Comparison of different remote sensing methods for glacier mapping in Afghanistan

Glaciers are important sources of fresh water particularly in arid regions which have low summer precipitation. Moreover, retreating glaciers can cause serious hazards by destabilizing slopes or causing outbursts of glacial lakes. Therefore glacier monitoring is an essential task for water resources and risk management. Recently, efforts have been made to monitor glaciers using manual or semi-automated remote sensing techniques. However a particular challenge remains: as glaciers retreat they commonly develop a surface debris layer that optically is similar to zones that have not been glaciated or that are truly deglaciated: the debris cover on the glacier surface has a similar reflectance to surrounding moraines in the visible to near-infrared wavelength region. In other hand, where debris cover develops, it may insulate ice from solar radiation and diurnal temperature rises, and this will also reduce melt. Therefore, debris cover on glacier boundaries critically hinders the global inventory of glaciers. To overcome the challenges this study uses a multiple band ratio approach. The method was tested for delineating three glaciers in Afghanistan at different scales and locations to map both clean ice and debris-covered ice. We used Landsat Enhanced Thematic Mapper Plus, and a 5-meter resolution digital surface model DSM data to extract the morphological parameters. Since clean glacier ice has a high reflectivity in the visible to near-infrared wavelengths, at first we used NDIS to extract the clean ice area, but It was found that the NDSI method for glacier mapping is less sensitive to cast shadows and steep terrain. Similarly, a slope parameter has tested to map the debris cover ice area but it did not map areas with gentle slopes correctly.

Nonetheless, NIR and SWIR were identified as potential candidates for distinguishing between glaciers in shade and clean ice for the debris free case; and a combination of those bands in three different ratios and thresholds was applied successfully (Red/SWIR>= 1.5, Pan/SWIR>0.1, and NIR/SWIR>1). With regards to debris-covered ice the thermal infrared bands show potential in resolving such ambiguity, as considerable temperature differences are found to exist between debris covered ice and surrounding moraines. However, we found that thermal infrared bands have too coarse a resolution (60m) for valley glaciers. Hence, we developed a new band ratio image combining thermal infrared and panchromatic bands to better distinguish periglacial debris and supraglacial debris. This new band ratio image is given by (PAN-TIR)/(PAN+TIR), and is named as normalized supraglacial debris index (NSDI).

Accuracy assessment was carried out through comparisons of the classified maps with a manual delineation done using 1-meter high resolution RGB image with same temporal resolution. The accuracy assessment shows that the results from the proposed method are in good agreement with the manual delineation. The proposed synergistic approach therefore appears useful in the accurate mapping of debris-covered glaciers in Afghanistan.