Impact of climate change on flood frequency using different climate models and downscaling approaches

The analysis of the climate change impact on flood frequency represents an important issue for water resources management and flood risk mitigation. However, for small/medium catchments (<1,000 km²), the spatial-temporal resolution of global circulation models (GCMs) output is not adequate (>40,000 km²) and downscaling procedures are required. In this paper, two different GCMs selected among the Coupled Model Intercomparison Project 3 models, the Hadley Center Coupled Model, the Parallel Climate Model, and two statistical downscaling approaches, (1) delta change, and (2) quantile mapping, are compared. For the generation of long hourly time series of rainfall, temperature, and discharge, stochastic weather generators coupled with a continuous rainfall-runoff model are employed. Therefore, the frequency of annual maxima rainfall and discharge is projected for the future period 2070-2099 over three small subcatchments in the Upper Tiber River Basin, central Italy. Results reveal that both the GCMs and downscaling methods play a significant role in the determination of the climate change impact for future scenarios, mainly in terms of annual maxima values. By comparing the future (2070-2099) with the baseline period (1961-1990), all GCMs project a decrease of mean annual rainfall (~30%) and an increase of mean annual temperature (~40%). However, in terms of annual maxima (of rainfall and discharge) the results are found to be dependent on the selected GCM and downscaling method. On one hand, through the application of the delta change method, both GCMs project a decrease in the flood frequency curves. On the other hand, if the quantile mapping downscaling method is considered, the Hadley Center Coupled Model 3 projects a decrease in the frequency of annual maxima discharge; the opposite occurs for the Parallel Climate Model. The hydrological characteristics of the study catchments play an important role in the assessment of the climate change impacts. For that, the need to use ensemble GCM results and multiple downscaling methods is underlined.