Complete Mining-Induced Subsidence Basin Reconstruction via Improved RIME-Based Probability Integral Parameter Inversion and SBAS-InSAR Residual Fusion

Large-gradient mining subsidence is difficult to reconstruct completely using small baseline subset interferometric synthetic aperture radar (SBAS-InSAR), because decorrelation and phase-unwrapping errors underestimate central subsidence, whereas the probability integral method (PIM) is sensitive to parameter inversion accuracy. This study proposes a basin-reconstruction framework combining PIM parameter inversion based on an improved rime optimization algorithm (RIME) with SBAS-InSAR residual fusion. Sobol initialization, a nonlinear adaptive search factor, and a stagnation-triggered perturbation improve RIME convergence and inversion accuracy. The inverted PIM field provides a physically constrained baseline, while normalized subsidence intensity regulates the SBAS-InSAR–PIM residual contribution for local correction. For a thick-coal-seam working face in northern Shaanxi, the improved RIME achieved a root-mean-square error (RMSE) of 66.7 ± 1.3 mm, equivalent to approximately 1.9% of the maximum measured dip-direction subsidence, together with a coefficient of determination (R2) of 0.990 ± 0.001. Its area under the convergence curve was 87.99% and 44.96% lower than those of particle swarm optimization (PSO) and the original RIME, respectively. PIM predicted a maximum subsidence of 3400 mm, whereas SBAS-InSAR detected 190 mm. With an optimal weight-shape parameter of 2.9, the fused result achieved an RMSE of 43 mm and a mean absolute error (MAE) of 25 mm, reducing the PIM-only errors by 33.85% and 51.92%, respectively.