Assessing the analog method to downscale daily precipitation in the Pampas region.

POGGI, M MERCEDES; BETTOLLI MARIA LAURA

La Paz

Workshop; CORDEX Central America and South America Training Workshop on Downscaling Techniques; 2018

CORDEX-WCRP

There is a growing need for climate information from various productive sectors in order to plan their activities and implement adaptation measures to reduce the impacts of climate variability and climate change, especially of extreme events. In Argentina, having estimators or predictors of daily precipitation extremes would provide more and better products for specific applications. Despite the merits of empirical statistical downscaling (ESD), these techniques have been little explored for climate change applications in Argentina. In this context, the aim of this work is to evaluate the potential of the analogue method for climate change assessments in the Pampas region. To this end, the analogue method was calibrated and validated using observed daily precipitation data from 25 meteorological stations as predictands and NCEP Reanalysis 2 daily fields as predictors. The statistical model was calibrated for the 1979-2000 period and validated for the independent period 2001-2014. It was then applied to a set of eight GCMs from the Coupled Model Intercomparison Project Phase for the historical run in the period 1979-2005 in order to evaluate the added value of the ESD method. The downscaling performance depended on the season under consideration. The highest skill was found for winter probably due to winter precipitation is mostly controlled by large-scale mechanisms that are well captured by the analogue method. However, some aspects such as wet day intensities, wet-wet and dry-wet transition probabilities, probability distributions and extremes values were quite well represented during the warm season where local processes control precipitation. Spatial distributions of precipitation extremes characterized by the 95th percentile were also adequately reproduced. The method failed to reproduce the interannual variability of the seasonal precipitation amounts in all seasons of the year but it showed a better performance for the seasonal frequencies. All GCMs considered tended to precipitate too frequently at low intensity and less frequently at high intensity over the region. Even if this characteristic was observed for all seasons, it is more accentuated in winter. Extreme precipitation values were highly underestimated with a large inter-model dispersion. The GCM downscaled precipitation series showed considerably better agreement with the observed precipitation statistical aspects, thus stressing the added value of the ESD method.