Large-scale vegetation assessments in southern Africa: concepts and applications using multi-source remote sensing data

Abstract

The interdisciplinary project structure of BIOTA Southern Africa opened the opportunity for applying integrated concepts for the spatiotemporal assessment of arid and semi-arid southern African ecosystems, including the characterization of inter-annual vegetation dynamics and large-scale land cover mapping. Due to existing high uncertainties in mapping arid and semi-arid environments, the studies on remote sensing-based vegetation mapping aimed to develop and apply land cover classifi cation techniques to derive adapted and standardized maps covering large areas along the BIOTA transect. The application of machine learning classifi cation and regression techniques proved to be useful for both fractional and categorical semi-arid land cover mapping. Key improvements were achieved by mapping vegetation types in Namibia on a national scale using time series data from the Moderate Resolution Imaging Spectroradiometer (MODIS). Synergies of multitemporal remote sensing and botanical fi eld surveys yielded a fl exible vegetation type map for the northeastern Kalahari in Namibia based on the United Nations (UN) Land Cover Classifi cation System (LCCS). The development of fractional land cover maps in north-eastern Namibia showing the percentage cover of woody vegetation, herbaceous vegetation, and bare land surface allowed for a realistic and accurate spatial description of complex and fi ne-structured semi-arid vegetation types. Time series of MODIS vegetation indices were used to map and analyse annual and inter-annual vegetation dynamics along the BIOTA Observatory transects.