Hydrologic models developed for mountain regions typically exclude deep groundwater flow, assuming it is a negligible component of the water budget. This study was carried out in a small (4.74 km²) snowmelt-dominated headwater catchment in the Okanagan Basin, British Columbia. To complement the established snow, soil water and stream monitoring networks, two deep groundwater wells were drilled at high elevation ∼3 m apart (W1-50 m and W2-30 m), and one in the valley bottom (W3-30 m). To explore the groundwater flow dynamics, estimate recharge, and quantify the contribution of deep groundwater to streamflow, a coupled land surface – subsurface model was developed using MIKE SHE. Over the simulation period (2005-2010), mean annual recharge to the bedrock was 19% of annual precipitation, ranging between a negative recharge or net loss of water from the saturated zone (-7% in 2008-09) to a net gain (+46% in 2006-07). Recharge begins in Jun-Jul and ends between Sep-Dec, with the majority of recharge occurring from Jul-Aug. Baseflow (i.e. the exchange between the bedrock and the stream) varied from 0.2 to 1.6% of measured streamflow between Apr and Jun, was relatively constant at 14% from Oct to Mar, and as high as 36-45% during Aug and Sep (average annual baseflow was 14% of streamflow). To allow for deep groundwater flow to exit the catchment, a specified outward flux of 2% of annual precipitation was assigned along each side of the lower catchment through the full bedrock thickness (220 m). Flux values as high as 6% of annual precipitation did not compromise the calibration. Water samples collected in 2007 and 2008 from the groundwater wells, soil piezometers, and at different locations along the stream network, as well as snow and rain samples, were analyzed for major ion chemistry, oxygen-18 and deuterium. Most waters have a Ca-Mg-HCO3 or Ca-Na-HCO3 water type and relatively low electrical conductivity (34-144 μS/cm), reflecting the low total dissolved solids of the waters. All waters have a meteoric origin and most cluster along the local meteoric water line close to the snow samples, reflecting a snowmelt origin. Samples collected in the valley (W3) consistently have a different chemical composition and a more depleted isotopic signature, reflecting the possible influence of a nearby fracture zone.