Modeling the Geographic Distribution of Prosopis africana ( G . and Perr . ) Taub . in Niger

Prosopis africana is a species of great socio-economic importance, threatened with extinction from its natural habitat in Niger due to overexploitation. The main objective of this study is to determine the potential geographic distribution of P. africana in Niger. Climatic and botanical data has been collected and used to model the distribution, on the basis of principle of maximum entropy (MAXENT) using MAXENT 3.3.3k, DIVA-GIS 7.5, and ArcGIS 10.0. programs. Rainfall and temperature are the most significant variables in the distribution of P. africana in Niger. Thus the southern band of the country (from the sudanian zone to the sahelio-soudanian zone), the wettest, is the area conducive to the development of P. africana (128,692.32 km2 in total, 10.16% of the territory). Given the extent of this area revealed by this study, a reforestation policy implementation of P. africana would allow to restore its stands in Niger.


Introduction
In the third world countries, wood plants figure prominently in the socio-economic life of the people, given the many products and services (wood, food and Pharmacopeia) that they provide (Matthias, Henri, & Félix, 2000;Larwanou, Oumarou, Laura, Dan Guimbo, & Eyog-Matig, 2010;Chidumayo, Okali, Kowero, & Larwanou, 2011;Priso, Nnanga, Etame, Din, & Amougou, 2011;Laouali, Dan Guimbo, Larwanou, Inoussa, & Mahamane, 2014).They are thus subject to overexploitation and a threat of disappearance of their natural habitats, due to high population growth and climatic conditions increasingly unfavorable to their development (Ozer, Hountondji, Niang, Karimoune, Laminou, & Salmon, 2010).Indeed, the natural range of the species is influenced, among others, by climatic variables such as precipitation, temperature and wind speed (Intergovernmental Panel on Climate Change [IPCC], 2007;Chidumayo et al., 2011).However, it should be noted that according to the quality of products supplied, woody species do not have the same importance and therefore do not have the same degree of exploitation.
Characterized by a low capacity of natural regeneration (Ahoton, Adjakpa, M'po Ifonti & Akpo, 2009;Niang-Diop, Sambou & Lykke, 2010;Laouali, Dan Guimbo, Youchaou, Rabiou & Mahamane, 2015a) and a low fruit production (Laouali, Dan Guimbo, Chaibou & Mahamane, 2015b), P. africana is represented in Niger by generally isolated individuals whose spatial distribution across the territory remains unknown.But a good knowledge of the spatial distribution and the habitat of species helps to develop effective strategies for biodiversity planning and conservation (Scheldeman & Van Zonneveld, 2010), forecasting the effects of environmental change and zonation of plant diversity (Schmidt, König, & Müller, 2008).Among the methods that allow knowing the distribution of species, there is modeling.

Study AREA
This study covered the entire territory of Niger, 1,267,000 km 2 between 11° 37' and 23° 33' north latitude and between 0° 10' and 16° 00' east longitude.Niger is bounded on the north by Algeria and Libya, to the east by Chad, to the south by Nigeria and Benin, and to the West by Burkina Faso and Mali (Figure 1).
The climate is generally warm and dry but with a variability according to a north-south aridity gradient for distinguishing a saharan zone in the north (about 2/3 of the territory) with an average annual rainfall of less than 200 mm, a sahelian zone with an average annual rainfall between 200 and 600 mm and on the south a soudanian zone whose average annual rainfall exceeds 600 mm.The passage from one zone to another is marked by a climatic transition band which implies variability in the spatial distribution of vegetation.So, from north to south, the vegetation varies from steppe to dry woodland, through savanna and Combretaceae thickets.
During the year, the mean temperature varies from 9.9 to 32.63° C in the north, from 19.44 to 32.78° C in the southeast and from 25.84 to 33.57° C in the southwest of the country.
The main types of soils in Niger are: • raw mineral soils in the north, the least watered area of the country; • low developed soils in the south of Tenere, in the west and center of the country; • sub-arid soils rich in organic matter, more watered than the previous, located from west to east in the central part of the country; • soils with sesquioxides of iron, rich in swelling clay, located in the southern part of the country; • vertisols especially in the east in the Lake Chad basin (Saadou, 1990).
About geomorphology, Niger has a fairly wide variety of landforms: structural forms marked by the lithological and tectonic data and climatic forms related to the erosion action.

Plant Material
The genus Prosopis is represented by 45 species of trees and shrubs native to North America, Central America, South America, Africa and Asia (Akaaimo & Raji, 2006).The unique tropical african species of the genus is Prosopis africana, native to the Sahel region of West Africa.Its natural distribution covers from Senegal to Ethiopia in the north, Guinea to Cameroon in the south and Uganda to Egypt in the East (Weber, Larwanou, Abasse, & Kalinganire, 2008).
The tree, with a height of 12-15 m, sometimes even 20 m, is characterized by a straight and cylindrical trunk, sometimes up to 110 cm in diameter, an open crown and drooping light green foliage (Arbonnier, 2000).Its root system is of the tap type (Weber et al., 2008).

Presence Points of the Species
A series of botanical surveys in different regions of Niger has enabled to establish a database on the occurence of P. africana in Niger.Other databases such as that of the Garba Mounkaila laboratory of Abdou Moumouni University of Niamey, and GBIF (Global Biodiversity Information Facility) (www.gbif.org,accessed on 2/11/2014) and previous works (Weber et al., 2008;Larwanou et al., 2012) have been exploited.According to all these bases, the species occurs mainly in the sahelian and sudanian climatic zones with a higher concentration in the south-central region of the country.So, 234 presence points of the species, of which 203 in Niger and 31 distributed among eastern Burkina Faso, northern Nigeria and northern Benin were selected for this study (Figure 1).The presence points of the species have been extended beyond Niger administrative borders to improve the accuracy of the model.According to several authors (Fitzpatrick & Hargrove, 2009;Scheldeman & Van Zonneveld, 2010), it is recommended to use an area that covers to the fullest the natural distribution area of the species under the same climatic conditions.Two types of climatic data (precipitation and temperature monthly mean, divided into 19 variables) of the 1950-2000 period were selected and used in this study.The data in raster CLM with a spatial resolution of 2.5 minutes (5 km x 5 km = 25 km 2 the pixel) have been downloaded from WorldClim archives (www.worldclim.org,accessed on 9/21/2014).Other environmental factors such as soil factors are decisive in the distribution of a species but at finer scales; do not take this into account in this study results from the tolerance of P. africana for most soil types (Weber et al., 2008).

Data Analysis
To generate the theoretical model of distribution of P. africana, data on the presence of the species and the 19 climatic variables were treated with DIVA-GIS 7.5, MAXENT 3.3.3k.and ArcGIS 10.0 softwares used in the same frame by several authors (Hijmans et al., 2005;Phillips et al., 2006;Scheldeman & Van Zonneveld, 2010;Young, Carter, & Evangelista, 2011).2013)), only one was selected based on its ecological importance for the species, because a strong correlation between these variables would introduce a bias in the model (Warren, Glor, & Turelli, 2010;Lee, Hanneman & Hackenbrook, 2011;Dormann et al., 2013;Maël, Catherine, Beatriz, & Vincent, 2014).
From the variables used and the information layer on the occurrence of P. africana, the MAXENT program generated a raster on the potential distribution of the species and a file containing the various statistical results of the model.Finally, the superposition of an information layer of administrative boundaries in ArcGIS, as had been done by Scheldeman and Van Zonneveld (2010), Young et al. (2011), Hijmans, Luigi, andPrem (2012), enabled to map and to locate the potential distribution areas of P. africana.
To determine the individual contribution of each variable, a Jackknife test (Young et al., 2011), which consists in spread by turns the variables and generating each time a model with the remaining variables then a model with the only discarded variable, was performed (Figure 2).
Cross-validation was also used in this study.Cross-validation is a process of evaluating the efficiency of a model that consists of split the presence points in two parts to calibrate and to test the model.The calibration of a model is to adjust the parameter values (climatic variables and presence points) in order for the simulated data correspond to the actual values (Guisan & Zimmermann, 2000;Dzotsi, 2002).In this study, 75% of the species' presence points were used to calibrate the model, while 25% were used to test it.These same rates have been used by Doko et al. (2008), Young et al. (2011), Fandohan et al. (2013), Gbesso et al. (2013) to calibrate and test their models.To strengthen a model, it is important to generate it from an average of multiple repetitions of cross validation (Young et al., 2011).Cross-validation was repeated 10 times in this study.About the predictive ability of the model, it was evaluated by using the parameter AUC (Area Under the Curve).The AUC is equal to the likelihood that a randomly selected presence point is located in a raster cell with a higher probability value for species occurrence than a randomly selected absence point (Phillips et al., 2006).

Contribution of Climatic Variables and Performance of the Model
After the correlation test, only five (5) climatic variables were selected because of their low correlations (| |≤ 0.7), and used for modeling.These are BIO3 (isothermality), BIO9 (average temperature of the driest quarter), BIO10 (average temperature of the warmest quarter), BIO12 (annual precipitation) and BIO18 (precipitation of the warmest quarter).Among these variables, the study reveals that BIO12 and BIO10 have the highest levels of contribution to the model (Table 1).The cross-validation, with an AUC of 0.953 and the Jackknife test (Figure 2) confirm the contribution to the model of BIO12 that has the highest gain when used in isolation and decreases the gain the most when it is omitted from the analysis.
Figure 2. Jackknife test results on the 5 variables.The blue bar shows the gain obtained if a variable is used alone and the remaining variables are excluded from the analysis and the green bar shows how much the total gain is reduced if this variable is excluded from the analysis

Potential Distribution Areas of P. africana in Niger
The study reveals that favorable habitats for the development of P. africana in Niger, 128,692.32 km 2 in total, 10.16% of the territory, are allocated between sahelian, sahelo-soudanian and soudanian zones (Figure 3).The most favorable habitats cover an area of 41,304.32km 2 , 3.26% of the territory and are divided mostly between the sahelo-soudanian and soudanian zones, the wettest of the country.The averagely favorable habitats for the development of P. africana, with an area of 57,138 km 2 , 4.51% of the national territory, are located mostly in the sahelo-soudanian zone.The less favorable habitats, with an area of 30,250 km 2 , 2.39% of the territory, are mainly located in the sahelian zone.

Discussion
The modeling of species distribution is very important in ecology and models can be seen as a simplified view of reality.However, the choice of data and the methodological approach affect the quality of results (Elith et al., 2006;Schmidt et al., 2008).
This study is based on a set of existing data and additional data collected in the field to fill the gaps in spatial coverage and take account of environmental variability as suggested by Hirzel and Guisan (2002) that compared the effectiveness of different sampling strategies for modeling species distribution.For these authors, systematic sampling is the best strategy for predicting the geographic distribution of species and it is the same type of sampling that was applied in this work.The low concentration of presence points of the species in neighboring countries of Niger is due to the lack of sources because field missions have not gone beyond the country's borders.Even within the country, the spatial heterogeneity of the presence points observed in the field would be partly related to the sampling effort.This could indeed vary depending on the accessibility of areas and / or the collectors' alertness, and thus constitute a limit that must be interpreted with caution, as suggested Ndayishimiye (2011).Nevertheless, the different localities surveyed cover the entire study area.The concentration of presence points of P. africana in the southern part, the wettest of the country, may be related to rainfall which is also one of the main factors determining the vegetation distribution in tropical zones (Saadou, 1990;Mahamane, Morou, Zaman-Allah, Saadou, Saley, Bakasso, & Jauffret, 2012).However, the spatial heterogeneity of the presence points of the species observed at the different territories in this area also suggests an influence of other factors such as human pressure, in addition to eventual sampling bias.Indeed, nearly 90% of Niger's population lives in the southern agricultural (Institut National de la Statistique [INS], 2014), the area where much of the woody vegetation is located.This demographic concentration, coupled with a shortage of farmland becoming poorer for lack of fallow could only lead to an impoverishment of the population and pressure on forest products.
The largest contribution to the model of the annual precipitation (BIO12) and the average temperature of the warmest quarter (BIO10) reflects the actual climatic conditions of the species' high concentration area that is relatively rainiest than the north part.According to Weber et al. (2008), P. africana prefers an annual rainfall of between 600 and 1500 mm.The effectiveness of these two direct parameters in the model, especially when the species distribution modeling covers a large spatial scale, has been noted by several authors (Guisan & Zimmermann, 2000;Fandohan et al., 2013;Gbesso et al., 2013).The type of soil is not a limiting factor in the distribution of P. africana because according to Weber et al. (2008), this species tolerates most soil types, although it abounds in general on land fallow or sandy clay soils.This is corroborated by the diversity of soils in the species' distribution area found in this study.Indeed, the soils are mainly sandy clay in southern Dosso region (Gavaud & Boulet, 1967), dunal to south Maradi and sandy, sandy loam, clay loam in southern Zinder (Larwanou et al., 2006;Laouali et al., 2014).
The value of AUC (0.953), closer to its maximum (1) than its minimum (0.5, corresponding to a random prediction) indicates a good model performance (Swets, 1988;Phillips et al., 2006).A high AUC value does not necessarily reflect a good performance of a model (Phillips et al., 2006;Scheldeman & Van Zonneveld, 2010) when the predicted distribution area is small compared to the total area of the study.For this study, the predicted distribution area is relatively large compared to the area of the study.In addition, the use of repetitions and cross-validation in executing MAXENT yielded a model derived from the average, most powerful.
The resulting model suggests that P. africana is widely distributed from the soudanian zone to the sahelo-soudanian band of the country.This potential distribution area is wider than that suggested by Bonnet, Arbonnier and Grard (2008) which is limited to a thin strip in the south of the country.It is also wider than that observed in the field through the presence points of the species.However, the probability of encountering the species is higher in the southern strip and the extreme south-west, where conditions are more favorable for its development.Indeed, a species may be absent in areas where it should be if his natural habitat has been altered by human interference (Scheldeman & Van Zonneveld, 2010).Effectively, P. africana is a species subject to a strong human pressure in some areas of Niger because of the domestic use of its wood and its importance in the traditional pharmacopoeia (Laouali et al., 2014), which would have contributed to the heterogeneity observed in the distribution on the field.
This study allowed estimating the distribution area of P. africana in Niger and revealed heterogeneity in the organization of the species in the landscape.It thus showed a theoretical distribution area wider than that observed in the field.Given the extent of the area revealed by this study and the socio-economic importance of P. africana, restoration operations through a reforestation policy implementation must be carried out in all habitats favorable for its development.Other endangered species in Niger must also be the subject of similar studies to identify favorable habitats for their development and proceed to their restoration.

Figure 1 .
Figure 1.Presence points of P. africana in the study area The first step in the creation of the model consisted, using the DIVA-GIS program, to convert the CLM format of the imported climatic data in ASCII format to obtain environmental raster supported by the MAXENT modeling program.These rasters are made up of 19 climatic variables, some of which have been discarded after a Pearson correlation test.Indeed, among the variables that have a high correlation coefficient (| | > 0.7, as proposed byDormann et al. (

Figure 3 .
Figure 3. Potential distribution areas of P. africana in Niger

Table 1 .
The climatic variables used and their contributions to the model