Fluvial flooding is a recurring phenomenon in Rwanda due to a combination of climate, topography and human interventions. This poses a hazard to human lives and infrastructure, particularly in low-lying areas adjacent to major rivers. An example of such areas is the Masaka wetland, located in the southern edge of the capital Kigali along the Nyabarongo river. As the city expands into the vicinities of this location, there is growing concern regarding infrastructure damage and human losses associated with extreme discharges. Simultaneously, potential for agricultural development, for which flood duration is a critical factor, has also been identified. Hence, there is a need for increasing the available knowledge on the seasonal pattern of discharges at this location.
However, significant changes in this behaviour may occur as a result of reservoir commissioning in the Nyabarongo river catchment. In particular, the Shyorongi reservoir in the scope of the Nyabarongo II Multipurpose Development Project will regulate approximately 47% of the catchment of the Masaka wetland. Hence, its operation may result in significant shifts on seasonal discharges and associated flooding in this location, which consisted of the main topic of investigation in this research. For this purpose, time series of inflows to the reservoir, as well as contributions from non-regulated areas, have been generated by separately-calibrated lumped rainfall-runoff models based on the FlexTOPO framework, the period from January 1981 to August 2018. The operation of the reservoir was approximated by a standard operating policy. Two operational scenario families were devised: one maximising hydropower generation and one prioritising the reduction in peak discharges by limiting releases to a pre-defined threshold. For each scenario family, different parametrisations (target electricity output and maximum release, respectively) were tested, comprising a total of 50 scenarios. The resulting discharge time series of each scenario was compared to the baseline (no-reservoir) scenario according to two criteria: reductions in flood durations and reductions flood peaks. While flood durations were assessed on a relative basis, the effect on flood peaks was quantified by the expected shifts in the exceedance probability curve. This was estimated by a partial duration series (i.e. Peaks-Over-Threshold) approach, which was implemented to estimate peak seasonal discharges associated with return periods ranging from 2 to 500 years.
On specific years such as 2018, operation emphasising hydropower production had the secondary result of reducing peak discharges by up to 20%. However, over the majority of the simulated period, reductions were found to be insignificant, as reflected in the limited shifts in the fitted probability distributions. This type of operation also had generally insignificant effects on flood durations, regardless of the electricity output (and associated release) target. In contrast, it was observed that reservoir operation with the specific goal of flood mitigation may consistently result in 20% to 25% reductions of seasonal peak discharges. However, this performance was found to depend, to a large extent, on the specified reservoir releases in comparison to inflow hydrographs, which vary over different years. Therefore, the availability of a seasonal outlook on discharges may guide the definition of optimal release rules for each year. Additionally, reservoir rules that account for the pronounced seasonal variability in inflows may allow for significant reductions on both flood peaks and durations in comparison to the policies hereby simulated, and it is recommended that such optimisation be carried out for the Shyorongi reservoir. On relatively wet years, a trade-off was observed between the minimisation of flood peaks and of flood durations at the Masaka wetland. This is linked to the increment in discharge at the tail of the hydrograph caused by reservoir emptying, which may trigger the exceedance of the relatively low bankfull discharge threshold at this location. However, the simplified implemented operation rules adopted in this study have a limited capacity to represent practical reservoir operation. Therefore, there may be considerable potential for the optimisation of reservoir operation to improve performance in both flood mitigation and hydropower generation scenarios. However, due to the limited storage capacity of this reservoir, flood management in the Masaka wetland requires the adoption of complementary structural and non-structural measures according to the desired targets.
It is highlighted that several assumptions have been made concerning the conceptualisation of the catchment and characteristics of the Shyorongi dam and the Masaka wetland, as detailed in the body of the report. Additionally, the estimation of parameters for hydrological models and extreme value distribution functions inherently involves uncertainties. Therefore, the results and associated interpretations hereby presented should be interpreted in the context of a considerable amount of propagated uncertainty, and should not be adopted for design purposes.