Prediction of corn yield in New York state by harnessing satellite remote sensing and deep learning
No Access Until
The forecast of crop yield is of great significance to improve agricultural production efficiency. It not only affects the local economic development, but also is closely related to human life. With the significant growth of open-access high-resolution real-time remote sensing observations from satellite missions and the broad applications of machine learning techniques, the combination of the two technologies offers a promising solution for crop yield estimation. Based on the land surface temperature data and surface reflectance data obtained by Moderate-Resolution Imaging Spectroradiometer (MODIS), this study aims to estimate the corn yield in New York State at the county-level by using Convolutional Neural Network (CNN) integrated with a Gaussian Process (GP) model that can improve the accuracy of spatio-temporal structure modeling. My results show that the CNN-GP integrated model can improve the prediction accuracy by 18% relative to CNN alone, indicating the potential of harnessing satellite remote sensing and deep learning to inform decision making at both field and regional scales.