In　this　paper,　optimum　intensity　measures　for　probabilistic　seismic　demand　model　of　subway　stations　with　different　burial　depths　are　investigated.　A　three-story　three-span　subway　station　which　is　commonly　seen　in　the　metro　system　in　China,　is　selected　as　a　typical　case.　The　subway　station　is　embedded　in　two　generic　engineering　sites　corresponding　to　site　classes　II　and　III　specified　in　Chinese　code　for　seismic　design　of　urban　rail　transit　structures.　Five　different　burial　depths　(4　m,　8　m,　12　m,　16　m,　and　20　m)　are　considered　in　this　study.　Nonlinear　soil-structure　interaction　time　history　analyses　are　conducted　on　the　two-dimensional　numerical　model　of　the　subway　station　embedded　in　two　different　engineering　sites.　Based　on　the　theory　of　wave　propagation　in　homogeneous　and　elastic　media,　seismic　excitations　at　the　engineering　bedrock　for　the　two　generic　sites　are　back　calculated　from　an　ensemble　of　50　pairs　of　as-recorded　seismic　motions　at　bedrock　outcrop.　The　efficacy　of　twenty-one　commonly-used　scalar　intensity　measures　(IMs)　for　seismic　response　prediction　of　underground　structures　are　evaluated　by　the　metrics　of　efficiency,　practicality,　proficiency　and　sufficiency.　The　numerical　results　reveal　that　for　subway　stations　embedded　in　site　class　II,　velocity　spectrum　intensity　(VSI)　is　the　optimum　IM　with　different　burial　depths.　Furthermore,　for　subway　station　embedded　in　site　class　III,　VSI　and　peak　ground　velocity　(PGV)　are　the　optimum　IMs　with　a　burial　depth　of　4　m,　and　VSI　and　compound　velocity-related　intensity　measure　are　the　optimum　IMs　with　burial　depths　of　8　m-20　m.　Moreover,　for　the　three　well-recognized　amplitude　IMs,　peak　ground　acceleration　(PGA),　PGV,　peak　ground　displacement,　with　the　increase　of　the　burial　depth　of　underground　structures,　the　proficiency　of　PGV　is　gradually　enhanced　compared　with　PGA.　Finally,　based　on　the　theory　of　one-dimensional　wave　propagation　in　homogeneous　and　elastic　half-space,　the　preliminary　mechanism　explanation　of　the　optimum　amplitude　IMs　for　variation　of　burial　depth　is　given.