It’s just before noon on a cool August day at Millersville University in Lancaster, Pa.

As a gentle breeze blows in from the patio of his fourth-floor laboratory, Kyle Elliott settles into a chair to give his medium-range forecast on the Millersville University YouTube channel.

“It’s going to be sunny and comfortable.” Two words you often don’t hear together in August, but that’s what makes weather forecasting fun for Elliott. It’s unpredictable and, at least in Lancaster County, always changing.

“I just try to produce the best forecast possible,” he says.

Judging by the eight giant TV screens, and the numerous computers and other gadgets in his lab, Elliott, who heads the university’s Weather Information Center, has a lot of tools that help him determine what the weather is going to be like in the days, weeks and even months ahead. How are they helping him predict what this fall will be like? And what kind of impact could this have on your operation?

Here’s a look.

Weather history can repeat itself

Making a seasonal forecast, Elliott says, depends on looking at past data.

Right now, we are transitioning from an El Niño to a La Niña, climatological patterns that originate in the central and eastern Pacific but affect the weather around the world. Elliott looked at past years with similar conditions to see how it could affect the weather this fall.

LA NIÑA TRANSITION: An important element to Elliott’s fall outlook is the transition from El Niño to La Niña. Here, he points to cooling of Pacific Ocean water temperatures, a sign that the transition to a La Niña is happening.

The years 1998, 2005, 2007, 2010 and 2016 were the most similar. Using a program on his computer, he can put in a variable such as precipitation or temperature to find the “anomaly” — how much more or less precipitation occurred in those years, or how much colder or warmer it was — compared to relatively normal conditions.

"So, if I blend August through October of those five years, we see below-normal precipitation," Elliott says, adding that he also expects a warmer-than-normal fall.

“These transition years from El Niño to La Niña tend to have very hot summers that persist through September into early October," he says.

But there is an important caveat: the tropics. Right now, water temperatures in the Atlantic Ocean are unusually warm, which is good for tropical storm development. Another factor is the Bermuda high. If it drifts closer to the U.S., this could allow tropical storms to come up the Eastern Seaboard, just like what happened with Tropical Storm Debby earlier this month.

It’s getting warmer

Elliott says that data clearly show that summers are lasting longer — by weeks in some years — and falls and winters are delayed.

"You don't have the first freeze or frost in the area, on average, until late October or early November. One hundred years ago, it was early to mid-October," he says. "So, your growing season is extended by two to three weeks on average now."

The last freeze, which in some years was late April or May, is also happening sooner.

"Think of how many times you see a freeze warning after April 15. Frost advisories, yes. But a freeze warning? You rarely if ever see it anymore, and the average last freeze used to occur around May 1,” he says.

But since the mid-1990s, temperatures have skewed warmer.

These graphs show departures from normal, plotted by Millersville University, by month and year since 1990.

"People may not agree on ‘climate alarmism’ or to what degree climate change will alter our future, but something concerning is going on," Elliott says. “How much is manmade or natural is up for debate, but both are contributing to a warming climate.”

"I do think the data is a little better now than it was in the past. It's a little more accurate," he adds. “But it’s not enough to really affect the official records in terms of temperatures, precipitation, snowfall and those sorts of things.”

Technology isn’t perfect

Weather forecasting has come a long way from the days of Galileo’s thermoscope — in which he used the principle of buoyancy in a glass tube to measure temperature.

On the eight TV screens lining Elliott’s classroom wall, he can bring up the latest satellite images from space, sea-surface temperatures and a lot more.

He uses new tools like the national blend of models from the National Weather Service to help come up with as accurate a forecast as possible and to teach his students the art and science of forecasting.

But advances in technology still can’t replace the most important element: the human element.

Elliott explains.

"A clear night after a fresh snowfall, with a fresh snowpack on the ground, the blend might say 15 degrees. I've seen it do that in York where it predicts a low of 10. The low that night ends up being -5," he says. “Just because of the fresh snowpack on the ground, and then all the heat is getting radiated back to space. So, surface temperatures just bomb out at night. And as a meteorologist, we can account for that, and that's what clients are looking for.”

But he's not perfect. He gets it wrong, too.

One example was in 2018 when he was working for AccuWeather. Just before a November snowstorm that year, the models showed snow changing to sleet and rain in most of the region, and at Penn State where he was working at the time. The models, it turned out, were wrong. Very wrong. It stayed snow for much longer, creating havoc on roads throughout the region.

What the models did not account for, Elliott says, were the strong dynamics and cooling in the different layers of the atmosphere.

“And what we were seeing in the mid-levels of the atmosphere, 5,000 to 10,000 feet above the ground, was not matching modeled precipitation types. The robust dynamics suggested that surface temperatures could stay lower for longer, and that the precipitation could stay snow a lot longer than the models suggested. And both happened,” he says. "It's a sterling example of meteorologists not following their gut instinct and getting it wrong. Lesson learned: Follow your gut. What happens in the mid-to-upper levels of the atmosphere will always trump what a model suggests at the surface. Because precipitation intensity, type and location are controlled by and ‘follow’ the dynamics there.

"Everybody got it wrong!”

Some tools you can use

According to Yahoo Finance, the weather forecasting services market was estimated at $3 billion this year and is expected to grow to $5.8 billion by 2030.

So, if you want a forecast for a specific field at a given day and time, you can get it — just for a price.

There are many tools you can use on your own that are free. Be ready to spend some time learning what they do: