Characterization of catchment behaviour and rainfall selection for flash flood hydrological model calibration: catchments of the eastern Pyrenees

Abstract

Accurate flash flood prediction depends heavily on rainfall data quality and knowledge of catchment behaviour. A methodology based on global sensitivity analysis and hydrological similarity is proposed to analyse flash storm-flood events with a mechanistic model. The behaviour of medium-sized catchments is identified in terms of rainfall-runoff conservation. On the basis of this shared behaviour, rainfall products with questionable quantitative precipitation estimation (QPE) are excluded. This facilitates selection of rainfall inputs for calibration, whereas it can be difficult to choose between two rainfall products by direct comparison. A substantial database of 43 flood events on 11 catchment areas was studied. Nash-Sutcliffe efficiencies for this dataset are around 0.9 in calibration and 0.7 in validation for flash flood simulation in 250 km² catchments with selected QPE. The resulting calibration framework and qualification of possible losses for different bedrock types are also interesting bases for flash flood prediction at ungauged locations.