The Applicability of Two Statistical Downscaling Methods to the Qinling Mountains

The Qinling Mountains is not only the dividing line of northern China and southern China, but also the dividing line between monsoon climate of medium latitudes and subtropical monsoon climate in China, i.e., the dividing line between China's warm temperate and subtropical regions. It also has abundant natural resources because of its special geographical location and complex climate environment. In the context of global warming, impacts of climate change on forest ecosystems in the Qinling Mountains are of great significance. Global climate model which is widely used in large-scale climate simulation studies, cannot be applied in this region due to low resolution. Statistical downscaling model can be used to provide local-scale daily temperature and precipitation for studying climate change impacts of this region. Different statistical downscaling models have different principals, as well as different predictors. Therefore, it is necessary to compare different downscaling models and to select more appropriate downscaling model to obtain reasonable simulation results. Focusing on the future daily mean temperature and precipitation for the Qinling Mountains, the multiple linear regression and the ridge regression downscaling approaches based on ASD (automated statistical downscaling) model are implemented. Outputs from the general circulation model (MPI-ESM-LR) under RCP4.5 and RCP8.5 scenarios are analyzed. Simulation results of two statistical downscaling approaches during calibration and validation periods are analyzed and future climate change projections in periods of 2006-2040, 2041-2070 and 2071-2100 are generated. During the calibration and validation periods, both statistical downscaling approaches perform well in simulating the mean temperature and precipitation. However, the multiple linear regression perform better than the ridge regression, and the mean of simulated temperature is better than that of precipitation. Both statistical downscaling approaches project an increase for the mean temperature and its magnitudes depending on the emission scenarios, i.e., RCP8.5 resulting in a higher temperature than RCP4.5. The annual precipitation would slightly decrease but not statistically significantly, while the seasonal distribution of annual precipitation will change, a slightly increase in spring and a decrease in other seasons, especially in summer. In summary, the multiple linear regression is more suitable for statistical downscaling research in the Qinling Mountains.