Tree cover, which has been widely studied using various remote sensing techniques, serves an essential indicator of forest productivity and habitat quality. Spatial heterogeneity in tree cover has received less attention despite links with critical phenomena such as biodiversity, tree re-expansion into formerly deforested areas, and degradation of existing forest landscapes. Few remote sensing studies have explored how tree cover heterogeneity varies across space and time, or how it may determine local or large-scale forest dynamics. This study used a global vegetation time-series product to map spatio-temporal dynamics in global tree cover heterogeneity over a 35-year period from 1982 to 2016, with heterogeneity quantified using the diversity metric Rao’s Q. We first explored the underlying relationship between tree cover and its heterogeneity at landscape scale across the globe. We then investigated how tree cover heterogeneity varied across tree cover gradients, biogeographic biomes, stand history, land protection status, and forest landscape intact status. Finally, nine possible combinations of variation trends in tree cover and its heterogeneity were used to generate a new map of forest dynamics categories. Our results show that monotonic spatio-temporal changes in tree cover are not necessarily associated with changes in tree cover heterogeneity. This suggests that simply using variation trends of tree cover without considering the context of local spatial heterogeneity does not fully capture forest dynamics. We show that remotely sensed tree cover heterogeneity can easily distinguish tree plantations from primary and secondary forests and that temporal change in tree cover heterogeneity is sensitive to spatial heterogeneity of open forest landscapes. The new forest dynamics categories map captures ongoing woody expansion in the Sahel region, forest degradation in the Amazon and central Africa, and widespread forest regrowth in Europe and Asia. Thereby, it is clear that spatio-temporal variation in tree cover heterogeneity supplements existing remote sensing studies in depicting global forest degradation, succession, and recovery.