Abstract: In the southern region of the Ordos Basin, the thin, single-layer distributary channel sands in the Lower Shihezi Formation cause coupling between the sidelobes of the underlying Shanxi-Taiyuan coal seam reflections and the reflections of thin sand bodies within a limited seismic resolution, making it difficult to accurately analyze the reservoir response characteristics. Additionally, challenges in petrophysical analysis arise owing to the poor quality of the logging curves and insufficient shear wave data in this area block, which impede quantitative reservoir prediction. To address these issues, this study proposes a set of sand, mud, and coal three-phase petrophysical modeling techniques based on the Xu–White petrophysical model, employing a multiparameter fitting curve correction and stepwise fusion method. This approach effectively improves the quality of logging curves and enhances the accuracy of shear wave velocity prediction, thereby providing reasonable logging information for subsequent inversions. Furthermore, to address the reservoir prediction problems mentioned above, this study introduces a multidimensional facies-controlled pre-stack geostatistical inversion method suitable for the study area. First, based on the existing geological understanding and well log information, seismic attributes related to channel sands were analyzed, and a two-dimensional lithofacies probability density conforming to sedimentary variation patterns was extracted as a planar constraint for the Shihezi Formation reservoir. Then, using the coal seam inversion volume combined with Bayesian discrimination principles, a three-dimensional coal facies probability density was extracted as a spatial constraint for the underlying coal seam. Finally, both two- and three-dimensional constraints were comprehensively integrated to conduct a pre-stack geostatistical inversion. The prediction results obtained using this method comprehensively considered the coupling characteristics of the underlying coal seams and sandstone reservoirs, effectively reducing the uncertainty in predicting thin reservoirs and demonstrating good practical application with significantly improved drilling concordance.