ABSTRACT
The idea of crop models is to use information on environmental conditions, crop characteristics and crop management in order to make predictions of certain properties of the crop, in most cases the yield. Traditionally, crop models have been divided into two categories, namely statistical models and mechanistic models, while some are considered to be a combination of these, so-called semi-mechanistic models. The general argument has been that the statistical models are only valid within the data range where they have been parameterized, and in order to apply such models, calibrating them with local data is usually necessary. Therefore, models (or statistical relationships) obtained, for example, from on-farm trials may not apply in other sites or agro-climatic conditions (Burrell, 1991). In contrast, the parameters of mechanistic models are considered to remain relatively stable for a wide range of applications (e.g. Launay and Guerif, 2005; Bolker, 2008), and this is used as a justification for developing and using these, sometimes very complex, models in the context of crop research. However, in reality, this division is not as straightforward as it sounds, and we try to demonstrate this issue in this chapter.
