Analysis of NDVI Data for Crop Identification and Yield Estimation

Abstract

Crop yield estimation is of great importance to food security. Normalized Difference Vegetation Index (NDVI), as an effective crop monitoring tool, is extensively used in crop yield estimation. However, there are few studies focusing on the aspect of mixed crops grown together. In this study, a correlation-based approach for crop yield estimation is applied to three small counties (Jianshui, Luliang, and Qiubei) in the Nanpan River basin, Yunnan Province of China, and three main crops (paddy rice, winter wheat, and corn) in these areas are selected. Based on the correlation analysis between MODIS-NDVI data and crop yield, the crop planting areas as well as the best periods for a reliable estimation are identified. The best time is found approximately coinciding with the periods of heading, flowering, and filling of the crops. By Akaike's information criterion, the most fit regression models with extracted NDVI in the corresponding crop planting areas are determined. They work reasonably well in small regions, especially in the areas where crop types are unknown exactly. Further improvements to the regression models are possible by incorporating other physical factors such as soil types and geographical information.