Pemanfaatan Penginderaan Jauh dan Machine Learning Untuk Pemetaan Kedalaman Lahan Gambut Tropis Kepulauan di Pulau Bengkalis
Suryandaru Rizky Pangestu, Wirastuti Widyatmanti, S.Si., Ph.D.
2026 | Skripsi | KARTOGRAFI DAN PENGINDRAAN JAUH
Tropical peatlands are important ecosystems that serve as carbon reservoirs, water regulators, and habitats for various flora and fauna. Information on peat depth is a crucial parameter in peat ecosystem management as it relates to carbon stock estimation, fire vulnerability, and the establishment of protected areas. However, conventional peat depth measurements through field surveys have limitations in terms of coverage, time, and cost. Therefore, this study aims to develop a peat depth estimation model through the integration of remote sensing data and machine learning and to evaluate the performance of the Random Forest (RF), Cubist, and Quantile Regression Forest (QRF) algorithms on Bengkalis Island, Riau. Peat depth data were obtained from field surveys using the line transect method, while the predictor variables used include DEMNAS, Topographic Wetness Index (TWI), Euclidean Distance to the river, and Sentinel-2A imagery indices in the form of NDVI, NDWI, TCT Wetness, and NDSI. Model evaluation was conducted using the coefficient of determination (R²), Multiple R, Root Mean Square Error (RMSE), and Mean Square Error (MSE).
The results showed that the Cubist algorithm produced the best performance with an R² value of 0.538, a Multiple R of 0.733, an RMSE of 186.60 cm, and an MSE of 34,819.56, better than RF and QRF. The DEMNAS and NDWI variables were identified as the most influential factors in explaining variations in peat depth. Spatially, the distribution of peat depth follows the morphology of the peat dome, with the greatest depth in the central part of the island and becoming shallower towards the coast. The results of this study demonstrate that the integration of remote sensing data and machine learning can be an effective approach for mapping peat depth over large areas and support carbon stock inventory and sustainable peat ecosystem management.
Kata Kunci : kedalaman gambut, penginderaan jauh, pembelajaran mesin, cubist, random forest, quantile regression forest, Pulau Bengkalis