Tree cavities are an essential habitat component for wildlife species across diverse taxa, from insects to large mammals. Many of these species are imperiled by loss of cavities. Further, conservation action is hindered by limited information on the spatial distribution of cavities, largely due to difficulties in developing useful models of their presence or abundance. Accurately predicting the fine-scale, landscape-wide, spatial distribution of these important habitat features would greatly benefit conservation measures. In this study, we evaluated the efficacy of using remotely sensed data, including LiDAR, multispectral imagery, and SAR, to predict the locations of cavities suitable for nesting Wood Ducks Aix sponsa at fine scales (≤40 9 40 m) and across a broad landscape (~254 000 ha). We used Random Forest models to classify the presence-absence of cavities at four spatial scales of prediction (5, 10, 20, and 40-m pixels) as well as three groupings of predictor variables (LiDAR-derived metrics, spatial forest-inventory variables [derived via LiDAR or imagery], and ancillary remotely sensed data [derived via SAR or imagery]), and then compared the accuracy between models. The 20-m response-scale had the highest accuracy. Variables from each of the groupings were important and largely relied on multispectral imagery and LiDAR data. Our final model and predictive map had 84% overall Out of Bag accuracy and, when tested with an independent cavity dataset, correctly identified 80% of trees with suitable cavities as presences. The predictive map can be used by researchers and land-managers to determine how past management actions have affected cavity availability as well as the locations of habitat complexes for Wood Ducks and, potentially, other secondary-cavity-nesting wildlife. In addition, our analysis can serve as a methodological case-study for cavity-nesting wildlife in other regions, as the use of remotely sensed data like high-density LiDAR and multispectral imagery increases.