Alternative silvicultural treatments such as thinning can be used to restore forested watersheds and reduce wildfire hazards, but the hydrologic effects of these treatments are not well defined. We evaluated the effect of two shelterwood-with-reserve silvicultural prescriptions, one leaving residual trees evenly distributed (SE) and the second leaving residual trees in groups (SG), both with 50% basal area removal, on snow accumulation in lodgepole pine stands at the Tenderfoot Creek Experimental Forest, west-central Montana. The snow water equivalent (SWE) close to the seasonal peak was measured at >250 locations in the SE and SG treatments and a control in 2003, 2004 and 2005. Reduced interception in SE resulted in significant (P < 0.0001) 7.2 and 5.6 cm increases in SWE relative to the control in 2003 and 2004, respectively, and a 1.7 cm increase in 2005. Predictive models for the mean peak SWE in the control and the SE treatment were based on an inverse-exponential relationship between interception efficiency and mean storm intensity, and the average percent deviation for the two models was 9.4 and 7.6%, respectively. In the SG treatment, increased solar radiation and wind resulted in sublimation losses that offset gains due to reduced interception in the openings between groups. Consequently, SG always accumulated significantly less snow than SE (P < 0.0001), and the SG treatment and control means were not significantly different. Differences in snow accumulation between groups and openings and between the north and south sides of groups meant that the SWE in SG was up to three times more variable than in either SE or the control. The contrasting responses in the SE and SG treatments demonstrate that thinning can have substantially different effects on snow accumulation depending on the spatial arrangement of tree removal.