Calibration of the APSIM-Lucerne model for ‘Grasslands Kaituna’ lucerne crops grown in New Zealand

Abstract

This study used results from field experiments to calibrate the Agricultural Production Systems Simulator (APSIM) model for ‘Grasslands Kaituna’ lucerne (Medicago sativa L.) crops grown in the cool temperate climate of New Zealand. The field data allowed refinement of functions related to the seasonality of crop phenology and above- and below-ground growth. The model structure was adequate to accommodate the calibration of reproductive development in response to temperature and photoperiod in both release and calibrated versions (root mean squared deviation [RMSD] from 14% to 3% of the mean, respectively). Similarly, vegetative development, characterised by the rate of appearance of main-stem nodes, was simulated with acceptable accuracy (RMSD from 28% to 25% of the mean). Analysis showed a lack of current field data to accurately represent the number of main-stem nodes at the beginning of each regrowth cycle, which is required to further improve the accuracy of leaf appearance simulations. The simulation of leaf area index (LAI) was slightly more accurate when using a site-specific empirical canopy modelling approach for ‘Grassland Kaituna’ (RMSD = 34% of the mean) than with the more mechanistic approach in the release version used for Australian cultivars (RMSD = 40% of the mean). This is particularly important before canopy closure, when LAI has the greatest influence on light interception and biomass accumulation. ‘Grassland Kaituna’ showed a seasonal pattern of change in leaf size and specific leaf area that could not be parameterised using the approach in the release version. Finally, the model structure of the calibrated version was adapted to explicitly simulate the seasonal flow of biomass to shoots and roots. For that, the ‘shoot’ radiation use efficiency (RUE) was modified to a ‘whole plant’ RUE of 1.6 g/MJ total solar radiation and a seasonal rate of below-ground biomass turnover (i.e. respiration plus senescence) was incorporated to correct for the initial offset observed between simulated and observed root biomass. These changes improved the RMSD from 53% to 38% for shoot biomass and 29% to 18% of the mean for roots. This calibration illustrates the heuristic aspect of applying crop models to identify knowledge gaps for guiding future experimentation, including time of node appearance on basal buds and leaf area index dynamics of contrasting lucerne cultivars.