"International Journal of Geography, Geology and Environment"

Author(s): Sachin Chinchorkar

Abstract: Advanced monitoring techniques are needed to assess vulnerabilities and to preserve global food security in the face of climate change challenges to global agriculture. Combining satellite data with machine learning to detect and analyze agricultural stress due to climate variability, this research combines satellite data with machine learning. Four machine learning models, namely Random Forest (RF), Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Long Short Term Memory (LSTM), were applied on high resolution optical and SAR imagery with climate parameters. Prediction of crop health, as well as drought susceptibility, is attempted and the results indicate that CNN achieved the highest accuracy of 92.4%, followed by LSTM at 89.7%, RF at 85.3% and SVM at 82.5%. There is comparison with existing research that shows that deep learning models can perform better than traditional models in capturing the spatial and temporal dependencies. Moreover, the accuracy of vulnerability predictions differed between climate zones, with predictions between arid zones 10 percent less accurate because of the lack of data available. Therefore, these findings indicate the development of AI based solutions for agricultural monitoring and climate adaptation at a real time. Future research can further improve prediction accuracy by including IoT sensor data and the set of climate variables. This study offers a robust approach that will inform policy makers and agricultural stakeholders how to adapt to the climate related risk and farmers how they can continue their operations sustainably.

DOI: 10.22271/27067483.2025.v7.i3a.345

Pages: 01-09 | Views: 127 | Downloads: 65

Download Full Article: Click Here