The length of winter was revealed this morning by Punxsutawney Phil. Outlook: six more weeks. It occurred to me that I should add him as one of my weather forecasting sources, but he’s doesn’t do five-day forecasts, especially in the Seattle area.
The next Weather Rodeo will commence with forecasts rendered on Wednesday evening between 21:00 and 23:00. KIRO TV7 won Round One with the most accurate forecast. Do they have what it takes to pull off a repeat?
I have been working on a system to quantify the accuracy of a forecast. Suffice to say, this was not easy and what I have will certainly require revision. Since there’s no money involved, and I am supreme galactic overlord here, I reserve the right to tinker. Points will be awarded for sky conditions, events, and temperature. Here’s the gist:
- Events
-
- One point will be awarded for accurately predicting rain.
- Five points for accurately predicting any of the following conditions: snow, snow flurries, freezing rain, thunderstorms, water spouts.
- Any source accurately predicting frogs raining from the sky will automatically be declared the winner as soon as I emerge from my underground bunker.
- Three points will be awarded for accurately predicting none of the other events, e.g. the forecast is for no rain and there is no rain.
- Accurately forecasting multiple events will earn points, but inaccurately predicting them will lose the equivalent number of points.
Hey, this is Seattle, and it’s the middle of winter.
These are all rare and represent a gamble to predict.
- Sky Conditions
-
- One point will be awarded for accurately predicting: partly cloudy, partly sunny, mostly cloudy, overcast, or mist/fog.
- Two points will be awarded for accurately predicting mostly sunny.
- Three points for accurately predicting sunny.
- Multiple accurate predictions will earn more points while inaccurate ones will lose points.
(Sunny in Seattle… (snort)
Rationale: I hope to better measure the sense of “five-day optimism,” in the forecasts, but feel that predicting sunny this time of year shows the meteorological equivalent of cajones.
- Temperature
-
- 2 points : within 0 – 5°F
- 1 point : within 5 – 10°F
- 0 points : within 10 – 15°F
- -2 points : > 15°F
- A control case will guess “rain” and “mostly cloudy/overcast.” Temperature will be predicted solely by averaging weather from the last few years on that date.
Exact temperatures will be a tiebreaker for first place only. We’ll award points based on four buckets:
For example, if the forecast was for 45/35, and the actual was 43/37, the delta is 4 degrees (the absolute value of Hactual – Hforecast plus the absolute value of Lactual – Lforecast) and the forecaster earns two points.
(I tried a dart board, but my five year old can’t hit it yet.)
King County International Airport (aka “Boeing Field”), five miles from downtown Seattle, is again the designated forecasting point. Weather Underground will be the official, impartial keeper of hourly weather conditions.
I like the idea of adding Phil to the list. I’d bet anything he’s as accurate as the rest.
As indicated in the next entry, the rules for sky conditions have been modified:
One point will be awarded for accurately predicting partly cloudy, mostly cloudy, partly sunny, overcast.
Two additional points will be awarded for accurately predicting mostly sunny and/or fog.
Three additional points are awarded for accurately predicting sunny.
Phil will be judge emeritus.