GIS and remote sensing are essential tools in evaluating wildlife habitat quality. Ultimately, habitat evaluations should relate quality measures to population dynamics particularly when habitat conditions are changing. Following the 1980 eruption, populations of Columbian black-tailed deer (Odocoileus hemionus columbianus) (hereafter BTD) increased rapidly but quickly declined. Similar declines have been historically attributed to a predominance of even-aged closed canopy forest and low forage biomass. The link between forage and BTD population dynamics has been shown but to our knowledge no study has applied geospatial tools to quantify landscape forage conditions and relate this to population changes for BTD. We used 6 dates of post-eruption Landsat imagery (1984, 1988, 1991, 1996, 1999, and 2002) to map forest successional patterns and estimate subsequent changes in winter forage for a portion of the Mount St. Helens blast zone. We used forest maps as inputs into a GIS model estimating the supportable density of deer as an indicator of nutritional carrying capacity (NCC). To simulate potential road and cover effects on habitat use, we reduced habitat values based upon published effects of distance to roads and the ratio of forage to cover vegetation. We created 1000 randomly placed simulated home ranges to provide a bootstrap data set for model comparisons and to characterize potential uncertainty around model estimates. We compared model estimates to actual deer population estimates in the study area. Habitat models indicated a pattern of rapid forest succession and decrease in forage which closely followed the estimated population trends for the study area. Closed canopy forest dominated the landscape after 1991, and by 1996 comprised 70% of the winter range. The combined effects of roads and cover reduced estimated habitat values by over 70% (P < 0.02) in all years although NCC models alone were closer to actual population estimates in 1984 and 1988. By 1991, NCC reduced by road and cover effects was closer to actual trends although NCC was most highly correlated (r = 0.72) with estimated deer numbers. Confidence intervals around NCC estimates indicated a considerable amount of variability in all years although a statistically significant decline in forage biomass was predicted between 1991 and 2002 (P < 0.03). Even given high variation, estimated declines in habitat quality were similar to actual population declines indicating a predictable link between habitat changes and population dynamics. Our models did detect relevant changes in habitat quality similar to population changes observed and indicate that management interspersion of early successional forests will be required for successful deer management in this region.