Regional-scale carbon flux estimation using MODIS imagery.

摘要

The National Aeronautics and Space Agency NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) platform carried by Terra and Aqua satellites, is providing systematic measurements summarized in high quality, consistent and well-calibrated satellite images and datasets ranging from reflectance in the visible and near infrared bands to estimates of leaf area index, vegetation indices and biome productivity. The objective of this research was to relate the spectral responses and derived MODIS products of ecosystems, to biogeochemical processes and trends in their physiological variables. When different sources of data were compared, discrepancies between the MODIS variables and the corresponding ground measurements were evident. Uncertainties in the input variables of MODIS products algorithms, effects of cloud cover at the studied pixel, estimation algorithm, and local variation in land cover type are considered as the cause. A simple "continuous field" model based on a physiologically-driven spectral index using two ocean-color bands of MODIS satellite sensor showed great potential to track seasonally changing photosynthetic light use efficiency and stress-induced reduction in net primary productivity of terrestrial vegetation. The model explained 88% of the variability in Flux tower-based daily Net Primary Productivity. Also a high correlation between midday gross CO2 exchange with both daily and 8-day mean gross CO2 exchange, consistent across all the studied vegetation types, was found. Although it may not be possible to estimate 8-day mean Light Use Efficiency reliably from satellite data, Light Use Efficiency models may still be useful for estimation of midday values of gross CO2 exchange which could then be related to longer term means of CO2 exchange. In addition, the MODIS enhanced vegetation index shows a high potential for estimation of ecosystem gross primary production, using respiration values from MODIS surface temperature, providing truly per-pixel estimates.