Browsing by Author "Second supervisor: Prof. Dr. Altangerel Lkhamsuren"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Inference of Subsidence Prediction Model Parameters Based on InSAR Data
    (GMIT, 2025-12-05) Suvd-Erdene Tsend-Ayush; First supervisor: Prof. Dr. Jörg Benndorf,; Second supervisor: Prof. Dr. Altangerel Lkhamsuren
    Mining and storage operations can cause ground subsidence that threatens infrastructure. Satellite InSAR measures millimeter‑level motion but only along the radar line of sight, which mixes vertical and horizontal components. This thesis builds a controlled synthetic framework to test how well three subsidence parameters - the influence angle β, convergence rate μ, and horizontal-tilt factor λ can be recovered from synthetic InSAR line-of-sight (LOS) data. A Gaussian influence function generates vertical, tilt, and horizontal fields that are projected into ascending and descending viewing geometries and perturbed with 1–2 mm Gaussian noise. A series of numerical experiments was performed, including baseline inversion, ASC-only vs. DES-only geometry tests, noise-level sensitivity, and multi-scenario horizontal-factor cases (λ = 0.3, 0.5, 1.0). Parameters are estimated with weighted nonlinear least squares and assessed with residual diagnostics, Monte‑Carlo uncertainty, and sensitivity tests for noise and geometry. Limited Bayesian (MCMC) and Extended Kalman Filter (EKF) trials demonstrate probabilistic and sequential extensions. Results show accurate recovery of (β, μ, λ) with residuals at the noise level (~1 mm). Histograms and Q–Q plots indicate near‑normal errors, and per‑orbit RMSE closely matches the standard deviation. Using both orbits (ASC+DES) reduces uncertainty relative to single‑track inversions and weakens parameter coupling, especially for λ. Monte‑Carlo spreads are narrow and scale with noise as expected. The Bayesian and EKF tests agree with the deterministic solution and provide credible intervals and time‑update capability. These findings support combined‑geometry InSAR as a practical basis for estimating subsidence parameters and suggest a clear path to applying the framework to real time‑series.