Abstract
Burned area detection with remotely sensed satellite data in agricultural landscapes is not only necessary for the estimation of global biomass burning emissions, but also has gained attention from managers interested in improved methods for the quantification of local scale emissions which affect air quality and human health. Mapping agricultural burned areas accurately, precisely and reliably, with methods that can be applied globally, is difficult because of the spectral and temporal characteristics of agricultural regions and prescribed cropland fires. These challenges have not been fully addressed by the scientific literature. Chapter 1 of this thesis presents an extensive literature review on the methods currently used for agricultural burned area mapping. Chapter 2 presents original research on the spectral characterization of agricultural burned areas, using field data and mixture models to analyze the response of spectral indices to the changes induced by fire and agricultural practices. The conclusions summarize the significance of the presented research for understanding the potential and limits of satellite data for agricultural burned area monitoring, and outline the directions for future work.