Quantification and Change Assessment Benjamin Aubrey Robson 2016 Dissertation date: 31st October 2016 of Debris-Covered Glaciers using Remote Sensing
Not peer reviewed
MetadataShow full item record
This thesis investigates how remote sensing data can be used to assess the changing state of debris-covered ice. The principal study areas are the Manaslu Region in Nepal (papers I and III) and the Hohe Tauern National Park, Austria (paper II). Clean glacier ice is straightforward to semi-automatically classify using multi-spectral satellite imagery owing to the strong spectral signature of clean ice in the visible and near-infrared sections of the electromagnetic spectrum. Since the ablation zones of clean ice glaciers are at the pressure melting point, a change in terminus position or glacier area can be directly linked to a change in climate. Debris-covered ice is however more complicated to map and to interpret temporal change. Supraglacial debris is spectrally indistinguishable from the surrounding paraglacial terrain, and requires auxiliary data such as a Digital Elevation Model (DEM), thermal band data, or flow data. Object-Based Image Analysis (OBIA) provides a framework for combining multiple datasets in one analysis, while additionally allowing shape, contextual, hierarchical and textural criteria to be used to classify imagery.
Paper I combines optical (Landsat-8), topographic (void-filled SRTM) and SAR coherence (ALOS PALSAR) data within an OBIA workflow to semi-automatically classify both clean ice and debris-covered ice in the challenging area surrounding Mount Manaslu in Nepal. When compared with manually delineated outlines, the classification achieved an accuracy of 91% (93% for clean ice and 83% for debriscovered ice). The classification was affected by seasonal snow and shadows while the debris-covered ice mapping was influenced by the datasets being temporally inconsistent, and the mountainous topography causing inconsistencies in the SAR coherence data. The method compares well with other automated techniques for classifying debris-covered ice, but has two additional advantages: firstly, that SAR coherence data can distinguish active ice from stagnant ice based on whether motion or significant downwasting has occured, and secondly, that the method is applicable over a large study area using just space-borne data.
Paper II explores the potential of using high-resolution (10 m) topographic data and an edge detection algorithm to morphologically map the extent of debris-covered ice. The method was applied in the Hohe Tauern National Park, Austria, using a 10 m DEM derived from airborne Light Detection and Radar (LiDAR) acquisitions. Additionally, the end-of-summer transient snowline (TSL) was also mapped, which approximates the annual Equilibrium Line Altitude (ELA).
Our classification was applied on three Landsat satellite images from 1985, 2003 and 2013 and compared the results to the Austrian Glacier Inventories from 1969 and 1998 to derive decadal-scale glacial changes. A mean rate of glacier area reduction of 1.4 km2a-1 was calculated between 1969 and 2013 with a total reduction in area of 33%. The TSL rose by 92 m between 1985 and 2013 to an altitude of 3005 m. By comparing our results with manually delineated outlines an accuracy of 97.5% was determined. When a confusion matrix was calculated it could be seen that the results contained few false positives but some false negatives which were attributed to seasonal snow, shadows and misclassified debris. Our results correspond broadly with those found in other areas of the European Alps although a heterogeneity in glacier change is observable. We recommend that future glacier mapping investigations should utilise a combination of both SAR coherence data and high-resolution topographic data in order to delineate the extent of both active and stagnant glacier ice.
Paper III investigates decadal scale changes in glacier area, velocity and volume in the previously undocumented Manaslu Region, Nepal. Between 2001 and 2013 the glacier area reduced by 8.2% (-0.68% a-1). Simultaneously, the glaciers lowered by -0.21 ± 0.08 m a-1 and had a slightly negative specific mass balance of -0.05 ± 016 m w.e a-1 although mass balances ranged -2.49 ± 2.24 to +0.27 ± 0.30 m w.e a-1 throughout the region. The geodetic mass balance for select glaciers covered by a Corona DEM between 1970 and 2013 was -0.24 ± 0.12 m w.e a-1 which became more negative (-0.51 ± 0.12 m w.e. a-1) between 2005 and 2013. Rates of surface lowering over debriscovered ice increasing by 168% between 1970 – 2000 (0.40 ± 0.18 m a-1) and 2005 – 2013 (1.07 ±0.48 m a-1). The rate of glacier melt varies due to presumed increases in debris thickness at the upper and lower boundaries of the ablation zone, while an area of enhanced glacier downwasting corresponds to the presence of supraglacial lakes and exposed ice.
The glacier velocity varies across the region. Many glaciers have stagnant sections towards the glacier termini, and a trend of ongoing stagnation is observable. No relationship exists between trends in glacier area and glacier volume or velocity, although a weak relationship exists between trends in the changes of volume and velocity. The rates of glacier area and velocity change appear to be similar, although the number of glaciers that had records of area, velocity, and volume was few. Our results are comparable to studies looking at mean surface lowerings and geodetic mass balances in other areas of the Himalayas, and point towards heterogeneous yet pronounced mass losses across the Himalaya region.