Building crop models within different crop modelling frameworks

Modular frameworks for crop modelling have evolved through simultaneous progress in crop science and software development but differences among these frameworks exist which are not well understood, resulting in potential misuse for crop modelling. In this paper we review differences and similarities among different developed frameworks and identify some implications for crop modelling. We consider three modelling frameworks currently used for crop modelling: CROSPAL (CROp Simulator: Picking and Assembling Libraries), APES (Agricultural Production and Externalities Simulator) and APSIM (Agricultural Production Systems sIMulator). The frameworks are implemented differently and they provide more or less flexibility and guidance, to facilitate assembly of crop model from model components. We underline the importance of systematic approaches to facilitate the selection of appropriate model structure and derive suggestions to facilitate it. We particularly stress the need for better documentation of the underlying assumptions of the modules on simulated processes and on the criteria applied in the selection of these modules for a particular simulation objective. Such documentation should help to point out the sources of uncertainties associated with the development of crop models and to reinforce the role of the crop modeller as an intermediary between the software engineer, coding the modules, and the end users, agronomists or crop physiologists using the model for a specific objective. Finally, the key contributions of modelling frameworks in the crop modelling domain are discussed and we draw conclusions for the prospects of such frameworks in the crop modelling field which should continue to reside on the principles of systems analysis but combined with up-to-date advances in software engineering technique