Yellowstone floods reveal how climate change has made forecasting difficult
Extreme weather increasingly difficult to predict as rare events become more likely
The Yellowstone National Park area's weather forecast the morning of June 12 seemed fairly tame: warmer temperatures and rain showers would accelerate mountain snow melt and could produce "minor flooding."
A National Weather Service bulletin recommended moving livestock from low-lying areas but made no mention of danger to people.
By nightfall, after several centimetres of rain fell on a deep spring snowpack, there were record-shattering floods.
Torrents of water poured off the mountains. Swollen rivers carrying boulders and trees smashed through Montana towns over the next several days. The flooding swept away houses, wiped out bridges and forced the evacuation of more than 10,000 tourists, park employees and residents near the park.
As a cleanup expected to last months grinds on, climate experts and meteorologists say the gap between the destruction and what was forecast underscores a troublesome aspect of climate change: Models used to predict storm impacts do not always keep up with increasingly devastating rainstorms, hurricanes, heat waves and other events.
"Those rivers had never reached those levels. We literally were flying blind not even knowing what the impacts would be," said Arin Peters, a senior hydrologist with the National Weather Service.
Modelling can't keep up with changes
Hydrologic models used to predict flooding are based on long-term, historical records. But they do not reflect changes to the climate that emerged over the past decade, said meteorologist and Weather Underground founder Jeff Masters.
"Those models are going to be inadequate to deal with a new climate," Masters said.
Scientists said the Yellowstone flooding was consistent with changes already documented around the park as temperatures warm.
Those changes include less snowfall in mid-winter and more spring precipitation — setting the stage for flash floods when rains fall on the snow, said Montana State University climate scientist Cathy Whitlock.
Warming trends mean spring floods will increase in frequency — even as the region suffers from long-term drought that keeps much of the rest of the year dry, she said.
Masters and other experts noted that computer modelling of storms has become more sophisticated and is generally more accurate than ever. But extreme weather by its nature is hard to predict, and as such events happen more frequently there will be many more chances for forecasters to get it wrong.
The rate of the most extreme rainstorms has increased by a factor of five, Masters said. So an event with a one per cent chance of happening in any given year — commonly referred to as a "one in 100-year" event — now has a five per cent chance of happening, he said.
"We are literally re-writing our weather history book," said University of Oklahoma meteorology professor Jason Furtado.
That has widespread implications for local authorities and emergency officials who rely on weather bulletins to guide their disaster response approaches. If they're not warned, they can't act.
But the National Weather Service also strives to avoid undue alarm and maintain public trust. So if the service's models show only a slim chance of disaster, that information can get left out of the forecast.
Another extreme weather event where the models came up short was Hurricane Ida, which slammed Louisiana last summer and then stalled over the Eastern Seaboard — deluging parts of Pennsylvania, New Jersey and New York with unprecedented rainfall that caused massive flooding.
The weather service had warned of a "serious situation" that could turn "catastrophic," but the predicted 8 to 15 centimetres of rain for New York, New Jersey and Pennsylvania was far short of the 23 to 25 centimetres that fell.
The deadly June 2021 heat wave that scorched British Columbia and the Pacific Northwest offered another example. Warmer weather had been expected, but not temperatures of up to 47 C that toppled previous records and killed more than 600 in B.C. alone.
In November 2021, B.C. also experienced "once-in-a-century" flooding that destroyed highways, forced nearly 20,000 people from their homes and killed thousands of animals. Heavy rainfall was anticipated, but no travel advisories or flood warnings were in place.
Scrambling in the dark to save lives
The surprise Yellowstone floods prompted a nighttime scramble to close off roads and bridges getting swept away by the water, plus rushed evacuations that missed some people. No one died, somewhat miraculously, as more than 400 homes were damaged or destroyed.
As rock slides caused by the rainfall started happening in Yellowstone, park rangers closed a heavily used road between the town of Gardiner and the park headquarters in Mammoth Hot Springs, Wyo. It later washed out in numerous places.
The rain and snowmelt was "too much too fast and you just try to stay out of the way," Yellowstone Deputy Chief Ranger Tim Townsend said.
If the road hadn't been closed, "we probably would have had fatalities, unquestionably," park Supt. Cam Sholly said.
"The road looks totally fine and then it's like an 80-foot drop right into the river," Sholly said.
Within a matter of hours on June 12, Rock Creek, which runs through the city of Red Lodge and normally is placid and sometimes just ankle deep, became a raging river. When the weather service issued a flood warning for the creek, the water already had surged over its banks and begun to knock down bridges.
By the time the warning was sent, "we already knew it was too late," said Scott Williams, a commissioner for Carbon County, Mont., which borders Yellowstone.
Weather service officials said the agency's actions with the Yellowstone flooding will be analyzed to determine if changes are needed. They said early warnings that river levels were rising did help officials prepare and prevent loss of life, even if their advisories failed to predict the severity.
Computer-based forecasting models are regularly updated to account for new meteorological trends due to climate change, Peters said. Even with those refinements, events like the Yellowstone flooding still are considered low-probability and so often won't make it into forecasts based on what the models say is most likely to occur.
"It's really difficult to balance that feeling that you've got, that this could get really bad, but the likelihood of it getting really bad is so small," Peters said.
He added that the dramatic swing from drought to flood was hard even for meteorologists to reconcile and called it "weather whiplash."
To better communicate the potential for extreme weather, some experts say the weather service needs to change its forecasts to inform the public about low probability hazardous events. That could be accomplished through more detailed daily forecasts or some kind of colour-coded system for alerts.
"We've been slow to provide that information," North Carolina State University atmospheric scientist Gary Lackmann said. "You put it on people's radars and they could think about that and it could save lives."
With files from Amy Beth Hanson