Change detection of permafrost thaw dynamics across resolutions: From low to high spatial and temporal observations on the Bykovsky Peninsula, Lena Delta, East Siberia

• Main Authors: Günther, Frank, Grosse, Guido, Overduin, Paul, Veremeeva, Alexandra, Grigoriev, Mikhail N.
• Format: Conference Object
• Language: unknown
• Published: 2016
• Subjects: Ice; lena delta; permafrost; Thermokarst; Siberia
• Online Access: https://epic.awi.de/id/eprint/40888/; https://epic.awi.de/id/eprint/40888/1/LPS16_poster_Guenther.pdf; http://lps16.esa.int/page_session185.php#2727p; https://hdl.handle.net/10013/epic.47885; https://hdl.handle.net/10013/epic.47885.d001

Now there are initial indications that ground ice in permafrost is thawing in response to rising temperatures in the Arctic. However, particularly in the light of the enormous area underlain by ice-rich continuous permafrost, still only few observations of permafrost-thaw related landscape dynamics exist. Thermokarst lake development, active layer detachments, widespread and irreversible land surface subsidence, and coastal erosion, all of these processes taken together present a sketch of current permafrost degradation with consequences for local hydrology, ecosystems, biogeochemical cycling, and sometimes communities. In this study, we focus not only on monitoring these processes, but also aim to develop a best practice strategy for remote sensing image data fusion, in order to combine 2D and 3D information from very high resolution (VHR) imagery of low acquisition frequency (GeoEye, WorldView, KOMPSAT, ALOS PRISM) with high temporal resolution data such as from the RapidEye satellite constellation. Local field measurements (meteorology, ground temperature, geodetic surveys) during several recent Russian-German Arctic expeditions complement our remote sensing studies and help differentiating factors causing relief and land cover changes. Our work aims at finding commonalities and differences of change or no change on yedoma uplands, slopes, and thaw depressions on the landscape scale using multi-temporal DEMs from historical aerial photographies, modern stereo imagery, and on-site geodetic surveys. This ground truth data provides the basis for calibration and ortho-correction of simultaneously acquired VHR data. More than 30 RapidEye scenes (level 1B) from 2014 to 2015 were adjusted using a bundle block adjustment procedure. Incoming raw data is constantly included only via common Tie Points, minimising processing efforts. Band metrics such as NDVI and PCA-analyses of multi-temporal image composites were used to differentiate seasonal from inter-annual processes and to identify signatures and trends typical ...