Application of Established Seismic-Constrained Surface TEM Inversion Framework to Gypsum Mine Roadway Detection

Abandoned gypsum mine roadways may evolve into structurally complex cavities and potential water-related hazards, while a single geophysical method may provide either insufficient structural resolution or ambiguous electrical interpretation. This study applies the established unidirectional cross-gradient formulation, in which a fixed acoustic impedance model derived from seismic data constrains transient electromagnetic (TEM) resistivity inversion, to a gypsum mining field case under a surface TEM configuration. The contribution is the application and evaluation under these geological and acquisition conditions rather than a new inversion algorithm. Qualitative comparison of the two- and three-dimensional synthetic results suggests closer alignment of the constrained resistivity anomalies with the prescribed geometry. The constrained TEM step requires approximately 1.4% and 2.1% more single-iteration CPU time than separate TEM inversion in the two- and three-dimensional tests, respectively; these timings measure computational cost rather than accuracy. The two-dimensional benchmark uses a 1.5 km/s low-velocity anomaly and a 10 Ω·m conductive anomaly of the same prescribed geometry as modality-specific contrasts rather than a calibrated petrophysical description of a single roadway filling. Application to TEM blocks 27 and 28 from a gypsum mine in Jiangsu Province yields roadway-related low-value structures that are qualitatively consistent with the available roadway layout and visually clearer than those in the separate TEM result. Because only two adjacent blocks are evaluated and the field TEM response is weak, the result is interpreted as a local qualitative demonstration of structural delineation. The low resistivity values are not used to infer water saturation or to distinguish water-filled from dry roadways. The results support further evaluation of seismic-constrained TEM inversion with quantitative synthetic benchmarks and independent field validation.