Wednesday, July 9, 2008

Ever noticed how weather forecasts are often wrong? Me too. Ever wonder if there was a way to determine which weather source was the most accurate? Me too. Ever spend over a week writing down forecasts from three sources and then putting them into a database so you can devise calculations to find the most accurate? Me too.

Those not prone to geekery, please stop reading now.

For each of the three local weather sources, two television stations and weather.com, I wrote down their forecast for the next seven days, including the current day, for both the high temperature and conditions. Eventually the conditions were broken down into Sunny, Partly Cloudy/Partly Sunny, Chance of showers/snow or scattered rain/snow or part day rain/snow, and Rain/Snow. There was a slight bit of human element in converting the various phrases for each source to these categories ("Partly Sunny with Scattered Thunderstorms" becomes, um, "Chance of Showers"). The day after I would get the conditions and high temperature from Weather.com.

The analysis was in two parts.

Temperature: One point was given for every degree they were off on their forecast seven days away, two points for every degree they were off on their forecast six days away, three points... etc.

Conditions: I gave much more weight to conditions then temperature. If they say it's going to rain in three days and you change your plans, well, that's much more serious than if it was four degrees warmer than forecast. Each of the assigned conditions were given a value based on severity - 1 point for Sunny, 2 for Partly cloudy, etc. The difference between the forecast condition value and the actual condition value was then multiplied by five for the seven day forecast, ten for the six day forecast, fifteen for the five day forecast, etc. The logic of this is that if they are wrong on conditions a week from now, well, no big deal. If they louse up tomorrow's forecast then someone needs more training.

The lower the number the better the forecasting abilities. Make sense? Should I get a life?

I found that my previously preferred source was the worst of the bunch. ACK!



WANE TV - 325

Indiana News Center - 277

Weather.Com - 273


Admittedly, this is only based on six full days of data and a handful of partial days. I intend to analyze a two week span soon. Results will be posted.

3 comments:

Me said...

Yet we haven't had time together in three weeks.

Uvulapie said...

Yet this was all done on non-personal time :)

Big Doofus said...

I'm a weather geek (did you know that?). I think it's rather hard to accurately predict what will happen, but you can do a pretty good job on your own of paying attention to weather patterns, etc.

Personally, I rely mostly on radar. I'm not too concerned about temperatures as I figure I can usually adapt unless I'm swimming or sledding. I watch for rain on the radar or the possibility of rain and I'm usually pleased with what I can see. Forecasts are for ninnys.

Start here:

http://radar.weather.gov/radar.
php?rid=iwx&product=N0R&overlay=
11101111&loop=no
(paste that together into one link)

The resources on this site are great (your tax dollars at work). You'll love it.