Crop rotation is defined as the more or less fixed pattern in the successions of crops on a certain field (Struik and Bonciarelli, 1997). The most important attributes that characterize a crop rotation are the crop species, the frequency of each crop and the crop sequence (Struik and Bonciarelli, 1997; Dogliotti et al., 2003). Crop rotation schemes affect the yielding ability of each individual crop, the physical, chemical and biological soil fertility, soil erosion, the existence of soil borne diseases, pests and weeds (Dogliotti et al., 2003). Hence, crop rotation considerations are usually tried to be incorporated in the bio-economical farm models (BEFM). The latter use Linear Programming (LP) models, in order to capture as much as realistic the interactions and the relation between the different elements that comprise the farming systems.
The aim of this research is to analyse and compare two methodologies that capture crop rotation consideration in BEFMs and suggest improvements wherever it was thought to be necessary. The first methodology makes use of a modular approach constructing crop rotations exogenously of an LP model, i.e. SmartFarmer (Dogliotti et al., 2003), in a separate model ROTAT (Dogliotti et al., 2003) that generates feasible crop rotation schemes subject to agronomic rules. The second methodology takes into account the crop rotation considerations endogenously in an LP model i.e. MGOPT_CROP (Schans, 1996), by capturing all rotational aspects such as cropping frequency, sequence and between crop operations in the formulation of various constraints. In order to analyse and test the ROTAT model a manual simulation run was conducted for a small example. The latter obtained different results compared to the ROTAT model’s run by using the same input of data. This difference implies that there is a mistake in ROTAT model. This mistake is suggested to be inside the code of ROTAT that was used to model the ROTAT algorithm and not in the algorithm. Hence, a further investigation of the ROTAT code is suggested. According to the analysis of the MGOPT_CROP model, it has been proven that the latter encountered inconsistencies in the formulation of crop rotation constraints, which resulted either in erroneous crop rotation schemes or infeasible solution in many cases. After the improvements that were made the model provided correct results, in terms of crop rotation aspects, for all cases that was tested during this research. Finally, both methods were compared in a case study. The outcome of the comparison and the analysis of the two methodologies resulted in the assessment of pros and cons of each methodology. On the one hand the advantages of the first method compared to the second are that in general it is easier to be followed and understood; the construction of an LP model is less laborious; the outcome of the first method allows the consideration of the complete cropping sequence over all years of the crop duration, while the second method considers only the sequence of crops for two successive years. On the other hand, the second method comprises a more direct approach since there is no intermediate step of first generating crop rotations and then choosing the optimal one.