An Ensemble of Global High-Resolution Products of Energy Fluxes over Land
Friday, 19 December 2014: 4:15 PM
We present an ensemble of global high-resolution energy flux products over land derived from upscaling FLUXNET observations with eleven machine learning methods using an array of remote sensing data. The products cover latent heat, sensible heat, ground heat fluxes, as well as net radiation over the period 2001-2012 at 8 daily temporal and 0.0833 degree spatial resolution. To account for the typical energy balance (EB) under-closure observed at flux sites, five different EB correction techniques were used that correspond to different hypothesis on the causes of the EB closure deficit. Hence, in total six different variants of sensible and latent heat flux products are available that allow quantifying the product uncertainty related to the assumptions on the EB closure. We evaluate our products using cross-validation at site level and against various independent observation-based data streams including latent heat fluxes derived from runoff, various precipitation products, and net radiation derived from Earth observations. The evaluation exercise also allows making inferences on the EB closure problem at FLUXNET sites, which remains a major concern for the community. Our products are a valuable source for a variety of applications including the evaluation or calibration of global land surface models and the assessment of the biophysical properties of terrestrial ecosystems.