Large-scale forest dynamics, mostly driven by climate change and disturbance activities, have great implications for forest biodiversity, and ecosystem structures and services. Therefore, monitoring these changes over space and time is critical to improve our understanding of forest responses to climate change and disturbance activities. In this dissertation, remote sensing data were employed in statistical and machine learning (ML) approaches to characterize forest disturbances (infestation and fire) and long-term phenological changes at the eastern United States. Further, vegetation phenological information were integrated with field measurements of meteorological parameters to evaluate hydrologic alterations of forested watersheds. More specifically, three objectives are the focus of this dissertation: (1) developing an accurate approach to characterize the spectral-temporal trajectory of forest disturbances using long-term satellite observations, (2) evaluating the impacts of both vegetation and snowpack seasonal dynamics on emergent watershed-scale hydrologic behavior over vernal and autumnal transition periods, and (3) mapping fire-induced changes in forested landscapes using the combined approach of remote sensing and ML techniques.In the first chapter, using time series analysis of remote sensing imagery, we evaluated the performance of well-known spectral indices in capturing vegetation dynamics following the two major disturbances, fire and hemlock woolly adelgid (HWA; Adelges tsugae Annad) infestation. Our results suggested that the overall performance of Normalized Difference Vegetation Index (NDVI) was the most accurate in detecting disturbances intensity, temporal dynamics, and recovery patterns. Results obtained from our second chapter demonstrated that the lengthened growing season and declined winter snowpack can significantly alter the low-frequency seasonal streamflow distributions by changing the seasonality of evapotranspiration. Our findings also suggested that with continuous declines in winter snowfall, the effect of snowpack dynamics on identifying the seasonal streamflow regimes has weakened, while vegetation phenology has exerted more dominant control. In the third study of this dissertation, we evaluated the performance of four ML classification techniques in estimating fire severity, using spectral indices, topographic parameters, and surface temperature (ST) variables from remote sensing data. We found that Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models had high predictive capability to detect fire severity in regions with mixes of low to medium fire disturbances. Results also indicated that spectral indices, especially Normalized Burn Ratio (NBR) and Tasseled Cap Wetness index (TCW), elevation, aspect and growing season ST had higher relative contributions to fire severity prediction than other variables.