I’m pleased to announce that, at my request, Sam just extended the GFS data out through 240 hours, which means that WXSIM’s longest forecasts (of 9 full days) are completely covered, even if the data is a few hours old. Of course, I don’t know if days 8-10 of the data are particularly reliable, but it’s certainly better than nothing, and it might be interesting to see what’s “on the horizon” in these forecasts.
I just ran a test forecast with the new data (which was posted less than 15 minutes ago) and it ran fine. Some users with older versions of WXSIM may have problems, though. The solution to that is to upgrade to the latest version - free, as usual, by using
Another change we are considering within a week or so is to start using the 0.25 degree resolution data instead of 0.50 degrees. This may improve accuracy in coastal or mountainous regions. The GRIB files we have to download are four times the size of the current ones, though, so we will do some testing.
Through all this, the McMahon data remains the same (180 hours, 0.50 degrees) and you can use that with older versions of WXSIM anyway. Chris has a pretty reliable system going there. Sam and I will do the experimenting and let Chris’ be the “old reliable” for now.
Hopefully these changes will open the door to another wave of (small, but still worthwhile) improvements!
[quote]But even weather models have their limitations: Any forecasts more than a week out are going to be less accurate than climatological forecasts on average, which we
Sam’s new data seems to run fine here in Sim 2015, Build 2.0. Could only run out 9 days due to WXSIM limit. The gfsadv files did contain data out 240 hours and the lat/long step size was 1 degree. A 0.5 degree step size still would be nice…I think.
The 1 degree step size to which you refer is the GFS-for-advection grid, which is in fact denser than the network of advection sites in a typical customization (usually about 200-250, out to about 1200 miles distance). The “home site” data is interpolated off the 0.5 degree data set, but we’re thinking of going to interpolating it off the 0.25 degree set (have to test some because those GRIB files are 4 times the size).
As for the forecast length, accuracy steadily fades out with time, but even at 10 days (or the 9 WXSIM actually goes out to), I think it’s a fair bit better than chance or climatology. I just wouldn’t count on it!
