Regional flood hydrology in a semi-arid catchment using a GLS regression model

• A regional flood frequency hydrology analysis is carried out in the Ebro catchment.
• Inter-site correlations of both sampling and modelling errors are accounted for.
• The regression model is adapted to the Spanish flood frequency recommendations.
• GLS regression models improve the existing OLS models at different spatial scales.
• Most GLS models removed almost completely the spatial correlation of residuals.

SummaryThe regional flood frequency hydrology of the 86,000 km2 and semi-arid Ebro catchment is investigated using an extended generalised least square model that includes separate descriptions for sampling errors and model errors. The Ebro catchment is characterised by large hydro-climatic heterogeneities among sub-regions. However, differences in flood processes among sites are better explained by a set of new catchment descriptors introduced into hydrological regression models, such as new characteristics derived from the slope of flow duration curves, the ratio of mean annual precipitation to extreme precipitations and the aridity index. These additions enabled a more direct link to be established between the general flow regime and the extreme flood characteristics through-out the entire catchment. The new regression models developed in this study were compared to a set of existing models recommended for flood frequency estimation in Spain. It was found that the generalised least squares model developed in this study improves the existing ordinary least squares models both at regional and trans-regional scales. An adequate description of flood processes is obtained and, as a direct consequence, more reliable flood predictions in ungauged catchments are achieved.