Why forecasts don't go much beyond a week

There is a good example this week of why weather forecasters have a propensity to get gray hair, and also why weather forecasts don't go much beyond a week.

On Wednesday, long range forecast models had gone a few runs in a row showing a decent sunny streak around the weekend of Sept. 25 and lasting into early the following week. At face value, we were probably talking sunshine and highs in the low-mid 80s.

Here is the model run from Wednesday evening, showing a forecast for 5 p.m. on Sunday, Sept. 26:

Note the really big ridge of high pressure over the West Coast with a decent thermal trough moving up the coastline. That is a pattern where you check record highs for the period and start thinking they are in jeopardy.

Now, here is the model run for that same Sunday evening, created on Thursday morning:

Whoa! It shows a big cold front draped over the Northwest with a steady heavy-to-moderate rain. If the first panel is sunny and 85, this one is driving rain and 60.

Look at the other major differences. The first panel has the ridge of high pressure over Nebraska and a big low over the northern end of Manitoba. The second panel has the high over New Jersey and the low approaching Greenland -- remember, this is forecasting the same time period, and created just 18 hours apart!

How could this be -- taking away our sunny and 85 and replacing it with another rainy weekend? (Wait, it's Seattle. We know the answer.)

Anyway, the way computer models work is that we take weather observations from around the globe from various sources -- such as weather instruments on the ground, ships at sea, weather balloons, satellites, pilot reports, etc. All that data then gets fed into the computer, and using what we know about how the planet and dynamics work, we apply incredibly complex mathematical equations to that data to try and figure out how the conditions right now will change over time.

The problem with forecast models is that the atmosphere is really, really big, and computers are still not yet fast enough to be able to account for its entirety.

Think about just how much air is on the planet -- the surface area of Earth is roughly 15.7 million miles and the troposphere, which is basically where our weather occurs, extends from the ground to roughly 40,000 feet (or roughly 7 1/2 miles) high. That's an incredible playground for weather to be created and move around.

To have a perfect computer model, we would need to know what is going on at every parcel of air at a given moment, have a perfect terrain map of every inch of the globe, and then have the perfect mathematical equations to calculate that data, and account for every nook and cranny on the planet that could affect the data.

But we don't know what is going on at every parcel of air. So what we do is take what observations we do have, and then extrapolate the data to fill in the gaps, which leaves us prone to errors if we make assumptions on what is going on in those gaps, and those assumptions turn out to be wrong.

Check out my blog entry "Forecast models: Just really expensive dart boards?" for more information on how models work.

But what is likely happening here is that the model is making an error in trying to calculate the atmospehre, and the error is getting magnified over time.

For example, we're going to build a computer prorgram that will predict the next number in a series of seven - much like computers trying to predict the next day's weather over the course of the week, which is of course, dependent on current weather. Mother Nature has secretly decided the perfect solution is to just take 2 and double it each day, so the actual series we are looking for is 2, 4, 8, 16, 32, 64, 128

But our forecast model gets an error and decides to triple the number on day 3. Now it's going to predict the series is 2, 4, 12, 24, 48, 96, 192. Note how even with just one error and the other 6 days running correctly, our forecasted numbers for the end of the "week" are a mess, even though it did the calculations correctly for almost every day of the week.

And that is how a sunny and 85 degree forecast can become a rainy and 60 degree forecast.

Then again, that is also why forecasts don't go much beyond what they do, as model errors get magnified more and more over time, so the farther you go out in your forecast, the greater the chance is for problems.

Now many of you are probably thinking 7 days is too far, and while beyond 3-4 days the score does decrease a bit, more times than not, the trend ends up being right, even if the exact high temperature forecast might not pan out. In other words, it is rare that a forecast of sunny and 85 on Day 7 ends up being rainy and 65 -- note that the example above is still an 11 day forecast as of this writing.

And that also gives the forecast plenty of time to revert back to sunny and 85 :)