The development of equation of sugarcane yield prediction using vegetation index from Sentinel-2 satellite imagery: A case study in Chaiwan District, Udon Thani Province

Abstract

Predictions of sugarcane production can estimate the amount of raw product prior to it entering the factory and the final
amount of cane sugar produced. Thus, the prediction can help the efficient management of cane sugar manufacturing.
The objective of this study is to create a simple linear regression equation to predict sugarcane production by using
vegetation indices from Sentinel-2 satellite images in Chai Wan district, Udon Thani province and the data of sugarcane
production from sample fields. The vegetation indices included Ratio Vegetation Index (RVI), Normalized Difference
Vegetation Index (NDVI), Normalized Ratio Vegetation Index (NRVI), Ashburn Vegetation Index (AVI), Soil Adjusted
Vegetation Index (SAVI), Transformation Vegetation Index (TVI) and Infrared Percentage Vegetation Index (IPVI).
A total of 100 sample fields (20x20 meter each) were used in this study: 70 of them were used to create the equation
and 30 of them were used to check the accuracy. The results revealed that the linear regression equation from NDVI,
NRVI, SAVI, TVI, and IPVI has the highest coefficient of determination and equals to 0.88. The linear regression
equation from NDVI, NRVI, and SAVI had the mean absolute percentage error equal to 1.82. The equations can further
be utilized to determine sugarcane production and manage cane sugar manufacturing in the future.