Abstract: As a key indicator of global climate change and an essential freshwater resource, the accurate acquisition of multiple physical parameters of glaciers holds significant importance for global climate change research, ecological conservation, and water resource planning. In China, glaciers are predominantly mountain glaciers distributed in high-altitude regions. Constrained by harsh environments and complex terrain, traditional in-situ detection methods fail to achieve large-scale continuous monitoring of internal glacier parameters. Satellite-borne glacier remote sensing, meanwhile, faces limitations in resolution and interference from complex ground clutter in mountainous glacier regions, and thus has yet to be operationalized. Airborne radar, with its superior spatial resolution and flexible detection capabilities, has become a critical technical tool for glacier monitoring and research. However, airborne detection of mountain glaciers still confronts challenges posed by undulating ice surfaces and complex subglacial topography: scattering clutter from the uneven ice surface interferes with radar signal interpretation and precise inversion of key parameters, while the intricate subglacial structure and scattering losses caused by ice surface topography interact with dielectric losses within the ice, impeding accurate inversion of glacier dielectric constants. To address these challenges, this study integrates airborne ultra-wideband radar detection data from mountain glaciers with the Pseudo-spectral Time Domain (PSTD) numerical simulation method. A coupled model of ice surface-subglacial dual interface topography and dielectric parameters is established. Through two-dimensional PSTD electromagnetic simulations, the interaction mechanism between topographic scattering and ice dielectric loss is elucidated. Furthermore, an inversion method for the imaginary part of the ice layer dielectric constant in measured regions is proposed based on dynamic range analysis. For the measured data from Laohugou Glacier No. 12, iterative optimization converges the estimated imaginary part value to 6.0×10⁻⁴, with a dynamic range difference of 0.61% from the measured mean value. The relative error between the estimated imaginary part and the theoretical mean is 21%. Cross-validation between simulation results and theoretical models demonstrates that this method effectively improves the inversion accuracy of glacier dielectric parameters in complex terrain by decoupling the synergistic interference between topographic relief and dielectric parameters, thereby offering a viable solution for studying internal dielectric properties of glaciers.