Hi Everybody,
I think I’ve got a new version worth testing, but still a bit for official release. I figured some of you would want to try it out!
First, here’s the main new feature:
The retrieval module (wret.exe) now has an ‘auto select’ feature which allows quick analysis of a very large number of old forecasts. This feature also allows you to save the results, which (if you have professional mode) can be used to fine-tune WXSIM’s forecasts! (in terms of temperature and humidity, anyway).
I have not yet written a thorough description of how to use this. There is a new blue-text help item in wret.exe that more or less explains it. I’ll supplement just a bit here.
When you open wret.exe, you’ll see some new buttons. Select a forecast from a while back (months ago, if you have it) and click Compare to Actuals. This sort of initiates things so the new feature can work. (You do NOT have to check the ‘Combine’ box.)
Next, click the ‘Set as Earliest Date’ button. After that, click on a more recent forecast (but at least four days old). Then, click ‘Auto Select’. A bunch of stuff will happen, including the display of a bunch of text data. In some of my testing, there have been formatting/scrolling problems on this form, and the data has been presented out of order, but it’s all there.
Several (hopefully somewhat self-explanatory) parameters will be listed, and you can view a couple of (hopefully somewhat self-explanatory) graphs. The really good part, though, is that you can click YES so the results will be saved as cor.txt.
This file can be read by WXSIM (if you check the new box that appears after you click READY/GFS Bias Factors button on the import form). From that point onward, WXSIM will record in the .wxf files what correction factors it just used, so they can be iteratively refined with future uses of the feature.
Roughly equal weights are given to the last 10 days of data, the last 40 days, and the whole group (months), so the values will slowly adjust, but will have enough there to keep them fairly stable. There’s a variable which is recorded if you don’t get enough model, etc. data for a “good” forecast, and such forecasts will (I think) be properly omitted from future analysis. Likewise, forecasts with average temperature errors over 10 degrees C (18 degrees F) will be omitted (as there’s almost certainly some error beyond noral forecasting errors).
After you run such an analysis, I think you may get an error if you go straight to the Compare to Actuals button again. however, you can use the Set as Earliest Date button, and select different ending times, and run it again.
By the way, your forecasts need to go out at least four days, because the analysis is made of days 2 through 4.
OK, to get the new files, just download these and use them to overwrite existing. I haven’t integrated them into an installer yet, so do this “by hand”.
www.wxsim.com/wxsim.exe
www.wxsim.com/wret.exe
www.wxsim.com/help2.txt
I wouldn’t be surprised if there are some bugs or confusing items, but it’s basically functional for me, here.
I plan to be out of town for a couple days, but I’m curoius for feedback when I return!
Tom