Identification of land-cover change and interannual climate variability in Africa from satellite imagery

This thesis concerns the monitoring of land-cover change over large areas, specifically continental Africa, using time series of satellite remotely-sensed data. Case studies of sites with known histories of land-cover change illustrate that human intervention obscures effects of climate change on vegetation as viewed from space. This is evident when change processes occur at finer spatial scales than the resolution of the remotely-sensed data.; The sensor spatial resolution, which determines the level of detail captured by a given observation, governs the types of processes that can be detected. Fine-resolution sensors (e.g. 30 m x 30 m pixels) detect variability in land cover due to both anthropogenic influences and fluctuations in local climate. Coarse-resolution sensors (e.g. 8 km x 8 km pixels) usually only capture climate variability at regional or global scales, though observations are collected with high frequency.; Further analysis focusing on land-cover change at these study sites derives several measures (or metrics) of change, from time series of coarse spatial resolution satellite data. Included among these are land-cover change vectors, which measure vector differences between annual time series of land-cover indicators, such as vegetation greenness or surface temperature, derived from satellite data. The metrics relate statistically to aggregate changes in land cover. They relate more strongly when vegetation greenness and surface temperature data are combined, and are most effective when combined in multivariate models.; Over the continent of Africa, comparison of three selected metrics to inter-annual climate variability shows that they relate both qualitatively and quantitatively. Stronger metric/climate couplings exist for some vegetation types than for others. Also, the effect of time lags on these couplings is explored.; Understanding how the land surface interacts with atmospheric and oceanic components of the global climate system is of interest in the global-change community. This community models climate change with mathematical representations of earth-system processes. It therefore requires data about the land surface to establish base-line conditions. Global land-cover information is a crucial but dynamic feature of these models. As such, the synoptic view of satellite remote sensing presents an attractive option for high-frequency monitoring of large areas of the earth's surface.