It’s the driest summer in recent memory, and for months, many Minnesotans have been praying for rain, cursing patches of dead grass that once passed for lawns and pulling out the trusty smartphone to check the weather forecast.
“No rain today,” it always seems to say. “Maybe in a couple days.” “Maybe next week.”
When those supposedly rainy days finally roll around? Often nothing. Or so it seems in this prolonged drought. So we drag the hose out yet again to keep the tomatoes on life support, thinking: “Is it just me, or are the forecasts broken?”
There are no public real-time records of how accurate the forecasts are, so it’s tough to say whether this year’s forecasts are more or less accurate than in years past.
In general, the National Oceanic and Atmospheric Administration says a seven-day forecast accurately predicts the weather about 80 percent of the time. A five-day forecast? About 90 percent of the time. But a 10-day or longer forecast is only right roughly half the time.
In Minnesota, forecasting is tougher to do in the summer, said Bob Weisman, a retired professor of meteorology at St. Cloud State University. “We, overall, are not as good at forecasting the weather during the summer, where thunderstorms control the weather pattern,” he said. (Weisman forecasts the weather for the St. Cloud area on a St. Cloud State website with the tagline, “Please note that I make the forecast, not the weather!”)
“Summer is the least accurate in any of the forecasting because on a thunderstorm day, the whole stage can be set by the thunderstorms that develop the night before, and we won’t have a really good feel for where they are until a few hours after sunrise that morning,” he said.
This particular summer, much of the weather came from the northwest — not an area where there are particularly many instruments measuring the upper atmosphere. Canada has fewer weather balloons measuring the air in the upper atmosphere than the U.S., and many of the observations from the Pacific off the Alaskan Coast come from commercial aircrafts.
“So, many of our weather systems that had the potential of producing thunderstorms were coming from a relatively data-sparse area. So it was really hard to pin down which storms had the chance to produce these big convective blobs,” Weisman said — particularly not more than a day or two in advance.
Perception vs. precipitation
Maybe the bigger problem with the forecasts is us. Regular people looking at weather apps are not always well-equipped to interpret what the forecast means.
There are two types of precipitation forecasts we typically see when we check our weather apps.
The first is the “probability of precipitation” model. That’s what you see when you open the weather app and it says there’s a 30 percent chance of rain on Saturday.
“Whoever is forecasting that is saying, every time that I say 30 percent chance of rain, three out of 10 times, it’s going to rain,” said Matthew Vaughan, assistant professor of meteorology at St. Cloud State University.
Then there’s the QPF forecast, or quantitative precipitation forecast. This is when meteorologists predict the amount of rain that will fall in a given place, Vaughan said.
In the summer in Minnesota, precipitation forecasts’ performance tails off two to three days out because of all the instability in the weather, which makes it easier for models to produce predictions of precipitation, said Michael Griesinger, lead forecaster at the National Weather Service in Chanhassen.
Based on climatology alone, the baseline chance of rain on a given normal summer day in Minnesota is usually around 30 percent. “In other words, each day through the summer sees precipitation about 30 percent of the time, so if you ever see a 20 percent chance of rain, we’re actually saying the chance of rain is BELOW normal,” Griesinger said. “You tend to see 30 percent-or-so precipitation predictions more often when you’re looking more than three days out. As you get closer and closer, those chances typically go up or down. This summer, they’ve more often gone down than up.”
But if you really, really want rain, and there’s a scant chance of it in the forecast, you might over-inflate the likelihood of a small chance of rain.
“A 20 to 30 percent chance of rain is still like, a 70 to 80 percent chance of not getting rain, so even though rain is possible in the forecast, there is still a chance of it not happening,” said Caleb Grunzke, a senior meteorologist at NWS in Chanhassen.
Apps enable our overestimating our handle on the weather, Weisman says, by giving us single-point temperature predictions and chances of precipitation, sometimes down to 5 percent intervals, which can give an illusion of precision that’s just that — an illusion.
“It’s being presented in a way that looks more accurate thank I think it is,” Weisman said — making it look like forecasters can predict the weather with precision far in advance, and then making it look like they’re wrong when the predicted weather doesn’t pan out.
Weisman said he’s started to provide some indication of the confidence level he has in his forecasts to make it clearer there may be some uncertainty.
“In my forecast discussion that I put out for St. Cloud, [there’s] a reliability level,” he said. “I’ll go out with a forecast out five days, but I’ll rank each period, tonight, tomorrow, on a zero to 10 scale of how reliable I think it is. And it is rare over the summer that I would give it more than a five or a four beyond tomorrow — especially if there’s a thunderstorm forecast, because it’s just not that sure.”
Trying to predict the weather more than a few days out? In the summer? In Minnesota? You might be better off just enjoying the present.