Abstract: Tight sandstone reservoirs as the product of complex geological settings often exhibit subtle petrophysical property variations, significant fluid heterogeneities, and ambiguous rock physical responses, which pose challenges to reservoir prediction based on conventional isotropic sensitive parameters. To address these limitations, this study proposes a hierarchical identification strategy based on anisotropic rock physical parameters sensitive to lithofacies. The investigation focuses on 20 core samples obtained from the fourth member of the Xujiahe Formation in JY area. A multi-scale analytical approach, incorporating imaging logging, core analysis, porosity and permeability measurements, and cast thin-section observation, is adopted for the identification and classification of four lithofacies types: gas-saturated porous, water-saturated porous, gas-saturated fractured-porous, and water-saturated fractured-porous, in terms of pore and fracture characteristics along with fluid states. As per the multi-directional acoustic tests on the samples using an improved omnidirectional ultrasonic anisotropy test system, gas-saturated fractured-porous samples are remarkably different from additional samples in significant velocity and amplitude anisotropies. A rock physics cross plot of P-wave anisotropy (ε) and amplitude anisotropy (ε_A) is constructed to quantify these characteristic differences. The values of ε and ε_A both exceeding 20% tend to indicate gas-saturated fractured-porous lithofacies. Based on the similarity coefficient spectrum of full-azimuth waveform, a new anisotropy parameter: slowness anisotropy (ε_slow), is formulated to characterize the directional variation of slowness. The measured ε_slow values of water-saturated porous samples consistently below 0.01 indicate the extremely low anisotropy of this lithofacies type. By further integrating λ_ρ < 60 and μ_ρ < 30 from the λ_ρ-μ_ρ cross plot, we achieve the effective differentiation and identification of all four lithofacies types. These findings offer a robust quantitative framework for tight sandstone reservoir prediction and fluid identification.