GIS and Logit Regression Model Applications in Land Use/Land Cover Change and Distribution in Usangu Catchment

Abstract

This study applied time series analysis to examine land use/land cover (LULC) change and distribution in Usangu watershed and multinomial logistic regression in the GIS environment to model the influence of the related driving factors. Historical land use/cover data of the watershed were extracted from the 2000, 2006 and 2013 Landsat images using GIS and remote sensing data processing and analysis techniques. Data was analyzed using ArcMap 10.1, ERDAS Imagine, SPSS and IDRISI Selva software. Eight factors likely to influence LULC change and LULC distribution were assessed. These include elevation, slope, distance from roads, distance from rivers networks, population density, Normalized Vegetation Index (NDVI), annual rainfall and soil types. Results show that LULC changes are mainly influenced by variations in annual rainfall, population density and distance from road networks. LULC distribution is determined mainly by terrain and edaphic factors namely elevation, slope and soil types. NDVI does not influence LULC change nor determine the LULC distribution, but can be used to show concentration of LULC types on a landscape. Combination of GIS, remote sensing and statistical analysis capabilities are powerful tools for assessing and model processes of land use change and their underlying causes in terms of time and space. It is concluded that ingeniously integration of remote sensing, GIS application combined with multi-source spatial data analysis give great possibility of quantifying and explaining the temporal and spatial LULC changes and distribution in a given watershed.