Fire weather index prediction model based on IoT system for peatland forest in tropical region

The Internet of Things (IoT) has revolutionized the way to collect and analyze data, allowing for more efficient and accurate predictions in various fields. IoT systems are becoming increasingly popular in monitoring and managing natural resources due to their ability to collect real-time data from remote locations. Peatland forests, in particular, require continuous monitoring to prevent degradation and mitigate the risk of peatland fires, making IoT systems valuable for conservation and fire prevention efforts. This thesis examines the current challenges in fire weather prediction and ground water level estimation for peatland forests. One major issue is the lack of integration of ground water level measurements into the existing Fire Weather Index (FWI) system, which hampers accurate assessment of fire risk in these unique ecosystems. Additionally, the absence of machine learning techniques hinders effective prediction of ground water levels and FWI values. By proposing innovative solutions, this study aims to enhance fire risk assessment models and improve decision-making processes for peatland forest management. In this thesis, a LoRa-based IoT system was deployed in Raja Musa Forest Reserve (RMFR) in Kuala Selangor, Malaysia, to monitor peatland forests and prevent fire. Furthermore, the IoT system's feasibility was verified by comparing the correlation between its data and data from the Malaysian Meteorological Department (METMalaysia). This thesis proposes an improved tropical peatland Fire Weather Index (FWI) system by integrating Ground Water Level (GWL) measured by an IoT system into the existing FWI system. Four objectives are pursued: 1) Integrate GWL into the Drought Code (DC) calculation of the existing FWI system; 2) Integrate GWL into the Duff Moisture Code (DMC) calculation of the existing FWI system; 3) Predict GWL using Machine Learning (ML) combined with measurement of the IoT system; 4) Predict the FWI using ML combined with measurement of the IoT system. The performance of the proposed model is evaluated using IoT measurement data and actual values published by METMalaysia. The results show that the IoT system data has a high correlation of 0.8 with METMalaysia's data. The proposed model using GWL for DC formulation improves the FWI system's accuracy. The proposed DMC model that uses GWL and relative humidity achieves a correlation of 0.9071 and a Mean Square Error (MSE) of 2.5633 with actual values. The MSE between predicted and real GWL values from the two sensor nodes deployed through neural network machine learning are 11.069 and 13.256, respectively. Finally, the proposed neural network-based models using four and nine fire factors as input parameters achieve a MSE of 1.537 and 1.116, respectively, and the correlation of the nine-input model is 0.890, while the four-input model is 0.852. This thesis contributes to the growing research on applying IoT and machine learning techniques for environmental monitoring and disaster prevention. These findings have significant practical value in predicting and reducing the risk of tropical peatland fires, which cause devastating damage to the peatland forest ecosystem.

Text FK 2024 32 - Declaration Form.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (414kB) Text FK 2024 32 - Full Text.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (8MB) Text FK 2024 32.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (8MB) Official URL or Download Paper: http://ethesis.upm.edu.my/id/eprint/18997

Actions (login required)

View Item