Dear members,

The discussion about genetic coefficients is getting interesting. Please,
let me remember two very simple principles about modeling: 1) Models are a
simplification of the real system; 2) Models are perfectible rather than
perfect. In that sense, genetic coefficients are an attempt to simplify
cultivar-specific responses to the environment. If the model would be
perfect and the genetic coefficients calibrated correctly then year after
year, site after site, we would obtain accurate predictions with our
models. In which case we would be running out of business, aren't we?

Therefore, if after calibration I found that my corn genotype requires 45
growing degree days to produce a new leaf (PHINT), but next year it will
require 40 and the following 50, it does not mean that my genotype is
changing every year the way it responds to temperature, photoperiod, etc.
It means that either 1) there are problems in the way I'm measuring and
calculating thermal time requirements for leaf appearance or, 2) the model
requires improvement for the desired level of accuracy (maybe both!). If I
use different PHINT each year I would get good fit between predicted and
measured values but I'll be hiding the problem.

Our knowledge would be better served by showing a weakness in the model and
exploring the nature of the problem than by improving predictions changing
inputs.

Best regards,


At 05:45 AM 7/17/2002 -0400, you wrote:
>Deae members,
>
>I've got so many answers and distributions about my last mail, that the
>genetic coefficients in CERES have to be constant or not..
>Whe tested this model in the collaborative project (Sonderforschungsbereich
>192) for 9 years (1991-1999) at the university in Kiel and we could not
>validate.
>I think it is not right to calibrate these coefficients for each year, to
>make the model work correctly, only for one year. Because this is an adaption
>to the environment conditions in this special year.
>Securely there are interactions between genetic coeffients and environment.
>And I think, these interactions have to be integrated in calculating
>predictions of CERES.
>Every genetic factor has an interaction to his environment - more or less -
>we know from researching of behaviour of genes. If you calibrate the model
>directly about these genetic coefficients, you are overlayling the other real
>parameters of of influence. Or what do you think?
>
>Best regards
>
>Peter
>---------------------------------------------------------
>Schoop, Peter, Dr. sc. agr., Dipl. Ing.
>Institut für Agri- und Environment Research
>Sonnenweg 1, 21647 Moisburg
>Germany
>Phone: +49-4165-224 700
>Fax:     +49-4165- 224 701
>E-Mail: [log in to unmask]
>Internet: www.n-prog.de, also in english!
>---------------------------------------------------------


Jon I. Lizaso
Dep. Agricultural & Biosystems Engineering
124 Davidson Hall
Iowa State University    515 294 7350 (Voice)
Ames IA 50011            515 294 2552 (Fax)