Satellite-Based Light Use Efficiency Model for Sugarcane Yield Estimation at Field Scale

Sugarcane is a major cash crop grown in India. Several empirical, semi-empirical and mechanistic models have been used to compute and predict sugarcane yield using satellite data. Reliable data predictions are necessary for important decision-making situations. This study attempted a semi-empirical Light Use Efficiency (LUE) model to compute the field-level sugarcane yield for a mill catchment in Uttar Pradesh, India. The phenology parameters, Start of Season and End of Season, for each field were derived by fitting the Sentinel-1 VH backscatter data using the Savitzky-Golay filter. Later these fields were grouped into four classes based on their key phenology parameters. Photosynthetically Active Radiation (PAR) was computed from INSAT- 3D data and calibrated based on the ground information. The fraction of absorbed PAR was computed using Sentinel-2 images and ground truth Leaf Area Index observations. The model was limited to using temperature and water stress factors. Water stress was computed using the Land Surface Water Index from Sentinel-2. ERA5's 2 m temperature data were used to compute the stress caused by temperature. LUE and Harvest index were used as constants. The study yielded very satisfying results. The observed and predicted data had an agreement index of 0.91 and a Root Mean Square Error of 46.6 q/ha. The results improved substantially after the calibration of the INSAT-3D insolation. This study overcomes the limitation of varying planting and harvesting dates which are usually not accounted for in LUE-based yield estimation.