Wildlife managers at Kenai Fjords (KEFJ) National Park, Alaska, are developing a comprehensive bear management plan. An important first step in this plan is to establish a baseline of bear abundance and inventory approaches for future monitoring efforts. In this study we use non-invasive genetic sampling and DNA-based capture–mark–recapture (CMR) analysis to estimate abundance of American black bears (Ursus americanus) using coastal habitats in the park. We used 2 multi-session CMR models in program MARK and 2 continuous-occasion models specifically designed for DNA-based capture data in the programs CAPWIRE and BAYESN. The latter 2 models maximize data from non-invasive sampling by allowing the entire sampling period to be considered a single continuous capture–mark–recapture occasion. Although all models yielded similar point estimates of population size, confidence intervals varied widely among methods. CAPWIRE yielded the most precise estimates of bear abundance. The number of bears estimated in each KEFJ bay ranged from 73 in the least populated to 324 in the most populated bay. Overall, the continuous-occasion CMR models provided the best estimates with our dataset, and we found that this approach offers a practical option when trapping seasons are constrained by biological or logistical factors.