Stochastic downscaling of climate model precipitation outputs in orographically complex regions: 2. Downscaling methodology

Abstract

[1] A new methodology of stochastic downscaling of climate model precipitation outputs to subdaily temporal resolution and in a multisite framework is presented. The methodology is based on the reparameterization for future climate of the Spatiotemporal Neyman-Scott Rectangular Pulses model. The reparameterization is carried out by estimating the model parameters as done for the calibration of the model for the historical climate and using future statistics that are obtained: (i) applying to the daily historical statistics a factor of change computed from the control and future climate model outputs and (ii) by rescaling the altered daily statistics according to the scaling properties exhibited by the historical raw moments, in order to generate the future statistics at the temporal resolutions required by the reparameterization procedure. The downscaled scenarios are obtained in a multisite framework accounting for cross correlations among the stations. The methodology represents a robust, efficient, and unique approach to generate multiple series of spatially distributed subdaily precipitation scenarios by Monte Carlo simulation. It presents thus a unique alternative for addressing the internal variability of the precipitation process at high temporal and spatial resolution, as compared to other downscaling techniques, which are affected by both computational and resolution problems. The application of the presented approach is demonstrated for a region of complex orography where the model has proved to provide good results, in order to analyze potential changes in such vulnerable areas.