A comparison of two pasture growth models with observed data from Central Waikato from 2000 to 2020

Abstract

Pasture is essential to New Zealand farming, but its seasonal and annual variability complicates feed budgeting, farm management, and research predictions. This study evaluated the AgPasture and McCall-Romera pasture models for predicting monthly growth ates and annual herbage accumulation using observed
data from Scott Farm, Newstead, New Zealand, from 2000 to 2020. Both models overestimated annual pasture yield, with AgPasture showing a slightly higher bias (Mean Bias Error (MBE) = 1256 vs. 1155 kg DM/ha/year). However, the difference was not statistically
or practically significant. Adjusting for differences in annual nitrogen use between the observed and modelled data improved accuracy. McCall-Romera performed slightly better (MBE = 692 vs. 793 kg DM/ha/year), but again, not statistically significant. Both models showed seasonal biases underestimating pasture growth in late autumn and winter and overestimating it in late spring and summer. In May, the MBE was -11 kg DM/ha/day for AgPasture and -6 kg DM/ha/day for McCall-Romera, whereas in December, the MBE was
46 kg DM/ha/day and 23 kg DM/ha/day, respectively. Despite increased variability over two decades (coefficient of variation for AgPasture increased from 67% to 74%; McCall-Romera 46% to 58%), prediction accuracy remained stable, demonstrating robustness
under changing climate conditions. While both models are suitable for predicting pasture growth, further calibration is recommended to improve seasonal accuracy and enhance research reliability.