Simulation of the Dynamics of Land Cover and Land Use in the Lobo River Watershed Upstream of Nibéhibé (Center-West of Côte d’Ivoire)

From medium-resolution satellite images (Landsat TM, ETM+ and OLI), the spatial dynamics of land cover and land use are highlighted. The objective of this study is to quantify the evolution of land use in the watershed of the Lobo River upstream of Nibéhibé between 1986 and 2019 in order to analyze the impacts of human activities on the landscape. The study method was based, on the one hand, on the processing of satellite images, for the analysis of the dynamics of land use and, on the other hand, on the CA-Markov model, for the prediction of land use by 2050. It emerged from this study that the land use maps produced made it possible to highlight the spatio-temporal dynamics of land use on the basin. For the period from 1986 to 2019, there is a decrease in the area of forests in favor of built-bare ground and crops and fallows. A land use scenario for the years 2019 and 2050 was simulated with an accuracy of 87.11%. The regressive trend in forests seems to continue in the future with current land use practices.


Introduction
The lands of the world are undergoing profound spatial changes. In West Africa and particularly in Côte d'Ivoire, these spatial changes which result in the degradation of the vegetation cover are increasing day by day.
Like the developing countries, Côte d'Ivoire has an economy heavily dependent on the agricultural and forestry sector, leading to a rapid decrease in forest areas [1]. In addition, studies carried out on the evolution of this forest cover vegetation cover of the Center-West area has not escaped this deforestation.
In the area of Lobo River watershed studied here, stripping of land for agriculture has become increasingly important in recent years [4]. Resources are under strong anthropogenic pressures because of the population explosion and a purely extensive cropping system. The increased expansion of cultivation areas and agglomerations leads to a gradual reduction in the extent of wooded areas and a destabilization of the soil structure [5]. Crop rotation and fallow also have major disturbances on these precariously balanced ecological systems. Changes in land use thus have a direct impact on land cover and on the landscape configuration of this environment.
For several decades, the modeling and projection of land use change has established itself as a relevant decision support tool. It allows the analysis of territorial planning policies in order to assess and anticipate their environmental impacts [6]. The originality of this research lies in the fact that the land use dynamics modeling will make it possible to follow the evolutionary trend of the landscape and to find acceptable rules to preserve natural resources, like forest and water resources. Therefore, we can wonder about the spatial translation of this degradation of the landscapes. The objective of this study is to assess the land use dynamics of the Lobo River Watershed upstream of Nibéhibé, in order to predict their future changes. More specifically, it will be a question of simulating changes in land use by 2050.
To simulate changes in land cover and land use, several models are used. Among the most widely used, mention may be made of the combination of cellular automata model (CA) and Markov model (CA-MARKOV). The work of [7] shows that the CA-MARKOV combination integrates the stochastic function of Markov Method with the stochastic spatial characteristic of Technology CA. CA-MARKOV a powerful tool that has enabled several authors including [8] [9] [10] to simulate land cover and land use changes.
To conduct such a study, it is important to use reliable quantitative and qualitative data. Under these conditions, satellite imagery and geographic information systems (GIS) present themselves as an opportunity for this study. Indeed, satellite imagery, thanks to its synoptic view, makes it possible to understand and map dynamic phenomena such as land use. As for GIS, they make it possible to organize and better structure information. In view of the prediction of land use of the Lobo river watershed upstream of Nibéhibé, the use of models is necessary. Under these conditions, CA-Markov model appears to be useful tools for this study.

Study Area
The Lobo River Watershed upstream of Nibéhibé is located in the center-west of the Côte d'Ivoire between 6˚45' and 7˚55' North latitude and between 6˚15' and  The Lobo river watershed upstream of Nibéhibé is an agricultural area. Agriculture is the main activity which occupies most of the active population in the watershed [8]. The rural environment is dominated by agriculture. It is an extensive, anarchic, rainfed and manual agriculture, which compensates for its weaknesses by the permanent conquest of new lands [4] [12]. Cash crops are cocoa, coffee, rubber, cotton, cashew and food crops. In terms of food crops, maize, rice, yams, plantains, chili peppers, eggplant and cassava are the main subsistence crops of the populations (Ministry of Agriculture).

Materials
The material used consists of satellite images and agricultural data. The satellite  [15]. For this, our data was selected during the great dry season (December to February), the period during which the cloud cover rates are the lowest. In addition, this choice contributes to reducing any seasonal effects [16].

Satellite Image Processing
Knowledge of the study area made it possible to opt for a supervised classification. The Maximum Likelihood algorithm was chosen for the supervised classification. This method is commonly used in the literature. The validation of the classification is obtained using the confusion matrix, the overall precision and Journal of Geographic Information System the Kappa coefficient (Table 1, Table 2). These tables display in the diagonal, the percentage of well classified pixels and off diagonal, the percentage of misclassified pixels. The confusion matrix shows a good classification of the images.

Analysis of the Dynamics of Land Use
Analysis of changes over the entire study period was done by post-classification comparison. It produces a change detection matrix resulting from the comparison between the pixels of two classifications between two dates [9]. From this situation, the global rate of change (T g ) and the average annual rate of spatial expansion (T c ) were calculated.
Global scale changes were determined by showing the areas of different land use units for each year. The changes have been determined between 1986 and 2019. This consisted of carrying out the ratio between the difference in areas and the initial areas for each period.
In a second step, we moved on to an in-depth analysis, evaluating the changes that took place within each land use unit taken in isolation. This analysis is done by calculating the rate of change (T c ) or average annual rate of spatial expansion, commonly used in studies of land cover change [17] [18]. These rates of change are evaluated from the formulas of Equations (1) and (2) below: T g = [(S2 -S1)/S1)] × 100 (1) T c = [(S2/S1)1/t -1] × 100 (2) where: T g = overall rate of change (%). T c = rate of change (%). S1 = area of the class at date t1. S2 = area of the class at date t2 (t2 > t1). t = number of years between the two dates.
Analysis of the rate of change values shows that positive values indicate "progression" and negative values "regression". Values close to zero indicate that the class is relatively "stable".

Simulation of the Dynamics of Land Use
Simulation consists of putting a model into action. In this study, the CA-Markov Markov transition zone file name, created by MARKOV, was inserted. Then the name of the group file which lists the compatibility images of the transition has been inserted (file containing the aptitude cards). The value 5 × 5 was used as an adjacency filter.

Model Calibration and Validation
Simulating the dynamics of land use on the horizon (2050) requires first the calibration of the model previously constructed with known data. For this, the image of 2019 being the most recent, was retained. It will be the subject of a first simulation-test, calibrated by two earlier dates (1986 and 2002). The images from 1986 and 2002 are used as a basis for extrapolating the states of future land use. This is a linear extrapolation, as the simulation is based on two points in time to calibrate the model. According to [21], calibration is the estimation and adjustment of the parameters and constraints of the model in order to improve the fit between the outputs of the model and a set of data. This step is fundamental, because the quality of the results obtained will depend on the correct configuration of the model. For validation, the result of the 2019 land use simulation is compared to the 2019 land use map resulting from the classification.

Results
The overall evolutionary trends observed in the dynamics of land use from 1986 to 2019 are represented by Figure 2 and Figure 3. Analysis of these figures indicates a decrease in the surface areas of water and forests, which respectively go from 0.63% to 0.29%; from 76.22% to 6.73%, and an increase in the areas occupied by degraded forests, built-bare ground and crops and fallows, which rose respectively from 18.73% to 26.27%; from 1.87% to 7.95% and from 2.56% to 58.76%.   Table 3.
Analysis of this table shows that annual decreases of 12.53% and 14.93% were observed for water areas and forest areas respectively. We also note an annual average increase of 1.83% and 9.48% respectively in the areas of degraded forests and built-bare ground as well as a strong annual average increase of 21.98% in the areas of crops and fallows of the study area.  Figure 5.
The visual analysis of the result of this simulation indicates that crops and fallow land and bare buildings will have a very high growth rate. Degraded forests will experience a strong decrease towards their disappearance. Forests and water will be disappearing as shown in Figure 6.
The land use simulation indicates an increase in built-bare ground and areas of crops and fallow respectively from 7.95% to 30.27% and 58.76% to 61.50% over the period from 2019 to 2050, while water, forests and degraded forests will experience a decrease, respectively, of up to 0.05%, 1.01% and 7.17% over the same period.

Discussion
The analysis of the dynamics of land cover and land use brought out the different evolutionary processes that took place within the landscape during the period from 1986 to 2019.    In the Lobo River Watershed upstream of Nibéhibé, rainfall has been decreasing like the African humid tropics since 1960 as observed by [22]. This decrease in rainfall could explain the decrease in water surfaces observed in the study area, which fell from 0.63% in 1986 to 0.29% in 2019. The increase in the area of degraded forests, which went from 18.73% in 1986 to 26.27% in 2019, comes from forest deforestation. Agriculture and bush fires, practiced in an anarchic manner in the area, lead to a sharp increase in cultivation and fallow areas, which rose from 2.56% in 1986 to 58.76% in 2019 and largely contribute to the reduction of forest surfaces and depletion of flora. There is also an increase in built-bare ground which went from 1.87% in 1986 to 7.95% in 2019, due to the increase in the population as predicted by the [23]. Indeed, after one or two years of food crops, the peasants abandon the cultivated plots to colonize new, more fertile forest lands, leaving the former fallow in order to restore their fertility. projections made according to [23] predict a population increase of 2.7%. In densely populated regions, all available space will be converted to agriculture in the near future. This will result in a high rate of deforestation in these areas and, the small forests areas remaining will certainly be converted into agricultural land. However, the best way to slow the rapid changes in land use is to propose a method of managing bush fires and an adapted intensification of agricultural production, i.e. an increase in yield per unit of land area, accompanied by sustainable planning of land use.

Conclusions
In the light of this work, we note that the processing of the images made it poss-