Near Saturation in 2D/3D vs 1D

A discussion forum for Hydrus-1D users.
Post Reply
swhitman
Posts: 9
Joined: Fri Apr 15, 2016 8:36 pm
Location: USA

Near Saturation in 2D/3D vs 1D

Post by swhitman » Mon Jul 18, 2016 8:35 pm

I am running some simple sin flux columns, in which the upper boundary is hourly step data generated from a sin function, and the lower is free drainage. The soil type is a silty loam (from the catalog). I am running in 2D/3D and 1D as part of a comparison study. I am having issues with simulation convergence in 1D when the -2cm air entry option is not used, but the 2D/3D (axisymmetric, 2D) model seems to handle this just fine. I have tried using extremely small time and space discretizations (1000 nodes evenly spaced, over 150 cm domain), (1e-20 days dt), and am still having issues. The simulation crashes when it nears saturation and produces Nan values thereafter.

Do the two codes handle near saturation differently with the standard VGM model? I am confused. I am reading through some of the publications listed in FAQs on the 2cm air entry, and it looks like this could improve the simulations for this soil type, but it is on the edge of being applicable. I also prefer not to use it for consistency because some of the other models I am running do not have that option.

Thanks,

Jirka
Posts: 4733
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Near Saturation in 2D/3D vs 1D

Post by Jirka » Mon Jul 18, 2016 10:35 pm

Both models solve the Richards equation and use various soil hydraulic property models pretty much the same. However, I would do expect problems when you get a fully saturated column, free drainage BC and a flux upper BC. The free drainage BC does not allow higher flux than Ks, and if you specify upper flux bigger than that, it is not physically possible. J.

swhitman
Posts: 9
Joined: Fri Apr 15, 2016 8:36 pm
Location: USA

Re: Near Saturation in 2D/3D vs 1D

Post by swhitman » Mon Jul 18, 2016 10:45 pm

The column would only be saturated in a transient nature at the top of the column during the simulation as the max of the input data is equal to Ksat, so I don' think the scenario you describe would ever occur in these simulations. The set up is exactly the same for the 1D and 2D models, and the 2D runs just fine without the 2cm air entry option, while the 1D does not. Incidentally, VS2D was having issues similar to Hydrus 1D, and I used H and Krel data from the 2cm air entry option in Hydrus for the VS2D simulation, and it ran much more smoothly, so I think the problem is related to transient near saturation at the top of the column (I can also tell this because it is crashing when the precip rate is close to Ksat (I am using the atmospheric with runoff option for the upper boundary, as opposed to specified flux, so it isn't being "forced" into the porous media).

I am mainly just still confused as to why there are such difference between 1D and 2D. Does the 2D model have more storage to buffer the inputs at the top?

Jirka
Posts: 4733
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Near Saturation in 2D/3D vs 1D

Post by Jirka » Mon Jul 18, 2016 10:57 pm

No, there are not buffer (or other tricks) in 1D or 2D. Did you try to disable the internal interpolation tables to make the calculations of hydraulic properties more precise.
Can you post (or email me) your 1D project? I do not see why there should be a problem. Silty loam (n=1.41) is usually quite easy to handle. J.

swhitman
Posts: 9
Joined: Fri Apr 15, 2016 8:36 pm
Location: USA

Re: Near Saturation in 2D/3D vs 1D

Post by swhitman » Mon Jul 18, 2016 11:45 pm

Yes I had tried disabling the tables by putting zero in both spaces, to little effect.

I have attached the 1D model as a zipped folder. Thanks for looking! I hope it isn't something silly, I've spent a lot of time trying to make sure it isn't anything obvious.
Attachments
columnv2_sin1_siltloam.zip
(134.97 KiB) Downloaded 96 times

Jirka
Posts: 4733
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Near Saturation in 2D/3D vs 1D

Post by Jirka » Tue Jul 19, 2016 12:42 am

I have made few changes to your projects (fewer nodes, larger time steps, etc). The main change stabilizing the solution is that I use much more strict iteration criterion. When your flux (e.g., 10.7615) is so close to Ks (10.8 cm/d), you need higher precision to get it right. J.
Attachments
Work.rar
(2.84 MiB) Downloaded 101 times

swhitman
Posts: 9
Joined: Fri Apr 15, 2016 8:36 pm
Location: USA

Re: Near Saturation in 2D/3D vs 1D

Post by swhitman » Tue Jul 19, 2016 1:20 am

Thank you for helping me with that. I had increased water content tolerance, but not to the extent you did, I'll keep that in mind for future efforts.

I am still confused why hydrus 2d will solve without these adjustments. Because I'm somewhat interesed in the differences between the codes, I'd love to hear your ideas for why 2d solved the problem more easily. The water content tolerance in the 2d model is .001 and the remaining iteration criteria are the same as 1d, interpolation tables are on, didn't have to do much to get it to run, spatial disc. is 2cm, and temporal disc. is min. 1e-8.

Is the solver more robust for 2d?

Jirka
Posts: 4733
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Near Saturation in 2D/3D vs 1D

Post by Jirka » Tue Jul 19, 2016 1:29 am

I guess as the saying goes "You get what you pay for". Obviously one of these two programs is free and one is not. J.

Post Reply