MULTI-TEMPORAL LAND COVER MAPPING OF THE KAKAMEGA FOREST UTILISING LANDSAT IMAGERY AND GIS

Authors

  • D. Kuria Department of Geomatic Engineering and Geospatial Information Systems, Jomo Kenyatta University of Agriculture and Technology, Nairobi
  • E. Mutange Department of Geomatic Engineering and Geospatial Information Systems, Jomo Kenyatta University of Agriculture and Technology, Nairobi
  • D. Musiega Department of Geomatic Engineering and Geospatial Information Systems, Jomo Kenyatta University of Agriculture and Technology, Nairobi
  • C. Muriuki Department of Geomatic Engineering and Geospatial Information Systems, Jomo Kenyatta University of Agriculture and Technology, Nairobi

Keywords:

Land-cover mapping, image classification, change detection

Abstract

Forest resources contribute significantly to the Kenyan economy. However, due to pressures exerted by the growing population, this scarce resource is seriously endangered. In particular, the Kakamega Forest has experienced serious degradation in the past, though some restoration efforts have also been put in place. In this research, we utilise time series Landsat imagery to characterise the changes and capture the trends in land cover changes. Three epochs are utilised, namely, 1986, 1995 and 2005. Pre-processing involved georeferencing and radiometric corrections. As a first step the time series imageries were evaluated via a threshold analysis distinguishing between ‘forest’ and ‘non-forest’. Subsequently, a supervised multispectral classification was performed distinguishing various land cover classes. Ground truthing for the historical imagery was done using aerial photographs, topographic maps and site visits. Actual land cover verification was based on amateur photographs taken in 1999 from an aircraft, and ground observations in 2008. For classification the maximum[1]likelihood decision rule was applied considering bands 3, 4, 5, 7 plus 7/2 for thematic mapper (TM)/enhanced thematic mapper plus (ETM+) imagery and 1, 2, 3 and 4 for Multi-spectral scanner (MSS) data, respectively. The classification results form a solid basis for a consistent and detailed evaluation of forest history between 1986 and 2005. Analysis results presented include graphs and pie charts of change in land cover class areas over time as well as such allowing for true change detection with transitions between the different classes. In this study, maximum likelihood supervised classification change detection techniques were applied to Landsat images acquired in 1986, 1995 and 2005 respectively. To map land cover changes in kakamega forest, a supervised classification was carried out on the six reflective bands for the three images individually with the aid of ground truthing data. Changes among different land cover classes were assessed. During the study period, a very severe land cover change had taken place as a result of agricultural and settlement. These changes in land cover led to vegetation degradation. The effects of restoration efforts are also captured in the research findings.

Downloads

Published

2010-01-10