Description

Glaciers are identified as “Essential Climate Variables” by the World Meteorological Organisation as changes in glacier area, elevation and mass are major indicators for climate change. Hence there is a tremendous need for a consistent decade-long time series of annual, seasonal and sub-seasonal glacier surface mass balance (SMB) of all glaciers worldwide as well as regular updates of glacier area (GA) inventories. Until today the most common methods to estimate SMB such as repeat digital elevation model (DEM) differencing, altimetry and gravimetry have only provided snapshots of SMB estimates for limited periods and regions. Other remote sensing (RS) based approaches utilise the correlation between glacier surface states (snow cover (SC), glacier facies (GF), albedo) and glacier surface mass balance (SMB, excluding frontal ablation rates) and can be divided in three sub-types, namely the equilibrium line altitude (ELA) method, the albedo method and the snow-map method [1]. Nevertheless, studies using these techniques have also been restricted to a very limited number of glaciers and short time periods due to the lack of SMB data for the SMB regression (SMB-R).

Objectives and Methodology

There is a long list of glacier surface characteristics which can be mapped from optical and SAR remote sensing. In this research thesis, the student will investigate deep learning techniques for automatically extracting the glacier areas and map glacier facies using temporal series of remotely sensed images. Current methods are capable of mapping glacier outline or calving front, GF, surface dry-to-wet and end of summer snow line (an estimation of the ELA), superimposed ice and firn line altitude, and the evolution of the firn area. From these parameters the onset and length of the melt and cold seasons can be calculated. The student will design a convolutional neural network for delineating the glacier extent and map its surface characteristics.

Further reading

[1] Rabatel et al.: Annual and Seasonal Glacier-Wide Surface Mass Balance Quantified from Changes in Glacier Surface State: A Review on Existing Methods Using Optical Satellite Imagery, Remote Sens., 9(5), 507, doi:10.3390/rs9050507, 2017.