top of page

Reviewing the January 12–13, 2021 Windstorm

Late on January 12th and early on January 13th, a surprise windstorm ravaged the Pacific Northwest, knocking out power to 559,000 customers in Western Washington, over 100,000 in Northwest Oregon, and another 100,000 over Eastern Washington. Winds gusted to 50 for the Seattle metro area—strong, but not out of the ordinary for a Pacific Northwest windstorm. Why was this system so much stronger than expected? Why were there so many outages despite strong but unremarkable winds, particularly for NW Oregon northward to central Puget Sound? And what lessons can we learn from this event so that we don’t get caught flat-footed during future windstorms?

This storm can best be described as a prolonged “atmospheric river” that was capped off by a stronger-than-expected, compact, and rapidly intensifying low-pressure system that rode along this atmospheric river and slammed into the Olympic Peninsula the night of Tuesday 1/12. An “atmospheric river” is a long, narrow, relatively stationary filament of moisture that extends from the subtropics to the mid-latitudes, and because warm air can hold more moisture than cool air, atmospheric rivers tend to be far wetter than systems that lack a subtropical connection. This atmospheric river brought 2–3 inches of rain for the Seattle metro area, 4–6 inches of rain over Southwest Washington, and as much as 10 inches of rain over the Olympics, Willapa Hills, Oregon Coast Range, and NW Oregon/SW Washington Cascades, resulting in widespread minor-to-moderate flooding on many rivers and numerous landslides across Western Washington and Oregon.

We knew that this atmospheric river was going to drop a tremendous amount of rain and cause problems with landslides and river flooding. However, the winds with this tempest far exceeded our expectations. Nearly every weather model was going for simply breezy conditions to accompany this atmospheric river. The one exception was the European model, which was adamant that we would have a moderate windstorm with gusts in the 35–50 mph range. The European model nailed these wind speeds, but we instead saw damage more representative of a storm that had 55–65 mph gusts. Why did we see such extensive damage despite unremarkable wind speeds?

The primary reason for the greater-than-expected damage was due to saturated soils from all the heavy rain experienced during the atmospheric river and first half of January, as these saturated soils lose their cohesion and allow trees to topple more easily under strong winds. Many Pacific Northwest trees (especially conifers) have shallow root systems, as the ample precipitation that we receive west of the Cascades means that these roots can get all the water they need without reaching deep underground. The combination of shallow-rooted, leaved conifers in loose soils means that these trees can experience a lot of force under only moderate wind and topple completely, resulting in significant disruption to the electrical grid.

An important lesson from this storm is that we can’t let any particular hazards from a storm “overshadow” and “minimize” other impacts from the storm, as the sum of the impacts can be greater than the parts. We had very high confidence in heavy rain that would cause significant flooding and landslides and we effectively communicated this threat. However, the focus on flooding and landslides unfortunately came at the expense of any focus on wind, when in reality, it should have amplified the focus on wind due to soil saturation. While it is true that there is only so much information TV meteorologists can put in a 2-minute TV newscast (as a KOMO News intern who has practiced in front of a green screen, I’ve experienced this first-hand!), we have other methods of disseminating this information, such as social media and weather blogs, and I don’t think there’s much of an excuse for not correctly highlighting the small but very real potential for a highly disruptive windstorm.

Forecasting is a probabilistic exercise by nature. When we run weather models, we look at all the different solutions, discern what solutions are most likely based on model consensus, prior model performance, and climatology, and emphasize these solutions while still highlighting the risks associated with the less probable, but still possible, solutions. Unfortunately, humans are poor at understanding probabilities, leading to an incongruence with the nature of forecasting and human psychology. Thankfully, I know we’ve made great strides in doing this over the past few years, and I’m optimistic that we can take these lessons and further improve our forecast communication when the next windstorm inevitably strikes.

~Charlie Phillips

Charlie Phillips is a Madrona resident who received his B.S. in atmospheric sciences from the University of Washington. He works in Portland as a meteorologist. Check out his weather website at


bottom of page