Towards a more objective evaluation of modelled land-carbon trends using atmospheric CO2 and satellite-based vegetation activity observations

[摘要] Terrestrial ecosystem models used for Earth system modelling show asignificant divergence in future patterns of ecosystem processes, inparticular the net land–atmosphere carbon exchanges, despite a seeminglycommon behaviour for the contemporary period. An in-depth evaluation ofthese models is hence of high importance to better understand the reasonsfor this disagreement.

Here, we develop an extension for existing benchmarking systems by making useof the complementary information contained in the observational records ofatmospheric CO₂ and remotely sensed vegetation activity to provide anovel set of diagnostics of ecosystem responses to climate variability in thelast 30 yr at different temporal and spatial scales. The selection ofobservational characteristics (traits) specifically considers the robustnessof information given that the uncertainty of both data and evaluationmethodology is largely unknown or difficult to quantify.

Based on these considerations, we introduce a baseline benchmark – a minimumtest that any model has to pass – to provide a more objective, quantitativeevaluation framework. The benchmarking strategy can be used for any landsurface model, either driven by observed meteorology or coupled to a climatemodel.

We apply this framework to evaluate the offline version of the MPI EarthSystem Model's land surface scheme JSBACH. We demonstrate that thecomplementary use of atmospheric CO₂ and satellite-based vegetationactivity data allows pinpointing of specific model deficiencies that would notbe possible by the sole use of atmospheric CO₂ observations.

[学科分类] 地球化学与岩石