Remote Sensing of Rapid Permafrost Landscape Dynamics
2017
The global climate is warming and the northern high latitudes are affected particularly rapidly. Large areas of this region, or 24% of the northern hemisphere, are influenced by perennially frozen ground or permafrost. As permafrost is predominantly dependent on cold mean annual air temperatures, climate warming threatens the stability of permafrost. Since large amounts of organic carbon are stored within permafrost, its thaw would potentially release large amounts of greenhouse gases, which would further enhance climate warming (permafrost carbon feedback).
Thermokarst and thermo-erosion are an indicator of rapid permafrost thaw, and may also trigger further disturbances in their vicinity. The vast Arctic permafrost regions and the wide distribution of thaw landforms makes the monitoring of thermokarst and thermo-erosion an important task to better understand the response of permafrost to the changing climate. Remote sensing is a key methodology to monitor the land surface from local to global spatial scales and could provide a tool to quantify such changes in permafrost regions. With the opening of satellite archives, advances in computational processing capacities and new data processing technology, it has become possible to handle and analyze rapidly growing amounts of data.
In the scope of the changing climate and its influence of permafrost in conjunction with recent advances in remote sensing this thesis aims to answer the following key research questions:
1. How can the extensive Landsat data archive be used effectively for detecting typical land surface changes processes in permafrost landscapes?
2. What is the spatial distribution of lake dynamics in permafrost and which are the dominant underlying influencing factors?
3. How are key disturbances in permafrost landscapes (lake changes, thaw slumps and fire) spatially distributed and what are their primary influence factors?
To answer these questions, I developed a scalable methodology to detect and analyze permafrost landscape changes in the ~29,000 km2 Lena Delta in North-East Siberia. I used all available peak summer data from the Landsat archive from 1999 through 2014 and applied a highly automated robust trend-analysis based on multi-spectral indices using the Theil-Sen algorithm. With the trends of surface properties, such as albedo, vegetation status or wetness, I was able identify local scale processes, such as thermokarst lake expansion and drainage, river bank erosion, and coastal inundation, as well as regional surface changes, such as wetting and greening at 30m spatial resolution. This method proved to be robust in indicating typical landscape change processes within an Arctic coastal lowland environment dominated by permafrost, which has been challenging for the application of optical remote sensing data. The scalability of the highly automated processing allows for further upscaling and advanced automated landscape process analysis.
For a targeted analysis of well-known disturbances affecting permafrost (thermokarst lakes, retrogressive thaw slumps and wildfires), I used advanced remote sensing and image processing techniques in conjunction with the processed trend data. Here I combined the trend analysis with machine-learning classification and object based image analysis to detect lakes and to quantify their dynamics over a period from 1999 through 2014 within four different Arctic and Subarctic regions in Alaska and Siberia totaling 200,000 km². I found very strong precipitation driven lake expansion (+48.48 %) in the central Yakutian study area, while the study areas along the Arctic coast showed a slight loss of lake area (Alaska North Slope: -0.69%; Kolyma Lowland: -0.51%) or a moderate lake loss (Alaska Kobuk-Selawik Lowlands: -2.82%) due to widespread lake drainage. The lake change dynamics were characterized by a large variety of local dynamics, which are dependent on several factors, such as ground-ice conditions, surface geology, or climatic conditions.
In an even broader analysis across four extensive north-south transects covering more than 2.3 million km², I focused on the spatial distribution and key factors of permafrost region disturbances. I found clear spatial patterns for the abundance of lakes (predominantly in ice-rich lowland areas), retrogressive thaw slumps (predominantly in ice-rich, sloped terrain, former glacial margin), and wildfires (boreal forest). Interestingly, apart from frequent drainage at the continuous-discontinuous permafrost interface, lake change dynamics showed spatial patterns of expansion and reduction that could not be directly related to specific variables, such as climate or permafrost conditions over large continental-scale transects. However, specific variables could get related to specific lake dynamics in within locally defined regions.
Trend datasets of vegetation status (NDVI) were combined with high-resolution detailed geomorphological land-cover classification information and climate data to map tundra productivity in a heterogeneous landscape in northern Alaska. After decades of increasing productivity (greening), recently tundra vegetation showed a reverse trend of decreased productivity, which is predicted to continue with increasing temperatures and precipitation.
In this thesis project I developed methods to analyze rapid landscape change processes of various scales in northern high latitudes with unprecedented detail by relying on spatially and temporally high resolution Landsat image time series analysis across very large regions. The findings allow a unique and unprecedented insight into the landscape dynamics of permafrost over large regions, even detecting rapid permafrost thaw processes, which have a small spatial footprint and thus are difficult to detect. The multi-scaled approach can help to support local-scale field campaigns to precisely prepare study site selection for expeditions, but also pan-arctic to global-scale models to improve predictions of permafrost thaw feedbacks and soil carbon emissions in a warming climate.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
0
Citations
NaN
KQI