Atmospheric Rivers in Machine Learning Weather Models Journal Article uri icon

Overview

abstract

  • Abstract; This study examines the representation of atmospheric rivers (ARs) in medium-range forecasts (1–10 days lead time) from machine learning weather prediction (MLWP) models. ARs pose high-impact consequences, both in terms of flooding and their contributions to water resources. As MLWP models proliferate, there is a pressing need to understand MLWP model skill for AR prediction. ARs provide a useful test case for MLWP models because they are filamentary features exhibiting large gradients in thermodynamic and kinematic fields that may not be well represented within the machine learning loss function framework. We evaluate and compare forecasts from two deterministic MLWP models, GraphCast and Pangu-Weather, with two operational numerical weather prediction (NWP) models, NOAA's GFS and ECMWF's IFS. We focus on wintertime ARs in the northern Pacific over lead-times from one to ten days and employ a combination of grid point-based as well as novel, object-based verification methods. We also use storm-relative composites of thermodynamic and kinematic fields to diagnose strengths or shortcomings in the emulation of physical processes. Results show that MLWP models can represent ARs well according to certain aggregate metrics, with some caveats and systematic differences compared to traditional NWP models. The ARs in MLWP models are slightly too weak and too small compared to both reanalysis and NWP models. MLWP models differ systematically from NWP models in the biases of dynamic and thermodynamic fields within strong ARs. Better understanding phenomena-specific traits of MLWP models can help improve both forecaster interpretation of their predictions, and inform improvements to both MLWP and NWP forecasts.

publication date

  • June 3, 2026

Date in CU Experts

  • June 15, 2026 10:36 AM

Full Author List

  • M. Swenson L; J. Moore B; Mahoney K

author count

  • 3

Other Profiles

International Standard Serial Number (ISSN)

  • 0882-8156

Electronic International Standard Serial Number (EISSN)

  • 1520-0434