Abstract: To address the issues that linear stress boundary conditions are widely adopted in current in-situ stress field inversion, and that the inversion data from boreholes far from the boundary in the calculation model of deep-buried long tunnels fail to truly reflect the actual stress field, this study proposes a zoned nonlinear inversion method, which can effectively characterize the heterogeneous variation law of the in-situ stress field.This method was applied to a deep-buried long tunnel project crossing faults. The most prominent feature of this project is that the tunnel passes through a narrow horst-type mountain (fault-block mountain), with discontinuous strata on both sides of the fault. Based on this, studies on the in-situ stress field analysis and rockburst prediction in the tunnel axis area were conducted.Results show that: 1) The zoned nonlinear inversion method outperforms the non-zoned linear inversion method in overall accuracy, effectively solving the trade-off problem of the latter in data adaptation between shallow and deep parts; 2) This method can effectively resolve the issue that it is difficult to determine the in-situ stress distribution characteristics of the granodiorite section in the middle of the tunnel by relying solely on measured in-situ stress data, providing reliable data support for rockburst prediction; 3) The in-situ stress disturbance in the fault influence zone is particularly significant, and due to the stress release effect, the stress magnitude in this zone is significantly lower than that of the rock masses on both sides; 4) The deep-buried long tunnel crossing faults exhibits prominent high in-situ stress and rockburst problems. Among them, the length of the moderate rockburst section reaches 2651.2 m (accounting for 53.1% of the total tunnel length), and the length of the severe rockburst section is 635 m (accounting for 13.2% of the total tunnel length). Targeted rockburst prevention and control measures should be adopted during the construction phase.