### The Difficulties of Weather Forecasting

Jan. 24th, 2013 06:30 pmThis used to be on another online forum, and because I and others have referred to it recently, I decided it should be on my blog:

Complaining about “missed” weather forecasts, or commenting about how being a meteorologist is “the only job where you get to be wrong half the time” are part of one of my biggest pet peeves, and this was the case well before I headed down this career path.

Try this at home. I dare you:

- Take a huge, not quite spherical ball.
- Cover it with a very irregular surface, about 3/4 of which is water.
- Cover that with a thin mixture of gases that gets exponentially thinner as you go up from the surface.
- Tilt it 23.5 degrees, with a bit of wobble.
- Spin it.
- Heat it irregularly via a huge ball of fusing gas about 93 million miles away.
- Send it around that ball of gas in an elliptical orbit so its distance to the heat source changes.
- Accurately tell me what all the gases covering the first ball are going to do at any given time and place in the future.

Go on. I’ll wait.

The other problems?

1. We do know the set of mathematical equations that describe how that mix of gases flows around our ball, with and without varying amounts of water vapor. Unfortunately, that set of equations cannot be solved exactly by any known means. So, researchers and forecasters build models that have to make approximations. Approximations introduce some error.

2. We don’t have accurate observations in enough places, or at often enough time intervals. I’ve heard via a third party someone (on faculty at Purdue, supposedly, but I don’t know whom) theorized that if we had observations about every square kilometer and every 300 meters in altitude through the troposphere (the bottom 11 km or so), we might have enough observations to calculate within the forecast models with an extremely high degree of accuracy.

Yes, forecasters make mistakes. Now you know why. And with advances in knowledge of how the atmosphere and clouds work, along with advances in computing power, forecasts in general for 72 hours now are as good as a 36-hour forecast was 15 years ago.

[Edited to add: Also, look back just a few months at Hurricane Sandy, particularly when and where it was forecast to make landfall several days before it did so. Nailed it.]

Originally published at Abnormality Locality. You can comment here or there.