Beta6!

Hi Everybody,

A new beta of WXSIM 2022 is ready, implementing ozone data from McMahon GFS! The temis data can still be downloaded and used (except probably not if you are using Windows 7), and is averaged with the GFS ozone data if both are present.

The new beta’s installer is at

www.wxsim.com/wxsimupg2022beta6.exe

I’m hoping this is just about it before an official release. Let me know how it goes!

Tom

Here are the changes to WXSIM since Version 2021, Build 1.1:

(1) WXSIM can now use European (ECMWF) model data, either alone or mixed with the standard GFS data. This is accomplished through changes in WXSIMATE, allowing download of certain now-freely-available ECMWF data, prepared for customized sites by Chris McMahon, in the same basic format as the existing McMahon and Bohler GFS data. WXSIM itself is changed only in having a button (‘Open euromix.exe’, on the Data Import form) for accessing the new program, euromix.exe (described in a new section, on page 49 of this manual). WXSIMATE can be set to obtain both GFS and ECMWF data, and then call on euromix.exe automatically, to mix the data according to user-defined weighting factors in euromix.exe. The only adaptation needed in WXSIM itself is to make sure the name of the file to import (generally wdata.txt) is the same as that being created by euromix.exe. This should be different from the ‘File to create’ in WXSIMATE. You might keep wdata.txt as the file to create, and then perhaps call the mixed-data file euromix.exe creates something like wdatamixed.exe, and import that into WXSIM. Euromix.exe also has a number of data visualization and analysis tools, for studying the two model outputs and their mixture.

(2) The main display text box on the output form has been expanded from 32 KB to 64 KB (I just learned this was possible), so twice as long a forecast can be displayed without getting the “TEXT BOX FULL” message. It should now be able to display all nine days of half-hourly output.

(3) WXSIM can now read 3-hourly ozone forecast values (in Dobson units) from the McMahon GFS data (possibly Bohler also at some later date). WXSIM interpolates among those to get a value at the exact forecast hour. The TEMIS (daily) ozone forecasts can still be used, as well, and if both data types are encountered, these will be averaged. This addition was prompted by difficulty downloading TEMIS data using WXSIMATE under Windows 7 (it still works fine in Windows 10), due to changes in security certificate at TEMIS.

(4) This change is actually in wret.exe, but is relevant to wxsim.exe: wret can now ignore dew point in its learned bias correction analysis. There are two check boxes (either of which can be used): one at the bottom of the Compare to Actuals plot form, and one on the bottom of the form were the details of the analysis appears. The motivation for this is that I was encountering intermittent outages of humidity readings on my school’s Davis Vantage II Pro station. I have since learned that I simply needed to change the backup battery, but I’ve seen bad humidity data from customers with other stations, as well, so this is a useful feature. NOTE: This option is largely redundant with the one discussed below, in Version 2017, Build 1.0, which allows WXSIM to ignore dew point (derived from relative humidity) corrections from both wret/autolearn and WXSIM-Lite. In fact, I implemented the new feature in wret (as well as a new one in WXIM-Lite) frankly having forgotten about the protection in WXSIM. I’m leaving the new ones in, however, so that the data presentations in those programs will not reflect bad humidity data.

Unfortunately, the program does not work.
#-o
Marian


If you’re running on Windows 7, make sure you have McMahon selected for GFS and untick the Temis selection. You’ll get Ozone data from GFS so you don’t need the Temis data.

Hi
For this beta, do we just install over the existing wxsim and run as normal?? as I was thinking of trying it out.

Thanks

If you’ve not use any of the previous betas then there’s a bit of extra setup to do, e.g. changing the filename to allow Euromix to process the data, but essentially it is just install over the top of the existing installation then make the file name changes.

I’ve just installed beta6 and the first run was successful. I ran this with just Ozone data from GFS, i.e. the Temis data download was turned off.

Reiterating what Chris said, the TEMIS data does not work with Windows 7, because of security certificate changes at that site which Windows 7 can’t handle (though certain browsers can get through to the site, even under Windows 7), which is why Chris and I worked to provide the new (and at least as good) alternative of the GFS model’s ozone data. If you use Windows 7, you can just uncheck that TEMIS ozone box at the bottom of WXSIMATE’s main form, and use the McMahon GFS source.

As for pros and cons of the two sources, the new GFS ozone data goes out through 10 days, while TEMIS only goes through 6. Also, the GFS data is 3-hourly, and WXSIM interpolates among these to provide smoothly, continuously changing values, while the TEMIS data is provided as whole-day averages. Those are points for GFS (and possibly reasons to uncheck TEMIS), BUT on the other side of the ledger, TEMIS is apparently dedicated to tracking and forecasting ozone levels, while perhaps (?) it’s more of a “side job” for GFS, so possibly TEMIS may be more accurate. If WXSIM “sees” both, it simply averages the two values, partially compromising the smooth-curve benefit of GFS, but probably slightly enhancing accuracy. In my testing so far, the two sources are remarkably close together AND don’t vary much, either day to day, or hour by hour. These data are used ONLY for WXSIM’s UV index, and do not affect the rest of the forecast.

With regard to connection issues, versions of Windows, etc. …

For about the last year, since I had a computer crash and had to update to Visual Studio 2019 (instead of the 2003 version I’d been using, for which I couldn’t find the 16-year old installation disks), WXSIMATE has apparently not worked in Windows XP. However, there are a few of you who still do or would like to use XP. For this reason, I’m including WXSIMATE Version 7.3, from July 2019 (right before my computer crash), as a legacy option. It does not have the latest changes (i.e. to URL’s) or features (the new ECMWF and ozone data types), but still seems to be functional otherwise. All the other programs in the WXSIM suite are written in Visual Basic 6.0 (from 1998), so they’re still fully compatible with XP. I definitely recommend using something later than XP, though, to take advantage of new features.

The Windows 7 issues, with certain URL’s, are mostly related to those sites changing their security certificate from TLS1.1 to TLS1.2. This affected some NOAA URL’s, but was readily fixable by making changes in the registry. I successfully did this myself (well, one of my school’s IT guys helped me a little!), as I’m still using an old Windows 7 laptop as a dedicated WXSIM and weather machine. Some of you did the same. Why the TEMIS site is different here is a mystery to me; the changes I made with re-enabled the NOAA URL’s didn’t work for TEMIS. THankfully, that doesn’t really matter now, with the new GFS data. Again, my recommendation is to use WXSIM on something newer than Windows 7, but then, I’m still using it myself!

The proxy sites not responding is something I suspect will be fixed soon (I’m talking with Sam Bohler about that), but in the meantime, I think all of the sites backed up by proxies are working fine directly, so just don’t use the proxies for now. Actually, the proxies seem to have been needed only on computers which were out of date with respect to Windows updates for the Net Framework. It is my intention to maintain them as options, since a need could arise in the future, but they’re generally not necessary now, as far as I can tell.

I’ll also give an update here on the effect of using the new ECMWF data, here in Atlanta, anyway. An earlier analysis I did showed ECMWF doing a bit better than GFS, and the 60:40 ECMWF:GFS mix doing even better than that. My latest analysis shows ECMWF must have gone through a “rough patch”, as WXSIM’s GFS-based mean absolute error (for days 2-4) was 3.02 F, while with ECMWF’s it was 3.28. That’s mildly surprising, because ECMWF is widely regarded to be a bit superior to GFS. BUT, this is not enough data (and I need to weight the whole analysis period evenly … this is tilted towards the recent ECMWF rough patch), AND the 60:40 mix, at 3.06, is significantly better than the simple weighted average (here, 3.15) of the two models used separately. What this implies is a “two heads are better than one” advantage. I was expected that to be the case, because with some degree of independence between sources being mixed, this should happen. What I think we will see over time, is that GFS and ECMWF will be very close overall (and I think ECMWF will edge GFS out), but a properly weighted mix will beat BOTH GFS and ECMWF. Also, I’m only talking about temperature here; precipitation, etc. may show advantages independently.

Let me know how it goes with this beta. If nothing bad shows up, I plan to release this officially within a day! :slight_smile:

Thanks!

Tom