HR: 0800h
AN: B51A-0052    [Abstracts]
TI: Evaluation of Multi-Sensor Semi-Arid Crop Season Parameters Based on NDVI and Rainfall
AU: * de Beurs, K M
EM: kdebeurs@vt.edu
AF: Virginia Polytechnic Institute and State University, Dept. of Geography, 115 Major Williams Hall, Blacksburg, VA 24061, United States
AU: Brown, M E
EM: molly.brown@gsfc.nasa.gov
AF: SSAI, Biospheric Sciences Branch, NASA Goddard Space Flight Cente, Greenbelt, MD 20771, United States
AB: In semi-arid regions in Africa, the amount of rainfall, its distribution throughout the season and the length of the growing season are important sources of interannual variability in food production. Staple cereal crops in West Africa such as millet and sorghum are often photo-period sensitive and thus a sowing delay in the Sahel is expected to translate into yield reduction. Metrics that estimate the beginning of the growing season in semi-arid monsoonal ecosystems have become a central part of crop models and early warning indicators of possible future reductions in yield due to growing season length. Land surface phenology measures based on satellite observed vegetation data can provide a unique source of information about the start of the growing season and improve estimation of variations in food production. Here we present a new land surface phenology model which is tuned to the semi-arid, monsoonal ecosystem of the West African Sahel. We implement this model on vegetation index data from AVHRR, SPOT-Vegetation and MODIS data, and evaluate the results along with a rainfall-based start of season estimate for the region. We focus on determining which of four different NDVI datasets are best able to capture ground observations compared to a standard rainfall-based SOS using observed emergence of crops in West Africa as a benchmark. The results reveal good agreement between the satellite-derived estimate of the start of season and the sowing dates reported. The best agreement is reached for the MODIS data with a spatial resolution of 8km and a temporal resolution of 16 days. The RMSE for all datasets varies between 12 and 26 days, with higher errors for the SPOT data and lower errors for the MODIS data. Shorter compositing periods only provide small improvements in precision.
DE: 0402 Agricultural systems
DE: 0438 Diel, seasonal, and annual cycles (4227)
DE: 0480 Remote sensing
SC: Biogeosciences [B]
MN: 2007 Fall Meeting