Page 1 of 1

terminating not converging Hydrus runs

Posted: Wed Oct 04, 2006 4:40 am
by besel
Dear Hydrus user,

I am running Monte Carlo simulations with Hydrus1D using randomly sampled soil parameters. I am just wondering whether there is any option to terminate model runs, when the numerical solution is not converging (because unsuited parameter combinations are used). This would reduce the computing time of MC simulations considerably since these model runs have large water balance errors anyhow.

Thanks,
Benny

Posted: Wed Oct 04, 2006 3:12 pm
by Jirka
Benny,

I'm sending you the version of the code that automatically stops and closes all the output files when it does not converge in the solution of the Richards equation (instead of trying to continue). Let me know if you would need the code to do something else in such case, such as sending some message, ...

Jirka

Re: terminating not converging Hydrus runs

Posted: Tue Aug 09, 2016 10:06 pm
by Green_roof54
Hi Jirka,

Sorry, I know that the discussion is old but I'm actually using the monte Carlo method to determine the sensitivity of van Genuchten model parameters. I used the home technique. I check the obs_node.out file. if the latter is empty, I run another set of parameters. The simulation takes very long time. Can you send please if it's possible the code to stop when it does not converge?

Thanks

Re: terminating not converging Hydrus runs

Posted: Wed Aug 10, 2016 9:29 am
by Jirka
The code automatically stops when it does not converge in 10 consecutive time steps. J.

Re: terminating not converging Hydrus runs

Posted: Wed Nov 06, 2019 4:19 am
by ritadg
Hi all!

I'm running HYDRUS-1D in Matlab and I'm also running Monte Carlo simulations using randomly generated soil parameters (given an upper an lower limit for each parameter).
Some runs takes ages (more than 20 minutes) for then get to a no converging solution.
Does anyone have this code that automatically stops and closes all the output files when it does not converge in the solution of the Richards equation ?

Or maybe someone have an idea of how to kill the run when longer than a certain time? (I've tried few options but it didn't work)

Thanks a lot!
Rita.

Re: terminating not converging Hydrus runs

Posted: Wed Nov 06, 2019 2:41 pm
by Jirka
The code should automatically stop when it does not converge in 10 consecutive time steps. I'm not an expert on MatLab, but I believe you have there a command that would kill the run if it does not finish before some specified time. J.

Re: terminating not converging Hydrus runs

Posted: Wed Jan 15, 2020 12:08 am
by ritadg
Thank you for your reply Jirka.
(It took me ages to see it, I'm sorry)