Ask A Met: How Can Two Forecasts For The Same Place Be Different?

This week’s question comes from Morning Brief reader Sarah who asks, “How can two or more weather forecasts from different sites have such conflicting forecasts for one area?”

Meteorologist Sara Tonks: Well, I’m going to start by talking about the difference between a forecast and a weather model.

This is something that we see a lot online with big storms, and even storms that don't actually end up happening. People will post weather models on their social media pages, and they'll be like, “Oh my God, this model shows this massive storm going to this place at this time.”

But checking one single model is not the same thing as checking a forecast.

We use weather models to make the forecast, but mistaking a single model for a forecast is one way to get very different results for one place.

These computer models are driven by equations. Just like how physicists use math to solve for how fast an object will fall considering wind resistance, meteorologists use math to predict what our atmosphere is going to do next.

You need a supercomputer to do these equations because they involve so much data. You've got data from satellites, you've got weather observations on the ground, you've got weather balloons and radiosondes, and all of that goes into the math.

But even with all of that data, the trick is that the atmosphere is very large, a bit random and chaotic, and since no model is completely perfect, they often have something that they do better than other models.

Do you know the scene of Jurassic Park when Jeff Goldblum’s character explains chaos theory? That comes into play here. Basically: a tiny difference can add up to a big difference.

Think about spaghetti models for tropical systems - all of those lines are different models and even different runs of the same models with different starting points and different settings...and different rounding, too. And the result can be all over the place, because no two models are exactly alike.

At the most base level, different apps and different sites use different models. And even they're all starting with the same data that's rounded and input in the same way, they're all coded differently.

How a model becomes a forecast

There are lots of ways a model becomes a forecast, and some of it is just meteorologists who know the region knowing what is and isn't likely to happen and adjusting the forecast accordingly.

We take into consideration what models are best at certain things. When I was a local meteorologist in Erie, Pennsylvania, for example, we would often just completely throw out some models because they don't understand lake effect snow at all.

Two of the biggest wide-scale global models are the American, which is known as GFS, and the European, which is known as the Euro.

The GFS model is notorious for predicting that in 10 days there's gonna be a major storm. Meteorologists will even make fun of it. It's like, “Oh, the GFS is going on a rampage again,” because the GFS said there's gonna be a major storm, but the Euro, because it's coded differently, doesn’t.

I don’t need to name names with other sites and how they make their forecasts. I do know that the Weather Channel app has consistently been ranked as the most accurate, partially because we're not just running a computer model. Don’t just take my word for it. This comes from an independent study.

We have a whole team of forecasters at the company that use the computer models and then go further into it. They’re constantly refining the models. They’re currently using AI to improve the forecasts, training AI models on past data to make connections between causes and events.

In the end, the differences between forecasts come down to the people making them.

Do you have a question to ask the meteorologists at Weather.com? Write to us and we’ll pick a new question each week from readers to answer.