"Safety is a value, not just a Priority" Geomechanically stable underground excavations require continuous assessment of rock mass behavior for maximizing safety. Optimizing pillar design is essential for preventing hazardous incidents and improving production in room-and-pillar mines. Maintaining regional and global stability is complicated for underground carbonate or stone deposits, where extensive fracture networks and groundwater flow become leading factors for generating unsteady ground conditions including karsts. A sudden encounter with karst cavities during mine advance may lead to safety issues, including ground collapse and outflow of unconsolidated sediments and groundwater. The presence of these eroded zones in pillars may cause their failure and poses a risk to the lives of miners apart from disrupting the pre-planned mining operations. A pervasive presence of joints and fractures plays a primary role in promoting structurally controlled failures in stone mines, which accelerates upon interaction with the karst cavities. The prevalent empirical and analytical approaches for pillar design ignore the geotechnical complexities such as the spatial density of discontinuities, karst voids, and deviation from the design during short-range mine planning. With the increasing market demand for limestone products, mining organizations, as well as enforcement agencies, are investing in research for increasing the efficiency of extracting valuable resources. While economical productivity is essential, preventing risks and ensuring the safety of miners remains the cardinal objective of mining operations. According to the Mine Safety and Health Administration (MSHA), since 2000, about 31% of occupational fatalities at all underground mines in the United States are caused due to ground collapse, which rises to 39% for underground stone mines. The objective of this study is to provide a reliable and methodological approach for pillar design in underground room-and-pillar hard rock mines for safe and efficient ore recovery. The numerical modeling techniques, implemented for a case study stone mine, could provide a pragmatic framework to assess the effect of karsts on rock mass behavior, and design future pillars detected with voids. The research uses data acquired from using remote sensing techniques, such as LiDAR and Ground-penetrating Radar surveys, to map the excavation characteristics. Discontinuum modeling was valuable for analyzing the pillar strength in the presence of discontinuities and cavities, as well as estimating a safe design standard. Discrete Fracture Networks, created using statistical information from discontinuity mapping, were employed to simulate the joints pervading the rock mass. This proposed research includes the calibration of rock mass properties to translate the effect of discontinuities to continuum models. Continuum modeling proved effective in analyzing regional stability along with characterizing the redistributed stress regime by imitating the excavation sequence. The results from pillar-scale and local-scale analyses are converged to optimize pillar design on a global scale and estimate the feasibility of secondary recovery in stone mines with a dominating discontinuity network and karst terrane. Stochastic analysis using finite volume modeling helped evaluate the performance of modified pillars to assist production while maintaining safety standards. The proposed research is valuable for improving future design parameters, excavation practices, and maintaining a balance between an approach towards increased safety while enhancing production. / Doctor of Philosophy / "The most valuable resource to come back out of a mine is a miner" – Anonymous. The United States accounted for 27% of the global limestone market share which was valued at US$58.5 billion in 2020 [148]. It is projected to reach a target of US$65.3 billion in 2027, growing even in midst of the COVID-19. As surface reserves deplete, much of the mineral demand gap is supplemented by mining underground deposits. Underground mines extract minerals from deep within the earth compared to surface mines. As a result, the miners experience a greater number of accidents in a constricted environment because of roof/tunnel collapse, fewer escape routes, ventilation, explosions, or inundation. According to the Mine Safety and Health Administration (MSHA), about 15% of all underground mine injuries in the US were caused by rockfalls since 1983. The majority of underground stone deposits are mined using the room-and-pillar mining method, which resembles a chessboard design where the light squares are mined, and the dark squares are left as rock pillars to support the tunnels. Limestone, a carbonate rock, contains a lot of fractures and joints (discontinuities). Erosion of rocks due to continuous water flow through the fractures leads to the formation of cavities known as karsts. Interaction of karsts with the prevalent fracture network increases rockfall risk during mining. The collapse of voids along with an inrush of filled rock-clay-water sludge can harm miners' lives, damage machinery, and stop further operations. Literature is scarce on topics that quantify the risk and disruption posed by these cavities in underground mines. Most rock classification systems cannot classify their effect because of the unpredictability and extensive analysis required. The objective of this research is to provide a reliable and methodological approach for designing pillars in underground hard rock mines for ensuring a safe working environment and efficient mineral recovery. This research starts with analyzing the strength of pillars, in which karst cavities were discovered while mining. The safety concerns often lead the miners to not excavate around the cavities and leave valuable resources unmined. Data from ground-penetrating radar and laser scanning surveys were used to characterize the voids and map the discontinuities. Discrete-element numerical modeling was used to simulate the pillars as an assembly of blocks jointed by the discontinuities. The simulation results help us understand the instability issues in the karst-ridden pillars and ways to improve upon the existing design. The findings were used to modulate the parameters for regional-scale models using finite volume modeling for less computationally intensive analyses and simulating rock as a continuum. The continuum models were highly effective in analyzing the instability issues due to the prevalent karstic network. This helps understand any alternative scenario that could have been implemented to optimize ore recovery while preventing risks. The results from the single pillar and regional analyses are combined to optimize pillar design on a global mine scale. This dissertation focuses on improving hazard mitigation in mines with unpredicted anomalies like karsts. Although this research is based on a specific mine site, it empowers the operators to explore the presented techniques to increase safety in all underground mines. The suggested methodology will help devise better strategies for handling instability issues without jeopardizing the mine operations. The primary motivation is to keep the underground miners safe from hazardous situations while fulfilling the secondary objective of maximizing mineral production.