Regional prediction of wetland degradation in South Africa

Abstract

Wetland assessments in South Africa continue to be undertaken at the scale of individual wetlands, while ultimately a landscape assessment is required for broader regional planning. The aim of this research was to develop probability models of degradation per wetland type using landscape-scale predictor variables. Condition data for a total of 463 seep, floodplain and valley bottom (channelled and unchannelled) hydrogeomorphic (HGM) units was combined with data for 21 potential explanatory variables for land cover, infrastructure and catchment physiography. Multiple logistic regression models to estimate the probability of degradation per HGM type were investigated at three different spatial levels (1,000 m buffer radius; 100 and 1,000 km2 scales). Models at the intermediate scale were most useful and only these are reported. Percent natural vegetation, road density, population density and altitude, together with anthropogenic land use (cultivation and forestry plantation area) were important predictors of wetland condition, and differed per HGM type. Regional patterns of degradation within quaternary catchments in KwaZulu-Natal were dissimilar for each HGM type. There is potential to refine and extend the models for application at a national level, and to predict actual condition scores using ordinal logistic regression based on larger datasets.