The methodology for selection, creation, and application of a global remote sensing validation dataset using high resolution commercial satellite data is presented. High resolution data are obtained for a stratified random sample of 500 primary sampling units (5 km × 5 km sample blocks), where the stratification based on Köppen climate classes is used to distribute the sample globally among biomes. The high resolution data are classified to categorical land cover maps using an analyst mediated classification workflow. Our initial application of these data is to evaluate a global 30 m Landsat-derived, continuous field tree cover product. For this application, the categorical reference classification produced at 2 m resolution is converted to percent tree cover per 30 m pixel (secondary sampling unit)for comparison to Landsat-derived estimates of tree cover. We provide example results (based on a subsample of 25 sample blocks in South America) illustrating basic analyses of agreement that can be produced from these reference data. Commercial high resolution data availability and data quality are shown to provide a viable means of validating continuous field tree cover. When completed, the reference classifications for the full sample of 500 blocks will be released for public use.
|Title||A global reference database from very high resolution commercial satellite data and methodology for application to Landsat derived 30 m continuous field tree cover data|
|Authors||Bruce Pengra, Jordan Long, Devendra Dahal, Stephen V. Stehman, Thomas R. Loveland|
|Publication Subtype||Journal Article|
|Series Title||Remote Sensing of Environment|
|Record Source||USGS Publications Warehouse|
|USGS Organization||Earth Resources Observation and Science (EROS) Center|