Submitted to: International Society for Optical Engineering
Publication Type: Abstract Only
Publication Acceptance Date: 5/15/2007
Publication Date: 9/17/2007
Citation: Doraiswamy, P.C., Akhmedov, B., Stern, A.J. 2007. Potential use of MODIS imagery for operational crop yield assessment [abstract]. International Society for Optical Engineering. 2007 CDROM. Interpretive Summary:
Technical Abstract: Monitoring crop condition and yields at regional scales remains a challenge. Ground-based sampling for assessment of crop yields at regional and national scales require enormous resources. Crop yield simulation models have shown great success in predicting crop yields at field and small scales; however, there is still a gap in using field-scale models for predicting yields at regional scales. Imagery from the MODIS sensor onboard the Terra satellite offers an excellent opportunity for daily coverage at 250 m resolution. Integrating MODIS derived parameters with crop yield simulation models were successful for small-scale. However, this method may not be practical for operational applications and at large scales. Timely and accurate assessment of crop yields is important for USDA’s operational program. This study focused on the major corn and soybean production states of Iowa and Illinois in the U.S. Corn Belt. The study objectives were; a) develop a decision tree classification algorithm to separate the crops during the crop season and b) evaluate a crop yield algorithm based on NDVI and surface temperature data from MODIS Terra. Classification accuracy for corn and soybean crop were found to be within 80-85 % of the Landsat-based classification. The crop yields for state and county levels were assessed for four years (2003-06). The state yield estimates were closely correlated with the USDA’s assessment. County level estimates were not as well correlated, however were within 20% of the USDA estimates. The remote sensing method is a bottom-up approach and the USDA’s estimate for county level is a top-down approach and may attribute to the differences in yield estimates.