A Four-Year Study of Data Quality in Large-Scale Smart Agriculture Networks

Abstract

Smart agriculture systems rely on continuous sensor measurements to support irrigation and crop management decisions. Over time, sensor drift, hardware aging, and environmental interference reduce data reliability. A monitoring network deployed across 620 hectares of farmland was studied over four growing seasons. The system included 1,340 temperature, humidity, and soil moisture sensors connected via LoRa gateways. Approximately 9.6% of sensors exhibited measurable calibration drift after two years of operation. A hybrid validation mechanism combining neighbor consistency checks and periodic manual sampling was introduced. After implementation, abnormal data reports decreased by 43%, and irrigation scheduling errors were reduced by nearly one third. Sustained data quality requires ongoing maintenance strategies rather than one-time system configuration.