- Main
Integrated InSAR, 3D volumetric inversion, and deep learning reveal ongoing oil induced subsidence in the East Coast of Lake Maracaibo
Published Web Location
https://doi.org/10.4995/raet.2027.26485Abstract
This study integrates Differential Interferometric Synthetic Aperture Radar (DInSAR) using the Small Base-line Subset (SBAS) approach, 3D volumetric inversion (DTerrain), and Long ShortTerm Memory (LSTM) neural networks to quantify and forecast land subsidence in the East Coast of Lake Maracaibo (COLM), Ven-ezuela. A dataset of 280 Sentinel-1 images (2018–2022) was processed to derive high resolution deformation time series. Line of sight (LOS) velocities indicate significant subsidence rates, reaching −5.91 ± 0.10 cm/yr in Lagunillas and −7.19 ± 0.29 cm/yr in Bachaquero. Decomposition into vertical and horizontal components reveals dominant vertical deformation, with maximum rates of −7.1 cm/yr in Lagunillas. The DTerrain inver-sion model identifies volumetric strain concentrations at approximately 1676 m depth, consistent with reser-voir compaction processes associated with hydrocarbon extraction. The estimated strain values (−600 µstrain/yr) for Lagunillas oil suggest significant porosity reduction within Miocene reservoir units. Addition-ally, the LSTM based forecasting model successfully captures the nonlinear temporal evolution of subsidence, projecting continued deformation at rates of approximately −4 cm/yr through 2027 under current conditions. These results confirm the persistence of subsidence in the COLM and highlight its strong linkage to reservoir subsidence. The integrated methodology provides a robust framework for geomechanical monitoring and risk assessment, with direct implications for infrastructure stability and flood hazard mitigation in subsidence prone coastal environments.
Many UC-authored scholarly publications are freely available on this site because of the UC's open access policies. Let us know how this access is important for you.