Risks of cryogenic landslide hazards and their impact on ecosystems in cold environments
Résumé
Research area: The research area is geographically located on the Bovanenkovo region, the northwestern part of Yamal Peninsula, West Siberia, Russia. Yamal Peninsula is a flat homogeneous lowland region with low-lying plains of heights lower than 90m. Such geographic settings create specific local environmental conditions in the region. Thus, Yamal is the worlds largest high-latitude wetland system covering in total 900,000 km² of peatlands, complex system of wetlands, dense lake and river network. Typical for this region are seasonal flooding, active erosion processing, permafrost distribution and intensive local landslides. Dominating vegetation types are typical tundra species: heath, grasses, moss, and lichens), and woody plants (shrubs and willows. the main environmental problems of this region are climate change and active landsliding, which affect on the landscapes and cause changes in vegetation coverage, i.e. land cover types. Research problem: The specific environmental problem of Yamal region is cryogenic landslides. The processes of superficial cryogenic landslides are especially active in tundra. Distribution of the permafrost, which serves as a shear surface for sliding, highly contributes to the landslide formation in this area. Cryogenic landslides developed on the fine-grained, saline marine sediments are common destructive disastrous geomorphological hazards on the Yamal Peninsula covering ca 70% of the area).
Methods: The research main method consists in Landsat TM image classification, spatial analysis and thematic mapping, technically performed in ILIWIS GIS software. The choice of Landsat scenes for land cover mapping is explained by their well-known advantages of application in geosciences and cartography, almost 40 year old history of the image record, and free availability. The images have a time span of 23 years: 1988 and 2011, taken in growing season, i.e. summer to early autumn, with clearly visible vegetation coverage.
The research workflow.
1. Data capture and converting import of .img file into ASCII raster format (GDAL). After converting, each image contained collection of 7 Landsat raster bands.
2. Pre-processing: Enhancement of visual color and contrast. Geographic referencing of Landsat scenes included setting of UTM projection (Universal Transverse Mercator), Eastern Zone 42, Northern Zone W, WGS 1984 datum (Georeference Corner Editor).
3.Creating Area Of Interest (AOI). The area of interest (AOI) was identified and cropped on the raw images This area shows Bovanenkovo region in a large scale which best represents typical tundra landscapes.
4.Supervised Classification. The images were classified by Minimal Distance method. This method is based on the spatial analysis of spectral signatures of object variables, i.e. vegetation types at various landscapes. The classes sampling was performed using Sample Set tool in ILWIS GIS. Training pixels for each land cover type were selected as representative samples and stored as classification key. They have contrasting colors, visually visible and distinguishable on the image.
Land Cover Types: The defined classes include following landscapes types: shrub tundra, willows, tall willows, short shrub tundra, sparse short shrub tundra, dry grass heath, sedge grass tundra, dry short shrub tundra, dry short shrub sedge tundra, wet peatland, peatland (sphagnum). The pixels were associated with land cover classes using their digital numbers, similar to key samples.
Results: Willows covers 2750,57 ha in 2011, which is more than in 1988, when it covered 1547,52 ha (both ’tall willows’ and ’willows’ classes). Noticeable is increase in tundra vegetation: ’short shrub tundra’, ’sparse short shrub tundra’ and ’dry short shrub tundra’ have more areas covered in 2011 comparing to 1988: almost 5442,00 ha vs 1823,00 ha. Increase of wooden vegetation class goes along with shrunk of grass and heath areas: ’dry grass heath’ occupied area of 3335.39 ha in 1988, while currently it covers 1204.94 ha. Slight decrease can be noticed in the ’peatlands’ and ’wet peatlands’ classes: 3958.40 ha against 2765.41 ha in 2011 by ’wet peatlands’, and 625.71 ha in 1988 versus 488.69 ha by ’peatland (sphagnum)’ class.
Domaines
Vision par ordinateur et reconnaissance de formes [cs.CV] Biodiversité et Ecologie Ordinateur et société [cs.CY] Milieux et Changements globaux Synthèse d'image et réalité virtuelle [cs.GR] Interface homme-machine [cs.HC] Recherche d'information [cs.IR] Modélisation et simulation Traitement des images [eess.IV] Traitement du signal et de l'image [eess.SP]
Origine : Fichiers produits par l'(les) auteur(s)