Improvement of cotton production in Africa using a mechanistic crop model could be accomplished through crop cultivation decision aid and production predictions.

The nature of tropical environments requires certain adaptations of the GOSSYM model. At the field level, spatial crop heterogeneity poses the problem of the determination of the simulated object: an average plant, or typical plant, and its significance in regards to the dynamics of the entire vegetative cover. Some of these spatial crop heterogeneities come from the variable conditions of the specific local environment: chemical and physical soil parameters or interactions between plants, within-plants, and with weeds. In order to integrate these idiosyncrasies, GOSSYM needs to be improved in regards to photosynthesis, water and nutrient absorption.

These improvements assume improvements in the modeling of morphogenesis and partitioning can be better represented. The partitioning of water and nutrient stresses can then be better localized in terms of plant architecture. This research areas are considered top priority and they are a part of a European project presently in collaboration between France, USA, Cameroon, Thailand and Spain.

Assuming that the GOSSYM model has been effectively tropicalized, what kind of applications can we foresee in an extension context? The first potential application concerns the prediction of yield at a regional scale. The GOSSYM model will then have to function based on heterogeneous, multi-field samples. This poses the problem of the sampling system, the collection of data necessary to run GOSSYM, and the precision of this data taking into account the objectives of maintaining an appropriate balance of the cost/precision ratio, and of accounting for the heterogeneity of the physical and human factors.

The second potential application is to implement, based on GOSSYM simulation runs, personalized advice for field-cluster-specific management in discrete regions. The constraint of the heterogeneity and limited monetary resources lead us to envision the utilization of GOSSYM by varying scenarios to be applied to management clusters within discrete regions based on ecological conditions to obtain production objectives. Further, GOSSYM can become a good training tool for extension agents in the diagnosis of cultivation practices. In each case, the challenge will be to find an equilibrium between the necessary simplification of model inputs and the inclusion of spatial, plant, and human heterogeneity.