I been meaning to set up WxSimLite for a long time and I’ve finally started but I’m confused about the startup requirements! There’s a 4 day period and a 2 week period where different things happen. As far as I know I’m currently between 4 days and 2 weeks so I can’t do the analysis stuff yet.
I have it scheduled to run every 6 hours and I can see new files being created in the wxsimlitedata directory. There are 23 files there, so I’m guessing I’m about 6 days in now?
However, when I look at WxSimLite it says “12 good forecasts” with 11.98 effectively used. I assumed that it would be saying 23 good forecasts by now? If not, why are 11 of the 23 forecasts not good? Also, do I have to wait until I have 56 (14 * 4) good forecasts before I can do analysis? That’s going to take more than two weeks if half the forecasts are considered “not good” for some reason.
Hopefully someone can demystify me based on their own experience.
It’s a long time since I started with lite. I’ll havd a look later but I have a feeling that it needs days not just forecasts.
Anyway I’ll say more later when I am on my PC.
Stuart
1 Like
Yes I did take a look and it needs 14 days of forecasts, so 12 days and 23 forecasts are not enough. Taking in to account it stops at 4 days before today you need at least 18 days of forecasts, a few more would not hurt.
I obviously have years of data now and limit my scans to 70 days and these scans finish 4 days before today so in my case it goes back 74 days and scans forwards stopping at -4 days before today. Obviously the more forecasts to play with the better so I would probably say if you could wait say 3 weeks then it might be better.
I schedule Lite run each time I run a forecast (not analysis) and import that data together with the WXSimate data into WXsim. I run analysis 3 times per week in the early hours on Mondays Wednesdays and Fridays and this data also gets used by WXSim.
Hope this helps a bit.
By the way I run the whole WXSim thing on Linux now under Wine and it behaves well 
Stuart
Thanks for that, but I’m now more confused!
I’m not sure what it’s doing in the first 4 days, but I assumed it was forecasting and doing something extra with those forecasts. It’s not clear from anything I can see whether the two weeks of data that are needed include those 4 days or not. I fully accept your point that the more data that’s available the better but it would be good to at least start doing some runs with WxSimLite data even if I’m not using them as my “official” forecasts.
What’s puzzling me more is that I assumed that the gfsdat*.txt files in the wxsimlitedata directory are the ‘forecasts’. I now have 24 files in that directory and the count is growing every 6 hours. However, WxSimLite just says I have 12 good forecasts and I’ve not seen this number increasing recently. I’ve no idea what counts as a good forecast, nor what I can do, if anything, to influence a forecast to be good.
On the grounds that there are 4 forecasts per day, then 12 good forecasts equates to just 3 days of forecasts which means I’ve still a long way to go to get to 14, or 18, days and I’m seemingly not getting any closer every 6 hours when there’s a new forecast run because no new forecasts seem to be good. Having said that, it doesn’t say anywhere that I need 14, or 18, days of good forecasts!
Chris I hope you remembered that if you click on blue text it pops up an explanation of the item. It does try to explain a lot about each option
I just ran a full scan manually having set the number of span days to 10 and it used 39 good forecasts with an effective number of 21.00 which is about half of the 39 because I have set Weight Recent More. This was a two pass scan as it was a full scan.
I run 4 forecasts a day which gives me 40 forecasts in 10 days. Normally I currently use 70 for the span days value but that takes ages! The following image shows the help text for Number of good forecasts.
Hope this may get you further.
Stuart
I did remember - as soon as you said it 
I thought I understood what the blue text told me until you said that you’re seeing 39 good forecasts. The help text suggests it’s the number of forecasts in days 2 through 4 (so 3 days) with good station data, so with 4 forecasts per day that should only ever reach 12.
If you see 39 then I don’t understand why I’m not seeing 24 when I’ve got 24 forecast files in wxsimlite directory? I’m not aware of any gaps in my station data this year so that’s not the reason for the low count.
I have answered one question though - the scheduler window says that you need at least 2 weeks worth of raw forecast files (gfsdat*.txt) to turn on analysis, so I just need to reach 14 days not 18.
Actually I just looked at my wxsmlitedata directory and counted on 39 forecasts for the10 days so it used all of them. For some reason I only had 3 forecast files for the 28th Feb!
These scans are using both the wxsimlitedata directory files and data from your station logs - in my case CumulusMX data files - and I assuming in your case WD log files so to do the scan it must find both sets of valid data for the day in question. I suspect it is also looking at the *.wxf files in the WXSim directory so I’m assuming it needs to match these with the wxsimlitedata files and your WD logs in order to do the analysis.
This days 2 through 4 part I think refers to the fact that there is that each forecast file contains about 7.5 days of data (mine go up to +180 hours) and so I think it uses days 2 3 and 4 out of each forecast.
As I said it is some years since I set this all up so in some ways I’m having to do a lot of revision!
It’s late now so I’ll wait for tomorrow to see you you get on!
Stuart
I’ve just checked and I have a complete set of WXF files for at least the last 2 weeks and there are no station data gaps in that time, and now 26 GFSDAT*.txt dating back to 25th Feb.
This morning if I run a full scan it says I have 15 good forecasts. It was stuck on 12 from about lunchtime yesterday despite running Full Scan a number of times…including at about 23:00 yesterday. There can only have been at most 2 forecast runs since then so I don’t know why it’s suddenly jumped by 3 this morning.
The only way I can partly rationalise this is…
- If I count back 15 GFSDAT*.txt files it takes me to the last forecast on 28th Feb. There are some gaps in this, but I think a couple might be due to Bohler data not being available and others possibly because WxSimLite stopped running and I didn’t re-start it straight away.
- There are 4 incomplete days (12z on 25/02 to 00z on 28/02) before those 15 files which might count as the ‘startup’ 4 calendar days?
I suspect that unless Tom sees this and comments I’m just going to have to wait to see when it says I can turn on analysis.
Can I also please ask what your forecast scheduled times are set to? I don’t think I’ve changed mine from the default and they seem to be weirdly spaced - 6am, 11am, 3pm and 9pm (5/4/6/9 hour spacings). Are these supposed to match Bohler data availability times, or match my WxSim forecast times, or is it some other schedule?
I’m really not sure what to say. I think you need 14 days covered by Lite data, so at least 1 file per day, I normally have 4 files per day as I run Lite prior to every forecast.
My forecasts run at 02:00, 09:00, 15:00 and 21:00 which 99% of the time corresponds with both GFS and ECMWF data being on the same time interval. So my Lite runs are the first thing which is scheduled, followed by WXSimate and the WXSim forecast run.
One confusion factor with my timings of the Lite runs is that the 02:00 run creates the Lite file for the previous day as the GFS data is from the previous day so at 02:00 today 5th March it creates the gfsdat0dt=2025-03-04-18.txt file
Yes I think we need to see what happens when you have enough days covered to run analysis.
Stuart
1 Like
I’m thinking that the number of good forecasts only updates once per day, perhaps after it’s seen the whole set of actual data from the previous day? It’s currently still only showing 15 even though there have been two more scheduled WxSimLite runs since I last looked.
Having watched a little longer I’m now sure that the good forecasts value only increases each day. It’s only going up by 3 per day which seems odd, but maybe it’s because the first forecast each day relates to GFS data from the previous day?
Also one of the prompts in the apps suggests that two weeks relates to having gfsdat*.txt files covering at least a two week period. These files are generated during the first four day period so it looks like the wait time is 4+10 days rather than 4+14 days. I’m on day 10 now so I still have to wait a little longer. It would be nice if WxSimLite could use WxSim .wxf files, I have loads of those.
Yes that makes sense. I think what Lite is doing is comparing the forecast data with the actual data from your station and I’m guessing that there may be insufficient data in the wxf files to do it all.
Stuart
Perhaps ask tom why this is and he ight make a small change to the program so that you can use the Wxsim wxf files
That’s possible although there’s a lot more data in a wxf than a gfsdat file. My wxf files are about 40 times the size the the gfsdat files.
I think it would take more than a few days for Tom to make those changes and I should hit my two weeks of data before then.
Yes Tom is likely not to be happy about changing Lite, after all it works fine once you have sufficient data! I was speculating as to why it uses so many files but it is what it is!
Stuart
