Mathias,
Thanks for your kind response. You just spilled the beans! When you
doubt that the soil water is correct why do you want to move further
until it is fixed. This explains my worries!
Also, I did not see in any DSSAT manual  that one need switch off some
modules for calibration! (It was many years back I read those books, if
I am wrong you can correct me). If I am correct switch on all the
modules, get the soil water correct first, then move on to nitrogen, and
finally to the crop.
Also, you need to use the DSSAT v4. instead of v3.5. Certainly there is
something better in the newer versions than the old ones always. I am
not sure, if you can download newer versions free when you have the old
version. Please check with the vendors!
Sasi



Saseendran S Anapalli,  Ph. D.
Agricultural Systems Research Unit
USDA-ARS
2150 D Center Ave.
Fort Collins, CO-80526


-----Original Message-----
From: Matthias Langensiepen [mailto:[log in to unmask]]
Sent: Friday, March 07, 2008 12:02 PM
To: Anapalli, Saseendran
Subject: Re: weakness of CERES-wheat

Dear all,

a brief response to Dr. Anapalli's recent mail: We did use the
model "correctly" by strictly following the prescribed procedures
for running the DSSAT 3.5 manual. It took us more than 3 months
alone to calibrate model. After completing this exercise we
came to the conclusion that this procedure is indeed "the most
cumbersome and subjective method of curve fitting that can be
imagined" to speek in Corneluis Teunis de Wit's words. I am aware
of two more large project which currently try in vain to get
the model "right".

Whether the model or the validation set is wrong is an
old argument between model developers and testers. To
my understanding validation data cannot be made subject
to dispute, because they represent reality. It is the
model (fed with appropriate data according to procedures
prescribed by the develeopers) which must compute reality
as good as possible. If it doesn't we have to look for
possible errors which leads to further model improvements.
The soil water routine of CERES is a key candidate and we have
just completed another study which focussed on
this problem. We will inform you about its outcome
when the final results become available.

Matthias


Anapalli, Saseendran wrote:
> Dear Mathias,
> Thanks for taking time to write to the DSSAT list. What you wrote
really
> makes sense, and hope all plant scientist would vote for it.
> Nevertheless, we also need understand the limitations of
> numerical/mathematical models, especially when deal with biological
> systems and their complex interactions with abiotic processes as we do
> in cropping system models.
>         To my understanding, the DSSAT models have mainly three
> components, the water, nitrogen and the crop. For modeling the crop,
one
> need make sure that the first two components works reasonably well for
> the system before modeling the third (the crop). I could not see any
> efforts in the study to do this or see any results presented. I expect
> the least a table showing the crude water balance simulated to prove
> that you got the soil water/nitrogen correct.  For that reason, it is
> left to anybody's guess what those components were doing! It is also
> possible that one do not want to present all the results when we
publish
> papers in journals that charge for pages.
> Notwithstanding, what I see in the study is that the calibration runs
> were made turning off the water and nitrogen modules. To my
> understanding, it is in the field experiment used for calibration one
> should make sure that there were no abiotic stresses but not in the
> model!
> If one is one hundred percent confident that the field experiments are
> free of water, nitrogen, and temperature stresses, simulations
> neglecting (turning off) those processes can be a reasonable
> approximation. But, in the studies presented I could not see
> experimental treatments!  However, I did notice that the optimum N
> treatments were defined as treatments with different levels of N and
> manure levels. Here also, I do not see anything mentioned about water
> levels in the treatment. Did you totally forget about water and
nitrogen
> interactions? Keeping in view of the heterogeneous behavior of the
> various processes and their interactions in the system do you honestly
> think these treatments are be free of abiotic stresses for turning off
> the switches in the model, unless we stretch our imaginations really
> thin?
>
> What I am trying to say is not that the model is or would work
> wonderfully well in all the environments and circumstances, as you
> pointed out and we all know that there are many unknowns and
> uncertainties in the system that we are not able to grasp yet to model
> them correctly. But, given the limitation, at the least the models
> should be used correctly to point out their limitations, and this
would
> certainly appreciated by one and all.
> Hope you appreciate the errors in your simulations.
> Sorry for taking your time and patience.
> Best wishes
> Sasi
>
>
> Saseendran S Anapalli,  Ph. D.
> Agricultural Systems Research Unit
> USDA-ARS
> 2150 D Center Ave.
> Fort Collins, CO-80526
>
>