A Framework to Identify the Uncertainty and Credibility of GCMs for Projected Future Precipitation: A Case Study in the Yellow River Basin, China

Abstract

General circulation models could simulate precipitation under climate change and have been recognized as a major tool to project future water resources, but huge inherent uncertainties mean that their credibility is widely questioned. The current analysis mainly focuses on some aspects of uncertainty and few on the whole chain process to yield a more reliable projection. This study proposes a framework to identify the uncertainty and credibility of GCMs, consisting of downscaling, uncertainty analysis (model spread and Taylor diagram), ensemble analysis (grid-based weighted Bayesian model averaging), credibility analysis (signal-to-noise ratio), and probability projection. Based on five selected climate models from the Coupled Model Intercomparison Project Phase 5 (CMIP5), the uncertainties and credibility of simulated precipitation in the Yellow River of China were analyzed. By comparing the models’ output with the observation in the historical period of 1986–2005, we can see that large uncertainty exists among models’ annual precipitation. For different-class precipitation, the uncertainties of the five models are small in relatively weak rain, but large in heavy rainfall, which indicates more risk in future projections and the necessity to explore their credibility. Moreover, in such a large-span basin, GCMs show vast spatial differences in space and even opposite trends in some regions, demonstrating the limits of Bayesian model averaging (BMA) on multi-model ensemble due to one weight group overall whole basin. Thus, a grid-based weighted Bayesian model averaging (GBMA) method is proposed to cope with the spatial inconsistencies of models. Given the multi-model ensemble results, the future precipitation changes of the periods of 2021–2050 and 2061–2090 are projected, and the probability and credibility of future precipitation changes in terms of spatial distribution are identified. Model credibility identification could allow for more reliable projections of precipitation change trends, especially for different spatial regions, which will be very valuable for decision-making related to water resource management and security.