The Toughest Places to Forecast Weather in the U.S.

Jon Erdman
Published: March 7, 2018

Weather forecasts of a particular storm – snowstorms, hurricanes, or severe weather outbreaks – can be challenging enough, but there are certain parts of the country where even day-to-day forecasting is difficult.

Despite several decades of forecast improvements and investments in technology and research, including improved numerical modeling, forecasting temperatures or simply whether it will rain or snow in some areas can be surprisingly tough.

Predicting how weather will interact with mountains and bodies of water can present the biggest challenges even for the most seasoned meteorologists.

To assess the difficulty of forecasts, we examined data from the National Weather Service's Digital Forecast Database and internal verification data of The Weather Company forecasts, provided by Forecast Watch.

For rain or snow, we simply examined how often a simple forecast of precipitation or no precipitation was correct. 

As it turns out, "there is a very strong seasonal signal in this (precipitation) skill," according to Dr. Peter Neilley, director of weather and forecasting technologies at The Weather Company, an IBM Business.

In essence, the areas of greatest and least precipitation forecast errors vary by time of year. To illustrate that, we examined the difference between winter and summer in recent years.

Great Lakes

Below is a loop of National Weather Service precipitation forecast error – specifically, the chance of precipitation in a 12-hour period – from a forecast three days out during January 2015, 2016 and 2017. The orange, red, and purple Xs indicate the greatest forecast error; green squares show the least error.

National Weather Service Digital Forecast Database errors in 12-hour probability of precipitation (PoP) for 72-hour forecasts in January 2015, 2016, and 2017. Greater errors shown by orange and red X's. Lower error depicted by green squares, according to legend in lower left.

While some regions vary from year to year in January, one trend emerged: Great Lakes snowbelt forecasting is tough.

"Forecasting the exact location and amount of snow from a lake-effect snowband a day or two in advance" is the essence of the winter challenge in the Great Lakes snowbelts, according to Tom Niziol, winter weather expert at The Weather Channel and former meteorologist-in-charge at the NWS office in Buffalo, New York.

These lake bands, at times, may be only 10 to 15 miles across. If the wind direction is slightly different than expected, an area of forecasted heavy snow may see only partly to mostly cloudy skies.

Temperature forecasts can also be head-scratching around colder bodies of water, such as the Great Lakes.

"Predicting the location of the lake-breeze boundary in early summer can mean as much as a 15 to 20 degree temperature difference over the course of a few miles," Niziol said. "It can turn shirt-sleeve weather into sweater weather in no time at all."

Satellite image of Winter Storm Juno's snow cover on Jan. 28, 2015 at 1:50 p.m. ET.


Nothing can be more pressure-packed than the potential for a high-impact snowstorm in the heavily populated Northeast Megalopolis, home to more than 50 million.

"Meteorologically, the I-95 corridor will always be a winter weather challenge as it is almost always the battleground between snow, ice, rain," said Gary Szatkowski, retired former meteorologist-in-charge of the National Weather Service office in Mount Holly, New Jersey, near Philadelphia.

Szatkowski issued an apology after an expected heavy snowstorm, Winter Storm Juno, missed Philadelphia but clobbered other parts of the Northeast, in particular, New England, in late January 2015.

"The meteorologist has to draw those lines somewhere, and they won't be perfect," Szatkowski said. "Small changes in timing and location of weather events can have huge changes in the impacts. That's the nature of the beast, here."

In this aerial view looking south toward downtown Atlanta, the ice-covered interstate system shows the remnants of Winter Storm Leon Wednesday, Jan. 29, 2014, in Atlanta.
(AP Photo/David Tulis)

The Atlanta Wedge

Cold air nosing southwestward down the eastern slopes of the Appalachians often hangs up near or over the Atlanta metro area, home to 5 to 6 million.

When that so-called wedge of cold air is below freezing, snow and ice can cripple travel in a hilly area less accustomed to winter driving with fewer plows and salt/sand trucks than, say, in the Midwest. This was the case with the infamous Winter Storm Leon in January 2014. 

"I've never seen any two (wedges) alike," said veteran Atlanta radio meteorologist Kirk Mellish. Mellish said forecasting the line between rain and snow is tough enough elsewhere, but in Atlanta, "the wedge just throws another monkey wrench in the equation."

Cold air occasionally will park itself only over the northeast or east sides of the metro area. So freezing rain may fall, say, in Gainesville, Georgia, while mainly rain falls over the rest of the metro area. 

"The models are so often useless with them (the wedges) that there must be a lot of nowcasting in those situations," said Mellish.

The wedge isn't just a forecast challenge in winter.

"It can also be a plus, when it so often 'kills off' severe storms heading our way from Alabama, so we either don't get them or they are much weaker if we are shielded by the cooler more stable wedge air mass."

Mellish also mentioned there are a few occasions where the wedge front can also enhance severe weather, however.

"That, right there, is a great example of the complexity."

Total estimated snowfall from Winter Storm Vexo in mid-April 2016.

The High Plains/Rockies

Backing up to the base of the Rocky Mountains, northern Colorado's Front Range features elevations ranging from roughly 5,000 feet along the Interstate 25 Denver-Boulder-Fort Collins urban corridor to 7,000 to 9,000 feet in the western foothills.

That makes forecasting there complicated, including for winter storms.

"We see a lot of localized, microscale features that are driven by terrain," said Denver-based meteorologist Noah Brauer, contributor to "Even high-resolution models have difficulty picking out heavy-banded snow events."

Perhaps the biggest challenges are fall and spring snowstorms, when snow levels may be just high enough to spare parts of the metro from heavier accumulations. Just a few miles away and a thousand feet higher in the foothills, feet of snow may down trees and trigger power outages. 

If that wasn't enough, the High Plains and Rockies see a wide range of temperatures from season to season, day to day, or even within a day itself. With this larger range comes the chance of larger errors.

Strong downsloping winds descending the Rockies into the adjacent Plains can send temperatures skyrocketing within an hour or less.

"Areas closer to the foothills such as Boulder, Colorado, might be at 55 degrees while low-lying areas to the east such as Greeley can be in the teens," said said Denver-based meteorologist Noah Brauer. These cities are only about 35 miles apart.

One other such Chinook wind event sent temperatures rising 49 degrees in just 2 minutes in Spearfish, South Dakota, in January 1943.  

Powerful Arctic cold fronts, however, frequently plunge south from Canada down the High Plains in the colder months, sometimes faster than numerical model forecasts. 

One rather egregious example was on Nov. 11, 1911, when Oklahoma City set both a daily record high (83 degrees) and daily record low (17 degrees) on the same day. 

How cold air occasionally gets drawn into Oregon's Willamette Valley, including the city of Portland.

Oregon's Willamette Valley

The Portland-Salem-Eugene corridor has nearby mountains, an ocean and a river gorge to challenge forecasters.

"The Columbia River Gorge is the only sea-level gap through the north-south Cascade Mountain range," said Mark Nelsen, chief meteorologist at KPTV Fox 12 in Portland. "It's the only place colder air from inland areas can surge into western Oregon."

Nelsen says when that cold, dense air spills into the Willamette Valley, it's hard to kick out.

"Weather models almost always push it out too quickly," he said.

If that cold air is below freezing and a moist Pacific frontal system pushes in, there could be a snow or ice storm, but sometimes, only in parts of the Willamette Valley.

"I have seen it 32 degrees in east Portland with a 40 mph cold east wind blowing while it's 50-55 (degrees) in the west metro area with a mild, southerly breeze," Nelsen said.

The winter of 2016-17 was particularly challenging.

"I've been forecasting here for 26 years, so I've seen a lot of different weather. This winter (2016-17) was the busiest and craziest I've seen," said Nelsen. "We had four different ice storms and four separate snow events in Portland. That included two snowstorms that pretty much shut down the city."

Southeast in Summer

How many times have you heard a forecast of scattered thunderstorms in the summer only to find your outdoor picnic didn't have to be postponed because thunderstorms rumbled in the next county?

In the summer months, the most frequent challenge in many parts of the country is attempting to forecast which areas have a better chance of thunderstorms on any given day.

Not so much the long lines of thunderstorms, known as squall lines, and other clusters of thunderstorms, known as mesoscale convective systems, that can produce severe weather and flash flooding.

Rather, the more hit-or-miss variety storms can give meteorologists summer headaches.

Even with today's high-resolution numerical models, it's still exceedingly difficult to pinpoint exactly where thunderstorms may occur on any given day on the scale of, say, a few counties. 

For now, the best meteorologists can do is a forecast of scattered thunderstorms.

As you can see in the July 2015 and 2016 forecast error maps below, this is a typical problem in the Southeast, including Florida.

National Weather Service Digital Forecast Database errors in 12-hour probability of precipitation (PoP) for 72-hour forecasts in July 2015 and 2016. Greater errors shown by orange and red X's. Lower error depicted by green squares, according to legend in lower left.

Coastal California

Forecasting in California is challenging? 

Hear me out on this.

First, consider the still relatively data void area upstream known as the Pacific Ocean. 

Data from satellites and trans-Pacific flights does help numerical forecast models, but the lack of surface data still lends itself to uncertainties in forecasting strong Pacific storms.

Visible satellite image of the southern half of California on Oct. 18, 2011 at 8:30 a.m. PDT. The cloudy "marine layer" is shown hugging the California coast in white.

However, my first of many humblings as a meteorologist was my first job out of college for a private weather company that provided forecasts for several California newspapers, including the Los Angeles Times.

These included highly detailed temperature forecasts for California's many microclimates. 

Often, a persistence forecast, one similar to what happened the previous day, worked well.

But at the edge of the layer of cloudy, cool air from the Pacific – the marine layer – is where temperature forecasting was difficult. 

A seemingly tiny error in placement or persistence of the marine layer could mean a large temperature forecast bust.

Even in this age of supercomputers capable of quadrillions of calculations per second and new satellites capable of imaging the entire hemisphere every 15 minutes, there remain many forecast challenges today.

Jonathan Erdman is a senior meteorologist at and has been an incurable weather geek since a tornado narrowly missed his childhood home in Wisconsin at age 7. Follow him on Facebook and Twitter.

MORE: Largest Temperature Swings in Each State

The Weather Company’s primary journalistic mission is to report on breaking weather news, the environment and the importance of science to our lives. This story does not necessarily represent the position of our parent company, IBM.