Computer models key to multi-day forecasts

Published 7:00 am Tuesday, February 17, 2015

Often in this column, I describe in general terms what to expect in the upcoming week’s weather. I base my description on information issued by the National Weather Service (NWS) on Monday morning, including their official forecasts for Pearl River County, their scientific discussions of the “why” behind the forecasts, and the weather patterns being predicted by the computer weather models run by the NWS late Sunday night.

The NWS issues forecasts twice per day covering the upcoming seven days. Many private forecasting companies, such as The Weather Channel and Accuweather, produce even longer-range forecasts and make them available on the Internet.

How much confidence should you put in a seven day forecast?

Sign up for our daily email newsletter

Get the latest news sent to your inbox

Common sense, and actual verification statistics, tell us that on average we can trust the shorter range forecasts more than the longer range forecasts. This is especially true in the more variable the weather patterns, such as ours during the non-summer seasons. (During the summer, most of us can usually make a pretty accurate 10-day forecast by simply predicting, “Hot with a chance of afternoon thundershowers.”)

Before computer weather models became operational in the 1960s, weather forecasters did not have much skill forecasting beyond the next day or two. But, as scientists’ understanding of the physics of the atmosphere has increased, they have built increasingly sophisticated computer weather models. More and better weather data from satellites are input into the models, and the computers on which the models run have become bigger and faster allowing the models to resolve finer scale details and to produce longer range predictions. Forecasters can now look out a couple of weeks into the future.

When model forecasts up to one week in advance are compared with what actually occurs, the models demonstrate a level of accuracy far beyond what is achievable by forecasters without the models.

But, those models are not perfect representations of the atmosphere and the underlying ocean and land. Also, the observations used as input to the models contain errors, and there are not enough of those observations to capture all that is going on in the real atmosphere. Combine this with the chaotic fluctuations inherent in the atmosphere and ocean, and it is not surprising that every model’s forecasts become less accurate, on average, the further into the future the model predicts.

In future columns, we will examine how forecasters use predictions from a variety of models to come up with their forecasts, as well as the statistics for how accurate those forecasts actually are for temperatures, precipitation, and wind. We will also explore how the needs and perceptions of different users can make a huge difference in how good a forecast needs to be for that user to consider it “correct.”

This week the models and the forecasters who use them are predicting dry weather and colder than normal temperatures on Mardi Gras through Thursday with freezing temperatures likely Tuesday night and Wednesday night, and possibly Thursday night also. As warmer air returns on Friday through Sunday, so will a chance of showers.

By Skip Rigney.