High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

AgriFieldNet: A Multi-Source Remote Sensing and Machine Learning Framework for Soil Moisture Estimation and Irrigation Advisory

Author(s):

Prathamesh Nibandhe , PES Modern College of Engineering, Pune; Yash Dhore, PES Modern College of Engineering, Pune; Aashi Jaitly, PES Modern College of Engineering, Pune; Jayashri Phatangare, PES Modern College of Engineering, Pune; Shripad Bhide, PES Modern College of Engineering, Pune

Keywords:

Soil Moisture, Sentinel-1, Sentinel-2, Random Forest, Google Earth Engine, ERA5-Land, Sensor Fusion, Irrigation Advisory

Abstract

Agricultural irrigation decisions depend strongly on soil water availability, yet conventional field monitoring is spatially sparse and difficult to scale. This paper presents AgriFieldNet, a web-based framework that estimates plot-level soil moisture from multi-source remote sensing and environmental data and converts the estimate into a water deficit class and an irrigation advisory. The system fuses Sentinel-1 radar backscatter, Sentinel-2 optical bands and vegetation indices, ERA5-Land meteorological variables, and MODIS leaf area index in a single Random Forest regression model with eighteen features. A historical pipeline operates on a prepared Berambadi dataset, while a live pipeline retrieves the latest suitable observations through Google Earth Engine. Evaluated with a plot-wise train and test split, the radar-only model gave a negative coefficient of determination, the radar plus optical model reached 0.34, and the full fusion model reached 0.53 with a mean absolute error of 6.78 and a root mean square error of 8.45. Humidity was the most influential feature. A successful live run confirmed end-to-end operation. The results are prototype-scale and require field validation before operational use.

Other Details

Paper ID: IJSRDV14I90004
Published in: Volume : 14, Issue : 9
Publication Date: 01/12/2026
Page(s): 1-5

Article Preview

Download Article