Improved Techniques for Crop Classification using MODIS Imagery

Abstract

Brazil has become a major player in world soybean markets, second only to the U.S. Brazil Crop area is about 10 million hectares and is now rapidly expanding into the Brazilian savannah (Cerrado) and the Amazonian region where forested area is being converted to cropland. There is a need for accurate updated information on the newly expanded agricultural areas in Brazil and the current total production. The objective of this research was to develop an operational method for assessing soybean crop area that would facilitate developing remote sensing based algorithms for assessing crop yields in major producing areas. The moderate resolution imaging spectroradiometer (MODIS) onboard the Terra satellite offers a good potential for assessing crop area as well as provide opportunity to retrieve crop condition parameters that can be used to assess crop yields. A three-year MODIS data set was acquired for the study and this research describes the methods used for processing the 8-day composite reflectance data from bands 1 and 2 and its use in developing the classification of soybean crop area in four major soybean producing areas in Brazil. The results suggest methods that can be used for operational application of MODIS 250m data for classification as well as potential use in crop yield assessment.