Here’s the abstract of the paper I mentioned in the previous post.
A.Britta K. Sannel and Ian A. Brown. (2010) High-resolution remote sensing identification of thermokarst lake dynamics in a subarctic peat plateau complex. Canadian Journal of Remote Sensing, 36:S26-S40.
Methods to quantify the dynamics of thermokarst lake margins in subarctic permafrost peatlands have been examined using historical aerial photographs and QuickBird imagery from the Hudson Bay Lowlands in west-central Canada, spanning a time period of 52 years (1954–2006). The goal of this study was to develop a method for detection of metre-scale changes in thermokarst lake extent using a time series of high-resolution imagery. The method should be applicable to a variety of lake forms, transferable to other locations, and sufficiently robust as to support different data types. Semi-automatic remote sensing techniques such as unsupervised and supervised classification and texture and high-pass filtering were tested, evaluated, and rejected. According to an experiment of manual digitalization of shorelines by multiple operators, the relative uncertainty for lakes surrounded by peat plateau was ±1.5 m. The uncertainty was reduced to ±0.6 m when binary encoding of transects perpendicular to the shoreline was used to refine the manual delineation. This proved to be the most accurate method to detect small-scale changes in lake extent. An increased understanding and quantification of thermokarst dynamics in permafrost peatlands is important for predicting future scenarios of greenhouse gas emissions from these ecosystems under changing climatic conditions, and our method supports such goals.