HR: 17:30h
AN: B22E-07    [PDF]
TI: Validation of the MODIS GPP algorithm (MOD17A2) using eddy flux tower data
AU: * Heinsch, F A
EM: faithann@ntsg.umt.edu
AF: The University of Montana, NTSG/College of Forestry and Conservation, 32 Campus Dr, Missoula, MT 59812 United States
AU: Kimball, J S
EM: johnk@ntsg.umt.edu
AF: The University of Montana, Flathead Lake Biological Station, 311 BioStation Lane, Polson, MT 59860 United States
AU: Zhao, M
EM: zhao@ntsg.umt.edu
AF: The University of Montana, NTSG/College of Forestry and Conservation, 32 Campus Dr, Missoula, MT 59812 United States
AU: Running, S W
EM: swr@ntsg.umt.edu
AF: The University of Montana, NTSG/College of Forestry and Conservation, 32 Campus Dr, Missoula, MT 59812 United States
AB: We compare satellite-based regional calculations of gross (GPP) primary production from MODIS (MOD17A2) to tower eddy CO$_{2}$ flux-based estimates across a diverse range of landcover types and climate regimes represented within the Ameriflux network. Recently revised input data, including meteorology from the NASA Data Assimilation Office (DAO; GEOS-4.02) as well as leaf area index (LAI) and fraction of PAR (fPAR) from the MOD15 algorithm (Collection 4), have resulted in improvements of GPP estimates at these sites when compared to previous productivity estimates. Three possible sources of error within the MODIS GPP algorithm are explored. We evaluate the sensitivity of MODIS outputs to input meteorology by using both DAO and site-based daily weather information to calculate the MODIS GPP estimates. Comparing these results provides an independent assessment of the accuracy of the MOD17 algorithm phenology and the ability of the DAO data to capture local meteorology is also tested. The MOD15 LAI/fPAR outputs are also compared with site data to determine additional sources of error. Finally, vegetation parameters within the MOD17 GPP algorithm are examined to determine the effects of these variables on GPP results. Our analysis indicate that MODIS and tower based estimates compare favorably in temperate regions, but that results vary for other regions such as boreal and Arctic regions. The DAO meteorology appears to be responsible for much of the difference between tower-based and satellite-based estimates of GPP, but additional sources of error arise from surface heterogeneity and cloud cover.
DE: 1615 Biogeochemical processes (4805)
DE: 1640 Remote sensing
DE: 1694 Instruments and techniques
SC: Biogeosciences [B]
MN: 2003 Fall Meeting