Is it correct to estimate all these parameters (D, S_max, K_a, K_d) simultaneously?

A discussion forum for Hydrus-1D users.
Post Reply
aajoker
Posts: 2
Joined: Wed Mar 15, 2017 6:04 am
Location: China

Is it correct to estimate all these parameters (D, S_max, K_a, K_d) simultaneously?

Post by aajoker » Fri Mar 15, 2019 9:16 am

Hi All,

I've fitted my data by the model including both time- and depth- dependent retention. In the previous project, the parameters S_max, K_a and K_d were estimated. Recently, I'm working on the transport of swimming bacteria, which may exhibit different dispersion process. Thus the value of Dispersity D should also be estimated for each breakthrough curve.

However, when I started to estimate Dispersity D simultanously (tick the Disp. option), their results were unsatisfactory with large values of Standard Errors and Sum of Squares. And vastly different results were obtained with different initial estimate values. Is it feasible to estimate all these parameters (D, S_max, K_a, K_d) simultaneously?

Any thoughts on this would be much appreciated!

Regards,

Jirka
Posts: 4712
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Is it correct to estimate all these parameters (D, S_max, K_a, K_d) simultaneously?

Post by Jirka » Fri Mar 15, 2019 4:06 pm

Normally, we try to estimate dispersivity from a separate experiment using a tracer. This is because all processes not considered by the model will be lumped into this parameter.

However, if your model is correct, then I do not see a reason why it would not be possible to estimate these four parameters from the breakthrough curve (BTC) since each of them affects a different part of BTC. Dispersivity affects the slope of increase and decrease of BTC, S_max the slope of the peak of BTC, attachment (K_a) the level of the peak, and detachment (K_d) the tailing after BTC. It will all obviously depends on the quality of your data, which should be reflected in the estimated parameter error and the 95% confidence intervals (should be small).

J.

aajoker
Posts: 2
Joined: Wed Mar 15, 2017 6:04 am
Location: China

Re: Is it correct to estimate all these parameters (D, S_max, K_a, K_d) simultaneously?

Post by aajoker » Sat Mar 16, 2019 3:11 am

Jirka wrote:
Fri Mar 15, 2019 4:06 pm
Normally, we try to estimate dispersivity from a separate experiment using a tracer. This is because all processes not considered by the model will be lumped into this parameter.

However, if your model is correct, then I do not see a reason why it would not be possible to estimate these four parameters from the breakthrough curve (BTC) since each of them affects a different part of BTC. Dispersivity affects the slope of increase and decrease of BTC, S_max the slope of the peak of BTC, attachment (K_a) the level of the peak, and detachment (K_d) the tailing after BTC. It will all obviously depends on the quality of your data, which should be reflected in the estimated parameter error and the 95% confidence intervals (should be small).

J.
Thanks for your detailed and helpful reply!

Post Reply