Up-scaling carbon and water fluxes over Alaska by multi-site eddy fluxes and satellite remote sensing data

The regional carbon and water fluxes was estimated by using observed fluxes by the eddy covariance method and a support vector machine (SVM) based regression model with satellite remote sensing over Alaska. According to the site scale evaluation, information about the landcover was a most important variable in the prediction of GPP and NEE compared with leaf area index (LAI), land surface temperature (LST), and solar radiation. Our predicted fluxes were consistent with the field observations at the site scale and carbon exchange by top-down approach, CarbonTracker, at the regional scale. On the other hand, predicted fluxes during winter periods had considerable uncertainties due to data quality and availability of both field and satellite observations. The predicted GPP, RE, and NEE for entire Alaska were 281, 203, and 78 g C m-2 y-1. The interannual variations were consistent with an anomaly in air temperature between June and August, and pacific decadal oscillation (PDO) index for August; Alaska acted larger carbon sink in warmer years. Decadal trends in GPP and NEE were strongly influenced fire disturbance, indicating that it was important to consider the disturbance history in evaluating long-term trends in satellite VIs and carbon fluxes.