Hi, All,

IM(NS)HO It depends on the question(s) that you want to answer. If you want
to simulate the exact outcomes of a set of agronomic treatments, then you
need some good crop phenological data so that you can rigorously assess the
genetic coefficients of the germplasm that you are growing/simulating, and
you need some good soil data. You also need good met data. It is a rare
luxury that you have all these numbers, so you often have to resort to some
sort of kludge. For example, you can get daily met data from NASA, but the
satellite only passes each 10 days so that there is a lot of interpolation
in them. Also the satellite only measures water vapour concentration in the
atmosphere, not actual precipitation.

Alas, many of us want to use simulation to answer broader questions, like,
what effect could modern agricultural technologies have on global food
security in 2050? Integrated soil ferility management? Precision
agriculture? Organic farming? Drought tolerant cultivars? On a global
scale, the niceties of the characteristics of a particular soil get lost in
the general assessment. We can estimate the relative proportions of which
soil texture by depth occur in a given 5 km square pixel, and generate
weather data using DSSAT's WGEN, or better, extract climate nornals from
WorldClim and use an external weather generator like MarkSim, which also
incorporates the effects of climate change estimated by 6 GCMs and 3 SRESs.
We have to accept that DSSAT does represent the response of crop yield to
whatever external factor we apply, and that the relative proportions of
different soils in a particular pixel are correct (or at least that the
errors cancel out).

Getting to the detail, for CERES-rice you specify the components of water
management, puddling, bund height, added water amount or flood depth. The
FILEX RMZ8601.RIX in the distribution package gives an example in the
section IRRIGATION AND WATER MANAGEMENT. The codes are described in the
file DETAIL.CDE in the DSSAT45 root directory. You can specify percolation
rate (IR008), so  saturated hydraulic conductivity is irrelevant.

IRRI produced their own simulation model, ORYZA2000.  It is a LOT more
complicated than CERES-rice. It will be included in DSSAT 4.6, but is not
in the current version (4.5.1.023). You can get it as a stand alone,
www.knowledgebank.irri.org/oryza2000.

There is a DSSAT training workshop May 20-25 at the University of Georgia -
Griffin Campus. It would be useful for you to attend if you could. The
details are on www.dssat.net

Cheers.

Myles Fisher
Emeritus Scientist
Centro Internacional de Agricultura Tropical CIAT
Cali Colombia


On Thu, Jan 31, 2013 at 12:23 AM, Sommer, Rolf (CIAT-Kenya) <
[log in to unmask]> wrote:

> Hi Joy (?),
>
> IMHO model evaluation is always necessary, unless you apply the model
> within very narrow boundaries. For example, you have wheat cultivar x in
> location y on one field, and would like to run it on a neighbored field
> with same soil properties. Then I would assume that now further calibration
> or evaluation is required.
>
> But I'd say if you change soils from, say, loam to clay, you would need to
> have some data for calibration and evaluation.
>
> My experience tells me that especially clayey soils are difficult to
> simulate, as there might be quite some uncertainty e.g. in regard to field
> capacity, permanent wilting point and infiltration characteristics.
> Not sure though how CERES-Rice simulates inundated soils, and water
> requirements. I would assume that you have to set saturated hydraulic
> conductivity at the bottom of your profile, and that this most likely
> drives infiltration and water redistribution, saturation and thus overall
> water requirements. But, as I said, I am not sure.
>
> Regards,
> Rolf
>
>
> > -----Original Message-----
> > From: DSSAT - Crop Models and Applications
> > [mailto:[log in to unmask]] On Behalf Of Joy Mondejar
> > Sent: Mittwoch, 30. Januar 2013 20:01
> > To: [log in to unmask]
> > Subject: model evaluation
> >
> > Dear all:
> >
> > Is it always necessary to undergo CERES-Rice model evaluation before we
> can
> > proceed to model application and simulate rice yields?
> > I mean for each location that we want to predict rice yields, do we
>  need to
> > evaluate the model and get the site soil profile and soil surface data,
> crop
> > management data from a field experiment, and observed experimental data
> > from the experiment?
> >
> > I ask this since I am thinking if I can use a default  clay soil profile
> available in
> > DSSAT to simulate rice yields for 4 different locations.
> >
> > Thanks for the help and advice.
>