Attributing the Variability of Gross Ecosystem Exchange (GEE) Using Environmental Data
Vineet
Yadav, The University of Michigan, vineety@umich.edu
Kim
L.
Mueller, The University of Michigan, kimlm@umich.edu
Anna
M.
Michalak, The University of Michigan, amichala@umich.edu
(Presenting)
Deborah
Huntzinger, The University of Michigan, dnhuntzi@umich.edu
(Presenting, Not an Author)
This research presents and applies a new adaptation of geostatistical regression (Mueller et al., Global Biogeochem. Cycles, [2010] and Yadav et al., Biogeoscience Discussions, [2010]) that can be used with eddy-covariance data to investigate relationships between carbon flux and environmental variables at multiple timescales. Regression methods used in previous eddy-covariance studies are limited because they (i) commonly consider only one covariate at a time (ii) do not include an objective approach for selecting variables to include in the case of multivariate regression models, (iii) do not account for the uncertainty associated with a selected set of variables, and (iv) include model assumptions, such as that of uncorrelated regression residuals, that are not consistent with available data. The GR approach improves on these methods by using the Bayes Information Criteria to select the set of variables to be used in the regression, and assessing the uncertainty associated with the selection of these variables. The method also accounts for temporal correlation in regression residuals. The GR method is applied to four sites in North America to identify the relationship between Gross Ecosystem Exchange (GEE) and eco-meteorological variables at multiple timescales. Results indicate that the dominant explanatory variables are different for each site, and that the inferred relationships vary across temporal scales. Overall, this work supports the use of the proposed approach, suggests that relationships important at one temporal resolution should not be “scaled-up or down” within biospheric models without evaluating the persistence of the relationship across temporal scales.
Presentation Type: Poster
Poster Session: Carbon Cycle Science
NASA TE Funded Awards Represented:
Michalak, Anna
Contstraining North-American Fluxes of Carbon Dioxide and Inferring their Spatiotemporal Covariances through Assimilation of Remote Sensing and Atmospheric Data in a Geostatistical Framework