An evaluation of selected perennial ryegrass growth models for development and integration into a pasture management decision support system

Abstract

Four perennial ryegrass growth models were evaluated for their suitability to form the basis of a herbage growth model (HGM) for a decision support system (DSS). The successful candidate had to be suitable for further development to meet the specification of the DSS and following redevelopment it would then be integrated into a pasture management decision support system for dairy production. The models selected for evaluation were the Irish produced Brereton model (Brereton et al. 1996), the LINGRA model (Schapendonk et al. 1998), produced in the Netherlands, and a version of the English Johnson & Thornley (1985) model, developed for field use at the Northern Ireland Plant Testing Station (Laidlaw & Gilliland 2000). The fourth model was a version of the LINGRA model, simplistically adapted by the authors to take account of reproductive growth (LINGRA_Rep). The performance of the models was tested using the mean squared prediction error (MSPE) against a total of 28 seasons' growth data, collected from two sites; i.e. at the former Grassland Research Institute at Hurley, England and the Northern Ireland Plant Testing Station at Crossnacreevy. The Brereton model, when validated against the Hurley dataset, had the lowest MSPE of the four models, but had the highest MSPE against Crossnacreevy data. The PTS model did not perform as well as expected considering its mechanistic basis. Equally, the performance of the LINGRA model was poor at both sites. However, the LINGRA_Rep performed well, having the lowest MSPE at Crossnacreevy and second lowest at Hurley. The LINGRA model was selected for development as the final HGM given that it proved suitable for adaptation and by making even simple adaptations, as in LINGRA_Rep, its performance could be substantially improved. Therefore, it was considered that it possessed the greatest potential for further development.