Dear All:
 
After reading Dr. Matthias Langensiepen's response and his conclusion about
the DSSAT model and its inability to predict yield, I would like to make
some comments.  Please keep in mind that my comments are not being made to
criticize anyone, but just want to expose my feelings about the DSSAT
models.
 
I don't think it is fair to assume (or conclude) that just because a model
did not perform well for a particular situation, this same model will not
perform well for other situations.  As far as I know, nobody ever claimed
the DSSAT models to be universal.
 
A lot of researchers (and other users) tend to criticize the more simple
and functional models just because those models did not work for their own
conditions.  Even though I don't have a lot of experience with the DSSAT
models (as well as mechanistic models), I do know that one of the main
reasons for the poor performance of the DSSAT models, is the lack of
realistic inputs that are fed into the models.  This is especially true for
the soil inputs (they are hard to measure and difficult to understand!).
Moreover, the models still require some type of local evaluation with
measured data, especially with respect to the cultivar coefficients.  In
many cases the cultivar coefficients need to be redefined for local
cultivars and many people seem to forget this. Soil water measurements
(reliable ones) are also very critical for evaluating the model.
 
A lot of people believe that simple and functional models do not require
accurate inputs.  In fact, they require the best input you can get in order
for them to work properly (measured in the field).  If you have good
inputs, at least you have a better chance to understand why a model did not
work for your situations.
 
It is common to see many users underestimating the importance of some input
sections of the DSSAT models.  In my opinion, every input section required
by the model should be treated as the most critical for the good
performance of the model.  We should avoid making too many "assumptions" or
using "mean" values from the literature.  As we all know, Garbage in =
Garbage out.  Therefore, let's take care of this problem.  On the other
hand, a lot of researchers are using excellent inputs and still getting
poor simulations.  This shouldn't be a surprise for anyone working with
models.
 
Sometimes people make more assumptions in the DSSAT models than in the
mechanistic models.  Again, the DSSAT models need good inputs to work
reasonably well.  What I really like about the DSSAT models is that these
models are being validated and improved every single day, at different
parts of the world.
 
I appreciate your patience in reading this message.  I am not claiming that
the comments above are the truth in crop modeling, but are just my feelings
about the DSSAT models.
 
Best regards,
 
Miguel