Posted Fri, 21 Aug 2015 21:33:52 GMT by adacovsk

So I am trying to estimate my flow flux boundary conditions via FePEST... I have flow going into the model and flow going out of the model and solving the inverse problem. However, since there is flux in/out we have negative values being written to the fpi file... I am thinking that negative values for ifm-implemented parameters in FePEST cannot be handled? I am using relative change limit. I'm not too good with this, so I don't know if it is a bug or my limited knowledge. :^)


Posted Mon, 24 Aug 2015 08:04:26 GMT by Björn Kaiser
Parameters which can have negative values can be estimated by PEST. The only problem is, is that you cannot log transform them.

However there is a way to still use it. Set the initial value to a positive value. Then please make the lower and upper bounds positive, but set the SCALE option to -1. In this way, the parameter is set to the negative of its value, just before it is set to the FEFLOW model.

As log transforming a parameter is recommended you probably should do this.
Posted Mon, 24 Aug 2015 20:09:14 GMT by adacovsk
Hi Björn, thanks for the suggestion, it's very clever! Never thought to scale by -1 and/or offset it!

Everything seems to be working, with a few minor hiccups. It seems that I am limited to a constraint of ~1e-8 for the flux and not smaller (really tiny model!). Also, I need to make 4 character names for the parameters or else scaling by -1 doesn't work? Might just be the delimiting causing an minor issue, oh well!



You must be signed in to post in this forum.