Glacial lake monitoring in the Karakoram Range using historical Landsat Thematic Mapper archive (1982 – 2014)

Abstract:

Glacierized regions are one of the most dynamic land surface environments on the planet (Evans and Delaney, In Press). They are susceptible to various types of natural hazards such as landslides, glacier avalanches, and glacial lake outburst floods (GLOF). GLOF events are increasingly common and present catastrophic flood hazards, the causes of which are sensitive to climate change in complex high mountain topography (IPCC, 2013). Inundation and debris flows from GLOF events have repeatedly caused significant infrastructure damages and loss of human lives in the high mountain regions of the world (Huggel et al, 2002).

The research is designed to develop methods for the consistent detection of glacier lakes formation during the Landsat Thematic Mapper (TM) era (1982 – present), to quantify the frequency of glacier lake development and estimate lake volume using Landsat imagery and digital elevation model (DEM) data.

Landsat TM scenes are used to identify glacier lakes in the Shimshal and Shaksgam valley, particularly the development of Lake Virjeab in year 2000 and Kyagar Lake in 1998. A simple thresholding technique using Landsat TM infrared bands, along with object-based segmentation approaches are used to isolate lake extent. Lake volume is extracted by intersecting the lake extent with the DEM surface. Based on previous studies and DEM characterization in the region, Shuttle Radar Topography Mission (SRTM) DEM is preferred over Advanced Spaceborne Thermal Emission and Reflection (ASTER) GDEM due to higher accuracy. Calculated errors in SRTM height estimates are 5.81 m compared with 8.34 m for ASTER. SRTM data are preferred because the DEM measurements were made over short duration making the DEM internally consistent. Lake volume derived from the Landsat TM imagery and DEM are incorporated into a simple GLOF model identified by Clague and Matthews (1973) to estimate the potential peak discharge (Q_max) of a GLOF event. We compare the simple Q_max estimates with those from a more complex model of lake outflow time-varying discharge using the approach developed by Ng et al. (2007).