Modelling ruminant grazing systems based on native grasslands. 1. Model description and evaluation of the herbage...

Increasing the sustainability of grassland-based livestock farms in the Pampas region requires ecological intensification to increase beef production while preserving and enhancing ecosystem services. This involves farm-context-specific adjustments that capitalise on the complex interplay between the high-diversity grassland and animal dynamics. Modelling is expected to provide insights into elaborate farm redesign plans. A persistent challenge in grassland modelling is achieving both generality and realism in heterogeneous, multispecies native systems, and most grazing models cannot jointly simulate herbage dynamics, animal performance, and multipaddock management under contrasting climates. The farm-scale PAmpa Sustainable grAzing Livestock Management (PASpALuM) model was developed to help address this limitation and provide a framework for quantifying the effects of different combinations of strategic, tactical, and decision-supporting techniques on herbage and animals. Here, we present the model and evaluate the herbage dynamics module. In an accompanying paper, we demonstrated the suitability of PASpALuM for estimating herbage intake, animal growth, and reproductive performance. Herbage height was used as the key state variable, given its ecophysiological relevance for light interception and biomass accumulation and its practical use in the Pampas region. The module was evaluated with 2-year experimental and 3-year on-farm datasets covering contrasting soils and weather conditions. RMSE for annual herbage accumulation rate, mass, and height was 7.8 and 5.9 kg DM/ha per day, 578 and 512 kg DM/ha, and 1.7 and 1.7 cm for the experimental and on-farm datasets, respectively. Model performance was stronger in autumn and spring and weaker in winter. Mean squared deviation decomposition for herbage accumulation rate showed that lack of correlation dominated the error structure, particularly on-farm (63-88%), reflecting the high natural variability in heterogeneous grasslands. Contributions of non-unity slope were small to moderate (0.5-38%), and systematic bias was low (2-23%). Modelling efficiency ranged from 0.6 to 0.8, indicating good to very good performance. Sensitivity analyses indicated that outputs were most sensitive to maximum accumulation rate, moderately sensitive to water stress effects, and minimally affected by height and temperature correction factors. Overall, the accuracy of the PASpALuM simulations offers scope to compare grazing strategies aimed at increasing herbage height and allowance under contrasting climate scenarios, supporting seasonal grazing management planning at farm and policy levels.