Hi all,

I agree with everybody's comments and feel the questions are the right ones to ask. It sounds to me that the questions and comments so far are:

1. Are genetic coefficients environmental (site, season) dependent?
2. Genetic coefficients are even model specific (for example, CERES-Maize and APSIM-Maize).
3. If genetic coefficients are not environmental dependent, why cannot we always use the same genetic coefficients we developed in one place to another?
4. Practically, what should we do?

Here I just want add some more comments along with some of them you all already made.

1. We, most of us as crop physiologists and modelers, try to define the FIXED genetic variation among cultivars using what we called "genetic coefficients (GC)". Theoretically and ideally, these coefficients should be very stable but not vary a lot as Ken said. However, most of genetic coefficients in current crop simulation models have not linked to the true genetic variation from genetics point of view. Most of them are actually phenotypic phenomena but reflect genetic variation to large extent regarding crop growth and development. Therefore they are very stable in most of environmental conditions but still have exceptions when cultivars grow outside the scope of environments they were developed and tested. For example, the potential maximum kernel numbers (G2 in CERES-Maize and head_grain_no_max in APSIM-Maize) for a specific hybrid would never be known for its absolute value since it is not the genetic fixed. We may find a new environment where we got another value that does not agree with ones we had before. Therefore, one should not expect that once you have a set of genetic coefficients in few environments (no matter how many years and locations you tested), you can use it universally without having any modification. The point I want to make is genetic coefficients are not 100% fixed but they are very stable if we generate them with sufficient datasets for the cultivars under different environments. From my experience, the genetic coefficients I generated for about 200 Pioneer commercial hybrids have been very stable thanks to the datasets of our multi-environment trials.

2. For different models, same genetic coefficients may not be exchangeable without any adjustment as Rolf observed, especially his example. This is because different model has been modified inside to model the same process but may use the coefficients in little bit different way. Here is another example, tt_flower_to_maturity in APSIM has the same concept as P5 in CERES-Maize, but it represents different period of thermal time in APSIM. Therefore, if one uses the value P5 directly for tt_flower_to_maturity in APSIM, you may get different result. Therefore, you needs to understand what the model you use and make reasonable adjustments of GC accordingly.

3. From the above, it is not difficult to understand why we cannot always expect the same set of genetic coefficients are universal. However, we also should not lose faith in them and change them every time when the prediction from models is different from observations.

4. Probably, the most difficult question is how we use the result from model practically to solve the problems in the real world. Some of my comments are:
        a. We probably should have a right level of expectation to models. As we all know, model is an purposeful abstraction of a real world. It reflects our understanding of the system we study and the predicting ability of a model depends on how much we know about the plant and its environment. Therefore, we cannot expect a model gives us the right answer all the time, rather we should trust the models such as those in DSSAT and APSIM, which is based on years of the sound scientific knowledge, understanding and findings from scientists from different disciplines, can give us very reasonable predictions to SUPPORT our decision (that is probably one way to see why we call DSSAT). We still need a second opinion from experts (for example extension specialists) to have a sensibility analysis (common sense) to validate or verify the model's prediction although I believe most of times both are agreeable.

        b. A good use of models such as recommendations or support of decision-making are based on the high quality of input data as many of you have mentioned from genetic coefficients generation, model validation and parameterization, to model application. The random errors and other abiotic and biotic factors that models have not considered are another set of killers to frustrate people and prevent us from leverage the power of crop models and simulation. While we never can eliminate these factors, we should do our best to reduce them to the minimums with our capacities. Furthermore, we should be aware that different environments for genetic coefficients generation are the key factor we need to consider. Number of years and locations are not relevant if the environments in these years and locations are similar. The bottom line is to optimize the set of genetic coefficients that reflect the cultivar's general genetic background and potential in the range of different environments you want to test, evaluate and use a specific cultivar.

To address Matthias's specific questions,

Can we have faith in such predictions ?
Yes, since the model has been developed, modified and tested for many environments.
No, this specific variety or the environment you want to recommend might be an exceptional combination.

I recommend you may make few runs using the coefficients you had already to see if the result makes sense to you and others. Especially when you talk and explain the simulated result to farmers, see what they think?

In case we don't, were should we start adjusting ?

You may need to conduct an field experiment. Make sure all input data such as soil you mentioned has high quality, see if you need to modify genetic coefficients since you believe the environment is different.

Is it really the model that needs to be
adapted first, or is it actually the
genotype that shows some degree of flexibility
in adjusting to the new environmental
conditions ?

The model may need to be adjusted too but that may be the last thing you want to do. Firstly you may try to modify genetic coefficients so that you optimize them for different environments including this new one.

Hope this helps rather than causes more confusion.



Jun

___________________________
Dr. Jun Wei, Research Scientist
Pioneer Hi-Bred International, Inc.
A DuPont Company
7250 NW 62nd Ave.
Johnston, IA 50131

Phone: (515) 334-6704
Fax:     (515) 334-6634
Email: [log in to unmask]
http://www.pioneer.com/



-----Original Message-----
From: DSSAT - Crop Models and Applications
[mailto:[log in to unmask]]On Behalf Of Matthias Langensiepen
Sent: Wednesday, February 04, 2004 4:31 AM
To: [log in to unmask]
Subject: Re: Genetic coefficients


Ken,

if cultivar coefficients are not site
or environment dependant we could perform
following comparative study:

A new wheat variety has been developed
in Sweden under temperate maritime climate
and terminal moraine soil conditions. To
be able to run the CERES-Wheat model
I would get in touch with some official
institution which performs standarized
breeding testing programs and ask for
their data. Such experiments are typically
carried out under optimum management
conditions, provide a wealth of data and
are thus suitable for performing the
calibration procedure (GENCALC). I would
re-run the calibration procedure to cover
experimental repetitions and different seasons.
If genetic parameters are not season specific
I would then possibly be able to obtain
relatively stable genetic parameters.
They would likely differ from those stated
in the "WINTER-EUROPE" standard set.

Now, only a few hundred kilometers apart
a Polish farmers association, located let's
say at Lodz, got interested in the high yield
potential feature of this new breed and asks
me to determine how this new breed will
possibly perform under their environmental
conditions. Since these conditions
are entirely different from those in Sweden
I would make the necessary modifications in
the soil.sol file and carefully check
the data that must be entered in the *.whx
file. Now, if genetic coefficients for
the new breed are universally applicable,
as you say, I would be able to run the
model under Polish conditions and base
my advise on these simulation outputs.

Can we have faith in such predictions ?

In case we don't, were should we start
adjusting ?

Is it really the model that needs to be
adapted first, or is it actually the
genotype that shows some degree of flexibility
in adjusting to the new environmental
conditions ?


Matthias Langensiepen
Modelling Plant Systems
Institute of Crop Science
Faculty of Agriculture and Horticulture
Humboldt-University of Berlin
Germany


This communication is for use by the intended recipient and contains
information that may be privileged, confidential or copyrighted under
applicable law.  If you are not the intended recipient, you are hereby
formally notified that any use, copying or distribution of this e-mail,
in whole or in part, is strictly prohibited.  Please notify the sender
by return e-mail and delete this e-mail from your system.  Unless
explicitly and conspicuously designated as "E-Contract Intended",
this e-mail does not constitute a contract offer, a contract amendment,
or an acceptance of a contract offer.  This e-mail does not constitute
a consent to the use of sender's contact information for direct marketing
purposes or for transfers of data to third parties.

 Francais Deutsch Italiano  Espanol  Portugues  Japanese  Chinese  Korean

            http://www.DuPont.com/corp/email_disclaimer.html