Twenty-first century regional temperature response in Chile based on empirical-statistical downscaling

Local scale estimates of temperature change in the twenty-first century are necessary for informed decision making in both the public and private sector. In order to generate such estimates for Chile, weather station data of the Direccion Meteorologica de Chile are used to identify large-scale predictors for local-scale temperature changes and construct individual empirical-statistical models for each station. The geographical coverage of weather stations ranges from Arica in the North to Punta Arenas in the South. Each model is trained in a cross-validated stepwise linear multiple regression procedure based on (24) weather station records and predictor time series derived from ERA-Interim reanalysis data. The time period 1979-2000 is used for training, while independent data from 2001 to 2015 serves as a basis for assessing model performance. The resulting transfer functions for each station are then directly coupled to MPI-ESM simulations for future climate change under emission scenarios RCP2.6, RCP4.5 and RCP 8.5 to estimate the local temperature response until 2100 A.D. Our investigation into predictors for local scale temperature changes support established knowledge of the main drivers of Chilean climate, i.e. a strong influence of the El Nino Southern Oscillation in northern Chile and frontal system-governed climate in central and southern Chile. Temperature downscaling yields high prediction skill scores (ca. 0.8), with highest scores for the mid-latitudes. When forced with MPI-ESM simulations, the statistical models predict local temperature deviations from the 1979-2015 mean that range between - 0.5-2 K, 0.5-3 K and 2-7 K for RCP2.6, RCP4.5 and RCP8.5 respectively.

Automated summary

Local scale estimates of temperature change are crucial for decision making, especially in Chile. Utilizing weather station data, empirical-statistical models were constructed based on large-scale predictors to estimate local temperature changes. The models showed high prediction skill scores and supported the main drivers of Chilean climate.