Assessing Suitability of Modis Data for Land Cover Variability Mapping: A Case of Mt. Kilimanjaro

Abstract

This study envisioned on exploration of the potential application of Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance for land cover mapping and change detection at a local scale. The study examined Mt. Kilimanjaro land cover from 2004 to 2013. Support Vector Machine classification learning algorithm was adopted to map the land cover. Where by eight different land cover types namely, Savanna, Evergreen Forest, Mixed Forest, Settlements/Croplands/Natural Vegetation, Shrublands, Barren or Sparsely Vegetated, Snow and Ice, and Grasslands were deduced. The overall classification accuracies attained are 70.33%, 76.00% and 72.53% for 2004, 2008 and 2013, respectively. Generally, it has been observed that Snow and Ice covered a smallest area. Barren or sparsely vegetated show a decreasing linear trend from 2004 to 2013. Five of the land cover categories (i.e. Snow and Ice, Shrub lands, Mixed Forest, Evergreen forest and Grasslands) showed a “high- low- high†nonlinear change between the three studied epochs. The rest of LULC categories Settlements/Croplands/Natural Vegetation and Savannah showed a “Low- high- low†change. Finding indicates that land cover changes could be detected accurately at a local scale even when using medium resolution satellite imageries. It is recommended that causes of land cover variability should be investigated.

Keywords:MODIS surface reflectance, land cover mapping, land cover change, Mt. Kilimanjaro