A comprehensive benchmarking system for evaluating global vegetation models

[摘要] We present a benchmark system for global vegetation models. This systemprovides a quantitative evaluation of multiple simulated vegetationproperties, including primary production; seasonal net ecosystem production;vegetation cover; composition and height; fire regime; and runoff. Thebenchmarks are derived from remotely sensed gridded datasets and site-basedobservations. The datasets allow comparisons of annual average conditions andseasonal and inter-annual variability, and they allow the impact of spatialand temporal biases in means and variability to be assessed separately.Specifically designed metrics quantify model performance for each process,and are compared to scores based on the temporal or spatial mean value of theobservations and a "random" model produced by bootstrap resampling of theobservations. The benchmark system is applied to three models: a simplelight-use efficiency and water-balance model (the Simple Diagnostic BiosphereModel: SDBM), the Lund-Potsdam-Jena (LPJ) and Land Processes and eXchanges(LPX) dynamic global vegetation models (DGVMs). In general, the SDBM performsbetter than either of the DGVMs. It reproduces independent measurements ofnet primary production (NPP) but underestimates the amplitude of the observedCO₂ seasonal cycle. The two DGVMs show little difference for mostbenchmarks (including the inter-annual variability in the growth rate andseasonal cycle of atmospheric CO₂), but LPX represents burnt fractiondemonstrably more accurately. Benchmarking also identified several weaknessescommon to both DGVMs. The benchmarking system provides a quantitativeapproach for evaluating how adequately processes are represented in a model,identifying errors and biases, tracking improvements in performance throughmodel development, and discriminating among models. Adoption of such a systemwould do much to improve confidence in terrestrial model predictions ofclimate change impacts and feedbacks.

[学科分类] 地球化学与岩石