ForensicSAR approach for detecting precursor deformation prior to the collapse of Derna Dam in Libya Arliandy Pratama (a,b*), Wataru Takeuchi (b)
(a) Dept. of Civil Engineering, The University of Tokyo, Tokyo, Japan
*arliandyarbad[at]g.ecc.u-tokyo.ac.jp
(b) Institute of Industrial Science, The University of Tokyo, Tokyo, Japan
Abstract
On 10-11 September 2023, Storm Daniel triggered cascading failures of the Bu Mansour (Upper Derna) and Al-Bilad (Lower Derna) dams in Libya. We develop ForensicSAR, a workflow that densifies on-structure InSAR sampling and couples satellite-derived deformation metrics with lightweight finite-element (FEM) hypothesis testing. Dual-orbit Sentinel-1 time series (2019-2024) are decomposed into vertical (U) and east-west (E) components, and Tracy-Widom PS selection (TW-PSI) recovers ~40% more persistent scatterers over low-coherence dam bodies relative to ADI-PSI, stabilizing pre-event velocity, acceleration, and Δ-slope estimates. Within a 12-month pre-event window, we detect a localized precursor at Bu Mansour, concentrated over the crest and eastern sectors with elevated ∣-Δ-slope| and zscores, whereas Al-Bilad shows weaker or inconsistent signals. Preliminary 2D Mohr-Coulomb FEM sections reproduce crest settlement and core-face stress concentrations consistent with central-core weakening under elevated head, providing a physically plausible interpretation of the remote-sensing evidence. Overall, integrating TW-PSI-densified InSAR with lightweight FEM discriminates plausible failure mechanisms and yields screening-level triggers for structural-health-monitoring prioritization in data-limited dam inventories, offering a scalable approach to dam-safety management.
