A footprint-aware approach for model-data benchmarking across AmeriFlux sites

Abstract:

Large datasets of greenhouse gas and energy surface-atmosphere fluxes (e.g., FLUXNET2015, AmeriFlux BASE) have been widely used to force, parameterize, calibrate, and benchmark large scale models (e.g. land surface models, remote sensing-based models and data-drive upscaled products) applied on regularized grids. This study addresses one of the major challenges facing such integration of flux measurements and gridded products, that is, to what spatial extent do flux measurements taken at individual tower locations reflect model grid cells? We evaluate flux footprints – the source areas that contribute to measured fluxes – and the representativeness of these footprints for target areas (e.g., areas within a 1-km, 3-km radius around flux towers) often used in flux data synthesis and modeling studies. Utilizing the long-term and fine-resolution Landsat collections and multiple remote sensing-based models, we generated the spatiotemporal predictions of footprint-aware CO₂ and H₂O fluxes around the flux towers at selected AmeriFlux sites. We then evaluated the footprint representativeness by comparing the fluxes as seen by the footprints to that as modeled in the targeted cells. Our results showed that many tower sites had flux footprints only representative of limited areas surrounding the tower, suggesting high resolution spatial forcing data is needed. Moreover, the extent to which the flux footprints represented is highly site-specific and ecosystem-specific. Substantial bias was found especially at sites that had relatively limited fetch or temporally dynamic footprints, and those that were located within a relatively heterogeneous landscape, such as those at croplands and wetlands.