Predicting Groundwater Levels in Indonesian Peatlands Using Machine Learning

Abstract

Recently, the large amount of carbon dioxide released by wildfires in peatland forests has become a problem. To suppress the wildfire, it is important to control the groundwater level in the peatland within an appropriate range. If the groundwater level can be predicted accurately in advance, the groundwater level can be controlled effectively, and hence wildfires are prevented. In this study, we developed machine learning models to predict the groundwater level based on the measured data including groundwater level, canal water level, and meteorological observation data. In creating a model, we compared several machine learning models, such as LSTM and XGBoost, and selected the best model. As a result, it was confirmed that the model could predict the groundwater level with high accuracy.