Drought Frequency Assessment in Southern South America through the CMIP5 Climate Models

JUAN ANTONIO RIVERA; OLGA PENALBA; HERNÁN BECHIS

Buenos Aires

Taller; Taller de trabajo Adaptación al Cambio Climático en sectores clave en América del Sur; 2012

Universidad de Buenos Aires

Drought frequency estimation is a key variable for drought risk assessment. The aim of this research is to evaluate how well the global climate models (GCMs) represent the drought frequencies in Southern South America (south of 20°S). For that purpose, we used the Standardized Precipitation Index (SPI), which quantifies the number of standard deviations that the accumulated rainfall in a given time scale deviates from the average value of a location in a particular period. The SPI is one of the drought indices most widely used worldwide, and one of the best suited for the study area. In order to build this index, monthly rainfall data were obtained from the CLARIS LPB Data Base for 120 of its stations, which were subjected to quality control procedures and have less than 10% of missing values. The SPI was computed for the period 1979-2008 on a time scale of 12 months, which represents long-term droughts. This procedure was also applied to the simulated precipitation from 14 CMIP5 GCMs over the study area. Two types of comparisons were performed, the first one for drought frequencies without taking into account the different drought classes and the second one for the frequencies of moderate, severe and extreme drought events. This second comparison is important given that drought risk was evaluated through a weighted index based in drought frequencies, which was constructed as a sum of drought classes ?moderate, severe and extreme- with a weighting scheme. Most of the regions with moderate to high observed drought frequencies arelocated in the western and southern portions of La Plata Basin and over Patagonia region, while lower observed drought frequencies were obtained for Northern Argentina, Southern Brazil and Paraguay. This spatial pattern is barely reproduced by the modeled frequencies, and in some cases major differences exist. In order to analyze these differences, we performed a regional assessment of the SPI time series, which showed that the time series of the model outputs are inconsistent with the variabilities based on observed precipitation data. This also indicates that monthly and seasonal variabilities of precipitation are misrepresented by the models and these results contribute to the poor representation of the drought frequency patterns. These are the kind of variabilities that could be important for the forecasting of droughts on seasonal scales and could help to a seasonal risk assessment, which is important for the agricultural and hydrological sectors. Moreover, the spatial pattern of annual precipitation are poorly reproduced, with a subestimation of precipitation over La Plata Basin and an overestimation over Patagonia, leeward the Andes. These results prove that GCMs outputs are far to be used for drought risk assessment in the region, given the substantial differences in the frequency patterns. Also, future projections must be evaluated carefully in longer time scale resolutions.