various indicators and models help to
disentangle some of the complexity of lowland use in West
Africa. The results highlight that lowland use is primarily
associated with proximity to urban markets, the agroecological gradient, lowland development and migrants.
4.1. Market opportunity as driver of lowland use
The study supports Von Thünen’s predicted preference
for high value crops over rice with proximity to urban
markets. However, there are some important extensions to
the original model in the context of the West African
lowlands.
Rice compared to other crops differs in terms of market
integration and tradability, contributing to the non-significant role of closeness to urban markets on its cultivation.
Across the research sites, rice is an important food crop,
often grown to cover at least partly the consumption needs of
farm households (e.g. De Groote et al., 1998), whereas other
crops cultivated in lowlands, particularly vegetables, are
primarily produced for the market (e.g. potato around
Sikasso, Kone et al., 1998). Rice is also a tradable
commodity and as such subject to competition from
imported rice in each of the sites (Coulibaly et al., 1998;
ICEF et al., 1999; ICEF and ENSEA, 2002; Lançon et al.,
2004), whereas for most other lowland crops such
competition is either absent or less pronounced. This
underlines that proximity to urban markets may have
ambiguous effects on the intensification of rice production,
particularly when production costs are high and produce
quality unfavorable.
Despite the importance of market opportunity as a driver
of lowland use, there are several challenges in terms of
measuring it empirically. Proximity to urban markets was
successfully used as proxy for exposure to market
opportunities. The present study substituted temporal
proximity for spatial proximity to accommodate the
condition of road networks in West Africa. This could be
further refined by including access to transport. However,
proximity remains an imperfect proxy for market opportunities. Proximity ignores the extent of ‘market pull’ as
influenced by the size of the urban market and linkages to
market networks. Satellite markets, decentralized traders
and/or trading networks can compensate for remoteness of
production sites and influence the development of agricultural production systems and their profitability. Around
Daloa, rice traders bulk rice at the village level and sell it in
Daloa town. In Sikasso, town-based traders buy potatoes
directly from lowland producers at the village level but also
provide complementary services (e.g. seed on credit,
214
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
fertilizer, transport, etc.). Proximity is also associated with a
number of other urban factors including land scarcity. To
accommodate the potentially confounding effects, proximity to the market could be substituted by the notion of
‘market access’ (Snrech, 1995; Erenstein et al., 2006).
Subsequent research could attempt to make this concept
operational and subsequently consider market access and
rural population density in site selection so as to disentangle
the relative contribution of these two drivers of land use.
4.2. Modifiers of lowland use
The models highlight that lowland use is associated with
the agro-ecological gradient, lowland development and
migrants. A common thread linking these variables is that
they modify resource scarcity and, by extension, lowland use
incentives. During the prolonged dry season in the Guinea
savannah, lowlands with their shallow water table represent
a precious option for irrigated crop production. In the more
humid regions further south this comparative advantage
diminishes because precipitation and rainy seasons are more
extensive. Lowland development raises the expected returns
to lowland use by reducing crop production risks and
increasing productivity and potential land use intensity.
Migrants typically have a more restricted access to land, but
also tend to have the necessary skills for making use of the
lowland, thereby increasing their incentives for lowland use.
The agro-ecological gradient in West Africa has a strong
influence on agricultural lowland use. The gradient is linked
to three modifiers of lowland use: the relative value of
lowland cropping with respect to other livelihood strategies;
the biophysical productivity of lowland cropping; and the
access rights to lowlands. The gradient is thus a reflection of
bio-physical and socio-economic factors and their interrelationships. Disentangling the contribution of these
modifiers and other site effects is a challenge that merits
further follow-up research. The present study thereby
highlights some of the limitations of a ‘‘key-site’’ approach,
whereby ‘‘representative’’ lowlands are singled out for indepth bio-physical and socio-economic characterization and
subsequent research.
A significant role for physical lowland development was
to be expected. Yet, there are a number of issues with these
significant investments of scarce public resources in West
Africa. For one their effectiveness, as there is often only
partial surface water control, i.e. lowland users cannot
irrigate or drain their fields when they would like to but
depend on other users or specific irrigation and drainage
plans. Surface water control was found to be virtually absent
in the NG savannah site, where variable flooding levels in the
rainy season prevail even in developed lowlands and water
availability is limited for gravity irrigation in the off-season.
The extent and intensity of lowland use after development
often falls short of expectations, particularly in the forest
zone. Lowland development increases the returns to crop
cultivation, but often leaves lowland entitlements in the hand
of specific ethnic groups which in recent years has generated
stakeholder conflicts in West Africa (Bouju, 1998; LavigneDelville, 1998). Land disputes have been common in the
forest zone of Côte d’Ivoire, including the Daloa site
(Lavigne-Delville and Boucher, 1998), exacerbated by the
ethnicity of lowland use within the socio-political context.
Lowland development reduces risk for production of
lowland rice, but also implies rice increasingly displaces
non-rice lowland crops and thereby reduces livelihood
diversification. These various issues question the social
returns to physical lowland development and merit closer
scrutiny in future lowland development efforts. There is a
need for more appropriate and cheaper irrigation technology
for lowlands.
4.3. Research approach
The present study shows the considerable scope for
predicting aggregate land use patterns by using aggregate
zone/village/lowland-level indicators. West African lowlands may be complex and challenging to characterize
(Windmeijer and Andriesse, 1993; Andriesse and Fresco,
1994), but the extent and intensity of lowland use can largely
be explained by a few key indicators: market access, agroecological zone, lowland development and migrants. These
indicators force us to think about the underlying drivers and
modifiers and how these modify the incentives to lowland
use. Their significance across contrasting sites also suggests
they are robust enough for extrapolation and allow for rapid
and enhanced targeting of lowland intensification efforts.
There is scope for improving the research approach and
indicators. Some of the proxy’s are imperfect and would
benefit from further refinement, as the above discussion of
market access and the agro-ecological gradient suggests.
The aggregation level facilitates data collection, but these
aggregate indicators however mask differences that exist
within lowlands and/or communities. These endogenous
factors modify the ability of households and/or individuals
to respond to market opportunities. There is significant
diversity amongst lowland users at the four sites. Migrants
are the main developers in the humid zone while further
north the indigenous populations cultivate most lowlands.
Women are responsible when rice is grown as a subsistence
crop during the rainy seasons, while men take over crop
cultivation like vegetables during the off-season when crop
production is market-oriented (Lavigne-Delville, 1998). The
analysis can be strengthened further by refining and adding
more detailed data, including the relative profitability of
crops and livelihood strategies and system linkages with
upland and livestock. The present study also did not consider
technology due to endogeneity reasons, but available
technologies influence crop choice through yield, resource
use, profitability, etc. (Erenstein, in press).
Although we have provided an historic perspective where
relevant, the analysis remains largely static, interpreting the
observed results of the drivers and modifiers of lowland use
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
at a given time. There is a need to further understand and
incorporate the dynamic elements of land use systems.
Policy and institutions can block or foster land use change,
but can also change and sometimes collapse. The recent civil
strife in Côte d’Ivoire is a case in point, significantly
upsetting market opportunities and overall law and order in
the study sites. West Africa has also seen significant changes
in rice policy environment (Pearson et al., 1981; Lançon and
Erenstein, 2002). Structural adjustment programs implied
the need to dismantle costly public support to the domestic
rice sector and have opened up regional markets to imported
rice. Increasing urbanization and changing food consumption patterns also implies changing market opportunities,
like the growing dairy demand of urban centers like Sikasso.
Finally, the rapidity of the urbanization in West Africa
(Kalasa, 1994) puts further strain on traditional land tenure
systems in the proximity of urban centers.
4.4. Development implications
In terms of a development agenda, a first lesson is that
lowland use in the forest zone of West Africa is not as
attractive as often assumed in agricultural development
spheres. Instead of a vast untapped agricultural development
potential (Andriesse and Fresco, 1994), there are various
underlying stringent constraints, including low productivity,
flooding risk and a widespread traditional preference for
upland cultivation. Significant levels of lowland use only
exist where there is significant pay-off to their cultivation.
The availability of shallow water during a prolonged dry
season is a case in point, explaining the widespread use of
lowlands in the savannah zone. Proceeding towards the drier
ecologies, lowland use not only becomes markedly more
widespread, land use intensity and its diversity also
significantly increases. It is thereby advisable to target
future lowland development efforts to the savannah zone
while ensuring equitable stakeholder participation and a
systems perspective, including market, crop-livestock and
upland-lowland inter-linkages.
For lowland development efforts to succeed there is a
need to make lowland use competitive vis-à-vis alternative
suppliers and attractive vis-à-vis alternative livelihood
strategies. Lowland use thereby stands to benefit from an
enabling environment for agricultural development in
general. A more intensive lowland use also implies the
need for targeted policies, such as facilitating the access to
appropriate technology (Erenstein, in press). This presents
particular challenges for lowland rice, as the extent of its
cultivation was not influenced by market proximity, underlining it often remains relatively extensive with considerable
self-consumption. There is a need to broaden the often
narrow association of lowlands with rice. The cultivation of
other crops in lowlands is significant, particularly outside the
forest zone. What is more, the close association of these
other crops with proximity to markets emphasizes their
market orientation and suggests they provide lucrative
215
livelihood diversification opportunities. The narrow focus on
physical lowland improvement in past development efforts
has proven both costly and inappropriate (Lançon and
Erenstein, 2002), and some of the issues were discussed
above. Lowland development investments did increase the
extent and intensity of lowland use, but by favoring the
cultivation of rice, had a negative impact on the diversity of
lowland use.
A key policy lesson is the importance of market access
for a more intensive use of lowlands. The models
unequivocally show that increasing travel time to the urban
market reduces the extent, intensity and diversity of
agricultural lowland use. Noteworthy is that this effect
was already observed over a relatively limited 30 km radius.
This provides a strong empirical basis to consider market
access in lowland intensification policies.
Scaling out is a final consideration because it is not
proximity to urban centers per se, but the associated market
access which is the key driver. Hence, the renewed interest in
rural infrastructure development is warranted. Enabling
market access to the rural masses beyond the urban
hinterland will be an important precondition to enable
agricultural intensification and poverty alleviation in West
Africa.
5. Conclusion
The study confirms the continued limited use of lowlands
in the forest zone of West Africa. The use of lowlands
significantly increases proceeding towards the savannah
zone, with increasing closeness to urban markets, with
lowland development and with migrant participation. An
important contribution of this study is in confirming
empirically that market opportunities are a major factor
driving lowland use. The study underscores the need to
qualify lowland use in West Africa, particularly in terms of
its extent, diversity and intensity. The study thereby
questions the often narrow association of lowlands with
rice and calls for equitable stakeholder participation and a
systems perspective in lowland development interventions.
Finally, the study shows the considerable scope for using
aggregate indicators to predict land use patterns to enhance
research and development targeting, and provided some
suggestions for their refinement.
Acknowledgements
The paper builds on work undertaken while the first two
authors were at the Africa Rice Center (WARDA). It draws
from a more elaborate study (Erenstein et al., 2004) within
the context of a BMZ-funded project implemented by
WARDA. The usual disclaimer applies. The paper benefited
from the assistance from several people, including WARDA
colleagues for the initial setup, fieldwork, research and
216
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
editorial support, IER Sikasso and ANADER Daloa for the
subsequent fieldwork and comments by reviewers.
References
Adam, M., 2001. Definition and boundaries of the peri-urban interface:
patterns in the patchwork. In: Drechsel, P., Kunze, D. (Eds.), Waste
Composting for Urban and Peri-urban Agriculture: Closing the Rural–
Urban Nutrient Cycle in Sub-Saharan Africa. CABI Publishing, Wallingford, UK, pp. 193–208.
Adams, W.M., 1993. Indigenous use of wetlands and sustainable development in West Africa. Geograph. J. 159, 209–218.
Ahmadi, N., Teme, B., 1998. Aménagement et mise en valeur des bas-fonds
au Mali: Bilan et perspectives nationales. Interêt pour la zone de savane
ouest-africaine. IER-CIRAD-CMDT-CBT, Montpellier, France.
Alla, D.A., Dynamisme de l’espace peri-urbain de Daloa – Etude Geographique. Thèse de doctorat 3ème cycle. Institut de géographie tropicale, Faculté des lettres, arts et sciences humaines, Université nationale
de Côte d’Ivoire, Abidjan, Côte d’Ivoire, 1991.
Andriesse, W., Fresco, L.O., 1991. A characterization of rice-growing
environments in West Africa. Agric. Ecosyst. Environ. 33, 377–395.
Andriesse, W., Fresco, L.O., van Duivenbooden, N., Windmeijer, P.N.,
1994. Multiscale characterization of inland valley agroecosystems in
West Africa. Netherlands J. Agric. Sci. 42, 159–179.
Bah, M., Cisse, S., Diyamett, B., Diallo, G., Lerise, F., Okali, D., Okpara, E.,
Olawoye, J., Tacoli, C., 2003. Changing rural–urban linkages in Mali,
Nigeria and Tanzania. Environ. Urbaniz. 15, 13–23.
Becker, L., Diallo, R., 1996. The cultural diffusion of rice cropping in Côted’Ivoire. Geograph. Rev. 86, 505–528.
Becker, M., Johnson, D.E., 1999. Rice yield and productivity gaps in
irrigated systems of the forest zone of Côte d’Ivoire. Field Crops
Res. 60, 201–208.
Becker, M., Johnson, D.E., 2001. Improved water control and crop management effects on lowland rice productivity in West Africa. Nutr. Cycl.
Agroecosyst. 59, 119–127.
Binswanger, H.P., McIntire, J., 1987. Behavioural and material determinants of product relations in land-abundant tropical agriculture. Ecol.
Dev. Cul. Ch. 36, 73–97.
Boserup, E., 1965. The Conditions of Agricultural Growth: The Economics
of Agrarian Change Under Population Pressure. Aldine, Chicago.
Bouju, J., 1998. Contrôle foncier et conflits pour les ressources: l’accès aux
bas-fonds amenagés comme enjeu de pouvoir local (Mali, Burkina
Faso). In: Ahmadi, N., Teme, B. (Eds.), Aménagement et mise en
valeur des bas-fonds au Mali. Bilan et perspectives nationales pour la
zone de savane ouest-africaine. Colloques. IER-CIRAD-CMDT-CBF,
Montpellier, France, pp. 95–108.
Bruinsma, W., 2001. Annotated Bibliography on Urban and Peri-Urban
Agriculture. ETC Urban Agriculture Programme, Leusden, The Netherlands.
Christiaensen, L., Demery, L., Paternostro, S., 2003. Reforms, remoteness
and risk in Africa: understanding inequality and poverty during the
1990s. Discussion Paper No. 2003/70. UN University World Institute for
Development Economics Research (WIDER), Helsinki. http://www.wider.unu.edu/publications/dps/dps2003/dp2003-070.pdf.
Club du Sahel, 2000. Urbanization, rural–urban linkages and policy implications for rural and agricultural development: case study from West
Africa. In: Understanding Rural–Urban Linkages and Rural Non-farm
Economies for Growth and Poverty Alleviation, Club du Sahel, Paris
http://www.oecd.org/dataoecd/40/21/2351008.pdf.
Coulibaly, C., Dakouo, C., Diarra, K., Fane, M., Kone, D., Lootvoet, B.,
Mesple-Somps, S., Traore, B., 1998. L’économie locale de Sikasso,
Mali. Club du Sahel, OECD, Paris.
De Groote, H., Djouara, H., Fomba, B., 1998. Les femmes et riziculture de
bas-fonds: les déterminants socio-économiques. In: Ahmadi, N., Teme,
B. (Eds.), Amenagement et mise en valeur des bas-fonds au Mali. Bilan
et perspectives nationales pour la zone de savane ouest-africaine.
Colloques. IER-CIRAD-CMDT-CBF, Montpellier, France, pp. 121–
132.
Diallo, D.C., 1998. Les enjeux socioculturels et économiques des bas-fonds
pour les femmes du Mali-Sud. In: Ahmadi, N., Teme, B. (Eds.), Aménagement et mise en valeur des bas-fonds au Mali. Bilan et perspectives nationales pour la zone de savane ouest-africaine. Colloques.
IER-CIRAD-CMDT-CBF, Montpellier, France, pp. 109–114.
Drechsel, P., Kunze, D., 2001. Waste Composting for Urban and Peri-Urban
Agriculture: Closing the Rural–Urban Nutrient Cycle in Sub-Saharan
Africa. CABI Publishing, Wallingford, UK.
Egziabher, A.G., Lee-Smith, D., Maxwell, D.G., Memon, P.A., Mougeot,
L.J.A., Sawio, C.J., 1994. Cities Feeding People: An Examination of
Urban Agriculture in East Africa. IDRC, Ottawa.
Ellis, F., Sumberg, J., 1998. Food production, urban areas and policy
responses. World Dev. 26, 213–225.
Erenstein, O., in press. Intensification or extensification? Factors affecting
technology use in peri-urban lowlands along an agro-ecological gradient
in West Africa. Agric. Syst.
Erenstein, O., Moussa, M., Oswald, A., Keijzer, P., 2004. Characterization
of peri-urban lowland use along an ecological and market access
gradient in West Africa. Project report—Factors influencing the development trajectories of lowlands in West Africa. Africa Rice Centre
(WARDA), Abidjan, Côte d’Ivoire.
Erenstein, O., Sumberg, J., Oswald, A., Levaisseur, V., Kore, H., 2006. What
future for integrated rice–vegetable production systems in West African
lowlands? Agric. Syst. 88 (2–3), 376–394.
FAO, 2002. Urban and Peri-Urban Agriculture. FAO, Rome.
Gobin, A., Campling, P., Feyen, J., 2002. Logistic modelling to derive
agricultural land use determinants: a case study from southeastern
Nigeria. Agric. Ecosyst. Environ. 89, 213–228.
Goldman, A., 1993. Agricultural innovation in three areas of Kenya: neoBoserupiantheoriesand regionalcharacterization. Econ. Geogr. 69, 44–71.
Henry, M.C., Rogier, C., Nzeyimana, I., Assi, S.B., Dossou-Yovo, J.,
Audibert, M., Mathonnat, J., Keundjian, A., Akodo, E., Teuscher,
T.M., Carnevale, P., 2003. Inland valley rice production systems and
malaria infection and disease in the savannah of Côte d’Ivoire. Trop.
Med. Int. Health 8, 449–458.
Hirose, S., Wakatsuki, T., 2002. Restoration of Inland Valley Ecosystems in
West Africa. Association of Agriculture and Forestry Statistics, Tokyo.
Humphreys, C.P., 1981. Rice production in the Ivory Coast. In: Pearson,
S.R., Stryker, J.D., Humphreys, C.P. (Eds.), Rice in West Africa. Policy
and Economics. Stanford University Press, Stanford, CA, USA, pp. 61–
105.
ICEF, ENSEA, 2002. L’économie locale du departement de Daloa. Club du
Sahel, OECD, Paris.
ICEF, ENSEA, IRD, 1999. L’économie locale de Korhogo et sa zone
d’influence. Club du Sahel, OECD, Paris.
Izac, A.M.N., Swift, M.J., Andriesse, W., A strategy for inland valley
agroecosystems research in West and Central Africa. RCMP Research
Monograph. IITA, Resource and Crop Management Program, Ibadan,
1991.
Kalasa, M.-B., 1994. Settlement patterns in West Africa. West Africa Long
TermPerspective Study. Workingpaper No.1.ClubduSahel,OECD, Paris.
Katzman, M., 1974. The Von Thünen paradigm, the industrial urban
hypothesis, and the spatial structure of agriculture. Am. J. Agric. Econ.
56, 683–696.
Kolawole, A., Scoones, I., Awogbade, M.O., Voh, J.P., 1994. Strategies for
the Sustainable Use of Fadama Lands in Northern Nigeria. CSER, IIED,
Zaria, Nigeria.
Kone, Y., Ahmadi, N., Mariko, D., 1998. La filiere de la pomme de terre au
Mali-Sud. In: Ahmadi, N., Teme, B. (Eds.), Aménagement et mise en
valeur des bas-fonds au Mali. Bilan et perspectives nationales pour la
zone de savane ouest-africaine. Colloques. IER-CIRAD-CMDT-CBF,
Montpellier, France, pp. 141–154.
Lançon, F., Erenstein, O., 2002. Potential and prospects for rice production
in West Africa. Paper presented at Sub-regional workshop on harmo-
O. Erenstein et al. / Agriculture, Ecosystems and Environment 117 (2006) 205–217
nization of policies and coordination of programmes on rice in the
ECOWAS sub-region, FAO-RAF, Accra, Ghana, February 25–28, 2002.
WARDA, Bouake.
Lançon, F., Erenstein, O., Toure, A., Akpokodje, G., 2004. Qualité et
compéitivité des riz locaux et importé sur les marché urbains ouestafricains. Cahiers Agric. 13, 110–115.
Lavigne-Delville, P., 1998. Logiques paysannes d’exploitation des basfonds en Afrique soudano-sahelienne: quelques reperes pour l’intervention. In: Ahmadi, N., Teme, B. (Eds.), Amenagement et mise en
valeur des bas-fonds au Mali. Bilan et perspectives nationales pour la
zone de savane ouest-africaine. Colloques. IER-CIRAD-CMDT-CBF,
Montpellier, France, pp. 77–93.
Lavigne-Delville, P., Boucher, L., 1998. Dynamiques paysannes de mise en
culture des bas-fonds en Afrique de l’Ouest forestière. In: CheneauLoquay, A., Leplaideur, A. (Eds.), Les rizicultures de l’Afrique de
l’Ouest. CIRAD, Montpellier, France, pp. 365–376.
Lele, U., Stone, S., 1989. Population pressure, the environment and
agricultural intensification: variations on the Boserup hypothesis.
MADIA Discussion Paper No. 4. The World Bank, Washington, D.C.
Manyong, V.M., Smith, J., Weber, G.K., Jagtap, S.S., Oyewole, B., Macrocharacterization of Agricultural Systems in West Africa: An overview.
Resource and Crop Management Monograph No. 21, IITA, Ibadan,
1996.
McMillan, D.E., Sanders, J.H., Koenig, D., Akwabi-Ameyaw, K., Painter,
T.M., 1998. New Land is not enough: agricultural performance of New
Lands settlement in West Africa. World Dev. 26, 187–211.
Moustier, P., Mbaye, A., de Bon, H., Guerin, H., Pages, J. (Eds.), 1998.
Agriculture periurbaine en Afrique subsaharienne – Actes de l’atelier
international 20–24 avril. CIRAD, CORAF, Montpellier, France.
Munroe, D.K., Southworth, J., Tucker, C.M., 2002. The dynamics of landcover change in western Honduras: exploring spatial and temporal
complexity. Agric. Econ. 27, 355–369.
Nelson, G.C., 2002. Introduction to the special issue on spatial analysis for
agricultural economists. Agric. Econ. 27, 197–200.
Pearson, S.R., Stryker, J.D., Humphreys, C.P., Rader, P.L., Monke, E.A.,
Spencer, D.S.C., Craven, K., Tuluy, A.H., McIntire, J., Page Jr., J.M.,
1981. Rice in West Africa. In: Policy and Economics, Stanford Universities Press, Stanford, CA, USA.
Pingali, P., Bigot, Y., Binswanger, H.P., 1987. Agricultural Mechanization
and the Evolution of Farming Systems in Sub-Saharan Africa. Johns
Hopkins University Press for the World Bank, Baltimore/London.
Randolph, T.F., Djayeola, M.B., Kamara, D.M., Gaye, M., 2000. A survey
of Irrigated Rice Schemes in Côte-d’Ivoire. WARDA, Bouake.
Raunet, M., 1985. Bas-fonds et riziculture en Afrique: approche structurale
comparative. L’agronomie tropicale 40, 181–201.
Raunet, M., Bas-fonds et riziculture. Actes du séminaire Antananarivo
Madagascar, 9–14 décembre 1991. CIRAD, Montpellier, France, 1993.
Richards, P., 1986. Coping with Hunger: Hazard and Experiment in an
African Rice Farming System. Allen and Unwin, London.
Ruthenberg, H., 1976. Farming Systems in the Tropics. Clarendon Press,
Oxford.
217
Saidu, M.M., 1994. Land and water rights in Fadama lands of northern
Nigeria. In: Kolawole, A., Scoones, I., Awogbade, M.O., Voh, J.P.
(Eds.), Strategies for the Sustainable Use of Fadama lands in northern
Nigeria. CSER, IIED, Zaria, Nigeria, pp. 91–96.
Sakurai, T., 2002. Land tenure systems and adoption of water control
technologies in lowland rice production. In: Sakurai, T., Furuya, J.,
Takagi, H. (Eds.), Economic analyses of agricultural technologies and
rural institutions in West Africa—achievements, challenges and application to rice farming research. Proceedings of JIRCAS international
workshop, Tsukuba, Japan, 12–13 July, 2001. JIRCAS Working Report.
Japan International Research Center for Agricultural Sciences, Tsukuba, pp. 107–125.
Shikano, K., 2002. Ecological anthropological study on daily herding
activities of pastoral Fulbe in Central Nigeria. In: Hirose, S., Wakatsuki, T. (Eds.), Restoration of Inland Valley Ecosystems in West
Africa. Association of Agriculture and Forestry Statistics, Tokyo, pp.
303–367.
Smit, J., Ratta, A., Nasr, J., 1996. Urban Agriculture: Food, Jobs and
Sustainable Cities. UNDP, Habitat II, New York.
Smith, J., Socioeconomic Characterization of Environments and Technologies in Humid and Sub-humid Regions of West and Central Africa.
Resource and Crop Management Monograph No. 10. IITA, Ibadan,
1992.
Snrech, S., Les Transformations De L’Agriculture Ouest-Africaine: Evolutions 1960–1990. Defis Pour L’Avenir. Implications Pour Les Pays
Saheliens. Club du Sahel, Paris, 1995.
Spencer, D.S.C., Byerlee, D., 1976. Technical change, labor use, and small
farmer development: evidence from Sierra Leone. Am. J. Agric. Econ.
58, 874–880.
Staal, S.J., Baltenweck, I., Waithaka, M.M., DeWolff, T., Njoroge, L., 2002.
Location and uptake: integrated household and GIS analysis of technology adoption and land use, with application to smallholder dairy
farms in Kenya. Agric. Econ. 27, 295–315.
Tachibana, T., Shinagawa, M., Sakurai, T., 2002. Lowland rice cultivation in
Ghana: a field survey around an inland market, Kumasi. In: Sakurai, T.,
Furuya, J., Takagi, H. (Eds.), Economic Analyses of Agricultural
Technologies and Rural Institutions in West Africa—Achievements,
Challenges and Application to Rice Farming Research. Proceedings of
JIRCAS international workshop, Tsukuba, Japan, July 12–13, 2001.
JIRCAS Working Report 25. Japan International Research Center for
Agricultural Sciences, Tsukuba, pp. 127–140.
Terrin, F., Etude agro-socio-economique des bas-fonds peri-urbains de
Bouaké (Côte-d’Ivoire). Memoire de fin d’études. L’institut supérieur
d’agro-developpement international, Cergy Pontoise, France, 1995.
Tiffen, M., 2003. Transition in Sub-Saharan Africa: agriculture, urbanization and income growth. World Dev. 31, 1343–1366.
Turner, B.L., Kates, R.W., Hyden, G., 1993. Population Growth and
Agricultural Change in Africa. University of Florida Press, Gainesville.
Windmeijer, P.N., Andriesse, W., 1993. Inland Valleys in West Africa: An
Agro-Ecological Characterization of Rice-Growing Environments,
ILRI Publication 52. ILRI, Wageningen.