A Satellite Approach to Estimate Land–Atmosphere $\hbox{CO}_{2}$ Exchange for Boreal and Arctic Biomes Using MODIS and AMSR-E

Northern ecosystems are a major sink for atmospheric $\hbox{CO}_{2}$ and contain much of the world's soil organic carbon (SOC) that is potentially reactive to near-term climate change. We introduce a simple terrestrial carbon flux (TCF) model driven by satellite remote sensing inputs from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Advanced Microwave Scanning Radiometer for EOS (AMSR-E) to estimate surface ($≪ 10$-cm depth) SOC stocks, daily respiration, and net ecosystem carbon exchange (NEE). Soil temperature and moisture information from AMSR-E provide environmental constraints to soil heterotrophic respiration $(R_{h})$ , while gross primary production (GPP) information from MODIS provides estimates of the total photosynthesis and autotrophic respiration. The model results were evaluated across a North American network of boreal forest, grassland, and tundra monitoring sites using alternative carbon measures derived from tower $\hbox{CO}_{2}$ flux measurements and BIOME-BGC model simulations. Root-mean-square-error (rmse) differences between TCF model estimates and tower observations were 1.2, 0.7, and 1.2 $\hbox{g} \cdot \hbox{C} \cdot \hbox{m}^{-2} \cdot \hbox{day}^{-1}$ for GPP, ecosystem respiration $({\rm R}_{\rm tot})$ and NEE, while mean residual differences were 43% of the rmse. Similar accuracies were observed for both TCF and BIOME-BGC model simulations relative to tower results. TCF-model-derived SOC was in general agreement with soil inventory data and indicates that the dominant SOC source for $R_{h}$ has a mean residence time of less than five years, while $R_{h}$ is approximately 43% and 55% of $R_{\rm tot}$ for respective summer and annual fluxes. An error sensitivity analysis determined that meaningful flux estimates could be derived under prevailing climatic conditions at the study locations, given documented error levels in the remote sensing inputs.