Land cover change analysis within the GLOWA Volta basin in West Africa using 30-meter Landsat data snapshots

Kurzfassung

Data on land cover and land cover change can be effectively inferred from multi-temporal satellite remote sensing data. Land cover change data is imperative for regional hydrological models and water budget estimates within large river basins or catchments. Especially in Africa, however, rigorous and standardized land cover change data is missing to parameterize hydrological models that are also used in a predictive capacity in the context of global change models or studies. Within the GLOWA Volta hydrological project in West Africa, land cover change was mapped for the Volta River basin, an area of 400.000 km2. We used 26 Landsat tiles from 1990 and from the period 2000 and 2001 correspondingly. A unique contextual fuzzy convolution filter was used to extract the linear land cover classes, whilst non-linear classes were mapped using a neural network classifier respectively for both time frames. The Landsat tiles were further augmented with daily and well corrected 250-meter MODIS timeseries observations, for the period from 2000 onwards. Data on soil texture and digital elevation model variables on streamflow and pit areas were also used. For the land cover definitions we adopted the standard Land Cover Classification System (LCCS) legend from the Food and Agricultural Organization (FAO). The change was calculated for each thematic LCCS land cover code and socio economic data from Ghana was used to verify the change vector for selected ‘hot spots of change’ areas. Mean accuracy for the change matrix and for all classes, based on the thematic accuracy of two mapped time frames respectively, was calculated at 54 percent certainty. We found that most ‘negative’ change (18 percent of the total mapped land cover) occurred from woody shrub to herbaceous vegetation and from woody shrub to herbaceous managed respectively. The second most dominant change (15 percent) was from woody closed to sparse woody. The third most dominate change (10 percent) occurred from woody closed to herbaceous. The fourth most change (6 percent) occurred from woody shrub to regularly flooded herbaceous managed (managed wetlands). An overall of 2 percent of the total land cover recorded a ‘positive’ change from managed areas in 1990, being both herbaceous and woody managed, to woody shrub (not managed) in 2000/2001. The ‘positive’ areas occurred in southern Ghana and predominately around urban areas. From the socio economic data analysis in the ‘hot spot’ areas in north eastern Ghana, most ‘negative’ change could be validated as being caused by the expansion of agricultural areas. The second most change in the hot spot area was caused by the expansion of rural clusters or urban areas, which was, area wise, directly related to the thinning of woody cover between 1990 and 2000/2001. We found the rates of deforestation and land conversion to agricultural areas to be dramatic in the GLOWA Volta basin, especially within the woody open to sparse (savanna) biome. Land cover conversion and dynamics can be effectively mapped using 30-meter Landsat data and other moderate resolution optical data feeds in conjunction with socio economic data. Especially the changes within wetland areas and along river systems as well as the thinning of woody cover densities at these measured rates influence or alter the flow dynamics of water and water availability in the GLOWA Volta basin.