Ongoing anthropogenic changes, such as land-use change, climate changes, and direct effects of CO2, can alter vegetation structure, composition, and growth rates. If these changes are substantial, they may redistribute forest biomes, with downstream feedbacks onto terrestrial carbon storage, climate, and availability of natural resources. Species and ecosystems at the edge of current distributions, such as at the boundary between savannas and closed forests, may be the most vulnerable to these environmental changes. In this dissertation I explore the long term effects of anthropogenic changes on the structure, composition, function of vegetation at a savanna-forest boundary. We use historical vegetation survey data from both the 1800's and today to test whether vegetation had multiple stable states within the same environmental space in the past, and to determine if long-term changes in land-use and management drove a vegetation state shift. We find that past vegetation had a multiple stable vegetation states of open savanna and closed forests likely maintained by disturbance-vegetation feedbacks. But ongoing fire suppression and logging have created and maintained modern vegetation in a single closed forest state. We show that disturbance is a strong predictor of aboveground biomass in the past, but only has a negligible effect on aboveground biomass on the present landscape. Strong residual taxa relationships that are unexplained by relationships with the environment were common in the past vegetation, but have since disappeared from the modern landscape. While it is likely that modern systems will be maintained as mostly closed forests by continued land management, future projections for the region indicate much hotter and potentially drier growing seasons, questioning how long vegetation will remain forested. Positive effects of increases CO2 on plant water use efficiency (WUE) could aid forest resilience to future drought conditions, but may be complicated by climate changes. To test whether combined changes in CO2 and climate have a positive or negative effect on growth, I use Bayesian models of annual tree ring growth and WUE. We find that CO2 increases plant WUE, driving precipitation sensitivity declines, but the negative impacts of increasing summer temperatures outweigh any positive impacts of higher WUE. Long-term ecosystem model runs simulate similar increase in WUE and declines in precipitation sensitivity by plant functional types over the 20th century, but do not always agree on either the direction or magnitude of overall growth changes or changes in temperature sensitivity. Overall, the historical perspective provided in this dissertation demonstrates that a combination of land-use, climate changes, and CO2 can drive very different vegetation structure, composition, and functional response to climates that would not be predictable from modern vegetation-environment relationships alone. As a result, future forecasting efforts might be improved by including the disturbance processes, species compositional feedbacks, and non-stationary climate responses highlighted in this dissertation.