Data Construction and Spatiotemporal Trend Attribution of Runoff over the African Continent (1981–2016)

Abstract

The availability of streamflow records in Africa has been declining since the 1980s due to malfunctioning gauging stations and data collection failures. Africa also has insufficient hydrological information owing to the allocation of few resources to research efforts. Unreliable runoff datasets and large uncertainties in runoff trends due to climate change patterns and human activities are major challenges to water resource management in Africa. Therefore, this study aimed to improve runoff estimates and to assess runoff trend responses to climate change and human activities in Africa during 1981–2016. Using statistical methods, monthly gridded runoff datasets were generated for the period of 1981–2016 from a modified runoff curve number method calibrated with river discharge data from 535 gauging stations. According to the cross-validation results, the constructed runoff datasets comprised the Nash and Sutcliffe coefficients ranging from 0.5 to 1, coefficients of determination ranging from 0.5 to 1 and percent biases between ±25% for a large number of stations up to 73%, 80% and 91% of the 535 gauged catchments used as references. Analysis of runoff trend responses to climate change and human activities revealed that land cover change contributed more (72%) to the observed net runoff change (0.30%•a−1) than continental climate changes (28%). These contributions were results of cropland expansion rate of 0.46%•a−1 and a precipitation increase of 0.07%•a−1. The performance and simplicity of the statistical methods used in this study could be useful for improving runoff estimations in other regions with limited streamflow data data. The results of the current study could be important to natural resource managers and decision makers in terms of raising awareness of climate change adaptation strategies and agricultural land-use policies in Africa.