Evaluation of climate model simulations in representing the precipitation non‐stationarity by considering observational uncertainties

The reliability of climate model simulations in representing the precipitation changes is one of the preconditions for climate‐change impact studies. However, the observational uncertainties hinder the robust evaluation of these climate model simulations. The goal of the present study is to evaluate the capacities of climate model simulations in representing the precipitation non‐stationarity in consideration of observational uncertainties. The mean of multiple observations from five observational precipitation datasets is used as a reference to quantify the uncertainty of observed precipitation and to evaluate the performance of climate model simulations. The non‐stationarity of precipitation was represented using the mean and variance of annual total precipitation and annual maximum daily precipitation for the 1982–2015 period. The results show that the spatial distributions of annual and extreme precipitation are similar for various observational datasets, while there has less agreement in the variance changes of extreme precipitation. Climate models are capable of representing the spatial distributions of the annual and extreme precipitation amounts at the global scales. In terms of the non‐stationarity, climate model simulations are capable of capturing the large‐scale spatial pattern of the trend in mean for annual precipitation. On the contrary, the simulations are less reliable in reproducing the change of extreme precipitation, as well as the trend of variance for annual precipitation. Overall, climate models are more reliable in simulating the mean of precipitation than the variance, and they are more reliable in simulating annual precipitation than extreme. Besides, the uncertainties of precipitation for both observations and simulations are much larger in monsoon regions than in other regions. This study suggests that considering observational uncertainties is necessary when using observational datasets as the reference to project future climate change and assess the impact of climate change on environments. Comparison of different datasets in representing PRCTOT and Rx1day indices over global land and monsoon regions using precipitation time series (the left column) and Taylor diagrams (the right column) for the 1982–2015 period: (a), (b) global land for PRCPTOT; (c), (d) global land for Rx1day; (e), (f) Monsoon regions for PRCPTOT; (g), (h) Monsoon regions for Rx1day. In the Taylor diagrams, the standard deviation is normalized, and the datas