Deterministic and probabilistic projections and their credibility in analyzing future precipitation variations in the Yellow River Basin, China

Abstract

It remains a key challenge to obtain reliable future precipitation estimates and their reliability under different climate scenarios. In this study, the deterministic projection of future precipitation in the Yellow River Basin (YRB) was obtained within the Bayesian model averaging (BMA) framework. A probability estimation method based on the BMA weighting scheme was proposed to obtain the probabilistic projection of precipitation. We also analyzed the credibility of these two projections. The results showed that four indexes projected by the BMA method showed an increasing trend with a higher probability. The probabilities of increasing with varying degrees were more than those for decreasing for all the precipitation indexes. The credibility of the precipitation estimation under specific climate scenarios was testified by the lower ED (the mean of long-term annual relative simulation deviation) and VD (the variance of long-term annual relative simulation deviation). The estimation based on the BMA model is more trustworthy than any other model. For the four precipitation indicators, the accuracy between the calculated VR (Variation range, to describe the interval of variation of the indicators) with the greatest likelihood and the actual VR was 38.31−53.74%. In 81.93−94.70% of grids, the deviations were smaller than one level. Both the deterministic and probabilistic projections have high geographic distribution and variation trend consistency.

Spatiotemporal Projections of Precipitation in the Lancang–Mekong River Basin Based on CMIP6 Models

“…MS is calculated on the basis of the simulation of each model without observation, thus facilitating assessment of the reliability of future projections. According to the principle of MS, if the results of each model are more similar, the projection is considered more reliable [111]. To verify the rationality of the MS, we examined the spatio-temporal relationship between the RMSE (based on the multi-model ensemble) and the MS (based on the single model) in the historical period.…”