The sublevel caving mining method is a mass production method with potentially very low operational costs. The success of this method is dependent on, among other factors, the cavability of the orebody and the overlying rock mass. However, caving of the surrounding rock mass also results in deformations in the cap rock as well as on the ground surface above the orebody being mined. From this follows that any existing infrastructure on the ground surface must be relocated as not to be affected by the mining-induced deformations.This thesis work was undertaken to bring about a better understanding of the rock mass behavior in the cap rock of non-daylighting orebodies, with particular application to the Printzsköld orebody as part of the LKAB Malmberget Mine. Rock testing, field observations and underground mapping was conducted to characterize the rock mass in the caving environment. A methodology for identifying the caving front based on seismic monitoring data was derived by studying the Fabian orebody (which has caved to surface), and using laser scanning data for validation. The methodology was then applied to the Printzsköld orebody to identify the caving front.Numerical modeling was performed for various scenarios of the rock mass as mining proceeded. Modeling included (i) stress analysis to understand stress changes and their effects on the rock mass behavior, (ii) discontinuum numerical modeling to quantify the influence of large-scale geological structures on the cave progression, and (iii) discontinuum cave modeling to simulate possible cave mechanisms in the cap rock more explicitly. Laser scanning together with seismic event data were used to calibrate the numerical models.The numerical simulation results showed that as mining progresses, the cap rock and hangingwall were exposed to stress changes that resulted in yielding. Two failure mechanisms were predominantly at play (i) shear failure (dominant in the cap rock) and (ii) tensile failure (dominant in the hangingwall). The presence of the large-scale structures affected thenearfield stresses through slip along the cave boundaries. The effect of the structures on the far field stresses were less significant.Discontinuum modeling to explicitly simulate failure and caving involved simulating the rock mass as a jointed medium using Voronoi tessellations in 2D, and bonded block modeling (BBM) in 3D. Both the 2D and the 3D modeling results showed fair agreement when comparing the inferred boundary of the seismogenic zone, with that identified from seismic monitoring data. Predictive numerical modeling was conducted for future planned mining to assess future cave development in the cap rock. The results from 3D modeling indicated that cave breakthrough for the Printzsköld orebody is expected when mining the 1023 m level, corresponding to approximately year 2022, as per current mining plans. The 2D model was non-conservative with cave breakthrough predicted to occur when mining the 1109 m level, corresponding to the year 2028.The estimated boundary between the seismogenic and yielded zones, as defined in the Duplancic and Brady conceptual model of caving, was coinciding with, or was close to, the cave boundary in the Printzsköld orebody. This may imply that in some areas the yielded zone was not present and that the Duplancic and Brady model may not be universally applicable. Additional work is required to verify this indication, as well as to fine-tune the modeling methodology.