On a recent Friday afternoon I received an email sent to a group of people that suggested going to a DC United game the following Saturday. The emailer wrote "weather looks like it's going to be great!" with a link to a 10-day forecast showing sun and mild temperatures with a 20% chance of rain. 8 days went by and the weather on that Saturday turned out to be less great than anticipated.

(from Jorge Quinteros on Flickr)

I'm really glad Greg Postel published this post over at Capitol Weather Gang about the accuracy of advanced forecasting. It's worth a read, but if you're not going to click through, the short-version of is that Greg uses empirical data and shows  that forecasting accuracy decreases the farther you get from the actual date in question. It's not a surprising finding, but one that's worth re-enforcing.

Capitol Weather Gang is the best weather source around. The reason it's so great is because they not only give a forecast, but explain why they think it's going to happen, what the alternatives might be, and how confident they are in the forecast. I've never seen TV forecasters go into this level of detail. I've certainly never seen it in a generic 10-day forecast.

From what I understand, weather forecasting is a specialized skill. It's not something that an untrained person can really do especially well. I'm confident that this is the reason most people misinterpret weather forecasts.

For example, what does it mean if a forecast says 30% chance of rain? Does it mean that 3 out of 10 days this forecast appears it will rain? Does it mean that 30% of the coverage area will see some rain? Does it mean that it will rain for 30% of the day? Further, how much rain is going to occur? If it rains lightly for 5 minutes and is sunny the rest of the time, does that satisfy the condition? How does this apply to percentages given in the hour-by-hour forecasts?

These are all questions that a trained meteorologist could answer but that a lay person probably could not. Still, people often look-up a forecast up to 10 days in advance and state with confidence that it's going to be correct. That's probably a mistake.

Since I bicycle a lot, weather is more important to my daily life than it is to someone who doesn't bike very much. The only thing I've found to be truly reliable is the radar. When I wake up in the morning, I check the radar, if it looks like I'm in the clear for the next hour, I ride to work. I do the same thing in the evening. A forecast could show 90% chance of rain in a given day and I might be perfectly fine to bike to and from work without getting at all wet.

0 comments: