Small coastal peat islands such as Bengkalis (Riau, Indonesia) are hydrologically sensitive systems where groundwater level (GWL) controls ecosystem stability. Both extremes are hazardous: prolonged low GWL elevates peat-fire, subsidence, and carbon-loss risks, while excessively high GWL can trigger bog-burst. This study developed a simple empirical model to predict daily GWL in degraded peatlands on Bengkalis Island using in situ GWL data from three sites (drained, undrained inland, and undrained coastal) and GPM satellite rainfall (October 2023-April 2025). Calibrated over one year and validated over the next seven months, the model performed well at drained and coastal sites (R ~0.82, MAPE ~14%), capturing seasonal dynamics. In contrast, its performance at the inland site was lower (R ~0.5) due to minimal water table fluctuation. Coefficient values indicate the strongest rainfall response and fastest losses at the drained site, negligible daily loss at the inland site, and intermediate behavior at the coastal site. Scenario simulations highlight management-relevant risks: 15 rain-free days cause GWL to drop below the critical -0.40 m fire-risk threshold at the drained site and coastal site, whereas undrained inland remains just above it; conversely, 60 mm/day of rain for four days can raise GWL to the surface at coastal site (bog-burst risk). The model provides a practical tool for informing rewetting strategies to manage fire and collapse risks in degraded tropical peatlands.