Dear Luis,
There are many options to set about your problem: you may use data and information from one field (all trials x all seasons) to calibrate the model and try to validate it at the second field. Alternatively you can use information from both fields restricted to years 1997 and 2000 to calibrate the model and then validate with 2001 data. Actually it is recommended to use the most accurate data for calibration (1997 or 2001 or field 1 data may be more reliable than the others). In southern Togo we have used one year data (many fields) to calibrate CERES-Maize and we could validate at a second site with different agro-environmental conditions. This allowed us to account for different soil, weather, and management conditions. The job is not so straightforward I should add. Next we used the validated model with long term weather data to derive recommendations on best combinations of sowing dates, cultivars, and nitrogen (paper in press).

Quite all that happens for a good calibration is make sure the basic data you use are the most reliable you have. The least information (field observation on crop response to drought, plant population...) may be useful and interfere. It is important to detect such interferences and account for them, sometimes by modifying the soil or the experimental file accordingly to reflect such differences if they cannot basically show up in genetic coefficients. It might not be useful to wait for multi-years' trial to validate a model- securely there are many things you can already do with your 2 fields x 3 years data set,

Best regards
Kofikuma

IFDC Africa Division
Crop simulation modeling (Soil fertility)
[log in to unmask]
www.ifdc.org 

----- Original Message ----- 
From: "Luis Alvarez" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Tuesday, June 03, 2003 4:41 AM


> Dear Modellers:
> During three rice crop seasons (1997 and 2000 rainy seasons and 2001 dry season) I have collected soil, weather, cultivar and management data in order to calibrate, evaluate or validate? CERES-rice.
> I have used the information of 4 tryals in 2 fields (rice farms) to calibrate, and used these information for model validation.
> I know that in the iterative process I only calibrate the model. With information of other places and years I could validate it.
> I would like to have your and Dr. K.J. Boote (I don?t have his e-mail) opinion about this problem. Should I use the same information to calibrate and validate or not?. Perhaps this process could take many years, so where are the advantages of modelling?. I know that I can use the validated model in many important features, but it is still expensive in time and money, and in our tropical environment events goes very fast.
> Thanks for your advise.
> Sincerely yours:
> Luis E. Alvarez Larrauri
> Universidad Ezequiel Zamora. UNELLEZ.
> Guanare. Estado Portuguesa. Venezuela.
>