Water Security Evaluation of Long-Term Operation of a System of Reservoirs Facing Climate Change

Código

I-EBHE0041

Autores

Renan Vieira Rocha, Ályson Brayner Sousa Estácio, Gabriela Pinheiro Feitosa, Clebson do Carmo Raimundo, Jorgiane Pires Bezerra

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

A critical question in the context of climate change is whether we can successfully supply water for economic activities and domestic uses in future scenarios. One way to address this question is to couple bias corrected precipitation from the Coupled Model Intercomparison Project (CMIP) with hydrological models, operate the reservoir network with the resulting streamflow series and analyze the impact on key values used for water management, such as the Q90 value, commonly used to base long term water grant decisions. The Q90 value represents the outflow that a reservoir can sustain without failing 90% of the time and reflects the long-term water availability. Comparing Q90 with demand scenarios, we can evaluate water security and the feasibility of new infrastructure or economic activities. While this methodological framework has been used in previous studies, there has been a lack of deep analysis of the representation of the statistical properties of observed streamflow series. When dealing with reservoir operation, the variability of the series is a critical aspect with a large impact in the resulting Q90 values. Our results indicate that the coupled approach diminishes the possible climate change impacts. The CMIP6 coupled resultsfor the reservoir system of the Banabuiú basin (Ceará, Brazil) showsthat even for a model in which the SSP scenario imposes a significant reduction in the mean annual streamflow value (-35%) the resulting Q90 was almost 50% higher than current values since the coupled model streamflow series fails to reproduce the high interannual variability, presenting a coefficient of variation (CV) of just 0,55, compared to the expected value of 1,65. The misrepresentation of local variability is also observed in the streamflow series of the historical period, therefore the variability reduction cannot be seen as a possible feasible scenario. To overcome this issue, we propose the use of synthetic streamflow series to generate scenarios. To ensure correct representation of the spatial correlation, we generated annual series with the use of a multivariate gamma distribution, constructed from the univariate marginal distributions and the use of normal copulas, and converted to monthly series using the method of fragments. The univariate distributions were fitted using the method of moments and the scenarios were generated modifying the mean value according to the percentual changes indicated by the coupled model streamflow series. We generated 5000-years series to accommodate the interannual variability and account for the region?s decadal variability. The methodology produced spatially coherent series and feasible scenarios of Q90 and can be a used in any other regions whereas the coupled approach cannot reproduce the local variability. For future works we recommend improving the methodology to ensure the right reproduction of the variability in different timescales and the analysis of different bias correction approaches, in this work the bias was removed through quantile mapping.