Wxsim 2024 v1.1 almost all forecasts start with very high or very low temperature

On 1-Jan-2025 I installed WXSIM 2024 V1.1 version and switched to CMX version V4 with numbers instead of monthnames in monthly logfilenames.
Since then almost all forecast runs have been provided a very high or low peak in the temperature for the first hours of the forecast (regardless if done at 21:00, 09:00, 15:00 or 18:0).
I am wondering if other users see the same phenomen or have an idea what could be the root cause:

  • new version WXSIM 2024 V1.1
  • my specific weather situation in Lower Austria in January
  • or a wxsim preference or setup that i changed or overlooked

3 in 1 forecast

As a consequence also my MAE increased: wxsim MAE chart

Examples of latest.txt
latest.txt

CALIBRATION RUN:

DATE    TIME   TEMP   WIND  R.HUM %SC  %VST VIS PC/HR RN TOT   WEATHER

25 Jan 10:47      0,9   10   83   100   13   5,7   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 11:17      1,0   10   82   100   18  13,8   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 11:47      1,2    9   81   100   18  13,6   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 12:17      1,3   10   80   100   18  13,5   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 12:47      1,4    9   80   100   18  12,8   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 13:17      1,5    9   79   100   18  11,4   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 13:47      1,6    9   79   100   18  11,7   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 14:17      1,7    6   79   100   17   6,8   4   0,0   DNS.OVCST CHNC. DRZL  
25 Jan 14:47      1,7    9   79   100   18  10,6   4   0,0   DNS.OVCST CHNC. DRZL  
 
FORECAST RUN:

DATE    TIME   TEMP   WIND  R.HUM %SC  %VST VIS PC/HR RN TOT   WEATHER

25 Jan 15:00     43,0   10   99   100   10   1,8   4   0,0   DNS.OVCST CHNC. DRZL  
No convective shower or thunderstorm activity expected
25 Jan 15:30     36,5    9  100    93   31   9,4  12   0,0   CLOUDY    LIGHT RAIN  
Numerous thunderstorms likely, some heavy or severe
25 Jan 16:00     27,9    6  100    97   40  12,1   7   0,2   CLOUDY    PROB. SHWR  
Scattered showers or thundershowers possible
25 Jan 16:30     22,7    5  100    99   38  13,1   5   0,2   CLOUDY    PROB. SHWR  
Convective showers very unlikely
25 Jan 17:00     18,8    4  100    99   39  14,4   2   0,2   CLOUDY    MOD. DEW  
No convective shower or thunderstorm activity expected
25 Jan 17:30     15,8    3  100   100   40  15,6   2   0,2   CLOUDY    MOD. DEW  
SURFACE WIND NOW FROM  E  
25 Jan 18:00     13,7    3  100   100   40  16,8   2   0,2   CLOUDY    MOD. DEW  
25 Jan 18:30     12,2    2  100   100   39  18,0   2   0,2   CLOUDY    MOD. DEW  
25 Jan 19:00     11,1    2  100   100   38  19,0   2   0,2   CLOUDY    MOD. DEW  
25 Jan 19:30     10,3    2  100   100   38  19,9   2   0,2   CLOUDY    MOD. DEW  
25 Jan 20:00      9,7    3  100   100   38  20,8   2   0,2   CLOUDY    MOD. DEW  
25 Jan 20:30      9,3    4  100   100   38  21,6   2   0,2   CLOUDY    MOD. DEW  
25 Jan 21:00      9,0    4  100   100   39  22,5   2   0,2   CLOUDY    LIGHT DEW  
25 Jan 21:30      8,9    5  100   100   39  23,4   2   0,2   CLOUDY    LIGHT DEW  
25 Jan 22:00      8,8    6   99   100   40  24,3   2   0,2   CLOUDY    LIGHT DEW  
25 Jan 22:30      8,8    7   99   100   40  25,1   2   0,2   CLOUDY    LIGHT DEW  
25 Jan 23:00      8,8    7   99   100   40  26,1   2   0,2   CLOUDY    LIGHT DEW  
25 Jan 23:30      8,9    8   99   100   40  27,8   2   0,2   CLOUDY    LIGHT DEW  
FORECAST RUN:

DATE    TIME   TEMP   WIND  R.HUM %SC  %VST VIS PC/HR RN TOT   WEATHER

20 Jan  9:00    -39,0   11  100    75   37   5,1   2   0,0   M.C.-CLDY LT. FROST  
No convective shower or thunderstorm activity expected
20 Jan  9:30    -28,1   15   71    74   56   6,2   1   0,0   P.-M.CLDY LT. FROST  
20 Jan 10:00    -15,6   19   30    91   60   7,0   0   0,0   P.-M.CLDY   
20 Jan 10:30     -8,9   21   20   100   57   7,9   0   0,0   M.CLOUDY    
20 Jan 11:00     -5,3   21   18   100   55   8,7   0   0,0   M.CLOUDY    
20 Jan 11:30     -3,1   20   16   100   53   9,5   1   0,0   M.CLOUDY    
20 Jan 12:00     -1,4   19   16   100   51  10,2   1   0,0   M.C.-CLDY   
20 Jan 12:30     -0,3   19   16   100   49  10,8   1   0,0   M.C.-CLDY   
20 Jan 13:00      0,5   18   17   100   46  11,3   1   0,0   M.C.-CLDY   

I am using a GFS:ECMWF ratio of 30:70 (euromix), data from McMahon


and the following setup
wxsim setup and MAE.

My first thought is that you have changed everything at once which makes life somewhat difficult to diagnose what might have gone wrong. I did the beta testing with CMX V4 and WXSim btw.

Forgive me if I’m suggesting stuff you have already done.

Firstly did you follow all the instructions to install CMX V4 and convert all the V3 files to V4 format? This is critical since WXSim only reads either V3 or V4 files a mix will not work.

The full instructions can be found at https://cumulus.hosiene.co.uk/viewtopic.php?t=22051

So my second thought is have you correctly adjusted all the WXSim suite of programs to correctly use the CMX V4 files, WXSimate, WXimLite and WRET needs to be double checked to make sure they are all reading V4 files. Also make sure the UNITs are set correctly in all programs.

These are the basics which need to be checked as I presume prior to CMX V3 and the updated WXSim it was all working.

If all this is good then we can move on.

Stuart

Hi Stuart,
thx for fast feeback.
I checked again in detail and I think i could narrow it down:

  • you are right, before upgrade of CMX and WXSIM all has been running fine for 2 years.
  • CMX migration to v4.3.2 was fine and I followed instrutions and migrated all datafiles. Now all logfiles have no monthname DEC (or in my German locale case DEZ) inside bUT the month number.
  • Because of the CMX V4 I also had to upgrade WxSim to get the option to select as Station Cumulus (with numerical month) so that all tools: wxim, wximate, wtret and wxsim-light found the converted CMX logfiles. Checked for all tools and for all tools in place.
  • I checked also for all tools WXSIM, wxsimate, wtret and wxsim-light that units are still in Celsius, mm, km/h, hpa.
  • the forecasts are basically working, but the entry point temperature is much too high or much too low.
  • I looked now in detail into WXSIM and WXSIMATE and there i see, that the temperature is too hight but more or less exactly by a factor of 10. So it looks like that WXSIMATE v9.2 does find the correct CMX monthly logfiles but multiplies the temperatue by 10, although unit is set to Celsius (and Fahrenheit conversion would be different). Also the starting dewpoint is multiplied by 10. That explains why I get either high plus or high minus values, depending on my actual temperature (if close to zero the error levels out). All the other values are imported correctly: wind, pressure and humidity.
  • after the WXIMATE import of course also the value in WXIM as start temperature is wrong (see 2nd scrrenshot)
  • see attached screenshots (please note that my forecast was running at 18:12 and I now checked the figures a little bit later at 18:41 and now my real temperature is lower with 4.4 and not 4.8 as it was as WXSIMATE was doing the import.
  • WXSIMATE imported the temperature as 48.0 degree celsius (and dew point as 45.0) instead of 4.8 (and 4.5).
  • That explains the resulting too high temperatures in the forecast latest.txt
  • But I have no idea, why WXSIMATE is making that error. Seems not to be the customizaton for units: both for display and Cumulus are set to Celsius (and would not result in a factor of 10).
  • Can there really be such an error in WXSIMATE, or am I using the wrong version of WXSIMATE?
  • attached also the converted V4 CMx monthly logfiles for 202412log.txt and 202501log.txt
    leo.cmx.monthly.logfiles.zip (345.0 KB)
    , but to me they look fine.
    regards
    leo

I’ll have another look tomorrow but I noticed that it seems you are using comma decimal point so it might be something to do with that. I know CMX has had some issues with that. Maybe check your CMX data log files to make sure that they are formatting correctly. Also cannot remember if WXSimate needs to be told about comma decimals, I don’t use commas but the point character.

Stuart

With CMX V4 migration the commas in the monthlogfiles are now all decimal point and no longer commas (see zip in my last post). Only in CMX frontend when using webtags without RC=y the commas appear.

But you are right, could be something related to the commas. But in the past the German locale has been no issue with WXSIM and why just with temperature and not with pressure?
I could not find a WXSIMate preference/paramter/setup that switches between comma/point.

I investigated if there are any WXSIM setups for switching between comma and decimal point, but could not find them. I think WXSIM derives the locale setup during startup of the program from the windows setup and does not offer an option of its own. Based on that commas are used in my userinterface of WXIM and also written into all files like details.csv or latest.txt.
That has been working fine and is still fine.
But CMX v4 changed not only the name of the monthly logfle but also the format of the date and seperation and comma (regardless now of locale for all CMX users). Maybe that has issues in WXSIM(ate) in windows locale with comma.

2 lines old format of CMX monthly file v3 Dez24log.txt

01.12.24;00:00;-2,3;98;-2,6;1;3;37;0,0;0,0;1030,7;760,4;25,4;48;2;-2,3;-2,3;0,0;0;0,00;652,94;-4,9;0;0,0;48;0,0;0,0;-2,3;-2,3
01.12.24;00:05;-2,4;98;-2,7;1;3;37;0,0;0,0;1030,3;760,4;25,4;48;2;-2,4;-2,4;0,0;0;0,00;652,94;-5,0;0;0,0;26;0,0;0,0;-2,4;-2,4

2 lines new format of CMX monthly file v4 202412log.txt

01/12/24,00:00,-2.3,98,-2.6,1,3,37,0.0,0.0,1030.7,760.4,25.4,48,2,-2.3,-2.3,0.0,0,0.00,652.94,-4.9,0,0.0,48,0.0,0.0,-2.3,-2.3
01/12/24,00:05,-2.4,98,-2.7,1,3,37,0.0,0.0,1030.3,760.4,25.4,48,2,-2.4,-2.4,0.0,0,0.00,652.94,-5.0,0,0.0,26,0.0,0.0,-2.4,-2.4

I am also somewhat surprised by this. I think we need Tom to look at this. He may have had reports not shown in forums. Can you email him with a link to this thread?

Stuart

Hi Stuart,
I sent an email to Tom with a link to this forum post.
Can you also send him an email about that link (last year some of my emails seem not to have reached Tom, because usually he responds).
much appreciated
regards
leo
I am thinking about also opening a post in the https://cumulus.hosiene.co.uk/ forum to ask if any other CMX v4 user who is also using WXSIM (and a locale with commas) had any issues so far …

Will do… might be an idea as I know there are a few WXSim users there.

Stuart

Just an idea, is it possible that your WXSimate is an older version, I know it says 9.2 but maybe worth checking. Mine is dated 22/06/2024 and is 353.5Kib in size. If different then download the latest wxsimupg file and re-install.

Stuart

Hi Stuart,
my WXSIMATE is also from 22-Jun-2024 and 353KB.
Leo

OK it was worth a check. Let’s see what Tom suggests.

Stuart

I disabled the local data import in WXSIMATE and WXSIM for time being as today WXSIM forecast already stopped and complained about unrealistic high temperature (8.0 x 10 = 80 degree celsius.

Thanks for letting me know about this! I think it will be an easy fix, as it’s evidently something about importing that data. Two things come to mind: (1) Fahrenheit Versus Celsius: make sure that there is complete consistency (I assume you want Celsius) across WXSIM, WXSIMATE (under Customize, for both ‘units in WD or Cumulus log files, and in Units to display here’), AND in Cumulus MX itself; (2) the comma/decimal point/regional settings (in Windows) issue. Indeed I’ve had to be vigilant about such things, and think I’ve addresed these peroperly in the past, but some new combination could have come up. In any case, here’s what I need to investigate (and surely fix) this: custinit.txt, wxsmini.txt, localdat.txt, localcal.txt, and whatever the Cumulus MX log file was. Email them to me or post here.

Thanks!

Tom

Tom,

thx for the fast feedback.

I have just started a new manual run with import from local station. The local station temperature is 1.4 degree Celsius, but WXSIMATE imported 14.0 as starting point.

I think it is not related to Fahrenheit/Celsius, but to my windows locale with commas.

Please find attached the requested files

Regards
leo.pws-suttenbrunn.202501270600.zip (164.4 KB)

Leo

Hi Marian,

Good to hear from you!

I just looked at your links, and copied the WXSIM and corresponding EC forecasts into a spreadsheet (attached). I see that WXSIM’s forecasts are a bit colder than the EC’s as of when I just checked the links (11:20 PM, Monday). WXSIM’s lows are 0.5 degrees colder than EC’s, and WXSIM’s highs are 1.8 degrees colder than EC’s.

Maybe something briefly went wrong with an earlier forecast?

Let me know if you see any more problems.

Best Regards,

Tom

(Attachment wxsimeccomp.xlsx is missing)

I found that I can’t upload a spreadsheet on the forum. Here’s an image of the spreadsheet I was trying to share.

(attachments)

HI Tom,

thx for investigating.

I am using Windows 11

Display Language of Windows = German; no need to change that

Region/County = Oesterreich = Austria (changing that alone had no impact)

Regional Format = Deutsch (Oesterreich) = German (Austria); As soon as I changed that to English (Europe) or English (US); WXSIMATE import was correct

Keyboard setting = us-en; changing has no impact

The core dependency is Regional Format = English (Europe). As Soon as I change that, WXSIMATE import is correct, but that also affects all other output:

WXSIM GUI shows now decimal point and also in lastret.txt the decimal output is used.

NOK:

OK:

Thanks for the reply. I will continue to follow what the values ​​of the two forecasts will be.
Regards,

Marian