Land cover is changing dramatically worldwide from both anthropogenic and natural drivers. In the United States, the rates and types of land cover change have varied temporally due to government policy, environmental regulation, global and national economic conditions, and regional weather and climate variability. Land cover changes can cause environmental degradation that affects long-term sustainability of human societies. Therefore, balancing the human need and environmental degradation requires explicit knowledge about environmental changes over multiple scales and perspectives. Remote sensing has been used as an effective tool to assess land changes across broad scales with multiple resolutions. However, the extraction of information from remotely sensed images is still challenged by the complex interaction between land cover heterogeneity and spatial as well as temporal resolutions. This dissertation aims at exploring such interaction in data classification and data fusion to better extract useful information about land cover. To achieve such goal, this dissertation first analyzes the impact of land cover heterogeneity in per-pixel and subpixel classification. Furthermore, this study also analyzes and proposes a data fusion method to better detect forest disturbances with high spatial and temporal resolutions. Using a high spatio-temporal resolution map of forest disturbances, this study suggests the use of temporal characteristics of disturbances to identify disturbance types. This study uses the South-Central United States as a case study for all experiments.