Simulations of plant productivity are affected by modelling approaches of farm management

Despite their wide range of applications, process-based plant (crop and grassland) growth models often fail to reproduce yields, particularly at farm, regional and larger scales. This is largely due to inadequate information about field management activities needed as input to these models. A promising approach to overcome this limitation is to link plant growth models with farm management models which allow the simulation of management activities considering farmers’ aims and constraints. Different approaches to model farm management are available, but tangible results to justify the choice for a specific approach are lacking. The objective of this work was to compare the effects of different approaches of modelling farm management on the simulation of grassland mechanized harvest dates and yields. Simulations were run with each approach for two grassland-based beef farms and 3years and compared with available data over 156 harvest events. Our results show significant differences in the accuracy of simulated harvest dates depending on the approach to model farm management. Approaches using fixed dates or optimal phenological stages determined by expert knowledge performed less accurate than the one using calibrated phenological stages. Best results were achieved with a detailed farm management model. The accuracy of simulated yields was less affected by the chosen farm management modelling approach. However, this differed depending on the climate and the timing of harvest, allowing to rank approaches according to their ability to simulate harvest dates and yields. We conclude that further investigation is required to generalize these findings to other farm types including arable farming, and to support the analysis, modelling and calibration of farmers’ management decision processes.