Weather Blog

Why do we forecast 7 days out?

Why do we forecast 7 days out?

Today is Part 3 of our three part series on frequently asked questions about TV weather forecasting. (Check the blog archive for Parts 1 and 2.  Tomorrow, Part 4 of our three part series: Why math is important :) )

Our Wednesday subject: Why we do the things we do.

First up is why we forecast out 7 days. It's a fair question because once you get beyond 3-4 days, the forecast tends to be more fluid and change more often. But yet, we "weather" on anyway with the attempt.

Believe it or not, our research says people want to know the weather that far out, even if it's in flux. It's true that there are many times when forecast models diverge so much beyond 3-4 days that the forecast beyond will move around a bit, but many times, it's not a radical shift. We also found that people are most interested in the weekend weather, even if only to get an early gauge, and this way, the weekend is always on the map.

Yes, there are times the weekend forecast will shift, (and people *really* notice when it does so) --- case in point, the forecast for this upcoming Tuesday/Wednesday -- but I think the majority of the time, it stays fairly consistent -- it's quite rare where a 5-6 day forecast will go from, say, sunny and 80 to pouring rain and 55 over the course of a week.

(Just like the adage "you guys are wrong 80% of the time" but it's those days (usually involving snow) where it goes wrong that "clouds" the judgment and people hardly remember that, for instance, this past week has gone just about exactly to plan. But no one remembers those :)

It's also why we try to do a more detailed weather forecast on our online discussion, and I'll try and make special mention when forecasts for the days at the end of the graphic have a low confidence

Why do you put so many cities on the map, yet they're all within a few degrees?

It's a hybrid answer of A) there are a lot of microclimates here, and a forecasted high for Seattle is sometimes several degrees different than a forecast for Bellingham, or Forks, or even Olympia. And B) People want to see their town and not just get lumped in with the bigger cities. Like, say, for instance, people in Lake Stevens want to see them on the map, and don't just want to glean off Everett.

It's easy for people in Seattle or Tacoma to wonder why we do it, but we hear from people in places like Sequim or Bothell or Enumclaw that appreciate it.

Why is it so hard to forecast around here?

There was some study I read years ago -- and I can't find it anywhere on Google so if anyone else finds it, let me know -- but it said that the Seattle area was the second-most difficult major U.S. city to forecast for, behind Washington, D.C.

Why? We have a lot of challenges stacked against us.

This region has a nearby ocean, an inland sea (Puget Sound), two mountain ranges (Cascades and Olympics), a rain forest and a desert, all within about 200 miles of each other.

Rainfall varies from 18 inches in Sequim to 100+ inches just a few miles away near Forks, and Seattle gets 38 inches.

That situation stands in stark contrast to areas in the middle of the country where it is flat for miles and weather systems move at more predictable rates.

Our other problem is that we don't have a lot of reporting stations out in the Pacific Ocean where our weather comes from, so the computer models fill the missing data points with an average between two known points.

If we have a small error in that averaged data, the error becomes magnified over time.

Plus, without those reporting stations, it's hard to see what's going on. A satellite helps a bit, but it's like trying to see inside a home from a plane -- you can only see the roof and not know what's going on inside.

So yes, forecasting is quite a challenge here.

Why do forecasts vary among weather sources. You all get the same data, right?

The best way to think about it is if you're standing in an art gallery, and there are 5 people looking at a painting of abstract art, and if you were to ask each one what they though the painting was representative of, you might get five different answers.

There are a host of different forecast models that are put out by various institutions. The National Weather Service puts out a few different computer forecast models -- Some are more designed for short-term forecasts, some are designed for more long-term forecasts. The Canadian government has their own models, as does the European Union.

To top that off, the University of Washington runs their own model now, and we have one that's exclusive to us (That "Futurecast" product we air) -- they both run on a much higher resolution and provide short term forecasts for the Pacific Northwest.

All of the models are available to everyone in the world (except the Futurecast, which is just for KOMO.) Each model takes a slightly different tact in trying to compute the atmosphere and say what's going to happen next.

Now, most of the time, the models are in general agreement, with just a few different tweaks here and there.

Think of it as asking four people how to get from West Seattle to a restaurant in Auburn. They might have minor differences in what streets they take to get there, but they'll usually at least get you to the restaurant, and then it's up to you to rely on experience and expertise as to which is the shortest route, based on traffic, etc.

With the models, you learn over time which ones perform better in different situations.

There are some days, and some weather patterns when it can become a meteorological. That's because none of the models are on the same page -- it's utter chaos.

It'd be like you want to go to Auburn, but one person gives you directions to Tacoma, the other to Arlington, the third to Bremerton, etc.

In that case, now you have my first example. Everyone in the weather community looking at a piece of abstract art and trying to figure out a forecast amid the model chaos. Some may be taking model A, some may be going for B, etc. Some may be going on just looking at a satellite photo and throwing the models out of the equation completely.