Twine, T.E., and C.J. Kucharik.  Evaluating a terrestrial ecosystem model with satellite information of greenness.  Submitted to Journal of Geophysical Research-Biogeosciences, Sept. 2007.

Abstract:
The Agro-IBIS model, a version of the Integrated Biosphere Simulator (IBIS) that includes cropping systems, was evaluated with vegetation greenness information from the AVHRR (1982-2000) and MODIS (2000-2002) sensors. Leaf area index (LAI) and fraction of absorbed photosynthetically active radiation (FPAR) products were used to evaluate the model's ability to represent LAI/FPAR magnitude and the onset and offset of growing seasons for crop, forest, grass, and shrub ecosystems of the central and eastern United States. Compared with MODIS, Agro-IBIS showed a mean bias in monthly-averaged LAI of 0.41 with a mean absolute percent difference (MAPD) of 8% in crops, a mean bias of –0.92 (MAPD of –46%) in forests, a mean bias of 0.31 (MAPD of 22%) in grasses, and a mean bias of 0.11 (MAPD of 60%) in shrubs. During the growing season the bias increased in crops (to MAPD of 44%) but decreased in forests (to MAPD of -13%). Comparisons with point measurements of LAI suggest that Agro-IBIS may perform better in crop ecosystems than what was found in the MODIS comparison. Similar bias patterns in FPAR occurred in all biomes. Results from this evaluation suggest that (1) the AVHRR record may be used to test a model's simulated growing season length, but not LAI magnitude, and (2) the MODIS products are useful in testing a model's simulated growing season length and LAI/FPAR magnitude at a regional scale but more point measurements, continuing validation of MODIS products, and Agro-IBIS improvements are needed to reduce uncertainty in simulated phenology.