Developing a methodology using multi spectral remote sensing data for mapping vegetation change - a key variable in soil erosion mapping

Abstract

This project investigates the use of multispectral remote sensing data for developing vegetation change detection methodology and creation of Geographic Information System (GIS) layers in an automated manner. The vegetation change detection GIS layer has the potential to be used in modeling the risk from soil erosion in Botswana, Africa. Soil erosion is widespread in eastern Botswana and adversely affects the rangeland where livestock is grazed and arable lands are used for crop production. Since there are no spatial and temporal data sets which explain the distribution of soil erosion in Botswana there is a need to develop an automated methodology to derive such GIS layers in near real time. The GIS layers can then be used for several reasons such as a) understanding the spatial and temporal distribution and b) modeling risk from soil erosion. The main objective of this paper is to develop a methodology to map changes in vegetation cover and to generate GIS layers in an automated manner. Standard image analysis and GIS routines were performed on time series multispectral landsat TM datasets in order to detect the changing vegetation in three sample study sites in Queensland Australia. Results from three different methods used for mapping changes in the distribution of vegetation from 1988-2004 clearly shows the potential of this methodology to be used in eastern Botswana for mapping changes in vegetation. This approach to map vegetation changes can prove useful in Botswana where there are no spatial and temporal datasets for showing vegetation changes. This methodology may also prove useful in automated mapping of many GIS layers that influence soil erosion and assist in modeling the risk from soil erosion in Botswana.