Evaluation of CMIP6 GCMs for Simulating Atmospheric Rivers: Relating The Model Skill For Key AR Variables to the Skill for Winds and Water Vapor

Abstract

Fifteen GCMs in the Coupled Model Intercomparison Project Phase 6 are evaluated for the skill in simulating the atmospheric river (AR) frequency (F_AR) and integrated vapor transport (IVT) during 1995–2014. All GCMs simulate well the annual and seasonal climatology of F_AR and IVT for both the global and East Asia domains. Large biases in F_AR and IVT occur in the same regions characterized by high AR activities including the midlatitude Pacific and Atlantic oceans, the Southern Ocean, and the tropical region from the eastern Indian Ocean to the western Pacific. The sign and magnitude of large model errors vary across the GCMs to result in small model-mean biases. The seasonal variation of the skill of individual GCMs is smaller than the variation of the skill across the GCMs, implying that the model skill varies more widely by the difference in model formulations than the response of individual GCMs to seasonal forcing variations. A novel method to relate the skill for simulating F_AR and IVT to that for winds and water vapor is introduced. The method shows that the vertical integration of the covariance of wind and water vapor in the definition of IVT can be well approximated by the multiplication of two separate functions obtained by vertically integrating either winds or water vapor, especially in the regions of strong AR activities. Spearman’s rank correlation in conjunction with this method suggests that the model skill for F_AR and IVT is significantly related only to that for winds.