Predicting and Managing Waterflooding-Induced Subsidence: Insights from Inglewood Oilfield with Applications to Indonesian Oilfields

Summary Waterflooding helps sustain reservoir pressure but may induce subtle land deformation that threatens infrastructure and public confidence. This study applies a physics-informed Random Forest (RF) model to forecast vertical and horizontal displacement at the Inglewood Oilfield, integrating 14 years of reservoir, injection, and geomechanical data. Engineered features such as ΔPressure, Annual Injection Rate, and Compaction Strain were derived from poroelastic and effective stress principles. The horizontal model achieved R2 = 0.90 (training) and 0.36 (validation), forecasting +0.0871 inches eastward displacement in 2024/2025. The vertical model produced R2 = 0.69 (training) and 0.41 (validation), forecasting −0.0437 inches of subsidence. These values remain within the ±0.6 in/year stability threshold confirmed by InSAR and monument data. Feature importance analysis confirmed cumulative injection volume and injection variability as key displacement drivers, consistent with poroelastic stress redistribution. The model's simplicity, interpretability, and reliance on widely available data make it highly transferable to data-limited fields in Indonesia, such as those in the Rokan and Cepu Blocks. This approach enables cost-effective displacement risk screening and injection optimization in mature fields. Future work should integrate spatially resolved data and structural features to enhance prediction and inform proactive reservoir management.