Adapting Functional Trait-Based Mechanistic Model for Subtropical Climates.

Simulation models are important tools for studying and predicting effects of variable environmental conditions and disturbances in ecosystems. However, many grassland models are highly demanding in terms of parametrization, limiting their use across sites and species. Recent studies suggest that plant functional traits may be a useful approach to simplify and generalize grassland models, creating a robust tool that can be easily transferred across species and environments. To date, little information is available on mechanistic, functional-trait based models in grasslands. Here we adapt a dynamic mechanistic model developed for perennial temperate grasslands in order to study the herbage production dynamics of subtropical, C4-dominated grasslands. Major changes were in parameters and equations accommodating for differences in photosynthetic pathway and environmental characteristics. Model outputs were compared with observed herbage accumulation rate data from a long-term grazing experiment in the Pampa grasslands. Simulated herbage growth showed similar behavior to field data across seasons, but the model generally overestimated herbage production. Moreover, the model was not able to simulate herbage production of slow-growing, tussock forming grasses. We simulated scenarios representing sequence of dry and wet years, and under various cutting frequencies. Herbage growth responded linearly to water limitation, quickly recovering after water reserves were replenished. Cutting reduced the amount of dead biomass but had little influence on herbage growth. Our results show that adapting a model across environments is not just a process of changing parameters. Although the model has previously been validated for temperate grasslands, further work is required in order to accurately represent herbage production in subtropical grasslands. Important physiological processes in one climate may not necessarily be relevant elsewhere, thus not properly represented in the original structure of the model. Furthermore, plant functional traits which are easily measured are also highly variable, leading to potential errors in the model outputs.