Five days seemed like a nice range because it’s what all of the services purport to predict. However, I’d like to better quantify the quality of forecasts and will likely do another weather forecast rodeo in March.
Next time, I want to better standardize terminology for the weather condition. For example, the Seattle PI/KOMO TV4 lists their forecasts as “cloudy,” which I think is incredibly wimpy.
Other weather sources provide a scale — sunny, mostly sunny, partly cloudy, mostly cloudy, overcast — which is a similar concept to the aviation weather forecasts using “clear”, “few”, “scattered”, “broken”, and “overcast” as an indicator of the percentage of cloud coverage. Hey, we love our clouds. To quote Steve Pool, “we average 226 cloudy days [..] and 81 partly cloudy days” a year.
Because I’m a math geek, I also want to come up with a formula that will assign a point value for the forecast precision. My thinking is there should be some enumeration for the types of weather conditions and points assigned to how far off they are. I would also like to have a bonus category for future prediction accuracy because, quite frankly, predicting weather five days in advance is pretty tricky.
Similarly, there should be some point “penalty” for radical changes in forecast. For example, going from “partly cloudy” to “snow” a couple of days later suggests the computer model may closely resemble a
Finally, I’m going to inquire about National Weather Service forecasts as they’re very broad for a test like this.
TUESDAY NIGHT…EVENING RAIN OR SNOW THEN RAIN… RISING TEMPERATURES IN THE MID 30S TO LOWER 40S. SOUTH WIND 15 TO 20 MPH.