Satellite based remote sensing for estimating crop yield: examining the use of various functional forms and vegetation indices

Thumbnail Image
Miralda, Shania
Journal Title
Journal ISSN
Volume Title
This research includes two studies on estimating crop yield using satellite-based vegetation indices. For the first study, the objective is to compare eight different functional forms for estimating U.S crop yield using satellite based NDVI. For the second study, the objective is to examine 10 satellite-based vegetation indices for estimating U.S crop yield. For both studies, corn, soybeans, spring wheat, and winter wheat data for crop yield (bushel/acre) were obtained from the USDA NASS from 2008 to 2019 for a total of 12 years covering all 48 states in the United States except Hawaii and Alaska (though different states are included, based on where the crops are grown). Data for the vegetation indices were obtained from the MODIS satellite using 250m resolution level and selecting for maximum Vegetation Index values. The methodology used regression with crop yield as the dependent variable. The main independent variable is the selected vegetation index (e.g NDVI, GOSAVI, etc). A time trend variable is also included, and dummy variables for U.S States. Results for the first study indicated that relationship between NDVI and crop yield was mostly nonlinear, and piecewise regression was generally found to be the most suitable functional form. Results for the second study showed that for all 10 indices analyzed, that RDVI, GOSAVI, and GSAVI provided better estimates of crop yield than the commonly used NDVI. These results should be useful in providing a better understanding of various functional forms and various satellite based vegetation indices for improving crop yield estimation.
Piecewise Regression, Satellite Data, Remote Sensing, Crop Yield Estimation, NDVI, Normalized Difference Vegetation Index, RDVI, Renormalized Difference Vegetation Index, GOSAVI, Green Optimized Soil Adjusted Vegetation Index, GSAVI, Polynomial Regression, GAM