Assessing hydrologic effects of developing groundwater supplies in Snake Valley required numerical, groundwater-flow models to estimate the timing and magnitude of capture from streams, springs, wetlands, and phreatophytes. Estimating general water-table decline also required groundwater simulation. The hydraulic conductivity of basin fill and transmissivity of basement-rock distributions in Spring and Snake Valleys were refined by calibrating a steady state, three-dimensional, MODFLOW model of the carbonate-rock province to predevelopment conditions. Hydraulic properties and boundary conditions were defined primarily from the Regional Aquifer-System Analysis (RASA) model except in Spring and Snake Valleys. This locally refined model was referred to as the Great Basin National Park calibration (GBNP-C) model. Groundwater discharges from phreatophyte areas and springs in Spring and Snake Valleys were simulated as specified discharges in the GBNP-C model. These discharges equaled mapped rates and measured discharges, respectively. Recharge, hydraulic conductivity, and transmissivity were distributed throughout Spring and Snake Valleys with pilot points and interpolated to model cells with kriging in geologically similar areas. Transmissivity of the basement rocks was estimated because thickness is correlated poorly with transmissivity. Transmissivity estimates were constrained by aquifer-test results in basin-fill and carbonate-rock aquifers. Recharge, hydraulic conductivity, and transmissivity distributions of the GBNP-C model were estimated by minimizing a weighted composite, sum-of-squares objective function that included measurement and Tikhonov regularization observations. Tikhonov regularization observations were equations that defined preferred relations between the pilot points. Measured water levels, water levels that were simulated with RASA, depth-to-water beneath distributed groundwater and spring discharges, land-surface altitudes, spring discharge at Fish Springs, and changes in discharge on selected creek reaches were measurement observations. The effects of uncertain distributed groundwater-discharge estimates in Spring and Snake Valleys on transmissivity estimates were bounded with alternative models. Annual distributed groundwater discharges from Spring and Snake Valleys in the alternative models totaled 151,000 and 227,000 acre-feet, respectively and represented 20 percent differences from the 187,000 acre-feet per year that discharges from the GBNP-C model. Transmissivity estimates in the basin fill between Baker and Big Springs changed less than 50 percent between the two alternative models. Potential effects of pumping from Snake Valley were estimated with the Great Basin National Park predictive (GBNP-P) model, which is a transient groundwater-flow model. The hydraulic conductivity of basin fill and transmissivity of basement rock were the GBNP-C model distributions. Specific yields were defined from aquifer tests. Captures of distributed groundwater and spring discharges were simulated in the GBNP-P model using a combination of well and drain packages in MODFLOW. Simulated groundwater captures could not exceed measured groundwater-discharge rates. Four groundwater-development scenarios were investigated where total annual withdrawals ranged from 10,000 to 50,000 acre-feet during a 200-year pumping period. Four additional scenarios also were simulated that added the effects of existing pumping in Snake Valley. Potential groundwater pumping locations were limited to nine proposed points of diversion. Results are presented as maps of groundwater capture and drawdown, time series of drawdowns and discharges from selected wells, and time series of discharge reductions from selected springs and control volumes. Simulated drawdown propagation was attenuated where groundwater discharge could be captured. General patterns of groundwater capture and water-table declines were similar for all scenarios. Simulated drawdowns greater than 1 ft propagated outside of Spring and Snake Valleys after 200 years of pumping in all scenarios.