Analysis of soil organic carbon and vegetation cover trends along the Botswana Kalahari Transect

Determination of trends in soil organic carbon (SOC) and vegetation cover along savanna ecosystem moisture gradients is critical to the understanding of ecosystem functioning and global change. Field results from 57 sites along the Botswana Kalahari Transect (BKT) showed general increases in both SOC and vegetation cover components along the temperature/moisture gradient. However, details in both SOC and woody cover trends revealed an area of relatively declining values in the central Kalahari. Image classification using TM single band data as input values confirmed the existence of an anomalously low woody cover area in the north-central Kalahari, within an area mapped as northern Kalahari tree and shrub savanna. It is postulated that the occurrence of such a zone of low shrubs in a gradient of otherwise increasing tree height and structural complexity may result mainly from edaphic factors, because of the co-incidence of near surface calcrete. However, the effects of fire and heavy grazing also influence the extent of low above-ground biomass, at least locally. Attempts were made to examine the extent to which soil organic carbon and vegetation components can be predicted using Thematic Mapper (TM) imagery, to increase potentially the number of sites in the BKT database. As a result of correlation and regression analyses, it was determined that the imagery-based SOC indexes of the literature were not useful in the Kalahari environment because of the extremely low SOC values encountered. Results of multiple regression analysis confirmed that woody vegetation cover could be predicted mainly through the use of TM3 single band radiance values throughout the BKT even during a non-drought period. Woody vegetation cover could be predicted to a lesser extent by the Transformed Normalised Vegetation Index. As high correlation coefficients were found between woody vegetation cover and SOC, the density of woody cover as derived from TM3 data may be taken as a surrogate for relative SOC abundance. Because this method is indirect, field checking will be required to verify results.