Regional Yield Estimation for Sugarcane Using MODIS and Weather Data: A Case Study in Florida and Louisiana, United States of America

Abstract

Sugarcane is an important sugar source in America and is mainly planted in the states of Florida and Louisiana. The purpose of this study was to predict the sugarcane yield in these two states from 2008 to 2016. Three statistical sugarcane yield models (i.e., the CNDVI, K–M, and SiPAR models) were applied to predict yield in America, using remote sensing and weather data. To verify the robustness of models, model parameters obtained in places (i.e., Reunion Island and Southwestern China) far away from America were used. The results showed that the SiPAR model outperformed the CDNVI and K–M models for yield prediction. Solar radiation was an important constraint factor to ensure the statistical model’s robustness under different conditions. The CNDVI model had the lowest robustness because of the absence of solar radiation, although it could reflect the yield trend to some extent. The K–M model failed to predict the low sugarcane yield, owing to the lack of consideration of temperature and soil water stresses. Florida had a low sugarcane yield in the west and southwest; however, Louisiana had high sugarcane yield in the same directions. This study demonstrated the robustness of the SiPAR model and investigated the sugarcane yield status in Florida and Louisiana. It can be a reference for similar studies in the future.