Surface subsidence monitoring in mining areas is regarded as an essential basis for parameter inversion of subsidence prediction. Although traditional leveling surveys provide high accuracy, their application is limited by sparse observation points and intensive workload. Interferometric synthetic aperture radar (InSAR) technology enables large-scale monitoring with high spatiotemporal resolution. However, decorrelation often occurs in areas of steep gradients and rapid deformation, which restricts its applicability. The surface subsidence caused by coal seam mining in the Linxi mine was taken as a case, and the spatiotemporal characteristics of monitoring accuracy between leveling and InSAR were compared. A fusion method of InSAR and leveling data based on surface subsidence velocity was then proposed to invert the parameters of the probability integral method(PIM). The results show that with the expansion of rapid subsidence zones, the area of InSAR decorrelation gradually increases; in regions of steep gradients, monitoring errors followed an “increase-stabilization” pattern, which is closely related to subsidence velocity variations induced by multi-panel mining. Based on subsidence velocity, incoherent and coherent regions were delineated, and leveling data was introduced to compensate for InSAR decorrelation in rapid subsidence zones, ensuring the continuity and accuracy of parameter inversion. Verification with field data demonstrates that the proposed method significantly reduces the workload of leveling surveys while improving inversion accuracy, indicating strong practical applicability.