Using machine learning algorithms to predict groundwater levels in Indonesian tropical peatlands

Area of Science:

• Environmental Science
• Hydrology
• Ecology

Background:

• Tropical peatlands are critical global carbon reservoirs and sinks, with Indonesia holding the largest share.
• Human activities like logging and canalization lower peatland groundwater levels, increasing wildfire risk, especially during dry seasons.
• Understanding spatiotemporal groundwater level changes is vital for peatland management and carbon cycle integrity.

Purpose of the Study:

• To model and analyze groundwater level dynamics in Indonesian tropical peatlands.
• To compare the performance of machine learning algorithms against traditional statistical models for groundwater level prediction.
• To identify key environmental drivers influencing groundwater levels in peatland ecosystems.

Main Methods:

• Utilized multilinear regression, random forest, and extreme gradient boosting models to simulate groundwater levels.
• Conducted spatial modeling of groundwater levels within a Central Kalimantan peat dome (2010-2012).
• Assessed model performance using R-squared and Root Mean Square Error (RMSE) metrics.

Main Results:

• All models demonstrated strong performance, with extreme gradient boosting achieving the highest accuracy (R² = 0.998, RMSE = 0.048 m).
• Extreme gradient boosting significantly outperformed random forest and multilinear regression in predicting groundwater levels near drainage canals.
• Elevation and precipitation were identified as the most influential factors affecting groundwater levels across space and time.

Conclusions:

• Machine learning, particularly extreme gradient boosting, offers a highly effective approach for modeling tropical peatland groundwater levels.
• Accurate groundwater level prediction near drainage canals is essential for mitigating wildfire ignition risks in these vital ecosystems.
• Spatio-temporal analysis of groundwater dynamics, considering factors like elevation and precipitation, is key to sustainable peatland management and carbon storage.