Identifying Regional Models for Flow Duration Curves with Evolutionary Polynomial Regression: Application for Intermittent Streams

Abstract

Regional models of flow duration curves (FDC) are ubiquitous components in water resources studies, particularly in ungauged sites or in catchments with scarce monitoring. The identification of such models in arid regions, in which intermittent and ephemeral streams are common, constitutes an additional challenge in statistical hydrology applications. In this paper, we propose an alternative approach for the modeling of FDCs in null-flow conditions. First, we evaluated a set of flexible distributions frequently used in hydrology for modeling the entire range of observed streamflows, with a focus on the proper representation of intermittence and the description of the lower flows. Next, regional models for the parameters of the distributions were identified with the evolutionary polynomial regression (EPR) technique, which, although providing structurally complex equations, proved able to uncover the relationships between the characteristics of such parameters and catchments and, more importantly, preserve them in a cross-validation procedure. From a practical perspective, the decision-making processes involving water resources planning and management could be more coherently addressed in ungauged sites under the proposed framework, as compared to other established yet simpler modeling techniques.