Estimating population abundance of wolves (Canis lupus) in densely forested landscapes is challenging because reduced visibility lowers the success of methods such as aerial surveys and enumeration of group size using radiotelemetry. However, regular population estimates of wolves are necessary for population monitoring and sustainable management. We used noninvasive hair snaring and spatially explicit capture-recapture (SECR) to estimate wolf abundance on Prince of Wales Island (POW), Alaska, USA, during 2012-2015. We monitored 36-82 hair‐snare stations weekly for 9-11 weeks during autumn. The noninvasive study area covered 1,683 km2 during 2012-2013 and was expanded to 3,281 km2 during 2014-2015. We identified 57 individual wolves during the study period using DNA from hair follicles genotyped at 10 microsatellite loci. We used population density estimates using SECR (2013: 24.5 wolves/1,000 km2 [95% CI = 14.4-41.9 wolves/1,000 km2], 2014: 9.9 wolves/1,000 km2 [95% CI = 5.5-17.7/1,000 km2], 2015: 11.9 wolves/1,000 km2 [95% CI = 7.7-18.5 wolves/1,000 km2]) to predict the autumn population for the POW management unit (2013: 221.1 wolves [95% CI = 130-378]; 2014: 89.1 wolves [95% CI = 49.8-159.4]; 2015: 107.5 wolves [95% CI = 69-167]). We detected and redetected more wolves and increased the precision of the density estimate after increasing the hair sampling intensity and sampling area in 2014-2015. Our results demonstrate that estimating wolf abundance using noninvasive sampling and SECR was feasible and reliably applied producing a statistically robust population estimate for monitoring wolf populations in densely forested areas. These methods have promise for application to widely ranging carnivores at population‐level scales and may be especially useful when regular density estimates are necessary for management and conservation.