Insider Insights: Articles

Close

Country

Confessions of a meteorologist: How often are we wrong?

What many people don’t realize, is that some parts of the forecast are more uncertain than others. Courtesy: Gail Cas

What many people don’t realize, is that some parts of the forecast are more uncertain than others. Courtesy: Gail Cas


By Gina Ressler
Meteorologist
@ginaressler
Tuesday, June 18, 2013, 12:08 PM

Canadians love to talk about the weather. It's a subject that simultaneously fascinates and humbles, but it's also a topic we love to complain about. So I’m sure you can imagine the reaction I get when I tell people that I forecast the weather for a living. Many smile and confess that they check the forecast religiously before leaving the house each morning. But others smirk and remark that they wish they had a gig where they could be wrong 50 per cent of the time and still keep their job. (If us meteorologists had a nickel every time we heard this joke!) We laugh it off, but it does beg the question: How often are we wrong? And does this mean we’re doing a bad job?

One computer model predicted Hurricane Sandy's landfall in the NE US last October over 7 days in advance. (file photo, OCt 28, 2012, courtesy: NASA)

One computer model predicted Hurricane Sandy's landfall in the NE US last October over 7 days in advance. (file photo, OCt 28, 2012, courtesy: NASA)

The answer actually depends on what we're forecasting. Let’s break it down a little. We know that weather forecasts are inherently uncertain. (We're predicting the future, after all!) But what many people don’t realize, is that some parts of the forecast are more uncertain than others. In fact, the amount of uncertainty or confidence in a forecast depends on a number of factors: the location, the time of year, the current weather pattern, and the skill of our computer models. On top of all that, forecast uncertainty increases as we go further out into the future.

Today, data from computer models form the basis of almost all weather forecasts. (We call this Numerical Weather Prediction, or NWP.) Over the last few decades, with better and better model data, forecast accuracy has improved by leaps and bounds. In fact, five-day forecasts are about as accurate as two-day forecasts were 30 years ago. That's pretty good. Even our long range (1-2 week) forecasting abilities are getting better. (One computer model correctly predicted Hurricane Sandy's landfall in the northeast US last October over seven days in advance. That's astonishing.) Seasonal predictions still have pitfalls, but skill is steadily improving thanks to research into atmosphere-ocean connections.

Great progress has been made in the field of Numerical Weather Prediction. But there's still work to do. One area that has lagged behind in forecast accuracy is summertime precipitation, or as meteorologists call it, summertime QPF (quantitative precipitation forecasts). Why is this?

A brief rainshower is an example of convective weather (courtesy: Christy Guertin)

A brief rainshower is an example of convective weather (courtesy: Christy Guertin)

In the cooler seasons, precipitation usually results from synoptic-scale systems. The term “synoptic” refers to a spatial scale of about 1,000 km (the size of Ontario). Large low pressure systems like Colorado Lows, Alberta Clippers, and Nor'Easters are all synoptic-scale weather systems. Our computer models can simulate these really well, often 5-7 days in advance. 

It’s a different story in the summer. 

For much of Canada, summertime precipitation is convective, meaning that it results from instability in the atmosphere.  on an otherwise sunny day, dangerous supercell thunderstorms, or large mesoscale convective systems. Summertime precipitation is often very localized. Thunderstorms may only be a few kilometers across, yet can bring up to 50 mm of rain in less than an hour. Our computer models simply don't have the resolution to pinpoint exactly when and where these thunderstorms may hit, and certainly not days in advance. If the computer models do resolve them, they’re often not in the right location, or the timing may be off. What the models can predict are the ingredients needed to produce thunderstorms: moisture, instability, and a lifting mechanism (or trigger). It's then our job as meteorologists to analyze the ingredients and create a “most likely” solution. This is why in the summertime, instead of focusing on exact locations and amounts, we focus on risk areas – areas where we see the necessary ingredients coming into place.

So if you see a 60 per cent chance of thunderstorms in your forecast, be on alert. You're in the risk area. If you don't end up seeing a thunderstorm, it looks like we got the forecast wrong. But there's a good chance thunderstorms developed somewhere in your area – in the risk area. If this is the case, then our forecast worked out. If no thunderstorms developed anywhere in the risk area, then you can complain! (We'll be complaining, too.)

Often times in weather we want one answer. Will it rain? Yes or no. But in weather forecasting, it's never that simple. Every forecast comes with a probability, or a level of uncertainty. Some areas (like summertime QPF) will continue to challenge us; our forecasts will never be perfect. But huge accomplishments have been made in the science of weather forecasting over the last few decades, and if that continues, maybe I'll be less afraid to tell people what I do for a living.

More by this author

Leave a Comment

What do you think? Join the conversation.
Default saved
Close

Search Location

POINTCAST

Look up Canadian postal code or US zip code

Close