Assessing the Efficacy of IoT-based Forest Fire Detection: a Practical Use Case

TL;DR

This work evaluates the efficacy of an IoT-based forest-fire detection system (SEA-HAZEMON) deployed on a LoRaWAN platform in Tak, Thailand, as a practical 6G-enabled use case. By integrating 2022 observations of $PM_{2.5}$, $CO$, wind data from a sensor node, and FIRMS fire hotspot data, the study adopts a data-driven approach to assess detection capabilities and sensor placement challenges. Findings show that PM2.5-based thresholds (e.g., $PM_{2.5} > 71.22 μg/m^3$ with 58% confidence within a $10$ km radius) yield limited reliability, with many fires either not triggering peaks or causing false positives due to wind and regional pollution sources. The results highlight the need for multi-sensor data fusion, sensor deployment strategies, and robust guidelines for 6G IoT fire-detection systems to be practically effective in rural and semi-urban contexts.

Abstract

The implementation of early warning mechanisms that can be used to detect forest fires in rural areas is essential to mitigate their deleterious effects, in particular by notifying local fire authorities to mount timely emergency responses. 6G-enabled Internet of Things (IoT) infrastructures are promising technological developments in that direction. However, in practice, the ability to detect forest fires in an effective way using distributed sensor nodes is challenging to achieve. In this short paper, we exemplify this challenge based on a case study that uses real data collected from the Low-Cost Internet of Things Sensor of Haze Air Quality Disasters in Communities in Thailand and Southeast Asia (SEA-HAZEMON) platform. The work is a preliminary step towards assessing the efficacy of a real-life fire detection system based on distributed sensor nodes. More generally, the objective is to develop a set of practical guidelines for the design of a 6G-enabled IoT-based fire detection mechanism.

Figures (7)

• Figure 1: Correlation matrix of pollutant levels and atmospheric parameters - Case of Moo5 sensor in the Tak province.
• Figure 2: Wind rose - Case of Moo5 sensor in the Tak province.
• Figure 3: Fire events mapped in relation to the Moo5 sensor node location.
• Figure 4: Evolution of $PM_{2.5}$ concentration level and identified fire events.
• Figure 5: Annotated time series of the $PM_{2.5}$ level during the double forest fires recorded in the Tak region on the 29th January, 2022.