request for help on modeling Uranine BTCs

A discussion forum for STANMOD users. STudio of ANalytical MODels is a software package for evaluating solute transport in porous media using analytical solutions of the convection-dispersion solute transport equation.


Post Reply
Elise
Posts: 3
Joined: Fri Sep 19, 2003 2:15 am
Location: Australia

request for help on modeling Uranine BTCs

Post by Elise » Wed Jan 09, 2008 2:19 am

Hi,

My student conducted tracer experiments at our field site using uranine water tracing dye and bromide. The field site is for managed aquifer recharge using treated effluent and the tracer experiments were for estimating the travel time through the 10m-thick unsaturated (sand) zone. The recharge rate is about 17 litres/minute.

We have several nice datasets of uranine BTCs that were obtained from sampling from suction lysimeters at several different locations and 5 different depths in the unsaturated zone.

My reason for writing is to ask if anyone could please advise us on which code would be most suitable to analyse the data. We thought about using CXTFIT to obtain estimates of v and D, but I realise now that it may be unrealistic given that it is an advection-dominated system. I started by reading the STANMOD and CXTFIT manuals, focusing on Chapter 7 and the inverse sections. Several years ago I took a shortcourse on HYDRUS-2D, but really I don't have a lot of experience using it, so it would be nice to communicate with a more experienced person(s) on these topics.

We'd first like to know whether we should be aiming to use Hydrus-1D or CXTFIT.
It's not clear to me how one would input the water flux in CXTFIT if we took this approach. I could probably obtain rough estimates for the soil hydraulic properties and attempt using Hydrus-1D, but I was hoping to keep this simple as a first attempt.

If anyone has an example that sounds similar to our problem, could they please share it so that we can understand how to set up the problem?

Thank you,
Elise

ntoride
Posts: 74
Joined: Fri May 21, 2004 12:50 pm
Location: Japan

Post by ntoride » Wed Jan 09, 2008 1:48 pm

Dear Elise: CXTFIT is based on analytical solutions for the CDE subject to steady state water flow with uniform water content. Hence CXTFIS is generally used for well-controlled experiments to determine solute transport parameters. HYDUS numerically evaluate the CDE with Richards equation. Although HYDRUS could be used for solute transport with transient water flow, it is necessary to give proper parameter vales for flow and transport with appropriate initial and boundary conditions. If any further questions, please do not hesitate to ask me. Nobuo

Jirka
Posts: 4716
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Post by Jirka » Wed Jan 09, 2008 5:35 pm

Elise,

I will just briefly expand on Nobuo’s answer. Is not clear from your emails whether the water contents and water fluxes were constant during the tracer experiment. If yes, and if you can assume that the soil profile is homogeneous, then you should be able to use CXTFIT (STANMOD). If answer to any of these questions is no, then you need to use HYDRUS-1D.

Jirka

Elise
Posts: 3
Joined: Fri Sep 19, 2003 2:15 am
Location: Australia

Post by Elise » Thu Jan 10, 2008 3:25 am

Jirka and Nobuo, thank you for your prompt replies.

It sounds like we could use CXTFIT (STANMOD), assuming that our sand is homogeneous and the water contents and fluxes are roughly constant. These are safe assumptions for several datasets obtained at shallow depths.

I'd be curious to compare with Hydrus-1D.

I ran a CXTFIT simulation, but please correct me if these inputs seem incorrect:
1. I set up an inverse problem with a deterministic equilibrium CDE.
2. selected time and position as dimensional (cm; days; mg/l)
3. selected flux-averaged concentration, Cf

4. I had "No constraints for parameter estimation" and "No estimates for total mass" and maximum iterations=100.

5. I requested fitted values for v and D and I provided initial estimates of 100 cm/day and 100 cm2/day for them.

6. I selected a "Multiple pulse input" with only 1 pulse:
Pulse 1 (0.2093 mg/litre-cm) with a final time of 0.295 days
*There is perhaps a better way of doing this?? I'm trying to create a pulse of 6.174 mg/l over a duration of 0.295 days, starting on day 0. The water flux is 100 cm/day, hence I divided my solute concentration flux by this value as it says in the manual for flux-averaged concentrations.

The remaining inputs were as follows: zero initial concentraion and no production;
Inverse data structure of Z,T,C and 13 data points.
The data are BTC data taken at a depth of 50cm between time 0.049 and 0.6423 days. I converted my observed concentration data to flux-averaged values (again by dividing through by 100 cm/day).

Lastly, I provided output structure as follows:
7 output positions; 10cm spacing, 0 as my initial position;
800 output times, 0.0125 day for time increment; 0 as my initial output time; and concentration vs time for output code.

=================
The final results indicate that after 19 iterations,
v is 0.7E-4 cm/day and D is 0.168E+4 cm2/day with a MSE of 0.398E-5 and R2 for regression of 0.964.

==================
Would you please correct me if you see any obvious errors in how I set up this problem? I've also attempted a Hydrus simulation, but I'll leave that for another time.

I appreciate any advice.
Thanks,
Elise

ntoride
Posts: 74
Joined: Fri May 21, 2004 12:50 pm
Location: Japan

Post by ntoride » Thu Jan 10, 2008 5:39 am

Elise,
Although it seems to be OK with your input setting, please send me input files (project folder + cxt file). I will have look at the input and output files. Nobuo

--------------------------------
Nobuo Toride ntoride@bio.mie-u.ac.jp

Elise
Posts: 3
Joined: Fri Sep 19, 2003 2:15 am
Location: Australia

Post by Elise » Thu Jan 17, 2008 12:48 am

Please find below a record of the e-mail between Nubuo and Elise regarding the request for help on modeling Uranine BTCs. I hope others will benefit for this very helpful advice. Kind regards, Elise

-----Original Message-----
From: Nobuo Toride [mailto:ntoride@bio.mie-u.ac.jp]
Sent: Friday, 11 January 2008 10:25 PM
To: Bekele, Elise (CLW, Floreat)
Subject: Re: Inverse modeling Uranine BTCs with CXTFIT

Elise:
Enclosed is a modified project. I corrected the settings for the boundary condition, inverse data, and output data.

>The water flux is 100 cm/day, hence I divided my solute concentration
>flux by this value as it says in the manual for flux-averaged
>concentrations.

As van Genuchten and Parker (1984) discussed, flux conc. is a correction to use a solution for a semi0infinite system for an effluent concentration. Hence you do not need to divide the concentration by the water flux.

> By the way, my initial attempts failed when I provided a full BTC
> dataset (including the descending limb). Is this to be expected? I am
> sending you an Excel file that shows the full BTC obtained for the Red
> Station, but I was only able to get CXTFIT results when I used the
> ascending limb of the BTC.

I could not find a problem in your project.

If any further questions, please do not hesitate to ask me.

Regards,

Nobuo
--------------------------------------------------------
Nobuo Toride, Faculty of Bioresources, Mie University
1577 Kurimamachiya-cho Tsu, Japan, 514-8507
Phone: +81-59-231-9588 Fax: +81-59-231-9604 E-mail ntoride@bio.mie-u.ac.jp
==================================================================
==================================================================

-----Original Message-----
From: Nobuo Toride [mailto:ntoride@bio.mie-u.ac.jp]
Sent: Thursday, 17 January 2008 9:26 AM
To: Bekele, Elise (CLW, Floreat)
Subject: Re[2]: Inverse modeling Uranine BTCs with CXTFIT

Elise:

> Would you please tell me if I should be concerned that the model does
> not entirely fit the peak concentration? The overall shape of the
> curve is well-defined, but there are 3 data points that define the
> peak that are not fit by the model.

As described in p.228- in Jury and Horton's soil physics textbook, the CDE uses an analogy of the molecular diffusion for the hydrodynamic dispersion. If the microscopic heterogenous solute transport is different from the assumptions for the CDE, you may observed the irregular transport in the observed BTCs.

The mobile-imobile model is a possible modification, and the stream tube mode assuming no solute mixing in the cross-section area is also a possible candidate for these heterogeneous behaviors. As mentioned before, you also need to consider that you assumed constant v and theta for the CDE fitting.

You can find a lot of studies in soil physics and hydrology literatures regarding the solute dispersion model. It really depends on the situation how much we should take into account for these heterogenous behaviors for the transport modeling.

If any further questions, please do not hesitate to ask me.

Regards,


Nobuo

--------------------------------
Nobuo Toride ntoride@bio.mie-u.ac.jp

Post Reply