Springer International Publishing

Sustainable farming and fire risk management using IoT and MQTT technologies for enhancing food security

The variation in climate is posing a growing threat of fire to agricultural land while jeopardizing food security and the sustainability of farming practices. Sustainable agriculture, which uses advanced technologies to optimize production, is particularly vulnerable to these risks and hence affects the concept of cognitive and smart farming. Early detection of fires in farming practices is insufficient, requiring IoT solutions for real-time monitoring and rapid response to protect crops and ensure food security. This paper proposes an innovative Internet of Things (IoT)-based system for fire detection, integrating flame and smoke sensors and a Raspberry Pi 3 B+ embedded device. The system architecture is based on three layers: (1) a local layer, (2) a Message Queuing Telemetry Transport (MQTT) broker, and (3) a user interface. The local layer is utilized for data collection and processing. The MQTT broker has HiveMQ, which is used to ensure efficient communication via the publish/subscribe protocol using Quality of Service (QoS) 0. The user interface is utilized for real-time access to critical information. The performance evaluation demonstrated the effectiveness of this system in monitoring environmental parameters with reliable data published via dedicated topics. The central processing unit (CPU) performance and memory usage have been optimized to ensure low energy consumption. The experimental results show that the system publishes up to 5400 messages on six MQTT topics via HiveMQ in 30 min, at a rate of one message every two seconds. The data revealed variations in flames (41.9% to 43.1%) and smoke (21.9% to 27.0%), with system parameters kept constant. The frequency of one message every 20 s reduces the publishing load to 180 messages over 10 min. Contrary to the empirical threshold approaches, detection thresholds are optimally determined using a Receiver Operating Characteristic (ROC) analysis, ensuring a robust tradeoff between sensitivity and false alarm rate. The experimental results show that the optimal thresholds are set at 50% for smoke and 60% for flame. The statistical evaluation of the system highlights high performance, with an accuracy of 0.929, a precision of 0.864, a recall of 1.000, an F1 score of 0.927 and a specificity of 0.870. In addition, the values of the area under the curve (AUC), reaching 0.947 for smoke and 0.967 for flame, confirm an excellent ability to discriminate between fire and non-fire states. These results demonstrate the robustness, reliability and relevance of the proposed system for smart agriculture applications. The proposed system offers farmers practical tools for anticipating and responding quickly to fire risk situations, thereby contributing to food safety and sustainability in the agricultural sector.