Snags create nesting, foraging, and roosting habitat for a variety of wildlife species. Removal of snags through postfire salvage logging reduces the densities and size classes of snags remaining after wildfire. We determined important variables associated with annual persistence rates (the probability a snag remains standing from 1 year to the next) of large conifer snags (greater than or equal to 23 cm diameter breast height (dbh)). Our study sites were located within two wildfires in western Idaho (Foothills fire of 1992 and Star Gulch fire of 1994). Study sites in the Foothills were partially salvage-logged (one-half of standing snags greater than or equal to 23 cm dbh removed), and sites in the Star Gulch burn were unlogged. Snags were monitored within 0.04 ha plots for 8-9 years beginning in 1994 in Foothills Burn and 1995 in the Star Gulch Burn. A total of 1131 ponderosa pine (Pinus ponderosa) and Douglas-fir snags (Pseudotsuga menziesii) were monitored during the study period. Data were collected on snag species, height, decay class, and diameter at breast height. We also collected stand-level data (e.g., slope, aspect) and quantified information on remotely sensed data (e.g., pre-fire crown closure, burn severity) at the pixel-level (30 m by 30 m) and within 1 km of the plot centers (landscape-level). We modeled annual snag persistence as a function of data on all three scales using non-linear mixed-effects models. Additionally, we expected that variables on multiple scales, including information on burn severity and pre-fire crown closure, would be influential in determining snag persistence. The best models of snag persistence were selected using an information theoretic approach (i.e., AICc). Small-scale variables (age, height, dbh, tree species, decay, and snag density) were best predictors for both wildfire areas. In contrast to our expectations that large-scale variables would influence annual persistence rates, these factors did not appear in top models. Persistence was shorter for ponderosa pine than Douglas-fir snags. Additionally, smaller snags in plots with fewer snags fell sooner than did larger snags in more dense stands. Age of snag was also an important variable predicting snag persistence (older snags are more likely to fall). Snag longevity (the total amount of time the snag remained standing) also varied between the two sites. The predicted half-life of a ponderosa pine snag was 7-8 years in salvage-logged plots and 9-10 years in unlogged plots. The predicted half-life of Douglas-fir snags was longer than ponderosa pine, at 12-13 years in the salvage logged burn versus 15-16 years in the unlogged burn. On the partially logged sites, the primary effects of salvage logging on snags appeared to be the reduction of the average snag size (diameter and height) and density, which in turn reduced the subsequent longevity of individual snags. We concluded that the main effects of the partial-salvage logging appeared to be a reduction in overall snag density and average snag size which in turn reduces average persistence time. Management practices that preserve dense stands of snags will promote the longer-term persistence of suitable snags as nesting habitat for cavity-nesting birds.