Evaluation of parametric and statistical approaches for the regionalization of flow duration curves in intermittent regimes

Flow duration curves (FDCs) are the most popular tools to estimate the amount of water available in a basin. They show the link between streamflows and the respective exceedance frequencies, but their use and reliability depends on the availability of observed data. With the aim of overcoming the lack of information about observed streamflow in many basins, several procedures for the regionalization of the estimate have been developed. In this paper the performances of seven models (five parametric and two statistical), two of these original, were analyzed. These models are able to describe the behavior of intermittent regimes, and were used for regional estimates of the FDCs of a southern Italian region (Calabria). The non-regionalized models applied to 19 gauged basins present throughout the region showed good performance. For the definition of the regional models a multivariate stepwise regression analysis was used, while a jack-knife procedure in the validation phase was performed. The comparison between cross-validated and observed FDCs allowed several synthetic performance indexes (some of which were used in a combined Taylor diagram) to be calculated, showing a good reliability of the models, especially in the case of the statistical approaches, thereby allowing compensation, at least in part, for the limited availability of long observed streamflow series in the region analyzed.