Regionalising a meso-catchment scale conceptual model for river basin management in the semi-arid environment

Abstract Meso-scale catchments are often of great interest for water resources development and for development interventions aimed at uplifting rural livelihoods. However, in Sub-Saharan Africa IWRM planning in such catchments, and the basins they form part of, are often ungauged or constrained by poor data availability. Regionalisation of a hydrological model presents opportunities for prediction in ungauged basins and catchments. This study regionalises HBVx, derived from the conceptual hydrological model HBV, in the semi-arid Mzingwane Catchment, Limpopo Basin, Zimbabwe. Fifteen meso-catchments were studied, including three that were instrumented during the study. Discriminant analysis showed that the characteristics of catchments in the arid agro-ecological Region V were significantly different from those in semi-arid Region IV. Analysis of flow duration curves statistically separated sub-perennial catchments from (sub-)ephemeral catchments. Regionalised parameter sets for HBVx were derived from means of parameters from the sub-perennial catchments, the (sub-)ephemeral catchments and all catchments. The parameter sets that performed best in the regionalisation are characterised by slow infiltration with moderate/fast "overland flow". These processes appear more extreme in more degraded catchments. This is points to benefits to be derived from conservation techniques that increase infiltration rate and from runoff farming. Faster, and possibly greater, sub-surface contribution to streamflow is expected from catchments underlain by granitic rocks. Calibration and regionalisation were more successful at the dekad (10 days) time step than when using daily or monthly data, and for the sub-perennial catchments than the (sub-)ephemeral catchments. However, none of the regionalised parameter sets yielded CNS ⩾ 0.3 for half of the catchments. The HBVx model thus does offer some assistance to river basin planning in semi-arid basins, particularly for predicting flows in ungauged catchments at longer time steps, such as for water allocation purposes. However, the model is unreliable for more ephemeral and drier catchments. Without more reliable and longer rainfall and runoff data, regionalisation in semi-arid ephemeral catchments will remain highly challenging.