Browse Category

verification

Snowfall Verification for Jan 26-27

Using another snowfall site from the NWS in addition to the site I linked in my previous verification post, I came up with this for the snowfall totals:

 
The most striking differences are the snowfall totals in WV and the snowfall totals from MD north of DC up through southeast PA and into central NJ. These areas were under-forecast by 2-4″, which was partly due to the better-than-expected snow totals from the first shot of precipitation (which was 1-4″ when I was expecting 0.5-2″) and partly due to the northern extent of the snowfall, which was about 50-100 miles further north than I (and the models) predicted for the 1-2″ contour. This snow also helped the second snowfall to start accumulating a bit faster since there was no loss due to melting (which would have been minor anyway). The higher totals in central NJ were due to better banding than I expected. Also, the conversion over to snow happened a little more quickly than forecast, which accounted for an additional 1-2″ of snow.

Overall, I’d give my forecast a B. The general contouring and cut-off areas were excellent, but higher snow totals brought down the grade.

Snowfall Verification for Jan 20-21

I found a link on NCDC that can help me do snowfall verifications. I haven’t found an archive of the data yet, but at least I found a way to easily verify my forecasts.

Here is the 3-day snowfall total map, ending at 15z today:

 
Here is the comparison, using my contouring (with some smoothing to the actuals):

 
The mixing in the southern parts of the contours proved to be further north than I had originally anticipated, which was due to better-than-expected daytime heating and a slower development of the coastal low. The western side of the Appalachians got a bit more snowfall than I expected as lake-effect and residual moisture was able to have more of an impact.

Overall, not a bad forecast, but certainly one that needed improvements in some areas. Overall, I would give it about a C grade.