Parameter estimates from multiple simulations?
Is it possible, through TOUGH2 or iTOUGH2, to perform parameter estimations based on multiple simulations that are coupled to one another? I know that for a forward simulation you can establish initial conditions based on a previous simulation on a subsequent simulation by hand using the SAVE file. But could I change aspects of subsequent simulations programmatically in a single TOUGH2 input file for inversion through iTOUGH2? If not, is there another way to go about estimating parameters based on multiple simulations?
5 replies
-
Mikey,
You can accomplish this in several ways. Just briefly:
(1) make a single TOUGH2 input file that contains multiple "models" (they may share the parameters to be estimated). You would need to have separate "meshes" with unique element names (all combined into the same ELEMEN/CONNE blocks). This makes it run slower, which may or may not be an issue. Problem 2 in the iTOUGH2 Sample Problem report and the WRR article on data-worth analysis are examples of this approach.
(2) Use iTOUGH2-PEST, which allows you to link as many TOUGH2 (and non-TOUGH2) models together as you wish (at the expense of convenience in defining parameters and observations). The iTOUGH2-PEST Manual (e.g., Section 5.4) describes this approach.
Hope this helps,
Stefan
-
Thanks Stefan! Is there any way to go about the first approach if I need to assign the values at the end of the simulation of my first "model" as the initial conditions for the simulation of my second "model"? It seems the only links between "models" here are the parameters to be estimated. If this can't work, I'll look into the PEST protocol.
-
Hold on... I just came across the RESTART TIME command. I think this may be what I was looking for.
-
Mikey,
RESTART will not allow you to take a SAVE file from one "model" as INCON for the other. iTOUGH2-PEST seems to be the preferred option for you.
Stefan
-
That's good to know in the future, but I'm afraid I may have overgeneralized my original question. Essentially what I was trying to do is to make minor modifications to the MESH between multiple simulations of the same model to alter boundary/initial conditions. I was able to run an iT2 forward simulation with this option, and it seems to work.