Simulating future occurrence and magnitude of flood events through bias-correction of hourly climate scenarios and semi-distributed rainfall-runoff modelling: application to the Panaro river (Northern Italy)

Assessing the impact of climate change scenarios on flood regimes is a crucial issue when evaluating the future resilience of flood protection systems. Mainly due to the high computational effort and to the scarcity of hourly climate projections, expected changes in future floods are often simulated by hydrological models on a daily basis, even for basins with short response times, where hourly simulations would be needed.

In this work, the expected occurrence and magnitude of future flood events is modelled through the coupling of bias-corrected local climate scenarios at hourly time scale and continuous rainfall-runoff modelling. The case study for testing the procedure refers to the Panaro river (one of the OpenAir Laboratories in the OPERANDUM H2020 project).

The analysis entailed first the collection, validation and spatialization of historical meteorological ground data over the catchment, to be used both to calibrate the hydrological model and in the bias-correction procedure.

Secondly, the precipitation and temperature timeseries available at hourly time-scale for a set of climate modelling chains based on the same RCM nested in different GCMs, under RCP 8.5, are identified in the EURO-CORDEX ensemble and processed. The comparison with observed spatial fields obtained from weather stations and from gridded E-OBS products allows assessing the biases affecting “raw” data. The Scaled Distribution Mapping (SDM) bias correction procedure is then applied to “adjust” the raw model output towards hourly observations in a post processing step. The strength of such a procedure relies on preserving raw climate model projected changes in the bias-corrected series and on avoiding assumptions about stationarity.

A semi-distributed, continuously simulating HBV-type rainfall-runoff model is parameterised, especially focusing on the reproduction of past flood events, and then run to reproduce the streamflow in the Panaro river, providing in input i) historical meteorological forcing based on ground stations, ii) raw and bias-corrected climate scenarios over the control period, iii) bias-corrected climate scenarios for the future decades. Finally, the flood events are extracted from the continuous streamflow simulations and the changes in the flood signals expected over the future decades are analysed, in terms of both peaks and volumes.