Weather Blog

UW's "Probcast" takes different tack on weather forecasting

UW's "Probcast" takes different tack on weather forecasting

Meteorologists over at the University of Washington are trying a slightly different approach to generating a forecast -- using what are called "ensemble" forecast models.

Any one forecast model is subject to errors, and each have their own bias. And when forecasting models begin to vary amongst themselves, the confidence in a forecast drops. I try to point out when this is the case on our online weather forecast discussion.

What the UW is doing is trying to capture the essence of that uncertainty, giving a solid forecast, but also giving a range of possibilities, with wider ranges given when uncertainty in the forecasting models is higher.

In their case, they are taking their high-resolution MM5 model, and then giving it slightly different scenarios based on current weather to get a range of what could happen.

Here is the link to their forecast page. It defaults to the location of the University of Washington, but you can put in your own zip code or click on the map to get a spot closer to where you live (unless, you live on the campus, in which case, you're saved a step. Just one more reason to tell your parents you want to go to UW, kids. "Mom, Dad, just think of the savings in carpel tunnel treatments!")

The model stretches across Washington, Oregon, and slim parts of southern B.C and western Idaho.

Note the forecasts are still given in the familiar "best chance" single number high and low temperature format, but are also given in a range of "as high as" or "as low as".

We've found that it's had a fairly good track record so far, but be aware that it is still experimental, so it's prone to some errors here and there. But just one more spot to check out as you poke around the 'net for a forecast.