Automatic design of basin-specific drought indexes for highly regulated water systems

Abstract

Abstract. Socio-economic costs of drought are progressively increasing worldwide due to undergoing alterations of hydro-meteorological regimes induced by climate change. Although drought management is largely studied in the literature, traditional drought indexes often fail at detecting critical events in highly regulated systems, where natural water availability is conditioned by the operation of water infrastructures such as dams, diversions, and pumping wells. Here, ad hoc index formulations are usually adopted based on empirical combinations of several, supposed-to-be significant, hydro-meteorological variables. These customized formulations, however, while effective in the design basin, can hardly be generalized and transferred to different contexts. In this study, we contribute FRIDA (FRamework for Index-based Drought Analysis), a novel framework for the automatic design of basin-customized drought indexes. In contrast to ad hoc empirical approaches, FRIDA is fully automated, generalizable, and portable across different basins. FRIDA builds an index representing a surrogate of the drought conditions of the basin, computed by combining all the relevant available information about the water circulating in the system identified by means of a feature extraction algorithm. We used the Wrapper for Quasi-Equally Informative Subset Selection (W-QEISS), which features a multi-objective evolutionary algorithm to find Pareto-efficient subsets of variables by maximizing the wrapper accuracy, minimizing the number of selected variables, and optimizing relevance and redundancy of the subset. The preferred variable subset is selected among the efficient solutions and used to formulate the final index according to alternative model structures. We apply FRIDA to the case study of the Jucar river basin (Spain), a drought-prone and highly regulated Mediterranean water resource system, where an advanced drought management plan relying on the formulation of an ad hoc "state index" is used for triggering drought management measures. The state index was constructed empirically with a trial-and-error process begun in the 1980s and finalized in 2007, guided by the experts from the Confederación Hidrográfica del Júcar (CHJ). Our results show that the automated variable selection outcomes align with CHJ's 25-year-long empirical refinement. In addition, the resultant FRIDA index outperforms the official State Index in terms of accuracy in reproducing the target variable and cardinality of the selected inputs set.

Climate Adaptation as a Control Problem: Review and Perspectives on Dynamic Water Resources Planning Under Uncertainty

“…Here the number of indicator variables is assumed to be a proxy for the complexity of the problem, since it will increase the size of the search space roughly exponentially depending on the policy structure. There may be opportunities to reduce the computational effort through aggregate proxy indicators that combine multiple sources of information while maintaining key dynamic signals (Zaniolo et al, ). Overfitting to training scenarios: As the problem complexity increases, the resulting policy should improve in validation, to a point. After that, the more complex policy will not generalize well to other scenarios.…”

Optimización del indicador de escasez en la cuenca del río Júcar

“…Pérez-Blanco y Gómez (2014) desarrollan una metodología estocástica para estimar la disponibilidad esperada de agua en la agricultura que resulta de las medidas de gestión publicadas en los PES de 2007. Zaniolo et al (2018), contribuyen con un marco para el diseño automático de índices de sequía para una cuenca hidrográfica. Elaboran un índice que representa un sustituto de las condiciones de sequía de una cuenca, combinando toda la información relevante disponible sobre el agua que circula en el sistema de recursos hídricos (RRHH) mediante un algoritmo de extracción de características.…”

Response to Reviewer #2

“…A drought is a slowly developing natural phenomenon that occurs in all climatic zones and can be defined as a temporary significant decrease in water availability (Tallaksen and Van Lanen, 2004;Van Loon and Van Lanen, 2012). Drought impacts can propagate to virtually every water-related sector, such as farming and livestock production, industry, power generation, and public water supply (Spinoni et al, 2016).…”

“…Moreover, it can be further worsened by an inadequate design and management of the water exploitation system and its operating rules (Mishra and Singh, 2010). Operational drought indicators aim at comparing water availability to human water needs and serve as a measure of water well-being, rather than a measure of natural fluctuation as in the case of meteorological, agricultural, and hydrological indicators (Sullivan et al, 2003;Rijsberman, 2006). In the computation of operational drought indicators, the available water is often represented by the streamflow, or a fraction of it, and the water need is usually quantified by a standard per capita or by a fixed nominal demand (Falkenmark et al, 1989;Raskin et al, 1997).…”