Abstract
Free Air CO2 Enrichment (FACE) experiments provide a remarkable wealth of data to test the sensitivities of terrestrial ecosystem models (TEMs). In this study, a broad set of 11 TEMs were compared to 22 years of data from two contrasting FACE experiments in temperate forests of the south eastern US—the evergreen Duke Forest and the deciduous Oak Ridge forest. We evaluated the models' ability to reproduce observed net primary productivity (NPP), transpiration and Leaf Area index (LAI) in ambient CO2 treatments.
Encouragingly, many models simulated annual NPP and transpiration within observed uncertainty. Daily transpiration model errors were often related to errors in leaf area phenology and peak LAI. Our analysis demonstrates that the simulation of LAI often drives the simulation of transpiration and hence there is a need to adopt the most appropriate of hypothesis driven methods to simulate and predict LAI. Of the three competing hypotheses determining peak LAI—(1) optimisation to maximise carbon export, (2) increasing SLA with canopy depth and (3) the pipe model—the pipe model produced LAI closest to the observations. Modelled phenology was either prescribed or based on broader empirical calibrations to climate.
In some cases, simulation accuracy was achieved through compensating biases in component variables. For example, NPP accuracy was sometimes achieved with counter-balancing biases in nitrogen use efficiency and nitrogen uptake. Combined analysis of parallel measurements aides the identification of offsetting biases; without which over-confidence in model abilities to predict ecosystem function may emerge, potentially leading to erroneous predictions of change under future climates.
