Context: Forest cover change analyses have revealed net forest gain in many tropical regions. While most analyses have focused solely on forest cover, trees outside forests are vital components of landscape integrity. Quantifying regional-scale patterns of tree cover change, including non-forest trees, could benefit forest and landscape restoration (FLR) efforts. Objectives: We analyzed tree cover change in Southwestern Panama to quantify: 1) patterns of change from 1998-2014, 2) differences in rates of change between forest and non-forest classes, and 3) the relative importance of social-ecological predictors of tree cover change between classes. Methods: We digitized tree cover classes, including dispersed trees, live fences, riparian forest, and forest, in very high resolution images from 1998-2014. We then applied hurdle models to relate social-ecological predictors to the probability and amount of tree cover gain. Results: All tree cover classes increased in extent, but gains were highly variable between classes. Non-forest tree cover accounted for 21% of tree cover gains, while riparian trees constituted 31% of forest cover gains. Drivers of tree cover change varied widely between classes, with opposite impacts of some social-ecological predictors on non-forest and forest cover. Conclusions: We demonstrate that key drivers of forest cover change, including topography, road distance and historical forest cover, do not explain rates of non-forest tree cover change. Consequently, predictions from medium-resolution forest cover change analyses may not apply to finer-scale patterns of tree cover. We highlight the opportunity for FLR projects to target tree cover classes adapted to local social and ecological conditions.