Addressing Challenges in Simulating Inter–Annual Variability of Gross Primary Production

A long‐standing challenge in studying the global carbon cycle has been understanding the factors controlling inter–annual variation (IAV) of carbon fluxes, and improving their representations in existing biogeochemical models. Here, we compared an optimality‐based model and a semi‐empirical light use efficiency model to understand how current models can be improved to simulate IAV of gross primary production (GPP). Both models simulated hourly GPP and were parameterized for (a) each site–year, (b) each site with an additional constraint on IAV ( C o s t IAV $Cos{t}^{\mathit{IAV}}$ ), (c) each site, (d) each plant–functional type, and (e) globally. This was followed by forward runs using calibrated parameters, and model evaluations using Nash–Sutcliffe efficiency (NSE) as a model‐fitness measure at different temporal scales across 198 eddy‐covariance sites representing diverse climate–vegetation types. Both models simulated hourly GPP better (median normalized NSE: 0.83 and 0.85) than annual GPP (median normalized NSE: 0.54 and 0.63) for most sites. Specifically, the optimality‐based model substantially improved from NSE of −1.39 to 0.92 when drought stress was explicitly included. Most of the variability in model performances was due to model types and parameterization strategies. The semi‐empirical model produced statistically better hourly simulations than the optimality‐based model, and site–year parameterization yielded better annual model performance. Annual model performance did not improve even when parameterized using C o s t IAV $Cos{t}^{\mathit{IAV}}$ . Furthermore, both models underestimated the peaks of diurnal GPP, suggesting that improving predictions of peaks could produce better annual model performance. Our findings reveal current modeling deficiencies in representing IAV of carbon fluxes and guide improvements in further model development.

Plain Language Summary Terrestrial vegetation assimilates and releases carbon dioxide through photosynthesis and respiration, respectively, and their net magnitude determines if vegetation can be a sink or source of carbon dioxide. We are interested in understanding what controls the inter–annual variability (IAV) of gross primary production (GPP) which represents photosynthesis, and how their representations can be improved in models simulating GPP. Here, we considered an optimality‐based model that can be applied equally well globally, and a data‐driven semi‐empirical model. We found both models better simulated diurnal and seasonal cycles than the IAV of GPP. Such differences probably stem from model parameters, as critical ecosystem functions they represent may not be well‐constrained or model structures may lack critical representations via inaccurate simulation of peak diurnal GPP and drought stress. The IAV of GPP was comparatively better simulated if model parameters were fine‐tuned with data from specific years. Another challenge is that IAV of GPP can also be observed due to disturbances, such as forest fire, and human management besides natural causes, which were also not represented in models. Our results suggest that learning the variability of model parameters over the years can be key to better simulation of the IAV of GPP.

Key Points We investigated the limitations of biogeochemical models in simulating inter–annual variability (IAV) of gross primary production (GPP) Capturing IAV of model parameters and diurnal GPP peaks can be key to understanding IAV of GPP Variability in model performance is jointly influenced by model types, parameterization strategies, and site characteristics