Diversions amount to 74 m3/s in the Baseline scenario, which is small compared to the evaporation losses from reservoirs and wetlands. However, diversions increase to 179 m3/s in the Moderate development scenario, and to 564 m3/s in the High development scenario. This means that irrigation levels under the High development scenario have a similar magnitude as evaporation losses that are already occurring from existing reservoirs. Under this scenario mean annual discharge decreases by −18% as compared to the Baseline scenario. 87% of the irrigation demand (Table 3) can be met by the simulated diversions (Table 5). Similar percentages are obtained in the see more Moderate development and Baseline scenarios

– albeit with much lower diversion amounts. Shortages for meeting irrigation demand occur when reservoir water levels fall below minimum operation levels. This situation occurs at Zimbabwean tributaries under all scenarios, but also in dry years at Kariba reservoir under the High development scenario. It is clear

that an implementation of irrigation projects will cause a decrease in discharge due to increased diversions. The impact of future climate is less clear, though. Contrasting results are obtained for the scenarios based on climate data of GCMs. For the near future (2021–2050) the scenario based on CNRM climate data projects an increase in discharge of +10%, whereas MPI projects a decrease of −14%. These differences

are even larger for the far future BIBF-1120 (2071–2100), with projected changes of +14% versus −18%. To disentangle the effects of changes in precipitation and temperature the last four scenarios listed in Table 5 present assessments for changes super-imposed on historic climate (delta-change approach). If temperature increases by +4 °C then discharge decreases by −16%. An even larger decrease in discharge of −32% is obtained for a reduction tuclazepam of precipitation by −10%. An increase in precipitation by +10% results in an increase of discharge by +43%. The percentage changes in mean annual discharge are not evenly distributed during a year, as evident in an analysis of seasonality in discharge (Fig. 10, top left). By far the largest differences to the Baseline scenario are obtained with the Pristine scenario, with a more pronounced seasonality. The main reason is that the Pristine scenario does not include any reservoirs. The reservoir operation results in a strong attenuation of the seasonal flood peak and an increase of discharge during the dry period. This is even clearer when analysing the distribution of flows (Fig. 10, top right). In the Pristine scenario high flows are increased, but low flows are much lower, even though the mean annual discharge is larger. For the High development scenario the magnitude of changes in seasonality and distribution of discharge are considerably smaller than for the Pristine scenario (Fig.