Long -Term Vegetation Dynamics over the Bani River Basin as Impacted by Climate Change and Land Use

Abstract

This study investigated the long-term trends in vegetation and rainfall and the extent and rate of vegetation change over the Bani river Basin at multiple spatial and temporal scales in relation to local and regional drivers. Monthly 8-km Normalized Difference Vegetation Index (NDVI) timeseries data from 1982 to 2011 was derived from 10-day Satellite Pour l’Observation de la Terre vegetation product (SPOT-VGT) at 1-km (1998-2011) and 15-day GIMMS (Global Inventories Monitoring and Modelling Systems) at 8-km satellite data (1982-2006). Gridded rainfall data at 8-km grid resolution was created from 40 meteorological stations and complemented with Tropical Rainfall Measurement Mission (TRMM) data. A Mann Kendall (MK) trend analysis was used to determine the trend for each dataset using monthly and annual time-series. This analysis produced some indicators like Kendall’s tau, p-value and Theil-Sen. The p-value estimator (p-value less than 0.07) was used in this study to show the significance of the trend. Trend analysis revealed that within the study area vegetation greening trends are mostly associated with areas where natural vegetation is still well represented. From the results 934 pixels (49% of the study area) showed a positive trend while 155 pixels (8% of the study area) showed a negative trend significant at p-value less than 0.07. During the same period rainfall had increased by about 17 mm, translating into a positive trend for almost the entire study area. Vegetation productivity in the study area is dependent on rainfall which varies greatly temporally and spatially. The linear Pearson correlation was used to estimate the relationship between NDVI and rainfall for every pixel at monthly interval for the growing season data. Comparing their long-term mean the result showed a good correlation between the two datasets with an R value of 0.98. Four (4) reference areas were used to explain and cross verify representative areas that exhibit either entirely negative MK-trends or entirely positive MK-trends over the monitoring period. These reference areas were selected based on their trend in rainfall and NDVI and their NDVI long-term departure. Free 30-meter Landsat images were acquired for the four reference areas for the following three intervals: 1984 and 1986, 1999 and 2000 and 2009 and 2010. Land Use/Land Cover (LULC) change was then quantified and the rate of land conversion was determined. LULC variables included urban, Cropland and natural vegetation (Shrublands, Steppe, Open Trees and Closed Trees). For the entire period, the class ‘Natural Vegetation’ decreased between 22.83% and 63.47% from its initial area for areas (1) and (2), while the iv decrease was 8.35% for area (3) and 13.39% for area (4). The class ‘Cropland’ increased for 564.86% in area (3); 62.17% in area (4); 35.79% in area (2) and 16.22% in area (1). To investigate whether there is a relationship between NDVI, rainfall and LULC change, LULC variables were correlated with long-term trend in rainfall and NDVI. The results showed there is a positive correlation between increases in rainfall and some land cover classes, while some classes such as settlements were negatively correlated with vegetation productivity trends. Croplands and Natural Vegetation were positively correlated (r=0.89) with rainfall while settlements have a negative correlation with NDVI time series trends (r=-0.57). Despite the fact that rainfall is the major determinant of vegetation cover dynamics in the study area, it appears that other human-induced factors such as urbanisation have negatively influenced the change in vegetation cover. The results provide spatially explicit and temporally good and rich information of vegetation productivity dynamics and its drivers at landscape scale. This is an important input for assessing the impact of climate change on vegetation for biophysical modelling. It also improves our knowledge of the drivers of vegetation productivity changes. The study suggests that NDVI can be useful for general vegetation cover monitoring and planning. Future studies need to also look at the effect of vegetation cover change in regard to other landscape components such as specifically population density and soil degradation