[Rasp] Several RASP days not updates since yesterday

[Ian]

I have not done a detailed debug, but it looked like the legacy model 
crashed, I assume because of the timestep issue, but it may have been 
another upstream issue.

Both models are running again this morning. I don't plan on changing any 
criteria for this reason.

But I do intend playing with the "beta" model to see if we can get a run 
with resolution to match or better the 1.3km we have on the legacy one. 
This could come in useful for the wave season. That version could 
possibly be even less stable, but it is worth the trade off for wave and 
foot launch pilots. :-)

Thanks

Ian


On 4/9/26 11:50, rasp.admin at stratus.org.uk wrote:
> Ah ok, so the GFS issue just happened to be at the same time.
>
> On the model crashing with time-step too long, the one I set up 
> originally for you is a compromise. It is possible to always make the 
> higher resolution runs complete (smaller time step) but will take 
> longer to finish (or be too late to be of use).
>
> What I always suggest is again as said, which is to use the lower 
> resolution model instead over. This will be almost the same for most 
> parameters.
>
> However if looking for wave, then the lower resolution will probably 
> not provide the detail to see on the plots. But for wind 
> speed/direction, thermal height, cloud base, temperature, it will be 
> about right. Where it probably won’t be right is across sharp ridges 
> or topographically complex areas as the topographic model used, 
> smooths these out.
>
> Ian: If you want something that checks the basic parameters against 
> airport/airfield METAR reports for a ‘quality’ check, let me know.
>
> Regards,
>
> Darren
>
>> On 9 Apr 2026, at 10:16, Ian <ian at zomerlust.org> wrote:
>>
>> 
>>
>> There was one run which failed to converge. Ie the model got into a 
>> loop where instead of each step getting closer to a solution the 
>> results started to diverge again. This leads to a model crash.
>>
>> It is caused by an anomaly in the input data - often associated with 
>> unusual weather events like "cut off lows" (but that was not the 
>> cause of recent crashes). This coupled with compromises made when the 
>> model was constructed makes the model run crash. It happens on a 
>> handful of runs a year.
>>
>> In order to avoid this, the model would have to have smaller "time 
>> step" intervals and/or a smaller "resolution ratio" ie 12km -> 4km -> 
>> 1.3km. That would make it run slower and use more computer time and 
>> electric power. The parameters selected were a compromise that work 
>> most of the time.
>>
>> When this happens I have seen the same date forecast crash on 
>> subsequent forecasts runs, ie 1 day ahead, night before etc.
>>
>> If you see this, have a look at the alternate model legacy vs beta. 
>> It is possible one crashes when the other manages to resolve.
>>
>> Regards
>>
>> Ian
>>
>>
>> On 4/8/26 20:40, AirSchool Paragliding wrote:
>>> Greetings all
>>>
>>> Just noticed that a few of the RASP days haven't been updated since 
>>> yesterday e.g. RASP 1.3km d0 and d2, RASP 2km d4 and d5
>>>
>>> Many thanks
>>> Ria
>> -- 
>> Rasp mailing list
>> Rasp at lists.zsd.co.za
>> https://lists.zsd.co.za/cgi-bin/mailman/listinfo/rasp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.zsd.co.za/pipermail/rasp/attachments/20260409/285ababf/attachment.htm>