• Thaware Kajal Bhaskar
• Dr. Mohamed Zeeshan R

Agriculture is the backbone of Indian economy and the issue of crop losses due to poor crop management is critical to India’s effort in improving food security. Thus agriculture and crop yield monitoring is very important for an agrarian country like India. Prediction of crop yield ahead of time will help the farmers and decision makers in better management of their crop by taking appropriate measures for storage and devise efficient policy to help the farmers of India. Accurate prediction of crop yield will also help the allied industries in strategizing their plans for production and demand. India is second largest producer of the sugarcane and it is also important for achieving India’s important targets of Biofuels Policy by 2030. This current research aims at assessing the efficiency of an open source cloud based remote sensing tool, Google Earth Engine (GEE), in identification of the sugarcane crop. The research is targeted on the pixel-based approaches to crop mapping of the study area and exploring the efficiency of GEE. The ground truth data is collected from the study area and the efficiency of Landsat 8 and Sentinel 2 satellite data is assessed for crop classification. The classification is done using GEE’s inbuilt Machine Learning classifiers, Classification and Regression Tree (CART), Support Vector Machine (SVM) and Random Forest (RF). This research proposes and implements a system to predict the yield of the sugarcane from the 5 years crop yield data. The empirical model is prepared using feature extraction and feature selection techniques of data pre-processing like Principal Component Analysis (PCA), Partial Least Square (PLS), Regularisation and Stepwise selection. The research also aims to find the best growth stage of the sugarcane crop that correlates most to the yield. This is done by minimising the least square loss function through iterative convergence using empirical equation with Normalised Vegetation Indices (NVI) and pre-defined Vegetation Indices (VI) as predictors. The empirical equations obtained provided very low residual values with predicted yield very close to observed ones and able to explain 73% of the variance in the data (R2 = 0.73) for first stage i.e. 3 months after planting the crop using Partial Least Square (PLS) and 72% of the variance in the data (R2 = 0.72) for fourth stage i.e. two months before harvesting using Stepwise Regression. Of the methods used, SVM has outperformed other methods as the relative deviation in estimating the area of the sugarcane of the study area is only +9.2% with overall accuracy and F1-score as high as 78% and 0.8 for Sentinel images. However many studies has shown different classifiers to work better than SVM, it is important to remember ‘one size does not fit all’ thus other classifiers could give better results based on the geographical location. The crop yield model discussed in this research will further improve in future as more dataset is available. The crop yield map is also produced to help the farmers of the study area in managing the resources efficiently for improving the yield.

Indian Agriculture, Sugarcane, Yield Prediction Model, Machine Learning, Vegetation Indices, Remote Sensing, Decision Support System

"Prediction of Sugarcane Crop Yield using Open Source Satellite Images ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 12, page no.g635-g646, December-2024, Available :http://www.jetir.org/papers/JETIR2412672.pdf

"Prediction of Sugarcane Crop Yield using Open Source Satellite Images ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 12, page no. ppg635-g646, December-2024, Available at : http://www.jetir.org/papers/JETIR2412672.pdf