Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Etats Unis → 2022 → Extending the Timeline of Land Cover Data to Refine Our Understanding of Grassland Degradation in Northern China

George Washington University (2022)

Extending the Timeline of Land Cover Data to Refine Our Understanding of Grassland Degradation in Northern China

Iacone Brooke

Titre : Extending the Timeline of Land Cover Data to Refine Our Understanding of Grassland Degradation in Northern China

Auteur : Iacone Brooke

Université de soutenance : George Washington University

Grade : Master of Science (MS) 2022

Résumé
Growing demand for resources from China’s grasslands and the impacts of climate change continue to threaten the long-term health of these vital ecosystems. High-resolution CORONA imagery acquired by the United States through spy missions in the 1960’s may provide critical insight into historic land cover conditions in prior decades, particularly in regions such as northern China where data from that era are scarce. Since the 1960’s, the region has implemented many policy changes that have led to periods of intensified cultivation, grassland degradation and desertification and, in recent years, restoration. Most large-scale studies of land cover change only extend back to the 1980’s and the advent of moderate resolution remotely sensed data. This timeline limits our ability to attribute changes prior to the 1980s or establish baseline conditions. CORONA imagery presents an opportunity to expand the timeline of available data on land use and land cover change in northern China. Here, we evaluate the feasibility and accuracy of using derived texture and spatial contextual features from CORONA imagery for classification. In this study, four CORONA images were georeferenced using ground control points and Landsat TM imagery. Then, contextual features based on image texture characteristics were derived at several scales from a mosaic of the georeferenced images and combined with a Landsat MSS NDVI to create a multi-band image. We used an unsupervised clustering algorithm, paired with cluster statistics to derive a classified map of historic land cover containing eight classes, including three distinct dune types. The results show that contextual features derived from Corona imagery are useful for extracting land cover types in far greater detail than other available datasets. This study represents the first land cover assessment of northern China for this time frame and will be valuable for future studies of land cover change.

Mots Clés  : Remote Sensing Corona Imagery Contextual Features Grassland Degradation Northern China Drylands

Présentation

Version intégrale

Page publiée le 18 décembre 2022