**2. Materials and methods**

#### **2.1 Study area**

The Mono Transboundary Biosphere Reserve (MTBR) is located at the southern border between Benin and Togo (6°8′52.8″–7°3′41.8″ North latitude and 1°24′18.2″– 1°30′0.0″ East longitude) and covers an area of 345.22 km2 (**Figure 1**). The reserve is located in the Dahomey gap, a corridor characterized by mosaics of dense semideciduous forests, Guinean savannas, marshy meadows, marshes, mangroves, and water plans, and mosaics of crops and fallows [16]. The reserve is characterized by a tropical humid climate with a succession of four seasons per year, two dry seasons (November to March and July to September) and two rainy seasons (March to July and September to November). Rainfall varies between 850 mm and 1250 mm per year, with an average monthly rainfall of about 222.57 mm during the long rainy season and 88.30 mm during the short rainy season (October). The average maximum temperature is 31.25°C between December and April, and the minimum temperature

**Figure 1.** *Location of the Mono Transboundary Biosphere Reserve.*

is 28° C between July and September. The area is home to different ecosystems (marine, terrestrial, and lagoon). The Mono River is the main river around which the reserve is built. Approximately two million people are riverine to the reserve, with 80% depending largely on the ecosystem services provided by the reserve (small-scale farming, small-scale fishing, and exploitation of wood and charcoal [22].

#### **2.2 Land use land cover (LULC) maps**

Two data sources were used to establish the baseline and the analysis of the LULC dynamics of the reserve. These include two Landsat satellite images (TM (1986) and ETM + (2000)) and a Sentinel 2A satellite image (2015 being the reference year). The scenes were chosen during the dry season with low cloud cover [23]. The interpretation was aided by additional data sources including the administrative maps of Benin and Togo, the GPS data from the field, and Google Earth Pro.

The radiometric correction of Landsat TM, ETM +, and Sentinel 2A images was used to correct for atmospheric bias and change from pixel value to digital count as a reflectance value. This operation is completed by mosaicking the two Sentinel 2A image scenes in order to obtain a single scene that can be used to extract easily the study area. The color composition of bands 4-5-7 of TM and ETM + images and bands 4-3-2 of Sentinel 2A image was chosen by selecting training sites because they present the best discrimination of LULC types [24]. About 100 plots representing all types of LULC chosen according to their spatial distribution on TM and ETM + color compositions of bands 4-5-7 and bands 4-3-2 of Sentinel 2A image were identified and delineated.

The spectral properties were used to classify the different LULCs of the image into thematic classes based on the supervised classification (due to a good knowledge of the study area) using the maximum-likelihood algorithm. The accuracy of the classifications was evaluated using a confusion matrix or contingency table obtained from field truth data and a representative of each LULC class. The validation of the classification was based on the calculation of two indices: the overall accuracy (the proportion of well-ranked pixels in percentage) and the Kappa index (the ratio between the well-ranked pixels and the total pixels surveyed) [25]. In addition, the field truth data were used for validation.

#### **2.3 Land use/land cover dynamic analysis**

Quantitative analysis of changes over the entire study period was carried out in order to identify the different changes in LULC classes based on change detection matrix resulting from the comparison between the pixels of the classifications of two dates [26]. This analysis was done by calculating the rate of change (Rc) used in LULC studies [27, 28] as follows:

$$Rc = \left[ (\mathbf{S2}/\mathbf{S1})\mathbf{1}/\mathbf{d} - \mathbf{1} \right] \times \mathbf{100} \tag{1}$$

(where: Rc = rate of change (%); S1 = area of the LULC class of the date d1; S2 = area of the class of the date d2 (d2 > d1) et d = number of years between the two dates). Positive values indicate a "progression," whereas negative values indicate a "regression." Values close to zero indicate a relative "stability" of the class.

The average annual rate of forest degradation [29] was evaluated using the following formula:

*How Far the Mono Transboundary Biosphere Reserve Protects Biodiversity in the Dahomey-Gap… DOI: http://dx.doi.org/10.5772/intechopen.112884*

$$ARD = \left(\mathbf{S2} / \mathbf{S1}\right) / \mathbf{d} \times \mathbf{100} \tag{2}$$

(where ARD = average annual rate of degradation (%); S2 = Total area of forest lost; S1 = Initial area of forest and d = number of years between the two dates).

The transition matrix was elaborated by superposing the LULC maps of 1986, 2000, and 2015 with the "Intersect polygons" algorithm of the Geoprocessing extension using ArcGIS 10.0. The transition matrix was used to highlight the different changes in LULC between two dates [30]. The matrix values were standardized 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 method [31].

#### **2.4 Futures scenarios**

The standard annualized transition matrices were used to further predict the proportion of each land cover class at any one time based on a Markovian chain model. Two different scenarios were assumed corresponding to each of the two Markovian matrices (1986–2000 and 2000–2015). The area expected for 2015 scenarios based on the 1986–2000 period was compared with the area of 2015 from the 2015 map using a chi-square (χ<sup>2</sup> ) test for the model validation.

### **3. Results**

#### **3.1 Land cover maps**

The results of the processing images of the year 2015 indicated an overall good accuracy (89.84%) and an estimated Kappa index of 0.88 with nine LULC units (**Table 1**; **Figure 2**) including forest, savannas, mosaic of crops and fallows, wetlands, plantations, urban agglomerations, and bare soil. Forests were composed of dense semi-deciduous forests, woodland, and gallery forests with an area of 15,740.91 ha


#### **Table 1.**

*Land use land cover classes in the Mono Transboundary Biosphere Reserve in 2015.*

#### **Figure 2.**

*Reference situation map of LULC in the Mono Transboundary Biosphere Reserve in 2015.*

(4.55% of the reserve); these ecosystems were in the form of fragmented islands dispersed within the reserve.

Savannas on drained soil had an area of 88,917.48 ha (25.71% of the reserve), holding tree and shrub savannas. Wetlands covered an area of 124,139.88 ha (35.90% of the reserve) and included mangroves, floodplain savannas dominated by *Mitragyna inermis*, marshy meadows, and water. The majority of wetlands and their associated plant communities were mostly located in the southern half of the reserve. Within these wetlands, mangroves that constitute particular ecosystems occupied an area of 25,941.87 ha (7.50% of the reserve area).

Plantations with an area of 27,113.13 ha (7.84% of the reserve) were composed of *Tectona grandis, Khaya senegalensis, Eucalyptus* sp., *Elaeis guineensis,* and *Cocos nucifera*. Mosaic of crops and fallows with a total area of 80,599.05 ha (23.31% of the reserve) consisted of areas of crops and areas previously cultivated and abandoned (fallows) or invaded by exotic species.

Urban agglomerations and bare soil with an area of 9249.03 ha (2.67% of the reserve) included towns and villages and areas with very low vegetation cover, including quarries (sand and gravel) and rocky outcrops.
