Vegetation structure mapping with remote sensing time series: Capabilities and improvements

Abstract

Mapping vegetation structure units is a critical task for biodiversity inventories, modeling, and conversation studies. In a larger framework, global change studies including climate modeling or land surface change detection demand detailed vegetation maps for improved results. This study uses an existing vegetation map of Namibia with 23 vegetation types and applies classification trees to multispectral MODIS time series. The results of the re-mapping show a general agreement with the initial map. However, significant shifts of vegetation type boundaries were detected for the Western Highland, Thornbush and Highland Shrubland and the Southern Kalahari. In addition, the fuzzy classification results indicate broad ecotones, in particular for all Kalahari vegetation types. Temporal plots of the NDVI for each vegetation type provide further information on vegetation patterns in Namibia