A Case Study of Forecast Uncertainty Prior to a High-Impact Landfalling Atmospheric River in California in January 2021

August 11, 2025

An article titled “A Case Study of Forecast Uncertainty Prior to a High-Impact Landfalling Atmospheric River in California in January 2021,” co-authored by CW3E’s Jason Cordeira, Brian Kawzenuk, Samuel Bartlett, Chad Hecht, Christopher Castellano, Shawn Roj, and F. Martin Ralph, was recently published in the American Meteorological Society’s Weather and Forecasting.

This study analyzes the forecast uncertainty associated with a high-impact landfalling atmospheric river (AR) that struck California between 26–28 January 2021. The AR delivered extreme precipitation, including 300–400 mm of rain and over 250 cm of snow in the Sierra Nevada (Figure 1), and triggered a destructive debris flow near Big Sur on the burn scar of the 2020 Dolan Fire. The event was characterized by a strong and long-duration integrated vapor transport (IVT) maximum that stalled along the Central California coast, creating a challenging precipitation forecast scenario.

Using deterministic and ensemble forecasts from both the GFS and ECMWF models, the authors examined lead times from 1–7 days in advance of landfall. The GFS provided earlier (~2–3 day) signals of landfall (Figure 2), while the ECMWF better captured the subsequent stalling behavior and associated rainfall distribution. Both models exhibited a dry bias, especially over coastal topography. A key source of forecast uncertainty was traced to differences in how models initialized the evolution of synoptic-scale features over the North Pacific, especially the development and interaction of two Rossby wave trains (RWTs) and upstream cyclogenesis involving remnant tropical moisture. Forecasts initialized on 21–22 January differed significantly in how they resolved these features, with ECMWF forecasts after 22 January more accurately capturing downstream impacts (Figure 3).

The results emphasize the importance of accurate upstream observations, such as those from AR Reconnaissance missions, in improving forecast reliability for high-impact West Coast weather events. The study highlights key objectives in CW3E’s 2025-2029 Strategic Plan to better understand the science of ARs and extreme precipitation, including physical processes, forecasting, and impacts.

Figure 1. (Figure 1 from Cordeira et al. 2025) (a) ERA5 integrated water vapor transport (IVT; shaded and vectors; kg m^–1 s^–1) and mean sea level pressure (contours; hPa) valid at 0300 UTC 28 January 2021, (b) ERA5 integrated water vapor (IWV; shaded; mm) and mean sea level pressure (contours; hPa) valid at 0300 UTC 28 January 2021, (c) NCEP Stage-IV quantitative precipitation estimate (QPE; shaded; mm) valid for the 72-h period ending 1200 UTC 29 January 2021, and (d) NOHRSC snowfall analysis (shaded; cm) valid for the 72-h period ending 1200 UTC 29 January 2021.

Figure 2. (Figure 9 from Cordeira et al. 2025) Left panels illustrate lead-time–coastal latitude “dProg/dt” analysis of the (a) GFS ensemble and (b) ECMWF ensemble forecast probability of IVT magnitude ≥250 kg m^–1 s^–1 (shading as a fraction) and ensemble mean IVT magnitude (contoured every 100 kg m^–1 s^–1 starting at 100 kg m^–1 s^–1) for locations along the West Coast of North America for forecasts all verifying at 0000 UTC 28 January 2021. Right map panels illustrate IVT magnitude shaded yellow at 250 kg m^–1 s^–1 and orange at 500 kg m^–1 s^–1 from the corresponding model analyses (0-h control forecast). The analysis is derived from the coastal locations drawn in the right panels; the locations are drawn black for IVT magnitudes <250 kg m^–1 s^–1 and red for IVT magnitudes ≥250 kg m^–1 s^–1 at 0000 UTC 28 January 2021. The black box on left panels denotes the 6-to-10-day lead time discussed in the text.

Figure 3. (Figure 12 from Cordeira et al. 2025) Model forecasts of IVT magnitude (kg m^–1 s^–1; shaded according to scale) and direction (vectors) with sea-level pressure initialized by the deterministic (a) ECMWF and (b) GFS models at 0000 UTC 21 January 2021 (top row in each panel) and at 0000 UTC 22 January 2022 (bottom row in each panel). Model forecasts are valid at 0000 UTC on 24, 25, 26, and 27 January following the columns labeled in each panel. Red and black dashed lines refer to features discussed in the text. For consistency, the cyclone that develops from the remnants of the tropical depression is labeled as “TD” even though it is no longer tropical.

Cordeira, J. M., Kawzenuk, B. K., Bartlett, S. M., Hecht, C., Castellano, C., Roj, S., & Ralph, F. M. (2025). A Case Study of Forecast Uncertainty Prior to a High-Impact Landfalling Atmospheric River in California in January 2021. Weather and Forecasting, 40(8), 1543-1561. https://doi.org/10.1175/WAF-D-24-0088.1