Ronnie,

first you need to know what kind of crops
you would like to simulate. The most
important crops are addressed in DSSAT.
In case they are not you could still try
to adapt the shell to include a new crop.
I personally know one example where CERES
was modified to include growth modelling of
canola crops. However, this may become
tricky and also involves some validation
work. You must decide wether you want
to spend these efforts, which consume
considerable time and need capacity as
well.

I really appreciate the minimum data
input concept of DSSAT for practical
reasons. As I understand it, the concept
is just suited for the particular purposes
you mentioned where available input
data is scarce. However, it is important
to have this minimum data input available.
The more assumptions you make the more
uncertain your results will get. Here is
what you will need according to my
understanding :

Standard weather information
(Tmax, Tmin, Rain, daily global radiation;
the letter is often not available. You could
try to use correlations between temperature
and solar radiation. If you cannot derive
them at your particular site you you could
try the nearest airport. They usually have good
weather data).

Soil information:
Particle distribution for classification and
water characteristics (%Vol, conductivity at saturation)
slope, area and an empirical root growth factor.
All soil information could be derived from
literature if data is not available. However this
introduces some uncertainties.

Crop information:
Standard phenology and yield components
under optimum growth conditions. You will
need those to get the genetic coefficients.

Initial Conditions:
Water distribution (%Vol), Nitrogen (NH4+
and NO3-), soil-pH and organic carbon content


I personally would not call DSSAT a bio-physical model.
This would imply that the processes underlying crop
growth would have been parameterized in great detail
(photosynthesis, carbon metabolism, enzyme kinetics etc.).
It is my impression that we need such detailed data and
models to derive simple and robust relations to make modelling
a more science based operation. However, frameworks of such
detailed interrelations would be never practically applicable.
A study was recently published by Robin Matthews which verifies
this impression.
(http://www1.silsoe.cranfield.ac.uk/iwe/People/RobinMatthews.htm)

According to my understanding the real strength of DSSAT is
twofold :

It is a decision making tool which is practically applicable. It is
based on long-term experience from many parts of the world. It
is simple and robust and therefore well suited for practical
purposes.

It forces one to think in system-wide dimensions, an approach,
which seemed to be uncommon until the development of
DSSAT.

I regret not to be able to attend next week's symposium
honoring Joe Ritchie. I should mention that I was also very
much influenced by his work. So let me send my applause
from this place.

Matthias Langensiepen


--
Matthias Langensiepen, PhD
Horticulture
University of Hannover, Germany