Answers are grounded in this paper and cite their sources.
Water security in Southern Africa is increasingly constrained by climate variability, rising atmospheric demand, land-use change, infrastructure regulation and fragmented hydrological information systems. This article develops and critically evaluates a data-driven database framework for water availability assessment using two contrasting hydrological environments: the Kafue Catchment System in Zambia and the Jonkershoek Experimental Catchments in South Africa. The framework, rather than a calibrated suite of predictive models, is the principal contribution of this article; the proposed framework integrates river discharge, reservoir and floodplain observations, meteorological variables, remote-sensing products, terrain and soil layers, land-use data, infrastructure information and socioeconomic indicators within a reproducible database architecture. To demonstrate the framework in practice, Random Forest (RF) baseline streamflow models were implemented for all four study catchments, the Kafue Catchment System (Zambia) and the three Jonkershoek Experimental sub-catchments (J1 Fynbos, J2 Pine and J3 Mixed, South Africa), using a common training period (January 1995–December 2017, n = 273 months) and an independent test period (January 2018–December 2024, n = 84 months). The Kafue model achieved NSE = 0.989 and KGE = 0.966, with seasonal regime and reservoir level as dominant predictors. J3 Mixed (NSE = 0.851, KGE = 0.919) and J2_Pine (NSE = 0.779, KGE = 0.858) achieved satisfactory performance, driven by temperature, NDVI and humidity. J1 Fynbos recorded NSE = 0.116, reflecting the process complexity of small fynbos catchments and pointing to the need for physics-informed predictors at that site. All four RF models outperformed their respective persistence baselines. An illustrative Composite Drought Index for the Kafue Flats identified eight high-risk months during 2018–2024, clustered in 2018 and 2022–2023. The analysis confirms that a harmonised database provides a strong foundation for streamflow prediction, floodplain monitoring and drought-risk assessment. Scientific and operational value depends on data provenance, quantitative quality metrics, governance alignment and appropriate model selection. Comprehensive multi-catchment calibration, independent validation and stakeholder co-design of dashboard outputs are identified as priorities for future work.
No takes yet. Share an insight, caveat, or question.
Wankie et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: