To better understand the dynamic mechanical properties of rock at different depths, a method to simplify the stress characteristics of the rock mass surrounding a circular cavern was developed, and a series of impact tests involving high-speed photographic monitoring were performed on deep marble drilled in situ in the China Jinping Underground Laboratory at nine preloaded uniaxial stress levels (corresponding to simulated depths of 0, 100, 300, 600, 700, 900, 1200, 1500 and 1800 m). The results indicated that the strain rate sensitivity of the dynamic strength of Jinping marble decreased with the simulated depth, and obvious ductile deformation and failure characteristics occurred under high-energy dynamic loading and great depths. According to the fragmentation records, cracks clearly initiated and propagated mainly along the axial direction , and the observed failure patterns were classified into three typical modes: striped fragmentation (depth<600 m and strain rate<75/s), cylindrical surface spalling (depth≥600 m and strain rate<75/s) and comminution failure (strain rate≥75/s). Additionally, energy characteristics analysis revealed that the dissipative energy density e f and energy absorption efficiency A C exhibited significant depth differences and strain rate correlations, which corresponded well to the failure characteristics of the studied marble. Finally, an improved microcrack model considering the strain rate and simulated depth was established to characterize the dynamic mechanical behavior of the investigated marble, and the theoretically predicted strength values agreed well with the experimental values. • The dynamic failure behavior of rock under coupled static-dynamic loading conditions corresponding to different depths is revealed. • Striped fragmentation failure and a high strain rate sensitivity of the dynamic rock strength occur in shallow marble, while cylindrical surface spalling failure and a low strain rate sensitivity of the dynamic rock strength are observed at greater depths. • An improved microcrack model for dynamic rock strength prediction is established, which overcomes the limitation of previous models that lack comprehensive consideration of static preloading and strain rate conditions.