Estimation of ground water level (GWL) for tropical peatland forest using machine learning

• Main Authors: Li, Lu, Sali, Aduwati, Liew, Jiun Terng, Saleh, Nur Luqman, Syed Ahmad, Sharifah Mumtazah, Mohd Ali, Azizi, Nuruddin, Ahmad Ainuddin, Amir Aziz, Nurizana, Sitanggang, Imas Sukaesih, Syaufina, Lailan, Nurhayati, Ati Dwi, Nishino, Hisanori, Asai, Nobuyuki
• Format: Article
• Language: English
• Published: Institute of Electrical and Electronics Engineers 2022
• Online Access: http://psasir.upm.edu.my/id/eprint/111527/1/Estimation_of_Ground_Water_Level_GWL_for_Tropical_Peatland_Forest_Using_Machine_Learning.pdf; http://psasir.upm.edu.my/id/eprint/111527/; https://ieeexplore.ieee.org/document/9967973/

The tropical area has a large area of peatland, which is an important ecosystem that is regarded as home by millions of people, plants and animals. However, the dried-up and degraded peatland becomes extremely easy to burn, and in case of fire, it will further release transboundary haze. In order to protect the peatland, an improved tropical peatland fire weather index (FWI) system is proposed by combining the ground water level (GWL) with the drought code (DC). In this paper, LoRa based IoT system for peatland management and detection was deployed in Raja Musa Forest Reserve (RMFR) in Kuala Selangor, Malaysia. Then, feasibility of data collection by the IoT system was verified by comparing the correlation between the data obtained by the IoT system and the data from Malaysian Meteorological Department (METMalaysia). An improved model was proposed to apply the ground water level (GWL) for Fire Weather Index (FWI) formulation in Fire Danger Rating System (FDRS). Specifically, Drought Code (DC) is formulated using GWL, instead of temperature and rain in the existing model. From the GWL aggregated from the IoT system, the parameter is predicted using machine learning based on a neural network. The results show that the data monitored by the IoT system has a high correlation of 0.8 with the data released by METMalaysia, and the Mean Squared Error (MSE) between the predicted and real values of the ground water level of the two sensor nodes deployed through neural network machine learning are 0.43 and 12.7 respectively. This finding reveals the importance and feasibility of the ground water level used in the prediction of the tropical peatland fire weather index system, which can be used to the maximum extent to help predict and reduce the fire risk of tropical peatland.