Dave's Sweater Now Tracks 17 High Country Towns

For four months this site graded every forecast anyone published for one town. Since late July it has done the same thing for eighteen: Boone plus 17 more places across Watauga, Ashe, Avery, Mitchell, Wilkes, and Yancey counties.

What a town gets

Each town now has its own page with a multi-source forecast computed at that town’s actual coordinates. Beech Mountain gets forecasts computed for 5,436 feet. North Wilkesboro gets forecasts computed for 1,001 feet. Those are two different forecasting problems, 4,400 feet apart, and on the day we switched the pipeline on, the two towns’ forecast highs were 26 degrees apart before lunch.

How much of a regional forecast is really local is a fair question to ask about anyone, us included, so here is the answer for the incumbent. On a single afternoon in July, Ray’s Weather gave Boone, Blowing Rock, West Jefferson, Banner Elk, and Beech Mountain the same sentence, word for word, wind and all: “Mostly sunny; Hazy skies at times; Warmer; NW wind 5-15 mph.” Those five towns span 2,400 vertical feet. Each page swapped in its own high, from 81 down to 72. Drop off the mountain to North Wilkesboro and the description does change, so the writing is regional rather than one-size-fits-all, but inside the High Country a town’s own words are its neighbors’ words. That is why we grade him on the high, the low, and the sky icon, which are genuinely per town, and leave his wind unscored rather than mark him wrong for a number he never claimed was Beech Mountain’s in particular.

Every town is also graded. The same 100-point rubric that has scored Boone forecasts since March now runs against each town’s own verified conditions, every day. A town’s scoreboard appears once it has nine scored days behind it, which is the same bar we made every forecaster clear before ranking them. Nothing gets ranked on a hunch.

The honesty rules came along

Two things we refused to do while multiplying. First, no borrowed coordinates. Every town’s pin was geocode-verified before a single forecast was pulled, because a forecast for the wrong valley is worse than no forecast. Second, no borrowed stations. Ray’s Weather publishes town numbers for most of this set, and where he genuinely has a station in a community, his column is there to be graded. Where he doesn’t (Seven Devils, Sugar Grove, and Wilkesboro), his column sits empty. We would rather show you a blank than dress up the next town over as coverage.

There is a third thing, and it is the one worth telling on ourselves about. The towns launched with Ray’s scores too low, because our own capture was throwing away the sky icon on his town pages and grading him as though he had made no precipitation call at all. That is a forfeit he had not earned. We fixed the capture, went back through every town day on record, and rescored. His town average went up ten points. Nobody asked us to look. The rubric is only worth anything if it is the same rubric when the correction runs the other way.

What we don’t know yet

Whether elevation makes a town harder to forecast is now an answerable question. The ridge towns and the river towns are being graded side by side for the first time, and a few more weeks of grading will say which forecasters handle the vertical better. We have a guess. We are not going to publish a guess.

The methodology is the same one, in public, that it has always been. Pick your town, check the scoreboard, and if you live somewhere we skipped, tell us and we will run the numbers on adding it.