mf2k can calculate sensitivity of the simulated equivalents to the parameters of interest using the SENsitivity package.

Work this exercise under C:\Temp\YOURNAME\MODFLOW\exer\senmf2k\

starting by copying all the working files from obsmf2k into senmf2k

1) create an SEN file and add it to your exsscal.nam file with a new unit number, see the MF2K OSP manual pages 67-75 or (OnLine Manual)
    calculate sensitivity for parameters K1, K2base, KVtardS, KRB, RCH1, and RCH2

Suggested SEN input:
6 0 60 6 #Number-Parameters #global-sensivity/estimation-control #startingunit-for-temp-sens-files (cannot include 96-99) #max-parameters
1 0 0 0 #flags controlling saving and printing of sensitivity arrays
K1 1 1 3.0e-4 3.0e-5 3.0e-3 1e-7 #ParamName #SensFlag #LNflag #Value #reasonable-lower #reasonable-upper #alternate-scaling-for-values-of-zero
K2base 1 1 4.0e-5 4.0e-6 4.0e-4 1e-8
KVtardS 1 1 1.0e-7 1.0e-8 1.0e-6 1e-10
KRB 1 1 1.2e-3 1.2e4 1.2e-2 1e-6
RCH1 1 0 2e-8 2e-9 2e-7 1e-12
RCH2 1 0 1e-8 1e-9 1e-7 1e-12

You may also want to add the GLOBAL package to your name file at this time. This will cause MODFLOW to separate some of the output into the lst file and some to the glo file. Note, the first file in the name file must be list or global, so add this to the top of your name file. My new nam file looks like this: exsscal.nam file

use the pcg2 solver (page 82-85 of the MF2K GWF manual) exsscal.pcg2 file
rather than the sip solver because it will be more effective for calculating sensitivities (see pages 70-71 MF2K OSP)
be sure to delete sip and add pcg to the nam file exsscal.nam file

Suggested PCG input:
1 50 1 #MAX-OUTER-ITER #MAX-INNER-ITER #MATRIX-CONDITIONING-FLAG
0.0001 0.0001 1. 0 1 0 1.0 #HTOL #QTOL #RELAX.97-1 #if-above-is2 #print-interval #amount-of-convergence-printing #damping

    RERUN your model and note the difference in the exsscal.lst output
    correct any problems and proceed exsscal.sen file

Review output of exsscal.lst with sen

view all the other output files (these are all of the form root._**, I chose sscal for my root) and determine what information you have

*notice the sensitivity of correlation to parameter values and closure
*notice the incredibly good fit can sometimes cause computational troubles, these are highly unlikely in field problems