REMOTE SENSING DATA

Remote sensing plays a fundamental role in our understanding of spatial patterns and extent of land use and can also aid our understanding of land use drivers. Co-PI Dar Roberts (UCSB), along with graduate student, Michael Toomey, has been active in the production and management of a gapless annual Landsat imagery archive (1984-2010) covering the majority of the state of Rondônia.

At the heart of the land cover monitoring scheme is a standardized protocol of robust classification techniques, as described in Roberts et al. (2002). Landsat 5 and 7 Images are first georeferenced to GeoCover base imagery and radiometrically intercalibrated. We perform spectral mixture analysis, using a four endmember model and decision tree classification. Lastly, multitemporal filters correct and reassign misclassifications.

New corrections have been implemented, compensating for misclassifications due to topography/solar illumination interactions and phenology and will be described in an upcoming manuscript. All finalized remote sensing products, including seven-scene mosaics for 85% of the Rondônia State, are available for download at UCSB.