A cloud platform for the analysis of iot data and satellite indices in agriculture

Abstract

Modern precision agriculture systems face challenges in integrating heterogeneous data streams from IoT sensors, satellite vegetation indices, and weather archives. Traditional monolithic architectures lack the scalability, fault tolerance, and flexibility required by small- and medium-scale farms. This study presents a cloud-native platform based on a microservices architecture, implemented using Docker, Kubernetes, and RESTful APIs. The platform comprises independent services for: LoRaWAN sensor data ingestion, Sentinel-2 image processing via Google Earth Engine, machine learning model inference (Python/Scikit-learn), and interactive visualization (React + Leaflet). Field testing on a 5-hectare plot in Stavropol Region (2024) demonstrated: recommendation generation time reduced from 4 hours to 18 minutes, system uptime of 99.7%, and monthly deployment cost of approximately 12,000 RUB (Yandex Cloud). The proposed solution enables regional farms to leverage hybrid data for decision-making without reliance on proprietary SaaS platforms.