Evaluating the agreement between measurements and models of net ecosystem exchange at different times and timescales using wavelet coherence: an example using data from the North American Carbon Program Site-Level Interim Synthesis

摘要

Earth system processes exhibit complex patterns across time, as do the modelsthat seek to replicate these processes. Model output may or may not besignificantly related to observations at different times and on differentfrequencies. Conventional model diagnostics provide an aggregate view ofmodel–data agreement, but usually do not identify the time and frequencypatterns of model–data disagreement, leaving unclear the steps required toimprove model response to environmental drivers that vary on characteristicfrequencies. Wavelet coherence can quantify the times and timescales atwhich two time series, for example time series of models and measurements,are significantly different. We applied wavelet coherence to interpret thepredictions of 20 ecosystem models from the North American Carbon Program(NACP) Site-Level Interim Synthesis when confronted with eddy-covariance-measurednet ecosystem exchange (NEE) from 10 ecosystems withmultiple years of available data. Models were grouped into classes withsimilar approaches for incorporating phenology, the calculation of NEE, theinclusion of foliar nitrogen (N), and the use of model–data fusion. Modelswith prescribed, rather than prognostic, phenology often fit NEE observationsbetter on annual to interannual timescales in grassland, wetland andagricultural ecosystems. Models that calculated NEE as net primaryproductivity (NPP) minus heterotrophic respiration (HR) rather than grossecosystem productivity (GPP) minus ecosystem respiration (ER) fit better onannual timescales in grassland and wetland ecosystems, but models thatcalculated NEE as GPP minus ER were superior on monthly to seasonal timescales intwo coniferous forests. Models that incorporated foliar nitrogen (N) datawere successful at capturing NEE variability on interannual (multiple year)timescales at Howland Forest, Maine. The model that employed a model–datafusion approach often, but not always, resulted in improved fit to data,suggesting that improving model parameterization is important but not theonly step for improving model performance. Combined with previous findings,our results suggest that the mechanisms driving daily and annual NEEvariability tend to be correctly simulated, but the magnitude of these fluxesis often erroneous, suggesting that model parameterization must be improved.Few NACP models correctly predicted fluxes on seasonal and interannual timescales where spectral energy in NEE observations tends to be low, but wherephenological events, multi-year oscillations in climatological drivers, andecosystem succession are known to be important for determining ecosystemfunction. Mechanistic improvements to models must be made to replicateobserved NEE variability on seasonal and interannual timescales.