Implementation and sensitivity analysis of a dam-reservoir model over Spain

Water resources are considered to be a major challenge for the coming century, particularly in the context of climate change and increasing demographic pressure. Water resources are directly linked to the continental water cycle and its processes are mainly described by hydrological models. ISBA-CTRIP, developed at the CNRM, is an example of a coupled land surface - river routing model used for this purpose. However, anthropogenic impacts on water resources, and in particular the effects of dams-reservoirs on river flows, are still poorly known and generally neglected in global hydrological models, including ISBA-CTRIP. This study focuses on the improvement of the CTRIP river routing model, recently upgraded to 1/12° resolution, by integrating the effects of man-made reservoirs. This work is in preparation for the upcoming SWOT mission, which will provide the data necessary to make improved global scale river and reservoir storage and flow estimates.

A parameterized reservoir model was developed based on Hanasaki's scheme (Hanasaki et al., 2006). The model differentiates between irrigation and non-irrigation reservoirs, computes the mass balance in the reservoir and calculates monthly releases based on inflows and water demands. Using a first default parameterization, the model is run on the highly anthropized river basins in Spain. An operating rule is determined for each of the 215 largest reservoirs and simulated outflows and water storage variations are evaluated against in situ observations over the overall period 1979-2014. Results reveal the positive contribution of the model in representing the seasonal cycle of discharge and storage variation, specifically for irrigation large-storage capacity reservoirs as the model succeeds in reproducing the seasonal shift between inflows and outflows caused by irrigation management rules. The Nash-Sutcliff Efficiency (NSE) median index for discharge was 0.68, which corresponds to an outflow representation improvement of 28%, if compared to the naturalized representation of river flows. For irrigation reservoirs, the improvement rate reaches 67% in the median. 

An exhaustive sensitivity analysis regarding the 7 parameters of the model was conducted on the performance of an NSE bounded version on outflows using the Sobol method. Following Saltelli's approach, sampling is performed using the probability density functions defined for each parameter input, and first-, second- and total-order Sobol indices are estimated. The study is carried out separately on irrigation and non-irrigation reservoirs. It is shown that the most influencing parameter is the threshold coefficient describing demand-controlled release level : in the median, ~54% and ~80% of the total variance, respectively in the two reservoir categories, is assigned to this parameter alone. On the other hand, parameters specifying the ideal reservoir filling level and the minimum release have less influence on monthly long-term mean outflows variance. The second-order Sobol indices revealed several interactions between parameters and explained the observed bias between first- and total-order indices.

The results highlight the importance of incorporating reservoir operation in large scale hydrological models and represent a very useful step to further improve river flow modeling, through calibration schemes and SWOT data assimilation, by targeting the most influencing reservoir model parameters.