Reliability of simulating internal precipitation variability over multi‐timescales using multiple global climate model large ensembles in China

Abstract

Single model initial‐condition large ensembles (SMILEs) are valuable means to study the role of internal climate variability (ICV) in climate change. However, the ability of SMILEs to represent the multi‐timescale ICV is not fully understood, especially in the region of China. The main objective of this work is to assess the ability of six advanced SMILEs with ensemble members ranging from 30 to 100 in simulating ICV of precipitation at various timescales in China, using a set of observational datasets as the benchmark. To have a first insight into the selected models, the mean states and trends of annual precipitation were evaluated at the very beginning. The ICV of precipitation was then calculated at multiple timescales by two widely used methods of representing temporal variability and intermember variability. The minimum required ensemble size for each climate model to robustly capture the true value of ICV is investigated at last. The results show that the multiple SMILEs can capture the basic spatial distribution pattern of ICV across timescales. However, the ICV represented by temporal variability underestimates observed ICV in most regions across timescales. The ICV represented by intermember variability is similar to that represented by temporal variability at the interannual and interdecadal timescales, but significantly larger at the multidecadal timescales. Within ±10% error limit, 20–30 members are sufficient for all climate models at the interannual timescale while ensemble sizes of 74% or more of the full sizes for each climate model are needed at multidecadal timescales.

Quantifying the Role of Model Internal Year-to-Year Variability in Estimating Anthropogenic Aerosol Radiative Effects

“…Compared to model diversity (i.e., model bias), internal variability dominates climate uncertainty over the first decadal prediction (i.e., relatively shorter timescale) horizons at regional to local scales [28][29][30]. Fortunately, this kind of uncertainty can be significantly reduced through the application of large ensemble simulations [31][32][33][34]. In future climate projections, internal variability also significantly impacts extreme events and associated predictions such as droughts and crop production [35][36][37].…”