PNNL

Abstract

Atmospheric CO2 flux inversions rely heavily on assumptions about errors in the prior mean flux estimate. If errors are correlated, inversions need to represent both the shape of the correlations and the magnitude of the errors. Some previous studies have used the differences in CO2 flux estimates produced by terrestrial carbon cycle models and eddy covariance measurements to constrain the flux error correlations. However, these studies used a year of fluxes averaged over at least a day. Since inversions are starting to resolve the daily cycle, we set out to examine the correlations at sub-daily time scales, as well as the correlations across years. To this end, we examine the autocorrelations in the difference between net ecosystem- atmosphere exchange measurements from 75 AmeriFlux towers and 500 m-resolution estimates of the CO2 flux at AmeriFlux tower locations from the Carnegie-Ames-Stanford Approach (CASA) terretrial carbon cycle model. We find that the daily cycle is prominent in these hourly autocorrelations and that these autocorrelations persist across years. We propose a family of functions to model these temporal correlations in atmospheric inversions, and use cross validation to determine which of the correlation functions fits autocorrelation data from towers not in the training set the best. Correlation functions with a component that attempts to model the daily cycle in the differences match correlations from other towers better than those without. Those models that reproduce the same correlation structures at one-year intervals while modulating the amplitudes of the correlations between those intervals improve the fit still further. Including a component that attempts to model the annual cycle in the model-data differences also improves the fit to independent towers, though not as much as the other functional components. In general, increasing the complexity of the correlation improves the quality of the fit.