Spatial structure in North American terrestrial biological carbon fluxes and flux model errors evaluated with a simple land surface model

Of the roughly 7 pG of carbon released annually to Earth’s atmosphere within anthropogenic carbon dioxide (CO2), terrestrial ecosystems remove roughly 25% via gross primary production in excess of ecosystem respiration. Ecosystem-atmosphere exchange (net ecosystem exchange, NEE) of carbon dioxide is well-constrained at the global scale and, at field observation locations, at scales on the order of 1 km2, but is poorly constrained at spatial scales in between. This makes it difficult to assess the ecological and atmospheric processes that interact to drive NEE, which in turn results in considerable uncertainty around future ecosystem CO2 uptake, and therefore future concentrations of atmospheric CO2, a primary driver of global climate change.;This dissertation seeks to constrain estimates of North American NEE by analyzing data from nearly 100 eddy covariance NEE measurement sites spanning the continent in conjunction with satellite observations of ecosystem behavior and a simple land surface model (vegetation photosynthesis and respiration model, VPRM). VPRM is optimized to the eddy covariance observations.;A spatial covariance function for VPRM NEE error is then estimated. This results in a characteristic length scale of roughly 400 km for VPRM NEE error spatial covariance and defines an error spatial covariance matrix for VPRM in North America. This new piece of information empirically constrains atmospheric inversions of CO2 concentration measurements, thereby reducing their uncertainty.;A statistical regression model is then fit to annually integrated VPRM NEE error spread as a function of VPRM NEE magnitude, air temperature, and precipitation. The regression model is cross-validated at 27 eddy covariance sites not used for model fitting and shown to provide a reasonable fit at these sites. This provides a method to estimate errors for annually integrated VPRM NEE, allowing consideration of VPRM NEE interannual variability in the context of VPRM errors.