Título del libro: American Society For Photogrammetry And Remote Sensing Annual Conference 2010: Opportunities For Emerging Geospatial Technologies
Título del capítulo: Classifying the land cover of Mexico in the framework of the North American Land Change Monitoring System
RENE ROLAND COLDITZ;
Autores externos:
Idioma:
Inglés
Año de publicación:
2010
Palabras clave:
Ancillary data; Automatic derivation; Automatically generated; Biodiversity analysis; Biogeochemical modelling; Boosted decision trees; Change detection; Commission for environmental cooperations; Global change; Global models; Land cover; Land transformation; Me-xico; Monitoring system; Multiple Classification; North American; Optical satellites; Research topics; Sample data; Study areas; Tropical ecosystems; Biodiversity; Decision trees; Forestry; Monitoring; Photogrammetry; Remote sensing; Landforms; Biodiversity; Decision Making; Forestry; Monitoring; Photogrammetry; Remote Sensing; Trees
Resumen:
Automatic derivation of land cover information from optical satellite data is one of the main research topics in remote sensing. Accurate land cover products are needed for biogeochemical modelling and biodiversity analysis. Land transformation processes are analyzed in the framework of global change studies. While many global land cover products fulfil the needs for driving global models, their accuracy is insufficient for regional to continental applications. The North American Land Change Monitoring System is a tri-national initiative to provide accurate, automatically generated annual land cover and land change products. Its first product, the land cover map of North America for 2005, has recently been completed and is published by the intergovernmental Commission for Environmental Cooperation. The paper presents the classification for the Mexican section. Although the smallest, it is without doubt the most complex portion of the study area due to the transition of temperate and tropical ecosystems and a very heterogeneous small-patch land cover assemblage. The classification is based on monthly MODIS composites and ancillary data. A large sample data base was built to train a multitude of boosted decision trees. The combination of multiple classifications assures map consistency, an important step to mitigate false change detection in the future. Class memberships of the decision trees were transformed to a discrete map, which is accompanied by a confidence layer. The discrete map assessment yielded 82 %.
Entidades citadas de la UNAM:
ISBN:
9781617389160
Editorial:
Tipo de producto:
Conference Paper
