LULC - Special topic

Land Use and Land Cover II

6207.1 - New global map of land cover pattern types based on the ESA CCI-LC data using a segmentation-classification technique

Thursday, July 6
10:30 AM - 10:50 AM
Location: Maryland B

Global land cover (GLC) map products support many studies involving the Earth surface, such as climate, ecology, hydrology, soil erosion, and atmospheric quality. The most recent GLC product is the 22-classes GLC raster map of the entire terrestrial land surface at 300 m resolution developed under the auspices of the ESA Climate Change Initiative (CCI). This fine-scale map provides a detailed description of land cover at local sites worldwide, but often, it is the mapping of spatial patterns of LC categories rather than mapping the categories themselves that provides the key insights. Although patterns are sometimes easy to see on a LC map, their delineation to produce a world-wide broad-scale map of such patterns (referred to as landscape pattern types or LPTs) is hard.
Historically, such broad-scale maps were derived manually, but the CCI-LC is just too large for manual mapping. This contribution describes the process of automatic worldwide mapping of LPTs from the CCI-LC raster. The resulting map is unique inasmuch as it identifies and delineates, for the first time, characteristic patterns of LC over the entire terrestrial landmass.
First, the 8-billion cells LC raster is transformed into a regular grid of square-sized blocks of its cells called motifles. Motifels are elementary units of LC pattern. Motifel’s pattern is characterized by the co-occurrence histogram of its constituent cells. A degree of dissimilarity between a pair of motifels is calculated using the Jenson-Shannon divergence between their corresponding histograms. Second, the grid of motifels is segmented using a custom algorithm in such a way as to optimize homogeneity of LC pattern within segments. Segmentation delineates the local patterns of LC but there are too many of them to yield a lucid map. Nevertheless, the segmentation vector provides a searchable database of all existing local patterns and can be used to identify all locations with a given pattern of LC. We create a global map of LPTs by clustering the segments using the hierarchical clustering algorithm (with Ward linkage). LPTs are highly generalized, their patterns are much less uniform than patterns within local segments but they are sufficiently uniform to constitute characteristic types of pattern.
The character of the map depends on the scale of the local pattern (the size of motifel). We choose 30 km (motifel of 100 x 100 cells) as the scale of the local pattern. With such choice the segmentation algorithm yielded 10,129 segments. After examining several possibilities we clustered these segments into 32 types. Because we used hierarchical clustering these types can be divided into 11 groups: evergreen broadleaved trees, deciduous broadleaved trees, evergreen needleleaved trees, deciduous needleleaved trees, grasslands, croplands, shrublands, sparse vegetation, bare land, permanent snow and ice, and water. The names of these groups indicate major component of patterns in their LPTs; there are from one to six LPTs different pattern types in a group each featuring named-after category of land cover. For example, the grassland group includes the following patterns: grassland matrix, grassland/cropland mosaic, grassland/forest mosaic, grassland/bare land mosaic, and grassland/shrub mosaic. The complete map can be downloaded from http://sil.uc.edu/cms/data/uploads/pdf/jakub/LandCoverPatternTypesMap-ICC2017.pdf. This map aims at supporting macroecologic applications, global environmental assessments, environmental education and many other, similar tasks.

Jakub Nowosad

Postdoc
University of Cincinnati

Dr. Jakub Nowosad is a postdoctoral fellow at the Space Informatics Lab. During his PhD he had worked on spatiotemporal predictions of pollen concentration of Corylus, Alnus, and Betula using machine learning and GIS. Currently, his research is focused on design and implementation of methods for segmentation and clustering of ecoregions. Jakub's areas of interest also include spatial analysis, statistics, and programming. He has an extensive teaching experience in the fields of spatial analysis, geostatistics, statistics, and machine learning.

Presentation(s):

Send Email for Jakub Nowosad

Tomasz Stepinski

University of Cincinnati

Dr. Stepinski is the Thomas Jefferson Chair Professor of Space Exploration at the University of Cincinnati.

Presentation(s):

Send Email for Tomasz Stepinski

Ammatzia Peled

Prof.
University of Haifa

Dr. Ammatzia Peled is a professor for GIS&RS at the University of Haifa, Israel. Ammatzia Served as the Chair of the ICA Commission on Incremental Updating and Versioning of spatial Databases. He served also as ISPRS Treasurer and Second Vice President and as ISPRS president of Commission VIII. In 2010, Prof. Peled was awarded the Eduard Tsiolkovsky Memorial GOLD Medal by the Russian Academy for Cosmonautics for “Outstanding contribution to Cosmonautics”. In 2013 he was awarded as a Professor Honoris Causa by the Siberia State Academy for Geodesy & Cartography and last July he was awarded as ISPRS Fellow.

Presentation(s):

Send Email for Ammatzia Peled


Assets

6207.1 - New global map of land cover pattern types based on the ESA CCI-LC data using a segmentation-classification technique

Slides

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for New global map of land cover pattern types based on the ESA CCI-LC data using a segmentation-classification technique