rsofun v5.1: a model-data integration framework for simulating ecosystem processes

Abstract

Mechanistic vegetation models serve to estimate terrestrial carbon fluxes and climate impacts on ecosystems across diverse biotic and abiotic conditions. Systematically informing them with data is key for enhancing their predictive accuracy and estimating uncertainty. Here, we demonstrate and evaluate the Simulating Optimal FUNctioning (rsofun) R package, which provides a computationally efficient implementation of the P-model for site-scale simulations of ecosystem photosynthesis and the acclimation of photosynthetic traits, complemented with functionalities for Bayesian model-data integration and the estimation of model parameters and prediction uncertainty. We estimated model parameters simultaneously from observed time series of ecosystem gross primary productivity (GPP), and from globally distributed data on leaf carbon-13 isotopic discrimination (Δ13C) and the ratio of the maximum biochemical rates of carboxylation to electron transport (). The multi-target calibration yielded unbiased predictions for all variables simultaneously and produced similar distributions of prediction–observation residuals for both calibration and out-of-sample test data, indicating that the model generalises robustly across diverse environments. We found that a step-wise approach to successive model integration and calibration yielded best results, and that correlations among parameters related to the representation of water stress effects underpinned non-robust parameter estimations. This likely indicates a dominant source of model structural uncertainty related to the representation of the response of photosynthesis to dry conditions in soil and air.