NOAA GFDL - Geophysical Fluid Dynamics LaboratoryGFDL - Geophysical Fluid Dynamics Laboratory

Predictable Patterns of Seasonal Atmospheric River Variability Over North America during Winter

April 17th, 2025

Key Findings

  • The authors identified 3 sources of seasonal predictability of winter atmospheric rivers using GFDL’s SPEAR prediction system: El Niño-Southern Oscillation; global warming; and convection over the western equatorial Pacific.
  • Together these three modes account for nearly all wintertime forecast skill in SPEAR.
  • They also found that wintertime atmospheric river and precipitation predictability over North America are driven by the same sources.

Joseph P. Clark, Nathaniel Johnson, Mingyu Park, Miguel Bernardez, Tom Delworth. Geophysical Research Letters. 10.1029/2024GL112411

Atmospheric rivers (ARs) are long and narrow atmospheric weather systems that carry large amounts of water vapor. ARs typically cover areas larger than 300,000 km2 and accompany significant precipitation changes over the U.S. They present both societal risks, such as flooding and extreme precipitation, and benefits – including contributions to water resources and snowpack.  Using GFDL’s SPEAR prediction system, the authors pinpoint three distinct sources of seasonal predictability of winter AR frequency.  They also found that wintertime AR and precipitation predictability over North America are driven by the same sources.

While previous studies have shown that SPEAR can predict seasonal AR variability over North America with reasonable accuracy, the underlying sources of this predictability remained poorly understood.  Using a large set of hindcasts, the authors were able to identify the sources of seasonal AR predictability and its relationship to precipitation.

The three sources are:

  •  El Niño-Southern Oscillation
  •  A long-term trend related to global warming
  • Equatorial heating over the eastern flank of the western Pacific warm pool

These three explain nearly all of the skill in wintertime AR frequency and  seasonal precipitation forecasts over North America. Understanding what contributes to the skill of our current operational forecast systems is an integral step toward improving seasonal precipitation forecasts. Better AR prediction is increasingly important to save lives and protect property.

Figure 1: The (a-c) first three predictable patterns of December-February AR frequency identified by Average Predictability Time analysis alongside the associated (d-f) precipitation anomalies. The time variation of these patterns is shown for each ensemble member in panels (g-i), while the prediction skill, represented as the correlation between SPEAR and ERA5 (black lines in g-i) is shown in panels (j-l).