Interannual Variations and Recurrent Seasonal Discrepancies in Land Surface Latent and Sensible Heat Fluxes from Satellite Data and a Land Surface Model

Interannual variability of land surface latent and sensible heat fluxes has been analyzed at seasonal and sub-seasonal time scales for flux estimates derived from satellite data and from a land surface model. The satellite-derived estimates were produced with a neural network operating on a combination of microwave, visible, and infrared satellite data products. Properties of these flux datasets were assessed by subjective and statistical methods, including comparisons with data from flux towers. The agreement with tower fluxes is closer for the satellite-derived fluxes than for the LSM fluxes with respect to overall temporal variability. For interannual variations of sub-seasonal fluxes, the satellite/NN and LSM fluxes have similar, moderate correlations (~0.4) with the tower fluxes. Driving factors contributing to the interannual variability and recurrent discrepancies between these flux estimates were identified. These factors include the sensitivity of satellite-derived fluxes to the satellite inputs and the responses of modeled fluxes to changes in soil moisture induced by prior precipitation.