End-to-End Agricultural Land Monitoring: Integrating Sentinel-1 SAR Processing with a Cloud-Deployed Smart Farming Progressive Web Application

Abstract

Agricultural monitoring in cloud-prone tropical regions demands solutions that are simultaneously data-robust and operationally accessible to rural users with limited connectivity. This paper presents the design, implementation and evaluation of an end-to-end smart agricultural monitoring platform that integrates Sentinel-1 Synthetic Aperture Radar (SAR) data processing with a cloud-deployed Progressive Web Application (PWA) for real-time field monitoring and automated alert generation. The system comprises three coupled layers: a SAR pre-processing and machine learning pipeline implementing ESA SNAP processing, supervised land cover classification achieving an overall accuracy of 75.7% (kappa = 0.72), and YOLOv8s object detection; a FastAPI-based RESTful backend serving Cloud-Optimised GeoTIFF (COG) raster layers, detection events, and asynchronous notifications via a Celery task queue; and a React and Leaflet PWA supporting field boundary digitisation, temporal detection visualisation, and offline-capable access suited to low-connectivity rural environments. The platform was demonstrated on the Ita-Egbe watershed of Ogun State, Nigeria, delivering classification maps, anomaly alerts, and a field management interface without requiring dedicated GIS software. The COG-based raster delivery architecture reduces tile request overhead compared to conventional WMS tiling servers, while the asynchronous task queue decouples computationally intensive inference from API response latency. The integrated architecture demonstrates how cloud-independent SAR acquisition can be coupled with modern web technology to deliver actionable decision-support tools to smallholder farmers and agricultural extension officers across sub-Saharan Africa.