Main content area

An intercomparison of multiple statistical downscaling methods for daily precipitation and temperature over China: future climate projections

Yang, Yi, Tang, Jianping, Xiong, Zhe, Wang, Shuyu, Yuan, Jian
Climate dynamics 2019 v.52 no.11 pp. 6749-6771
atmospheric precipitation, climate, climate change, climate models, cumulative distribution, frost, summer, temperature, winter, China
In this study, we use four statistical downscaling methods to statistically downscale seven Coupled Model Intercomparison Project (CMIP5) Global Climate Models (GCMs) and project the changes in precipitation and temperature over China under RCP4.5 and RCP8.5 emission scenarios. The four statistical downscaling methods are bias-correction and spatial downscaling (BCSD), bias-correction and climate imprint (BCCI), bias correction constructed analogues with quantile mapping reordering (BCCAQ), and cumulative distribution function transform (CDF-t). Though large inter-model variability exists in the distribution and magnitude of changes in projected precipitation, particularly for wet spell length (CWD), all downscaling methods generally project a consistent enhancement of precipitation in both summer and winter over most parts of China. For the arid and semiarid Northwest China, the shortened dry spell length (CDD) is accompanied by the pronouncedly intensified very wet days (R95p), as well as the increase in maximum 5-day precipitation amount (Rx5day). In contrast, southeastern regions may experience more consecutive dry days and more severe wet precipitation extremes. The projected changes from different downscaling techniques are fairly similar for temperature, apart from the diurnal temperature range for BCSD. Warming is projected across the whole domain with larger magnitude over the north and in winter under RCP8.5. More summer days and fewer frost days would appear in the future. The bias correction components of downscaling methods cause a higher degree of agreement among models, and the downscaled results generally retain the main climate change signal of the driving models.