Inverse modeling of HYDRUS-1D.

A discussion forum for Hydrus-1D users.
Post Reply
Mitra
Posts: 14
Joined: Thu Dec 06, 2018 3:44 pm
Location: Switzerland

Inverse modeling of HYDRUS-1D.

Post by Mitra » Tue Jan 14, 2020 3:04 pm

Dear Jirka

I am using HYDRUS-1D to inversely calibrate my hydraulic parameters and the longitudinal diffusivity against soil and tracer data. I have found reasonable numbers with low errors, still, I was interested to check the procedure that I use with you to make sure that I am doing it correctly as in the manual there isn't a step by step description of it. Here is what I do:
1. Entering the observation data in data for the inverse solution (with all manually determined weights set to 1 in this table)
2. starting from a sensible range and initial value for desired parameters to be calibrated.
3. running the inverse modeling, changing the initial value for inverse modeling to a value that falls within the 95% confidence interval, and rerunning the inverse modeling until I find the calibrated parameter and the initial value to stay within the 95% confidence interval.

4. Redoing steps 1-3 with trying different weights (in the weighting of inversion data: standard deviation, mean and no internal weighting) and initial values, or ranges. The results didn't vary much.

To do a check after I found the primary optimal parameters, I tried calibrating parameters only one at a time, and they didn't vary much. Given good statistics from HYDRUS-1D for the inverse modeling, the low variation of calibrated parameters under the above-mentioned conditions, and the good match between observation and my model, I concluded that the calibrated parameters are close to the real values.

I did the aforementioned based on what I understood from the manual, but to be sure that I am on the right track, I am thinking if you would recommend something more? or you would have any advice for me on this matter.

Thanks a lot for your time and care.

Sincerely,
Mitra

Jirka
Posts: 4824
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Inverse modeling of HYDRUS-1D.

Post by Jirka » Wed Jan 15, 2020 11:30 pm

It seems that you do all the right things.

Obviously, since HYDRUS-1D uses a local gradient optimization method (and not a global optimization method), you could do some more comprehensive analysis. See, for example, some recent papers by Brunetti:

Brunetti, G., J. Šimůnek, and P. Piro, A comprehensive analysis of the variably saturated hydraulic behavior of a green roof in a Mediterranean climate, Vadose Zone Journal, 15(9), pp. 15, doi: 10.2136/vzj2016.04.0032, 2016.

Brunetti, G., J. Šimůnek, and P. Piro, A comprehensive numerical analysis of the hydraulic behavior of a permeable pavement, Journal of Hydrology, 540, 1146-1161, doi: 10.1016/j.jhydrol.2016.07.030, 2016.

Brunetti, G., J. Šimůnek, H. Bogena, R. Baatz, J. A. Huisman, H. Dahlke, and H. Vereecken, On the information content of cosmic-ray neutrons in the inverse estimation of soil hydraulic properties, Vadose Zone Journal, 18, 180123, 24 p., doi: 10.2136/vzj2018.06.0123, 2019.

j.

Post Reply