- Email: [email protected]
land cover and prediction of future extension of bowé in Benin, West Africa
Land Use Policy 69 (2017) 85–92
Contents lists available at ScienceDirect
Land Use Policy journal homepage: www.elsevier.com/locate/landusepol
Mapping changes in land use/land cover and prediction of future extension of bowé in Benin, West Africa
MARK
⁎
Elie A. Padonoua,b, Anne M. Lykkec, Yvonne Bachmannd, Rodrigue Idohoue, , Brice Sinsinb a
School of Forestry and Wood Industry, National University of Agriculture, Kétou, Benin Laboratory of Applied Ecology, Faculty of Agronomic Sciences, University of Abomey-Calavi, 01 BP: 526, Cotonou, Benin c Department of Bioscience, Aarhus University, Vejlsøvej 25, DK-8600 Silkeborg, Denmark d Institute of Ecology, Evolution and Diversity, J.W. Goethe University, Max-von-Laue-Str. 13, 60438, Frankfurt am Main, Germany e Laboratoire de Biomathématiques et d’Estimations Forestières, Faculty of Agronomic Sciences, University of Abomey-Calavi, 04 BP: 1525, Cotonou, Benin b
A R T I C L E I N F O
A B S T R A C T
Keywords: Land use/land cover change Natural vegetation Bowé Ferricretes West Africa
Desertification and land degradation are worldwide problems affecting soil, vegetation and the livelihoods of rural populations. Bowal (plural bowé) is a particular form of degraded land that occurs in tropical regions and leads to the exposure of ferricretes, which are unsuitable for farming. Bowé are more common on farmland and degraded savanna. Changes in land use/land cover were used to map a region of 6.7 million ha in northern Benin, West Africa in 1975, 1990 and 2010. The changes observed during these periods (1975–1990, 1990–2010 and 1975–2010) were used to predict the occurrence of bowé in the period up to 2050 using Markovian chain analysis. The results showed a considerable change in land use/land cover during the three periods. The types of land on which bowé occur (farmland and degraded savanna) increased in northern Benin by 5.4% per year during the period 1975–1990 and 9.5% per year during the periods 1990–2010, while the natural vegetation (forest, woodland and tree savanna) decreased by the same amount. The future scenarios also predicted the same trend. In the period 1975–1990, 1.28 million ha (26%) of natural vegetation was converted to degraded savanna and farmland while 2.23 million ha (53%) of natural vegetation was converted to degraded savanna and farmland in the period 1990–2010. Based on the dynamics recorded during the period 1975–1990 and 1990–2010 respectively, a total of 1.28 million ha (26% of the natural vegetation that was present in 1975) and 1.29 million ha (31% of the natural vegetation that was present in 1990) will be converted to farmland and degraded savanna in the study area by 2050.Thus bowalization will persist and increase in the period up to 2050. The natural vegetation could disappear if protection and restoration measures are not taken. It is thus important to take measures to stop the degradation and to implement programs to restore soils on bowé based on the soil and water conservation techniques used on highly degraded West African soils, such as zaï pit and stone rows with grass strips. Some native plants species adapted to bowalization and resistant to climate change in northern Benin (e.g. Asparagus africanus, Andropogon pseudapricus and Combretum nigricans) should be used in association with soil and water conservation techniques on bowé.
1. Introduction Desertification and land degradation are worldwide problems affecting soils, vegetation and the livelihoods of rural populations (D’Odorico et al., 2013; Gao and Liu, 2010). Desertification and land degradation lead to increasing levels of poverty, starvation, land abandonment and migration out of the affected regions (Verstraete et al., 2009). Combating desertification and land degradation is crucial for reducing global poverty, biodiversity loss and human-induced global climate change (MEA, 2005).
The adoption of control measures for desertification and land degradation requires the identification and monitoring of early warning signs. Among the commonly used biophysical indicators are changes in land use/land cover (Vogt et al., 2011). Changes in land use/land cover are affected by the ways in which the biophysical attributes of the land are manipulated and by the intentions underlying these manipulations (Qingqing et al., 2012). By 2100, the impact of changes in land use/ land cover on biodiversity is likely to be more significant than the impact of global climate change, nitrogen deposition, species introductions and changing atmospheric concentrations of carbon dioxide
⁎
Corresponding author. E-mail addresses: [email protected] (E.A. Padonou), [email protected] (A.M. Lykke), [email protected] (Y. Bachmann), [email protected] (R. Idohou), [email protected] (B. Sinsin). http://dx.doi.org/10.1016/j.landusepol.2017.09.015 Received 29 November 2015; Received in revised form 6 September 2017; Accepted 9 September 2017 0264-8377/ © 2017 Elsevier Ltd. All rights reserved.
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
proposed for the Tropical Ecosystem Environment Observations by Satellites Project (TREES) phase II (Achard et al., 2002). The three land cover maps were digitized and then the different land cover types were delineated using ENVI 4.1 software. Color composition (RGB 4, 7, 5) was used to improve the differentiation of the land cover types on the screen. Bands RGB 4 (0.750–0.900 μm), 7 (2.090–2.350 μm) and 5 (1.550–1.750 μm) made it possible to emphasize the differences between the stages of succession in the forested areas in addition to the savanna and farmland areas. The interpretation was aided by three additional data sources: 1:200,000 land cover maps edited in 1975 and 1987 by CENATEL Benin (the National Center of Remote Sensing) and the 2007 National Forest Inventory database, from which the land cover classes of forest, woodland, tree savanna, degraded savanna and farmland were adopted. Bowé occurred most often on farmland and degraded savanna (Fig. 3); by contrast, there was no bowal in either forest or woodland. The polygons of the land cover classes in the sub-scene acquired in 1975 were labeled according to their cover class as shown in the land cover maps from CENATEL. Once the digital map of the coverage had been finalized, the polygons were copied and overlaid with the 1990 image. The segments that required modification were changed by adding, deleting or modifying the polygon boundaries to reflect the changes in land cover between 1975 and 1990. The same process was used to visualize the changes between 1990 and 2010. This methodology was used because it avoids the generation of false changes that may occur if the databases contain a spatial mismatch (Mas et al., 2004). Areas without any recognizable vegetation in the images (water bodies, clouds and infrastructure) were excluded to reduce bias. The classification accuracy was assessed by an error matrix (Congalton, 1991). Three hundred reference points (50–60 per class) were used and these were distributed across the study area using a stratified sampling scheme (Achard et al., 2002). The error matrix was normalized with an iterative proportional fitting procedure that forced each row and column to sum to unity using the program MARGFIT (Congalton, 1991). A kappa analysis was then performed with the Kappa program to determine the accuracy of the classification (Congalton and Green, 1999).
(Sala et al., 2000). The analysis of the dynamics of changes in land use land/land cover is therefore a fundamental tool in the adoption of strategies for the conservation of biodiversity (Verburg et al., 2009) and in planning appropriate management techniques for degraded land. The analysis of changes in land use/land cover has become a fundamental tool in assessing the environmental consequences of human activity (Kanianska et al., 2014; Verburg et al., 2011). Changes in land use/land cover have consequences for biodiversity (Brink et al., 2014), geochemical cycles (Powers, 2004) and water quality (Schippers et al., 2004). The dynamics of changes in land use/land cover are influenced by the type of land cover, ecological mechanisms of succession and regeneration, the physical components of the environment, socioeconomic activities and their cultural context, meteorological phenomena and natural disasters (Yu and Lu, 2011). Bowal (plural bowé) is a particular type of degraded land that forms on hardened ferruginous soils (ferricretes) in tropical regions with a unimodal pattern of precipitation. It is a very distinctive landscape. The name originates from the fulfulde language in Guinea (Aubréville, 1947). Bowalization occurs when ferricretes are exposed as a result of the erosion of the soil surface by a combination of a dry climate and deforestation (Padonou et al., 2014, 2015b). Bowé retain very little water, so any vegetation quickly desiccates and burns early in the dry season. Root growth is impeded and the trees are dwarfed, gnarled and widely scattered. As the soils rapidly absorb and re-radiate solar energy, bowé become extremely hot and barren in the dry season. The land cover on bowé is mainly grassland and savanna (Padonou et al., 2013; Zwarg et al., 2012). Thus, it may be possible to predict the occurrence of bowé in tropical regions with a unimodal pattern of precipitation and hardened ferruginous soils by analyzing changes in land use/land cover. In this study, we considered land cover to be the biophysical state of the Earth’s land surface and immediate subsurface, including the biota, soil, topography and groundwater. We analyzed changes in land use/ land cover using a previously published set of categories (Lambin et al., 2003). Changes in land cover include changes in the biodiversity, the actual and potential primary productivity and soil quality. 2. Materials and methods
2.3. Changes in land use and land cover 2.1. Study area The elaborated maps were incorporated into a geographical information system using Arcmap 10 (ESRI) software. An overlay analysis was performed to assess the pathways of change observed over the three periods. A mask was generated to eliminate areas without vegetation (water bodies and infrastructure) to allow a comparative analysis of the same area at different times (Hall et al., 1995). After subtraction of the masked areas, the resulting area was 6,735,489 ha. Categories of change were classified into three stages: cover loss; cover regeneration; and unchanged cover. Cover loss refers to land cover that was subjected to a change with a concomitant loss in a regressive manner (for example, from forest to savanna or from woodland to farmland). Pathways of change that flowed in the opposite direction indicated regeneration.
Data were collected in the municipality of Banikoara (11° 18′ N and 2° 25′ E) in northern Benin (Fig. 1). This area is considered to be the breadbasket of the country (Kokoye et al., 2013) and agriculture plays a major part in the livelihoods of local people. Decisions about the allocation of land among farms in this region are made by farmers on socioeconomic and demographic grounds, taking into account production factors, such as labor and capital (Kokoye et al., 2013). A household’s capital is an important influence on decision-making about the amount of land allocated to cereals, legumes and cash crops. The allocation of land to cash crops is also determined by the farmers’ access to credit. The natural vegetation of this region is characterized by a mosaic of woodland, dry forest, tree and shrub savanna and gallery forest (Adomou et al., 2006). This zone was selected for study because it is dominated by the ferruginous soils (Fig. 1) on which bowé occur (Padonou et al., 2014, 2015a). Bowé could occur anywhere within the study area.
2.4. Land use/land cover transition probability matrices Transition probability matrices were elaborated for the periods 1975–1990 (15.5 years), 1990–2010 (20.5 years) and 1975–2010 (35.5 years) to describe the changes in each land cover class. Each matrix represents either the probability of the persistence of each land cover category, or the probability of a transition to another land cover category from the first to the last year in the period. The matrix values were standardized using the procedure proposed by Rovainen (1996) to obtain annualized changes and to make comparisons. To annualize the matrix values, each probability matrix was used separately to compute the matrix’s eigenvectors and eigenvalues using the diagonalization
2.2. Land cover maps Three sub-scenes of Landsat imagery for the years 1975 (Landsat Multi-Spectral Scanner), 1990 (Landsat Thematic Mapper) and 2010 (Landsat Enhanced Thematic Mapper plus) were interpreted (Fig. 2). The images were all taken during the dry season to minimize any variation in phenology of the vegetation (Clerici et al., 2007). An onscreen visual interpretation was carried out by a method similar to that 86
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 1. Soil map of the study area (Volkoff and Willaime, 1976).
method (Çinlar, 1975).
chain model. Three different scenarios were assumed corresponding to each of the three possible Markovian matrices (1975–1990, 1990–2010 and 1975–2010). The validation of the model was assessed using a χ2 test at two levels to verify the accuracy. First, the area expected from the 2010 scenarios based on the 1975–1990 period was compared with the area
2.5. Future scenarios The annualized transition matrices were used to predict the proportion of each land cover class at any one time based on a Markovian
Fig. 2. Land cover maps of the study area for 1975, 1990 and 2010.
87
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 3. Frequency of vegetation types that turned into bowé based on changes in land use/land cover classes from 1975 to 2010 and fieldwork in 2011.
Most of the forest cover was lost from 1975 to 1990 (10% of the total area decrease), while the loss of woodland and tree savanna mainly occurred from 1990 to 2010 (7 and 14% of the total area decrease, respectively). The amount of degraded savanna increased from 1990 to 2010, while farmland areas increased by 13% of the total area from 1975 to 1990 and by the same amount from 1990 to 2010.
for 2010. Second, the area expected from the 1990 scenarios based on the total period 1975–2010 was compared with the area for 1990. 2.6. Rate of degradation The rate of degradation for each of the three periods was calculated based on the cover data of the land cover types, forest, woodland and tree savanna, using the formula proposed by the Food and Agriculture Organization (FAO 1996). We considered these three land cover classes because they converted to the degraded savanna and farmland on which bowé were more frequent (Fig. 3). The rate of degradation, D (% lost area/year), was calculated as:
D = ⎡1 − (1 − ( ⎣
A1 − A2 1/ t )) ⎤ A1 ⎦
3.2. Changes in land use The transition matrices for the time periods 1974–1990, 1990–2010 and 1975–2010 (Table 2) were used to obtain the probability of change (Table 3). During the period 1975–1990, farmland and tree savanna showed large persistence (98 and 97% per year, respectively). The persistence of woodland and degraded savanna was almost as large at 96% while the persistence of forest was 94%. Most of the transformed forest converted to woodland (3.6%) and tree savanna (1.5%). The woodland was mainly converted to degraded savanna (1.3%). The tree savanna class was transformed to woodland (1%) and degraded savanna (1%), farmland (0.7%) and forest (0.5%). The transformed degraded savanna was mainly converted to farmland (2%). The farmland was mainly transformed to tree savanna (0.9%) and degraded savanna (0.9%). During the period 1990–2010, farmland had a large annual persistence (98%) and the annual persistence of woodland and degraded savanna was almost as large (97%). The annual persistence of forest was 96%. Forest cover was transformed in degraded savanna (1.6%), woodland (0.9%), farmland (0.8%) and tree savanna (0.5%). Woodland was mainly transformed in degraded savanna (2.4%) and farmland (0.6%). Tree savanna was converted to farmland (1.8%), woodland (0.9%), degraded savanna (0.6%) and forest (0.1%). Degraded savanna was mainly converted to farmland (2.2%) and woodland (0.7%). Farmland was converted to degraded savanna (1.1%), woodland (0.7%) and tree savanna (0.1%). In the period 1975–2010, a large annual persistence was obtained for farmland (100%) and the annual persistence of the other land cover types (woodland, tree savanna and degraded savanna) was almost as large at 98%. Forest had an annual persistence of 96.3%. The forest cover was mainly converted to woodland (2.0%) and degraded savanna (0.9%). The woodland cover was mainly converted to degraded savanna (1.4%) and farmland (0.7%). The transformed tree savanna and degraded savanna were mainly converted to farmland (1.1 and 1.6%, respectively).
× 100
where A1 and A2 are the initial and final cover (forest, woodland or tree savanna), respectively, and t is the interval in years for which the change in land cover is being studied. 3. Results 3.1. Land cover maps The proportions of the five land cover types changed considerably from 1975 to 2010 (Table 1; Fig. 2). In 1975, woodland and tree savanna were the dominant land cover types, whereas farmland covered only 4% of the total area (Table 1). In 1990, woodland and tree savanna were still the dominant land cover types. However, the amount of forest cover had decreased, while the amount of farmland had increased. In 2010, farmland and degraded savanna were the dominant land cover types and the other types of land cover (forest, woodland and tree savanna) had decreased. The cover of forest, woodland and tree savanna decreased in the periods 1975–1990 and 1990–2010, while the cover of degraded savanna and farmland increased during the same periods. Table 1 Land use/land cover classes used in the analysis of change (area in ha). Land cover class
1975
(%)
1990
(%)
2010
(%)
Forest Woodland Tree savanna Degraded savanna Farmland Total
1,215,566 1,949,195 1,782,725 1,499,961 288,042 6,735,489
18 30 26 22 4 100
569,563 1,943,146 1,660,493 1,394,452 1,167,835 6,735,489
8 29 25 21 17 100
243,314 1,464,790 741,123 2,078,519 2,207,743 6,735,489
4 22 11 30 33 100
3.3. Future scenarios The difference between the observed probabilities in the 2010 land 88
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Table 2 Land use/land cover transition matrix (area in ha) for the observed time period (1975–1990–2010). Forest
Woodland
Tree savanna
Degraded savanna
Farmland
Total
1975 Forest Woodland Tree savanna Degraded savanna Farmland
1990 125,702 244,591 128,431 65,779 5060
648,707 843,007 283,700 164,203 3529
270,657 222,919 915,089 212,088 39,741
128,530 369,870 258,082 599,081 38,889
41,970 268,808 197,424 458,810 200,823
1,215,566 1,949,195 1,782,726 1,499,961 288,042
1990 Forest Woodland Tree savanna Degraded savanna Farmland
2010 132,763 35,722 39,035 28,724 7071
102,966 693,587 305,901 190,983 171,354
62,378 66,254 507,454 76,353 28,684
181,113 933,392 214,672 496,409 252,934
90,385 214,208 593,405 602,014 707,731
569,605 1,943,163 1,660,467 1,394,483 1,167,774
1975 Forest Woodland Tree savanna Degraded savanna Farmland
2010 106,955 96,565 31,279 970 567
667,582 430,414 186,443 141,933 28,265
103,435 176,620 417,945 22,115 11,275
300,655 1,369,485 174,235 254,845 1137
78,241 720,195 541,630 581,626 291,079
1,256,868 2,793,279 1,351,532 1,001,489 332,323
land covered by forest, woodland and tree savanna decreases, whereas the amount of land covered by farmland increases. Assuming the dynamics recorded in the first period (1975–1990), the amount of forest will decrease by 12% of the area recorded in 1975 by 2050. The amount of woodland will decrease by 10%, tree savanna by 3% and degraded savanna by 2% of the areas recorded in 1975 by 2050. However, the area of farmland will increase by 28%. Assuming the dynamics recorded in the second period (1990–2010), the amounts of land covered by forest, woodland and tree savanna will decrease by 5, 9 and 17% of their cover in 1990, respectively, while the amount of degraded
cover map and the simulated probabilities in 2010 based on the period 1975–1990 was not significant (χ2 < 13.05; p > 0.05). The difference between the observed probabilities in the 1990 land cover map and the simulated probabilities in 1990 using the period 1975–2010 was not significant (χ2 < 11.85; p > 0.05). Thus the model can be applied to simulate scenarios of change in land use and land cover. Future land cover will change depending on the dynamics considered (1975–1990, 1990–2010 and 1975–2010). The historical (1975–2010) and estimated future (2010–2050) changes in each land cover class are shown in Fig. 4. In all three time periods, the amount of
Table 3 Annual probability matrices. Annual probability matrix 1975–1990 Forest 1975
1990
Forest Woodland Tree savanna savanna Degraded savanna Farmland
0.940 0.008 0.005 0.003 0.001
Woodland
Tree savanna
Degraded savanna
Farmland
0.036 0.962 0.011 0.007 0.001
0.015 0.008 0.967 0.009 0.009
0.007 0.013 0.010 0.960 0.009
0.002 0.009 0.007 0.020 0.980
Annual probability matrix 1990–2010 Forest
Woodland
Tree savanna
Degraded savanna
Farmland
1990
2010
Forest Woodland Tree savanna Degraded savanna Farmland
0.962 0.001 0.001 0.001 0.000
0.009 0.968 0.009 0.007 0.007
0.005 0.002 0.965 0.003 0.001
0.016 0.024 0.006 0.968 0.011
0.008 0.006 0.018 0.022 0.980
Forest
Woodland
Tree savanna
Degraded savanna
Farmland
0.020 0.976 0.004 0.004 0.003
0.003 0.002 0.980 0.001 0.001
0.009 0.014 0.004 0.978 0.000
0.004 0.007 0.011 0.016 0.996
Annual probability matrix 1975–2010
1975
2010
Forest Woodland Tree savanna Degraded savanna Farmland
0.963 0.001 0.001 0.000 0.000
89
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 4. Simulation of the evolution of the five land cover classes under three future scenarios. Dynamics observed during the periods (a) 1975–1990; (b) 1990–2010; and (c) the total study period (1975–2010).
main land cover classes where bowé are common (Fig. 3). During the three time periods, the highest rate of degradation was for forest cover, for which 1,089,864 ha were lost during the first period (1975–1990) and 436,841 ha in the second period (1990–2010). This is about 72,658 ha/year (4.93% of the study area) during the first period and 21,842 ha/year (4.16% of the study area) during the second period. The rates of degradation of woodland and tree savanna increased in the second period. The loss of woodland cover in the first period was 861,597 ha (57,440 ha/year), while 1,213,853 ha (60,693 ha/year) were lost in the second period. For the tree savanna, 455,506 ha (30,367 ha/year) were lost during the first period, while 808,077 ha (40,404 ha/year) were lost during the second period. During the first period, 1,264,684 ha (26%) of the three land cover types (forest, woodland and tree savanna) were converted to degraded savanna and farmland, while in the second period 2,227,174 ha (53%) were transformed to degraded savanna and farmland.
savanna and farmland will increase by 9 and 22% of the total area, respectively. Forest, woodland, tree savanna and degraded savanna will decrease by 14, 11, 18 and 2%, respectively, while the area of farmland will increase by 46% by 2050 when considering the dynamics recorded for the whole study period (1975–2010). 3.4. Rates of degradation The rate of degradation of the natural vegetation (forest, woodland and tree savanna) is presented in Table 4. We considered only these classes because they convert to degraded savanna and farmland, the Table 4 Annual rate of degradation (% lost ha/year). Land cover class
1975–1990
1990–2010
1975–2010
Forest Woodland Tree savanna Total
4.928 0.021 0.472 5.422
4.163 1.403 3.953 9.520
4.492 0.813 2.477 7.782
4. Discussion The conversion of forest, woodland and tree savanna into degraded 90
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
example, linking the universal soil loss equation to a geographical information system (El Garouani et al., 2007) may improve our understanding of soil erosion in areas where bowé occur. The limits observed with the method considered in this study should be taken into account in similar studies. However, the trends obtained in this work are useful for understanding the pattern of bowalization in the study area.
savanna and farmland was the most common change observed in the study area. The study area is considered to be the main area of food and cash crop production (mainly cotton) in Benin. Cotton contributes 64% of export income, 90% of agricultural income and 24% of state income in the study area (Kokoye et al., 2013). The natural vegetation is therefore intensely exploited as farmland. The annual rate of change of land cover to farmland was stable during both time periods (1975–1990 and 1990–2010) studied here, while the amount of degraded savanna increased in the second time period (1990–2010). Agriculture in this region has been dominated over the past three decades by cotton production (Ton, 2004). When farmland becomes depleted and degraded, losing its fertility, it is usually abandoned and new areas of natural vegetation (forest, woodland or tree savanna) are converted to farmland. The former farmland is converted to degraded savanna. We found a transition from farmland to degraded savanna during these time periods, confirming the increase in degraded savanna in the second time period resulting from farmland abandoned in the first time period. In both time periods, farmland showed a high persistence, varying from 98 to 100% while natural vegetation land cover class was degraded in the area. This is in accordance with previous findings of a conversion of land cover from forest to farmland with a loss of biodiversity in this region (Houessou et al., 2013). The persistence of farmland is mainly a result of cotton production. The production of this crop is encouraged by the government, which provides credit in the form of seeds, fertilizers, pesticides and tractors to farmers in this region. This credit is the main determinant in the allocation of land to cotton production (Kokoye et al., 2013). Thus farmland persists and increases each year at the expense of forest, woodland and tree savanna. The area of natural vegetation (forest, woodland and tree savanna) that was considered degraded increased by 4.1% between the first and the second time periods (i.e. from 5.4% over the period 1975–1990 to 9.5% over the period 1990–2010). The land cover was converted to farmland and degraded savanna. Although not quantified or measured in the field, the increase in farmland likely resulted in the occurrence of bowé in the study area (Padanou et al., 2014). Among the important factors leading to the occurrence of bowé in the study area were the increase in farmland and the practice of animal-drawn tillage. The amount of land that could be affected by bowalization thus increased during the two time periods considered here. Future scenarios developed using the Markovian model with annualized transition matrices (1975–1990, 1990–2010 and 1975–2010) showed a decrease in forest, woodland and tree savanna and an increase in farmland and degraded savanna by 2050. Based on this model, the total cover of forest, woodland and tree savanna will decrease from 74% (in 1975) to 48% by 2050 and from 62% in 1990–31% in 2050 using the dynamics recorded for 1975–1990 and 1990–2010, respectively. This means that by 2050, 26% of the total area recorded as forest, woodland and tree savanna in 1975 will have been converted to farmland and degraded savanna if we assume the dynamics recorded from 1975 to 1990. Considering the dynamics of 1990–2010, 31% of these land cover types recorded in 1990 will have been converted to farmland and degraded savanna. As bowé are more common in farmland and degraded savanna (Fig. 3), bowalization will therefore increase in the future. However, because bowé occur after the erosion of topsoil, taking information on other variables that drive soil erosion into account in more sophisticated models will improve this forecast and enable researchers to outline normative scenarios to assess bowalization. Markovian models have been used in many analyses of changes in land use/land cover (Flamenco-Sandoval et al., 2007). They are useful for exploratory analysis and for depicting contrasting scenarios. The models are not spatially explicit and assume that the transition probabilities are homogenous over time. More detailed spatially explicit models on changes in land use/land cover may be used in future analysis to improve our understanding of the locations and pathways of land use/land cover change dynamics that induce bowalization. For
5. Conclusions Agriculture is the dominant form of land management in northern Benin. Over the period 1975–1990, 5.4% of the natural vegetation was converted to farmland and degraded savanna while over the period 1990–2010 this figure was higher at 9.5%. Thus the land was degraded and bowé occurred. Extrapolation of all the dynamics considered to 2050 showed an increasing rate of degradation (mainly to farmland) and a decreasing amount of natural vegetation. About 26 and 31% of the total area of natural vegetation (forest, woodland and tree savanna) will be converted to farmland and degraded savanna by 2050 if we assume the dynamics recorded from 1975 to 1990 and 1990–2010, respectively. Bowalization will increase in the future and limit the amount of land available for farmland. The natural vegetation could disappear entirely if protection and restoration measures are not taken. Protection measures are necessary to preserve natural vegetation, but it is also important to implement programs to restore and conserve soils on bowé. These programs should include the adoption of appropriate soil and water conservation techniques for highly degraded West African soils, such as zaï pits and stone rows with grass strips. These soil and water conservation techniques have been used for decades to reclaim marginal and degraded barren lands in the arid and semi-arid zones of West Africa (Sidibe, 2005; Fatondji et al., 2009; Bayala et al., 2011). Some native plant species are adapted to bowalization and are resistant to climate change in northern Benin (e.g. Asparagus africanus, Andropogon pseudapricus and Combretum nigricans). These species could be used on bowé in association with soil and water conservation techniques. Acknowledgments This work was supported by UNDESERT (EU FP7 243906), “Understanding and combating desertification to mitigate its impacts on ecosystem services” funded by the European Commission, Directorate General for Research and Innovation, Environment Programme. Additional funding was obtained from the Robert S. McNamara Fellowships Program and the International Foundation for Science. References Çinlar, E., 1975. Exceptional paperÇMarkov renewal theory: A survey. Manage. Sci. 21 (7), 727–752. Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., Richards, T., Malingreau, J.P., 2002. Determination of deforestation rates of the world’s humid tropical forests. Science 297, 999–1002. Adomou, A., Sinsin, B., van der Maesen, L., 2006. Phytosociological and chorological approaches to phytogeography: a meso-scale study in Benin. Syst. Geogr. Plants 6, 155–178. Aubréville, A., 1947. Les brousses secondaires en Afrique équatoriale. Bois et Forêts des Tropiques 2, 24–35. Bayala, J., Kalinganire, A., Tchoundjeu, Z., Sinclair, F., Garrity, D., 2011. Conservation Agriculture with Trees in the West African Sahel–a Review. ICRAF (Occasional paper, 14). Brink, A.B., Bodart, C., Brodsky, L., Defourney, P., Ernst, C., Donney, F., Lupi, A., Tuckova, K., 2014. Anthropogenic pressure in East Africa—monitoring 20 years of land cover changes by means of medium resolution satellite data. Int. J. Appl. Earth Obs. Geoinf. 28, 60–69. Clerici, N., Bodini, A., Eva, H., Grégoire, J.-M., Dulieu, D., Paolini, C., 2007. Increased isolation of two Biosphere Reserves and surrounding protected areas (WAP ecological complex, West Africa). J. Nat. Conserv. 15, 26–40. Congalton, R.G., Green, K., 1999. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices. CRC Press, Boca Raton. Congalton, R.G., 1991. A review of assessing the accuracy of classifications of remotely
91
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
distribution of bowal and differences in physicochemical characteristics between bowal and woodland soils in Benin, West Africa. Catena 124, 45–52. Padonou, E.A., Lykke, A.M., Bachmann, Y., Sinsin, B., 2015b. Bowalization: its impact on soil, biodiversity, and human livelihoods in West Africa. Adv. Environ. Res. 40, 187–198. Powers, J.S., 2004. Changes in soil carbon and nitrogen after contrasting land-use transitions in northeastern Costa Rica. Ecosystems 7, 134–146. Qingqing, Z., Hailiang, X., Jingyi, F., Pujia, Y., Peng, Z., 2012. Spatial analysis of land use and land cover changes in recent 30 years in manas river basin. Procedia Environ. Sci. 12, 906–916 (Part B). Rovainen, E., 1996. Analysis and standardization of transition matrices. In: In: FAO (Ed.), Forest Resources Assessment 1990. Survey of Tropical Forest Cover and Study of Change Processes 130. FAO, pp. 111–114 Forestry Paper. Sala, O.E., Stuart Chapin Jr., F., Armesto, J.J., Berlow, E., Bloomfield, J., Dirzo, R., HuberSanwald, E., Huenneke, L.F., Jackson, R.B., Kinzig, A., Leemans, R., Lodge, D.M., Mooney, H.A., Oesterheld, M.N., Poff, N.L., Sykes, M.T., Walker, B.H., Walker, M., Wall, D.H., 2000. Global biodiversity scenarios for the year 2100. Science 287, 1770–1774. Schippers, P., Vermaat, J.E., de Klein, J., Mooij, W.M., 2004. The effect of atmospheric carbon dioxide elevation on plant growth in freshwater ecosystems. Ecosystems 7, 63–74. Sidibe, A., 2005. Farm-level adoption of soils and water conservation techniques in northern Burkina Faso. Agric. Water Manage. 71, 211–224. Ton, P., 2004. La production du coton au Bénin. FUPRO-Bénin, Pays-Bas et Bénin. Verburg, P.H., van de Steeg, J., Veldkamp, A., Willemen, L., 2009. From land cover change to land function dynamics: a major challenge to improve land characterization. J. Environ. Manage. 90, 1327–1335. Verburg, P.H., Neumann, K., Nol, L., 2011. Challenges in using land use and land cover data for global change studies. Global Change Biol. 17, 974–989. Verstraete, M.M., Scholes, R.J., Smith, M.S., 2009. Climate and desertification: looking at an old problem through new lenses. Front. Ecol. Environ. 7, 421–428. Vogt, J.V., Safriel, U., Von Maltitz, G., Sokona, Y., Zougmore, R., Bastin, G., Hill, J., 2011. Monitoring and assessment of land degradation and desertification: towards new conceptual and integrated approaches. Land Degrad. Dev. 22, 150–165. Yu, L., Lu, J., 2011. Does landscape fragmentation influence sex ratio of dioecious plants? A case study of Pistacia chinensis in the Thousand-Island Lake region of China. PLoS One 6, e22903. Zwarg, A., Schmidt, M., Janßen, T., Hahn, K., Zizka, G., 2012. Plant diversity, functional traits and soil conditions of grass savannas on lateritic crusts (bowé) in south eastern Burkina Faso. Flora et Vegetatio Sudano-Sambesica 15, 15–24.
sensed data. Remote Sens. Environ. 37, 35–46. D’Odorico, P., Bhattachan, A., Davis, K.F., Ravi, S., Runyan, C.W., 2013. Global desertification: drivers and feedbacks. Adv. Water Resour. 51, 326–344. El Garouani, A., Chen, H., Lewis, L., Tribak, A., Abahrour, M., 2007. Apport de la Télédétection et du SIG pour le suivi spatio-temporel de l’occupation du sol et de l’érosion nette dans le bassin de l’Oued Tlata (Maroc), Acte des JSIRAUF, Hanoi. Fatondji, D., Martius, C., Zougmore, R., Vlek, P.L.G., Bielders, C.L., Koala, S., 2009. Decomposition of organic amendment and nutrient release under the Zai technique in the Sahel. Nut. Cycl. Agroecosyst. 85, 225–239. Flamenco-Sandoval, A., Martínez Ramos, M., Masera, O.R., 2007. Assessing implications of land-use and land-cover change dynamics for conservation of a highly diverse tropical rain forest. Biol. Conserv. 138, 131–145. Gao, J., Liu, Y., 2010. Determination of land degradation causes in Tongyu County, Northeast China via land cover change detection. Int. J. Appl. Earth Obs. Geoinf. 12, 9–16. Hall, C.A.S., Tian, H., Qi, Y., Pontius, G., Cornell, J., 1995. Modelling spatial and temporal patterns of tropical land use change. J. Biogeogr. 22, 753–757. Houessou, L.G., Teka, O., Imorou, I.T., Lykke, A.M., Sinsin, B., 2013. Land use and landcover change at W biosphere reserve and its surroundings areas in Benin Republic (West Africa). Environ. Nat. Resour. Res. 3, 87–101. Kanianska, R., Kizeková, M., Nováček, J., Zeman, M., 2014. Land-use and land-cover changes in rural areas during different political systems: a case study of Slovakia from 1782 to 2006. Land Use Policy 36, 554–566. Kokoye, S.E.H., Tovignan, S.D., Yabi, J.A., Yegbemey, R.N., 2013. Econometric modeling of farm household land allocation in the municipality of Banikoara in Northern Benin. Land Use Policy 34, 72–79. Lambin, E.F., Geist, H.J., Lepers, E., 2003. Dynamics of land-use and land- cover change in tropical regions. Annu. Rev. Environ. Resour. 28, 205–241. MEA, 2005. Ecosystems and Human Well-Being: Current State and Trends. Island Press, Washington. Mas, J.-F., Velázquez, A., Díaz-Gallegos, J.R., Mayorga-Saucedo, R., Alcántara, C., Bocco, G., Castro, R., Fernández, T., Pérez-Vega, A., 2004. Assessing land use/cover changes: a nationwide multidate spatial database for Mexico. Int. J. Appl. Earth Obs. Geoinf. 5, 249–261. Padonou, E.A., Assogbadjo, A.E., Bachmann, Y., Sinsin, B., 2013. How far bowalization affects phytodiversity, life forms and plant morphology in Sub-humid tropic in West Africa. Afr. J. Ecol. 51, 255–262. Padonou, E.A., Fandohan, B., Bachmann, Y., Sinsin, B., 2014. How farmers perceive and cope with bowalization: a case study from West Africa. Land Use Policy 36, 461–467. Padonou, E.A., Bachmann, Y., Kakaï, R.G., Lykke, A.M., Sinsin, B., 2015a. Spatial
92
Contents lists available at ScienceDirect
Land Use Policy journal homepage: www.elsevier.com/locate/landusepol
Mapping changes in land use/land cover and prediction of future extension of bowé in Benin, West Africa
MARK
⁎
Elie A. Padonoua,b, Anne M. Lykkec, Yvonne Bachmannd, Rodrigue Idohoue, , Brice Sinsinb a
School of Forestry and Wood Industry, National University of Agriculture, Kétou, Benin Laboratory of Applied Ecology, Faculty of Agronomic Sciences, University of Abomey-Calavi, 01 BP: 526, Cotonou, Benin c Department of Bioscience, Aarhus University, Vejlsøvej 25, DK-8600 Silkeborg, Denmark d Institute of Ecology, Evolution and Diversity, J.W. Goethe University, Max-von-Laue-Str. 13, 60438, Frankfurt am Main, Germany e Laboratoire de Biomathématiques et d’Estimations Forestières, Faculty of Agronomic Sciences, University of Abomey-Calavi, 04 BP: 1525, Cotonou, Benin b
A R T I C L E I N F O
A B S T R A C T
Keywords: Land use/land cover change Natural vegetation Bowé Ferricretes West Africa
Desertification and land degradation are worldwide problems affecting soil, vegetation and the livelihoods of rural populations. Bowal (plural bowé) is a particular form of degraded land that occurs in tropical regions and leads to the exposure of ferricretes, which are unsuitable for farming. Bowé are more common on farmland and degraded savanna. Changes in land use/land cover were used to map a region of 6.7 million ha in northern Benin, West Africa in 1975, 1990 and 2010. The changes observed during these periods (1975–1990, 1990–2010 and 1975–2010) were used to predict the occurrence of bowé in the period up to 2050 using Markovian chain analysis. The results showed a considerable change in land use/land cover during the three periods. The types of land on which bowé occur (farmland and degraded savanna) increased in northern Benin by 5.4% per year during the period 1975–1990 and 9.5% per year during the periods 1990–2010, while the natural vegetation (forest, woodland and tree savanna) decreased by the same amount. The future scenarios also predicted the same trend. In the period 1975–1990, 1.28 million ha (26%) of natural vegetation was converted to degraded savanna and farmland while 2.23 million ha (53%) of natural vegetation was converted to degraded savanna and farmland in the period 1990–2010. Based on the dynamics recorded during the period 1975–1990 and 1990–2010 respectively, a total of 1.28 million ha (26% of the natural vegetation that was present in 1975) and 1.29 million ha (31% of the natural vegetation that was present in 1990) will be converted to farmland and degraded savanna in the study area by 2050.Thus bowalization will persist and increase in the period up to 2050. The natural vegetation could disappear if protection and restoration measures are not taken. It is thus important to take measures to stop the degradation and to implement programs to restore soils on bowé based on the soil and water conservation techniques used on highly degraded West African soils, such as zaï pit and stone rows with grass strips. Some native plants species adapted to bowalization and resistant to climate change in northern Benin (e.g. Asparagus africanus, Andropogon pseudapricus and Combretum nigricans) should be used in association with soil and water conservation techniques on bowé.
1. Introduction Desertification and land degradation are worldwide problems affecting soils, vegetation and the livelihoods of rural populations (D’Odorico et al., 2013; Gao and Liu, 2010). Desertification and land degradation lead to increasing levels of poverty, starvation, land abandonment and migration out of the affected regions (Verstraete et al., 2009). Combating desertification and land degradation is crucial for reducing global poverty, biodiversity loss and human-induced global climate change (MEA, 2005).
The adoption of control measures for desertification and land degradation requires the identification and monitoring of early warning signs. Among the commonly used biophysical indicators are changes in land use/land cover (Vogt et al., 2011). Changes in land use/land cover are affected by the ways in which the biophysical attributes of the land are manipulated and by the intentions underlying these manipulations (Qingqing et al., 2012). By 2100, the impact of changes in land use/ land cover on biodiversity is likely to be more significant than the impact of global climate change, nitrogen deposition, species introductions and changing atmospheric concentrations of carbon dioxide
⁎
Corresponding author. E-mail addresses: [email protected] (E.A. Padonou), [email protected] (A.M. Lykke), [email protected] (Y. Bachmann), [email protected] (R. Idohou), [email protected] (B. Sinsin). http://dx.doi.org/10.1016/j.landusepol.2017.09.015 Received 29 November 2015; Received in revised form 6 September 2017; Accepted 9 September 2017 0264-8377/ © 2017 Elsevier Ltd. All rights reserved.
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
proposed for the Tropical Ecosystem Environment Observations by Satellites Project (TREES) phase II (Achard et al., 2002). The three land cover maps were digitized and then the different land cover types were delineated using ENVI 4.1 software. Color composition (RGB 4, 7, 5) was used to improve the differentiation of the land cover types on the screen. Bands RGB 4 (0.750–0.900 μm), 7 (2.090–2.350 μm) and 5 (1.550–1.750 μm) made it possible to emphasize the differences between the stages of succession in the forested areas in addition to the savanna and farmland areas. The interpretation was aided by three additional data sources: 1:200,000 land cover maps edited in 1975 and 1987 by CENATEL Benin (the National Center of Remote Sensing) and the 2007 National Forest Inventory database, from which the land cover classes of forest, woodland, tree savanna, degraded savanna and farmland were adopted. Bowé occurred most often on farmland and degraded savanna (Fig. 3); by contrast, there was no bowal in either forest or woodland. The polygons of the land cover classes in the sub-scene acquired in 1975 were labeled according to their cover class as shown in the land cover maps from CENATEL. Once the digital map of the coverage had been finalized, the polygons were copied and overlaid with the 1990 image. The segments that required modification were changed by adding, deleting or modifying the polygon boundaries to reflect the changes in land cover between 1975 and 1990. The same process was used to visualize the changes between 1990 and 2010. This methodology was used because it avoids the generation of false changes that may occur if the databases contain a spatial mismatch (Mas et al., 2004). Areas without any recognizable vegetation in the images (water bodies, clouds and infrastructure) were excluded to reduce bias. The classification accuracy was assessed by an error matrix (Congalton, 1991). Three hundred reference points (50–60 per class) were used and these were distributed across the study area using a stratified sampling scheme (Achard et al., 2002). The error matrix was normalized with an iterative proportional fitting procedure that forced each row and column to sum to unity using the program MARGFIT (Congalton, 1991). A kappa analysis was then performed with the Kappa program to determine the accuracy of the classification (Congalton and Green, 1999).
(Sala et al., 2000). The analysis of the dynamics of changes in land use land/land cover is therefore a fundamental tool in the adoption of strategies for the conservation of biodiversity (Verburg et al., 2009) and in planning appropriate management techniques for degraded land. The analysis of changes in land use/land cover has become a fundamental tool in assessing the environmental consequences of human activity (Kanianska et al., 2014; Verburg et al., 2011). Changes in land use/land cover have consequences for biodiversity (Brink et al., 2014), geochemical cycles (Powers, 2004) and water quality (Schippers et al., 2004). The dynamics of changes in land use/land cover are influenced by the type of land cover, ecological mechanisms of succession and regeneration, the physical components of the environment, socioeconomic activities and their cultural context, meteorological phenomena and natural disasters (Yu and Lu, 2011). Bowal (plural bowé) is a particular type of degraded land that forms on hardened ferruginous soils (ferricretes) in tropical regions with a unimodal pattern of precipitation. It is a very distinctive landscape. The name originates from the fulfulde language in Guinea (Aubréville, 1947). Bowalization occurs when ferricretes are exposed as a result of the erosion of the soil surface by a combination of a dry climate and deforestation (Padonou et al., 2014, 2015b). Bowé retain very little water, so any vegetation quickly desiccates and burns early in the dry season. Root growth is impeded and the trees are dwarfed, gnarled and widely scattered. As the soils rapidly absorb and re-radiate solar energy, bowé become extremely hot and barren in the dry season. The land cover on bowé is mainly grassland and savanna (Padonou et al., 2013; Zwarg et al., 2012). Thus, it may be possible to predict the occurrence of bowé in tropical regions with a unimodal pattern of precipitation and hardened ferruginous soils by analyzing changes in land use/land cover. In this study, we considered land cover to be the biophysical state of the Earth’s land surface and immediate subsurface, including the biota, soil, topography and groundwater. We analyzed changes in land use/ land cover using a previously published set of categories (Lambin et al., 2003). Changes in land cover include changes in the biodiversity, the actual and potential primary productivity and soil quality. 2. Materials and methods
2.3. Changes in land use and land cover 2.1. Study area The elaborated maps were incorporated into a geographical information system using Arcmap 10 (ESRI) software. An overlay analysis was performed to assess the pathways of change observed over the three periods. A mask was generated to eliminate areas without vegetation (water bodies and infrastructure) to allow a comparative analysis of the same area at different times (Hall et al., 1995). After subtraction of the masked areas, the resulting area was 6,735,489 ha. Categories of change were classified into three stages: cover loss; cover regeneration; and unchanged cover. Cover loss refers to land cover that was subjected to a change with a concomitant loss in a regressive manner (for example, from forest to savanna or from woodland to farmland). Pathways of change that flowed in the opposite direction indicated regeneration.
Data were collected in the municipality of Banikoara (11° 18′ N and 2° 25′ E) in northern Benin (Fig. 1). This area is considered to be the breadbasket of the country (Kokoye et al., 2013) and agriculture plays a major part in the livelihoods of local people. Decisions about the allocation of land among farms in this region are made by farmers on socioeconomic and demographic grounds, taking into account production factors, such as labor and capital (Kokoye et al., 2013). A household’s capital is an important influence on decision-making about the amount of land allocated to cereals, legumes and cash crops. The allocation of land to cash crops is also determined by the farmers’ access to credit. The natural vegetation of this region is characterized by a mosaic of woodland, dry forest, tree and shrub savanna and gallery forest (Adomou et al., 2006). This zone was selected for study because it is dominated by the ferruginous soils (Fig. 1) on which bowé occur (Padonou et al., 2014, 2015a). Bowé could occur anywhere within the study area.
2.4. Land use/land cover transition probability matrices Transition probability matrices were elaborated for the periods 1975–1990 (15.5 years), 1990–2010 (20.5 years) and 1975–2010 (35.5 years) to describe the changes in each land cover class. Each matrix represents either the probability of the persistence of each land cover category, or the probability of a transition to another land cover category from the first to the last year in the period. The matrix values were standardized using the procedure proposed by Rovainen (1996) to obtain annualized changes and to make comparisons. To annualize the matrix values, each probability matrix was used separately to compute the matrix’s eigenvectors and eigenvalues using the diagonalization
2.2. Land cover maps Three sub-scenes of Landsat imagery for the years 1975 (Landsat Multi-Spectral Scanner), 1990 (Landsat Thematic Mapper) and 2010 (Landsat Enhanced Thematic Mapper plus) were interpreted (Fig. 2). The images were all taken during the dry season to minimize any variation in phenology of the vegetation (Clerici et al., 2007). An onscreen visual interpretation was carried out by a method similar to that 86
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 1. Soil map of the study area (Volkoff and Willaime, 1976).
method (Çinlar, 1975).
chain model. Three different scenarios were assumed corresponding to each of the three possible Markovian matrices (1975–1990, 1990–2010 and 1975–2010). The validation of the model was assessed using a χ2 test at two levels to verify the accuracy. First, the area expected from the 2010 scenarios based on the 1975–1990 period was compared with the area
2.5. Future scenarios The annualized transition matrices were used to predict the proportion of each land cover class at any one time based on a Markovian
Fig. 2. Land cover maps of the study area for 1975, 1990 and 2010.
87
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 3. Frequency of vegetation types that turned into bowé based on changes in land use/land cover classes from 1975 to 2010 and fieldwork in 2011.
Most of the forest cover was lost from 1975 to 1990 (10% of the total area decrease), while the loss of woodland and tree savanna mainly occurred from 1990 to 2010 (7 and 14% of the total area decrease, respectively). The amount of degraded savanna increased from 1990 to 2010, while farmland areas increased by 13% of the total area from 1975 to 1990 and by the same amount from 1990 to 2010.
for 2010. Second, the area expected from the 1990 scenarios based on the total period 1975–2010 was compared with the area for 1990. 2.6. Rate of degradation The rate of degradation for each of the three periods was calculated based on the cover data of the land cover types, forest, woodland and tree savanna, using the formula proposed by the Food and Agriculture Organization (FAO 1996). We considered these three land cover classes because they converted to the degraded savanna and farmland on which bowé were more frequent (Fig. 3). The rate of degradation, D (% lost area/year), was calculated as:
D = ⎡1 − (1 − ( ⎣
A1 − A2 1/ t )) ⎤ A1 ⎦
3.2. Changes in land use The transition matrices for the time periods 1974–1990, 1990–2010 and 1975–2010 (Table 2) were used to obtain the probability of change (Table 3). During the period 1975–1990, farmland and tree savanna showed large persistence (98 and 97% per year, respectively). The persistence of woodland and degraded savanna was almost as large at 96% while the persistence of forest was 94%. Most of the transformed forest converted to woodland (3.6%) and tree savanna (1.5%). The woodland was mainly converted to degraded savanna (1.3%). The tree savanna class was transformed to woodland (1%) and degraded savanna (1%), farmland (0.7%) and forest (0.5%). The transformed degraded savanna was mainly converted to farmland (2%). The farmland was mainly transformed to tree savanna (0.9%) and degraded savanna (0.9%). During the period 1990–2010, farmland had a large annual persistence (98%) and the annual persistence of woodland and degraded savanna was almost as large (97%). The annual persistence of forest was 96%. Forest cover was transformed in degraded savanna (1.6%), woodland (0.9%), farmland (0.8%) and tree savanna (0.5%). Woodland was mainly transformed in degraded savanna (2.4%) and farmland (0.6%). Tree savanna was converted to farmland (1.8%), woodland (0.9%), degraded savanna (0.6%) and forest (0.1%). Degraded savanna was mainly converted to farmland (2.2%) and woodland (0.7%). Farmland was converted to degraded savanna (1.1%), woodland (0.7%) and tree savanna (0.1%). In the period 1975–2010, a large annual persistence was obtained for farmland (100%) and the annual persistence of the other land cover types (woodland, tree savanna and degraded savanna) was almost as large at 98%. Forest had an annual persistence of 96.3%. The forest cover was mainly converted to woodland (2.0%) and degraded savanna (0.9%). The woodland cover was mainly converted to degraded savanna (1.4%) and farmland (0.7%). The transformed tree savanna and degraded savanna were mainly converted to farmland (1.1 and 1.6%, respectively).
× 100
where A1 and A2 are the initial and final cover (forest, woodland or tree savanna), respectively, and t is the interval in years for which the change in land cover is being studied. 3. Results 3.1. Land cover maps The proportions of the five land cover types changed considerably from 1975 to 2010 (Table 1; Fig. 2). In 1975, woodland and tree savanna were the dominant land cover types, whereas farmland covered only 4% of the total area (Table 1). In 1990, woodland and tree savanna were still the dominant land cover types. However, the amount of forest cover had decreased, while the amount of farmland had increased. In 2010, farmland and degraded savanna were the dominant land cover types and the other types of land cover (forest, woodland and tree savanna) had decreased. The cover of forest, woodland and tree savanna decreased in the periods 1975–1990 and 1990–2010, while the cover of degraded savanna and farmland increased during the same periods. Table 1 Land use/land cover classes used in the analysis of change (area in ha). Land cover class
1975
(%)
1990
(%)
2010
(%)
Forest Woodland Tree savanna Degraded savanna Farmland Total
1,215,566 1,949,195 1,782,725 1,499,961 288,042 6,735,489
18 30 26 22 4 100
569,563 1,943,146 1,660,493 1,394,452 1,167,835 6,735,489
8 29 25 21 17 100
243,314 1,464,790 741,123 2,078,519 2,207,743 6,735,489
4 22 11 30 33 100
3.3. Future scenarios The difference between the observed probabilities in the 2010 land 88
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Table 2 Land use/land cover transition matrix (area in ha) for the observed time period (1975–1990–2010). Forest
Woodland
Tree savanna
Degraded savanna
Farmland
Total
1975 Forest Woodland Tree savanna Degraded savanna Farmland
1990 125,702 244,591 128,431 65,779 5060
648,707 843,007 283,700 164,203 3529
270,657 222,919 915,089 212,088 39,741
128,530 369,870 258,082 599,081 38,889
41,970 268,808 197,424 458,810 200,823
1,215,566 1,949,195 1,782,726 1,499,961 288,042
1990 Forest Woodland Tree savanna Degraded savanna Farmland
2010 132,763 35,722 39,035 28,724 7071
102,966 693,587 305,901 190,983 171,354
62,378 66,254 507,454 76,353 28,684
181,113 933,392 214,672 496,409 252,934
90,385 214,208 593,405 602,014 707,731
569,605 1,943,163 1,660,467 1,394,483 1,167,774
1975 Forest Woodland Tree savanna Degraded savanna Farmland
2010 106,955 96,565 31,279 970 567
667,582 430,414 186,443 141,933 28,265
103,435 176,620 417,945 22,115 11,275
300,655 1,369,485 174,235 254,845 1137
78,241 720,195 541,630 581,626 291,079
1,256,868 2,793,279 1,351,532 1,001,489 332,323
land covered by forest, woodland and tree savanna decreases, whereas the amount of land covered by farmland increases. Assuming the dynamics recorded in the first period (1975–1990), the amount of forest will decrease by 12% of the area recorded in 1975 by 2050. The amount of woodland will decrease by 10%, tree savanna by 3% and degraded savanna by 2% of the areas recorded in 1975 by 2050. However, the area of farmland will increase by 28%. Assuming the dynamics recorded in the second period (1990–2010), the amounts of land covered by forest, woodland and tree savanna will decrease by 5, 9 and 17% of their cover in 1990, respectively, while the amount of degraded
cover map and the simulated probabilities in 2010 based on the period 1975–1990 was not significant (χ2 < 13.05; p > 0.05). The difference between the observed probabilities in the 1990 land cover map and the simulated probabilities in 1990 using the period 1975–2010 was not significant (χ2 < 11.85; p > 0.05). Thus the model can be applied to simulate scenarios of change in land use and land cover. Future land cover will change depending on the dynamics considered (1975–1990, 1990–2010 and 1975–2010). The historical (1975–2010) and estimated future (2010–2050) changes in each land cover class are shown in Fig. 4. In all three time periods, the amount of
Table 3 Annual probability matrices. Annual probability matrix 1975–1990 Forest 1975
1990
Forest Woodland Tree savanna savanna Degraded savanna Farmland
0.940 0.008 0.005 0.003 0.001
Woodland
Tree savanna
Degraded savanna
Farmland
0.036 0.962 0.011 0.007 0.001
0.015 0.008 0.967 0.009 0.009
0.007 0.013 0.010 0.960 0.009
0.002 0.009 0.007 0.020 0.980
Annual probability matrix 1990–2010 Forest
Woodland
Tree savanna
Degraded savanna
Farmland
1990
2010
Forest Woodland Tree savanna Degraded savanna Farmland
0.962 0.001 0.001 0.001 0.000
0.009 0.968 0.009 0.007 0.007
0.005 0.002 0.965 0.003 0.001
0.016 0.024 0.006 0.968 0.011
0.008 0.006 0.018 0.022 0.980
Forest
Woodland
Tree savanna
Degraded savanna
Farmland
0.020 0.976 0.004 0.004 0.003
0.003 0.002 0.980 0.001 0.001
0.009 0.014 0.004 0.978 0.000
0.004 0.007 0.011 0.016 0.996
Annual probability matrix 1975–2010
1975
2010
Forest Woodland Tree savanna Degraded savanna Farmland
0.963 0.001 0.001 0.000 0.000
89
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
Fig. 4. Simulation of the evolution of the five land cover classes under three future scenarios. Dynamics observed during the periods (a) 1975–1990; (b) 1990–2010; and (c) the total study period (1975–2010).
main land cover classes where bowé are common (Fig. 3). During the three time periods, the highest rate of degradation was for forest cover, for which 1,089,864 ha were lost during the first period (1975–1990) and 436,841 ha in the second period (1990–2010). This is about 72,658 ha/year (4.93% of the study area) during the first period and 21,842 ha/year (4.16% of the study area) during the second period. The rates of degradation of woodland and tree savanna increased in the second period. The loss of woodland cover in the first period was 861,597 ha (57,440 ha/year), while 1,213,853 ha (60,693 ha/year) were lost in the second period. For the tree savanna, 455,506 ha (30,367 ha/year) were lost during the first period, while 808,077 ha (40,404 ha/year) were lost during the second period. During the first period, 1,264,684 ha (26%) of the three land cover types (forest, woodland and tree savanna) were converted to degraded savanna and farmland, while in the second period 2,227,174 ha (53%) were transformed to degraded savanna and farmland.
savanna and farmland will increase by 9 and 22% of the total area, respectively. Forest, woodland, tree savanna and degraded savanna will decrease by 14, 11, 18 and 2%, respectively, while the area of farmland will increase by 46% by 2050 when considering the dynamics recorded for the whole study period (1975–2010). 3.4. Rates of degradation The rate of degradation of the natural vegetation (forest, woodland and tree savanna) is presented in Table 4. We considered only these classes because they convert to degraded savanna and farmland, the Table 4 Annual rate of degradation (% lost ha/year). Land cover class
1975–1990
1990–2010
1975–2010
Forest Woodland Tree savanna Total
4.928 0.021 0.472 5.422
4.163 1.403 3.953 9.520
4.492 0.813 2.477 7.782
4. Discussion The conversion of forest, woodland and tree savanna into degraded 90
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
example, linking the universal soil loss equation to a geographical information system (El Garouani et al., 2007) may improve our understanding of soil erosion in areas where bowé occur. The limits observed with the method considered in this study should be taken into account in similar studies. However, the trends obtained in this work are useful for understanding the pattern of bowalization in the study area.
savanna and farmland was the most common change observed in the study area. The study area is considered to be the main area of food and cash crop production (mainly cotton) in Benin. Cotton contributes 64% of export income, 90% of agricultural income and 24% of state income in the study area (Kokoye et al., 2013). The natural vegetation is therefore intensely exploited as farmland. The annual rate of change of land cover to farmland was stable during both time periods (1975–1990 and 1990–2010) studied here, while the amount of degraded savanna increased in the second time period (1990–2010). Agriculture in this region has been dominated over the past three decades by cotton production (Ton, 2004). When farmland becomes depleted and degraded, losing its fertility, it is usually abandoned and new areas of natural vegetation (forest, woodland or tree savanna) are converted to farmland. The former farmland is converted to degraded savanna. We found a transition from farmland to degraded savanna during these time periods, confirming the increase in degraded savanna in the second time period resulting from farmland abandoned in the first time period. In both time periods, farmland showed a high persistence, varying from 98 to 100% while natural vegetation land cover class was degraded in the area. This is in accordance with previous findings of a conversion of land cover from forest to farmland with a loss of biodiversity in this region (Houessou et al., 2013). The persistence of farmland is mainly a result of cotton production. The production of this crop is encouraged by the government, which provides credit in the form of seeds, fertilizers, pesticides and tractors to farmers in this region. This credit is the main determinant in the allocation of land to cotton production (Kokoye et al., 2013). Thus farmland persists and increases each year at the expense of forest, woodland and tree savanna. The area of natural vegetation (forest, woodland and tree savanna) that was considered degraded increased by 4.1% between the first and the second time periods (i.e. from 5.4% over the period 1975–1990 to 9.5% over the period 1990–2010). The land cover was converted to farmland and degraded savanna. Although not quantified or measured in the field, the increase in farmland likely resulted in the occurrence of bowé in the study area (Padanou et al., 2014). Among the important factors leading to the occurrence of bowé in the study area were the increase in farmland and the practice of animal-drawn tillage. The amount of land that could be affected by bowalization thus increased during the two time periods considered here. Future scenarios developed using the Markovian model with annualized transition matrices (1975–1990, 1990–2010 and 1975–2010) showed a decrease in forest, woodland and tree savanna and an increase in farmland and degraded savanna by 2050. Based on this model, the total cover of forest, woodland and tree savanna will decrease from 74% (in 1975) to 48% by 2050 and from 62% in 1990–31% in 2050 using the dynamics recorded for 1975–1990 and 1990–2010, respectively. This means that by 2050, 26% of the total area recorded as forest, woodland and tree savanna in 1975 will have been converted to farmland and degraded savanna if we assume the dynamics recorded from 1975 to 1990. Considering the dynamics of 1990–2010, 31% of these land cover types recorded in 1990 will have been converted to farmland and degraded savanna. As bowé are more common in farmland and degraded savanna (Fig. 3), bowalization will therefore increase in the future. However, because bowé occur after the erosion of topsoil, taking information on other variables that drive soil erosion into account in more sophisticated models will improve this forecast and enable researchers to outline normative scenarios to assess bowalization. Markovian models have been used in many analyses of changes in land use/land cover (Flamenco-Sandoval et al., 2007). They are useful for exploratory analysis and for depicting contrasting scenarios. The models are not spatially explicit and assume that the transition probabilities are homogenous over time. More detailed spatially explicit models on changes in land use/land cover may be used in future analysis to improve our understanding of the locations and pathways of land use/land cover change dynamics that induce bowalization. For
5. Conclusions Agriculture is the dominant form of land management in northern Benin. Over the period 1975–1990, 5.4% of the natural vegetation was converted to farmland and degraded savanna while over the period 1990–2010 this figure was higher at 9.5%. Thus the land was degraded and bowé occurred. Extrapolation of all the dynamics considered to 2050 showed an increasing rate of degradation (mainly to farmland) and a decreasing amount of natural vegetation. About 26 and 31% of the total area of natural vegetation (forest, woodland and tree savanna) will be converted to farmland and degraded savanna by 2050 if we assume the dynamics recorded from 1975 to 1990 and 1990–2010, respectively. Bowalization will increase in the future and limit the amount of land available for farmland. The natural vegetation could disappear entirely if protection and restoration measures are not taken. Protection measures are necessary to preserve natural vegetation, but it is also important to implement programs to restore and conserve soils on bowé. These programs should include the adoption of appropriate soil and water conservation techniques for highly degraded West African soils, such as zaï pits and stone rows with grass strips. These soil and water conservation techniques have been used for decades to reclaim marginal and degraded barren lands in the arid and semi-arid zones of West Africa (Sidibe, 2005; Fatondji et al., 2009; Bayala et al., 2011). Some native plant species are adapted to bowalization and are resistant to climate change in northern Benin (e.g. Asparagus africanus, Andropogon pseudapricus and Combretum nigricans). These species could be used on bowé in association with soil and water conservation techniques. Acknowledgments This work was supported by UNDESERT (EU FP7 243906), “Understanding and combating desertification to mitigate its impacts on ecosystem services” funded by the European Commission, Directorate General for Research and Innovation, Environment Programme. Additional funding was obtained from the Robert S. McNamara Fellowships Program and the International Foundation for Science. References Çinlar, E., 1975. Exceptional paperÇMarkov renewal theory: A survey. Manage. Sci. 21 (7), 727–752. Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., Richards, T., Malingreau, J.P., 2002. Determination of deforestation rates of the world’s humid tropical forests. Science 297, 999–1002. Adomou, A., Sinsin, B., van der Maesen, L., 2006. Phytosociological and chorological approaches to phytogeography: a meso-scale study in Benin. Syst. Geogr. Plants 6, 155–178. Aubréville, A., 1947. Les brousses secondaires en Afrique équatoriale. Bois et Forêts des Tropiques 2, 24–35. Bayala, J., Kalinganire, A., Tchoundjeu, Z., Sinclair, F., Garrity, D., 2011. Conservation Agriculture with Trees in the West African Sahel–a Review. ICRAF (Occasional paper, 14). Brink, A.B., Bodart, C., Brodsky, L., Defourney, P., Ernst, C., Donney, F., Lupi, A., Tuckova, K., 2014. Anthropogenic pressure in East Africa—monitoring 20 years of land cover changes by means of medium resolution satellite data. Int. J. Appl. Earth Obs. Geoinf. 28, 60–69. Clerici, N., Bodini, A., Eva, H., Grégoire, J.-M., Dulieu, D., Paolini, C., 2007. Increased isolation of two Biosphere Reserves and surrounding protected areas (WAP ecological complex, West Africa). J. Nat. Conserv. 15, 26–40. Congalton, R.G., Green, K., 1999. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices. CRC Press, Boca Raton. Congalton, R.G., 1991. A review of assessing the accuracy of classifications of remotely
91
Land Use Policy 69 (2017) 85–92
E.A. Padonou et al.
distribution of bowal and differences in physicochemical characteristics between bowal and woodland soils in Benin, West Africa. Catena 124, 45–52. Padonou, E.A., Lykke, A.M., Bachmann, Y., Sinsin, B., 2015b. Bowalization: its impact on soil, biodiversity, and human livelihoods in West Africa. Adv. Environ. Res. 40, 187–198. Powers, J.S., 2004. Changes in soil carbon and nitrogen after contrasting land-use transitions in northeastern Costa Rica. Ecosystems 7, 134–146. Qingqing, Z., Hailiang, X., Jingyi, F., Pujia, Y., Peng, Z., 2012. Spatial analysis of land use and land cover changes in recent 30 years in manas river basin. Procedia Environ. Sci. 12, 906–916 (Part B). Rovainen, E., 1996. Analysis and standardization of transition matrices. In: In: FAO (Ed.), Forest Resources Assessment 1990. Survey of Tropical Forest Cover and Study of Change Processes 130. FAO, pp. 111–114 Forestry Paper. Sala, O.E., Stuart Chapin Jr., F., Armesto, J.J., Berlow, E., Bloomfield, J., Dirzo, R., HuberSanwald, E., Huenneke, L.F., Jackson, R.B., Kinzig, A., Leemans, R., Lodge, D.M., Mooney, H.A., Oesterheld, M.N., Poff, N.L., Sykes, M.T., Walker, B.H., Walker, M., Wall, D.H., 2000. Global biodiversity scenarios for the year 2100. Science 287, 1770–1774. Schippers, P., Vermaat, J.E., de Klein, J., Mooij, W.M., 2004. The effect of atmospheric carbon dioxide elevation on plant growth in freshwater ecosystems. Ecosystems 7, 63–74. Sidibe, A., 2005. Farm-level adoption of soils and water conservation techniques in northern Burkina Faso. Agric. Water Manage. 71, 211–224. Ton, P., 2004. La production du coton au Bénin. FUPRO-Bénin, Pays-Bas et Bénin. Verburg, P.H., van de Steeg, J., Veldkamp, A., Willemen, L., 2009. From land cover change to land function dynamics: a major challenge to improve land characterization. J. Environ. Manage. 90, 1327–1335. Verburg, P.H., Neumann, K., Nol, L., 2011. Challenges in using land use and land cover data for global change studies. Global Change Biol. 17, 974–989. Verstraete, M.M., Scholes, R.J., Smith, M.S., 2009. Climate and desertification: looking at an old problem through new lenses. Front. Ecol. Environ. 7, 421–428. Vogt, J.V., Safriel, U., Von Maltitz, G., Sokona, Y., Zougmore, R., Bastin, G., Hill, J., 2011. Monitoring and assessment of land degradation and desertification: towards new conceptual and integrated approaches. Land Degrad. Dev. 22, 150–165. Yu, L., Lu, J., 2011. Does landscape fragmentation influence sex ratio of dioecious plants? A case study of Pistacia chinensis in the Thousand-Island Lake region of China. PLoS One 6, e22903. Zwarg, A., Schmidt, M., Janßen, T., Hahn, K., Zizka, G., 2012. Plant diversity, functional traits and soil conditions of grass savannas on lateritic crusts (bowé) in south eastern Burkina Faso. Flora et Vegetatio Sudano-Sambesica 15, 15–24.
sensed data. Remote Sens. Environ. 37, 35–46. D’Odorico, P., Bhattachan, A., Davis, K.F., Ravi, S., Runyan, C.W., 2013. Global desertification: drivers and feedbacks. Adv. Water Resour. 51, 326–344. El Garouani, A., Chen, H., Lewis, L., Tribak, A., Abahrour, M., 2007. Apport de la Télédétection et du SIG pour le suivi spatio-temporel de l’occupation du sol et de l’érosion nette dans le bassin de l’Oued Tlata (Maroc), Acte des JSIRAUF, Hanoi. Fatondji, D., Martius, C., Zougmore, R., Vlek, P.L.G., Bielders, C.L., Koala, S., 2009. Decomposition of organic amendment and nutrient release under the Zai technique in the Sahel. Nut. Cycl. Agroecosyst. 85, 225–239. Flamenco-Sandoval, A., Martínez Ramos, M., Masera, O.R., 2007. Assessing implications of land-use and land-cover change dynamics for conservation of a highly diverse tropical rain forest. Biol. Conserv. 138, 131–145. Gao, J., Liu, Y., 2010. Determination of land degradation causes in Tongyu County, Northeast China via land cover change detection. Int. J. Appl. Earth Obs. Geoinf. 12, 9–16. Hall, C.A.S., Tian, H., Qi, Y., Pontius, G., Cornell, J., 1995. Modelling spatial and temporal patterns of tropical land use change. J. Biogeogr. 22, 753–757. Houessou, L.G., Teka, O., Imorou, I.T., Lykke, A.M., Sinsin, B., 2013. Land use and landcover change at W biosphere reserve and its surroundings areas in Benin Republic (West Africa). Environ. Nat. Resour. Res. 3, 87–101. Kanianska, R., Kizeková, M., Nováček, J., Zeman, M., 2014. Land-use and land-cover changes in rural areas during different political systems: a case study of Slovakia from 1782 to 2006. Land Use Policy 36, 554–566. Kokoye, S.E.H., Tovignan, S.D., Yabi, J.A., Yegbemey, R.N., 2013. Econometric modeling of farm household land allocation in the municipality of Banikoara in Northern Benin. Land Use Policy 34, 72–79. Lambin, E.F., Geist, H.J., Lepers, E., 2003. Dynamics of land-use and land- cover change in tropical regions. Annu. Rev. Environ. Resour. 28, 205–241. MEA, 2005. Ecosystems and Human Well-Being: Current State and Trends. Island Press, Washington. Mas, J.-F., Velázquez, A., Díaz-Gallegos, J.R., Mayorga-Saucedo, R., Alcántara, C., Bocco, G., Castro, R., Fernández, T., Pérez-Vega, A., 2004. Assessing land use/cover changes: a nationwide multidate spatial database for Mexico. Int. J. Appl. Earth Obs. Geoinf. 5, 249–261. Padonou, E.A., Assogbadjo, A.E., Bachmann, Y., Sinsin, B., 2013. How far bowalization affects phytodiversity, life forms and plant morphology in Sub-humid tropic in West Africa. Afr. J. Ecol. 51, 255–262. Padonou, E.A., Fandohan, B., Bachmann, Y., Sinsin, B., 2014. How farmers perceive and cope with bowalization: a case study from West Africa. Land Use Policy 36, 461–467. Padonou, E.A., Bachmann, Y., Kakaï, R.G., Lykke, A.M., Sinsin, B., 2015a. Spatial
92