PEST optimization

A discussion forum for users of the new HYDRUS 2D/3D. HYDRUS is a software package for simulating water, heat and solute movement in two- and three-dimensional variably saturated media. Happy Posting!
Post Reply
dustming
Posts: 31
Joined: Thu Jul 21, 2011 4:40 pm
Contact:

PEST optimization

Post by dustming » Fri Jul 29, 2011 3:33 am

Hi all

I try to use pest to do optimization, everything looks fine, but PEST always inform me that n,ks,and a in the Genuchten - Mualem has no effect on the observations,is there any one konw why? thanks,following is my pest control file:
pcf
* control data
restart estimation
18 390 7 0 1
2 1 single point 1 0 0
10.0 2.0 0.3 0.03 10
3.0 3 0.01
0.1
30 0.01 4 4 0.01 3
1 1 1
* parameter groups
b1 relative 0.01 0.0 switch 2.0 parabolic
n1 relative 0.01 0.0 switch 2.0 parabolic
k1 relative 0.01 0.0 switch 2.0 parabolic
r2 relative 0.01 0.0 switch 2.0 parabolic
r3 relative 0.01 0.0 switch 2.0 parabolic
r5 relative 0.01 0.0 switch 2.0 parabolic
r6 relative 0.01 0.0 switch 2.0 parabolic
* parameter data
b1 none relative 6.000000E-02 0.01 0.09 b1 1.0000 0.0000 1
n1 none relative 1.50000 1.2 4 n1 1.0000 0.0000 1
k1 none relative 20.0000 4 100 k1 1.0000 0.0000 1
b2 none relative 4.000000E-02 0.01 0.09 b1 1.0000 0.0000 1
n2 none relative 1.40000 1.2 4 n1 1.0000 0.0000 1
k2 none relative 10.0000 4 100 k1 1.0000 0.0000 1
b3 none relative 5.000000E-02 0.01 0.09 b1 1.0000 0.0000 1
n3 none relative 1.50000 1.2 4 n1 1.0000 0.0000 1
k3 none relative 15.0000 4 100 k1 1.0000 0.0000 1
b4 none relative 4.000000E-02 0.01 0.09 b1 1.0000 0.0000 1
n4 none relative 1.50000 1.2 4 n1 1.0000 0.0000 1
k4 none relative 7.0000 4 100 k1 1.0000 0.0000 1
r1 fixed factor 70.0000 60 70 r2 1.0000 0.0000 1
r2 fixed factor 38.130000 1 50 r2 1.0000 0.0000 1
r3 fixed factor 2.5000 0.1 15 r3 1.0000 0.0000 1
r4 fixed factor 65.0000 60 70 r5 1.0000 0.0000 1
r5 fixed factor 47.6400 1 50 r5 1.0000 0.0000 1
r6 fixed factor 3.838000 0.1 15 r6 1.0000 0.0000 1
b1,n1,k1,means n,ks,and a in the Genuchten - Mualem for layers 1,there is four layers in my case.
thanks very much!

FrankS
Posts: 26
Joined: Wed Jan 23, 2008 4:48 pm
Location: USA

Post by FrankS » Mon Aug 01, 2011 2:52 pm

Impossible to tell. Insufficient information provided. For example, what is the objective function? What is the modeling scenario, etc.

Cuneyd
Posts: 4
Joined: Thu Dec 20, 2018 4:17 pm
Location: Turkey

Re: PEST optimization

Post by Cuneyd » Thu Dec 20, 2018 7:21 pm

For very small parameter values log transform is recommended in PEST manual


Experience has shown repeatedly that log transformation of at least some parameters can
make the difference between a successful inversion process and an unsuccessful one. This is
because, in many cases, the linearity approximation on which each PEST iteration is based
holds better when certain parameters are log-transformed. It also follows from the fact that,
implied in log-transformation, is a normalization of parameters with respect to their innate
variability. The inversion process is thereby able to access, to some extent at least, the
benefits which follow from Kahunen-Loève transformation of parameters.

Page 93 at
http://www.pesthomepage.org/getfiles.ph ... stman1.pdf

Post Reply