Report-back:
Can you spot an atmospheric river?

December 2023 - Led by Ben Hatchett and Meghan Collins  

An ever-widening definition is making it more difficult, but ground-based observations can help discern impacts.

Atmospheric rivers are commonly associated with the most impactful winter storms along the west coast of North America. Skillfully forecasting winter storms is crucial for our communities to prepare for potential flooding and other impacts such as power outages, road closures, and landslides. 

But as storms are more frequently “tagged” as atmospheric rivers, we are seeing confusion grow about what exactly the term means for west coast communities, especially those in mountains. Because the precipitation phase in these storms is closely tied to the impacts – such as huge snowfall or major flood problems – correctly predicting whether precipitation will fall as rain or snow brings more clarity to both forecasters and the communities relying on accurate predictions. 

To see the rivers in the sky, we must go to space

Scientists first coined the term “atmospheric river” in the early 1990s after recognizing narrow, elongated plumes of concentrated moisture being transported over the Pacific and Atlantic oceans. Some of our readers likely remember when the term “pineapple express” was used to describe the most impactful of these events–the wet and often warm storms bringing days of continuous rainfall and high snow levels that caused both localized and widespread flooding, such as the New Year’s Flood of 1997 (photo to the left, courtesy of the California Department of Water Resources). 


As research efforts sought to better understand the pineapple express, they became famous for their recognized ability to break drought conditions and drive large variations in total precipitation from year to year. For many years, the term “atmospheric river” was synonymous with pineapple express and conjured up images of rain-on-snow, flooding farmland, power outages, road closures, and growing concerns about the capacity of aging and poorly maintained water infrastructure such as dams and levees (recall the Oroville Dam Crisis in 2017).

Let’s start with a quiz: Of the four satellite images below showing the total amount of water vapor in the atmosphere, in which do you see an atmospheric river? Storm 1 gives a hint of what to look for! 

Satellite-derived integrated water vapor (the total amount of water vapor in the atmosphere) during four storms (a-d) occurring between October 2021 and January 2022. Data provided by the University of Wisconsin’s Cooperative Institute for Satellite Studies MIMIC-TPW2 product.

Storm 1 shows the signature form of an atmospheric river that the definition was originally based upon. But in fact, all four of these storms could be, and were, identified as atmospheric rivers using the broadest definition of this type of storm. 

If each of these storms were an atmospheric river, did they all bring flooding and associated damage? 

Definitions aside, let’s explore the precipitation brought by each storm. Even though each of these storms could be counted as an atmospheric river, they produced widely varying precipitation totals. Storm 1 on October 22-24 was an extreme atmospheric river that brought very heavy rainfall and the tail end brought high elevation snow. On December 13-15, Storm 2 brought heavy snow and low elevation rainfall. Storm 3 on December 26-28 resulted in snowfall that was quite heavy across a large range of elevations. Finally, Storm 4 on January 3-5 did not deliver a great deal of precipitation due to the unfavorable wind orientation with respect to the mountains, despite a strong signal for an atmospheric river along the coast (Storm 4 above). 

Storms between October 2021 and January 2022 brought varying amounts of precipitation, yet all were termed atmospheric rivers.  

By comparing precipitation and snowfall amounts in the figure above, you can see how different the storm patterns actually were! The term “atmospheric river” was applied to both very impactful storms (Storm 1 and 3) and relatively less impactful (Storms 2 and 4). 

We have seen confusion created when all storms become termed atmospheric rivers–especially in the popular media–which can make it difficult for communities to prepare. If the total rain or snowfall of these storms differs greatly, yet we apply the same terminology to all storms, we risk the effects of “crying wolf” when community member’s expectations of a storm’s impact are greater than the storm’s outcome. Take this scenario, for example: A meteorologist gets a call prior to Storm 4: “I heard it is going to be an atmospheric river. Will it be that bad?” But Storm 4 vastly underperformed relative to expectations of what we used to think of as the major flood (or impact) producing storms. As a result, this can erode the trust of community members in the forecast over time. 

Further complicating the matter is that atmospheric rivers are often classified based on the amount of water being transported and not based on the forecast precipitation amount, let alone the phase of precipitation that is forecast. Precipitation falling as rain or snow present very different hazards depending on where and when it falls. For example, one inch water falling as rain on an extensive and ready-to-melt snowpack will have larger consequences than an inch of water falling as snow in a remote mountain region.

Can’t the experts just tell us whether it will be an atmospheric river? 

Consensus among experts is important for consistency in communication to support preparedness for any storm. But exactly how to define an atmospheric remains a lively discussion among atmospheric science researchers. The height of the gold spikes below can be interpreted as the probability experts will agree that a storm should be classified as an atmospheric river. These probabilities were developed by O’Brien et al. (2020) and were calculated through input from experts in meteorology using historic storms, and then applied to newly identified storms using machine learning. We thank Yang Zhou at Lawrence Berkeley National Laboratory for providing the atmospheric river detection probabilities. 

Each of the gold spikes represents the probability that a storm will be identified as an atmospheric river on a scale of 0 to 100%. This calculation is based on historic detection of storms and can essentially be interpreted as the likely “consensus” among experts (data from O’Brien et al. 2020).

As you can see, there is not complete consensus as to whether or not these storms should be defined (or detected) as an atmospheric river. Some events have a higher probability, others have much lower probabilities, and these probabilities do not correspond perfectly with precipitation amounts or precipitation phase. 

So what is missing to be able to differentiate and predict an impactful event? Precipitation phase (rain, snow, mixed) is one of the most important characteristics of large storms that needs to be simulated correctly, especially in mountains. Specifically, we need to know what elevation and when the precipitation actually transitioned from rain to snow to understand a storm’s potential impacts, regardless of whether it is termed an atmospheric river.   

Your observations offer a clearer picture 

This gets to one of the core contributions of the Mountain Rain or Snow project to improve our forecasts. Your observations help us understand the precipitation phase differences within and across storms. We can see with better accuracy the proportion of rain, snow, and mixed precipitation that fell during each storm, and very importantly, where the rain-snow transition took place on the landscape and when it changed. Counts of the observations of precipitation phase are shown in the chart below.

Here, we've added the counts of Mountain Rain or Snow precipitation phase observations
(black, purple, and red lines).

In any storm, the rain-snow boundary may change as the storm progresses. Before Mountain Rain or Snow, insights into the rain-snow transition could be derived from just one daily satellite snapshot.  

Comparing and contrasting the number of rain, snow, and/or mixed observations across the four storms (and other storms during the winter), we can use your observations to understand the nuances of each storm’s progression. For example, we saw that Storm 1 was largely characterized by high elevation rain until snow finally began later in the event (rather than snow up high throughout). Storm 2 had both rain, snow, and mixed observations throughout, indicating a variety of areas receiving different precipitation phases, but it finished very cold with snow into the valleys. Because it was very cold, Storm 3 had very few rain observations. Storm 4 was quite dry and had very few observations of precipitation at all.

Your precipitation phase observations help us understand what differentiates these events after they make landfall. The rain-snow transition remains very difficult to predict, and forecasters need to know what characteristics of big storms matter the most to produce the best forecasts of precipitation totals and where the rain will turn to snow. You are an important part of helping us to pin down these differences and improve our capabilities to communicate the important nuances of storms. Until these prediction tools are perfected, the next time you hear “another atmospheric river is approaching” in the news, wait for a little more information before deciding to start filling sandbags, waxing your skis, or needing to do both! 

Mountain Rain or Snow logo with graphics of mountains, a snow crystal, and a rain drop.