Plant photosynthesis and productivity on Earth in the high Carbon Dioxide 21st century

Understanding how plant productivity responds to CO2 is crucial to understanding Earth System dynamics and therefore, predicting the Earth System’s response to anthropogenic forcing of atmospheric CO2. Free Air Carbon dioxide Enrichment (FACE) experiments test the CO2 response of semi-natural forest stands over the course of a decade of CO2 enrichment and this Thesis informs and develops global carbon cycle modelling using FACE data.
Meta-analysis of FACE experiments showed maintained productivity gains, and no evidence of photosynthetic acclimate to elevated CO2, over nine years of enrichment. An artefact of FACE methods is that CO2 concentrations oscillate at high frequency (1 oscillation per minute) and high amplitude (400–900 µmol mol-1) with the potential to impact carbon assimilation. Chapter three demonstrated that carbon assimilation was increased in Quercus robur and Populus x euramericana compared to steady state CO2.
Simulation of the Oak Ridge and Duke FACE experiments showed that both the Sheffield Dynamic Global Vegetation Model (SDGVM) and the Joint UK Land Environment Simulator (JULES) could reproduce Net Primary Productivity (NPP) with a reasonable degree of accuracy once Vcmax was accurately parameterised. This research highlights the necessity of rigorous model testing with observed data and shows the need to develop a strong, cross model, benchmarking system.
A global meta-analysis assessed the response of Vcmax to leaf nitrogen and phosphorus showing that phosphorus reduced the sensitivity of Vcmax to nitrogen. Global simulation with the empirical Vcmax to leaf nitrogen and phosphorus relationship led SDGVM to over-predict Gross Primary Productivity (GPP) and biomass, yet lowered terrestrial CO2 sequestration over the course of the 20th and 21st century due to higher rates of soil respiration.
Model bias and compensating factors are highlighted and correction of parameterisation error showed that more explicit process representation is necessary in SDGVM. Areas highlighted for model development were: nitrogen cycle simulation; Vcmax and Jmax; parameterisation; experimental quantification of the effect of soil water stress on forest productivity and the simulation of biomass and mortality. Accurate global datasets of biomass, NPP and leaf traits will help to uncover model bias and compensating factors and will help to develop model processes.