Abstract
Abstract. Calculating the spatial-temporal distribution of supraglacial debris cover on glaciers is essential for understanding mass balance processes, glacier lake outburst floods, hydrological predictions, and glacier fluctuations that have attracted attention in recent years. However, due to the reflectance of supraglacial debris is similar to that of non-glacier slopes, mapping supraglacial debris cover based on optical remote sensing remains challenging. In this paper, we used NDSI and machine learning algorithm to extract debris cover on glaciers in Hunza Valley, Pakistan. Our result showed that the RF model has the best classification accuracy with kappa coefficient of 0.94 and overall accuracy of 96%. The debris-covered area increased by 21.31% from 1990 to 2019 (394.76 km2 – 478.88 km2) in the study area. Results and the method are of significance in the assessment of meltwater modeling for glaciers with debris cover.
Highlights
Glaciers, known as ‘alpine solid reservoirs’, are a promising natural freshwater resource and a sensitive indicator of global climate change (Kaab et al, 2012; Yang, 1995; Zemp et al, 2019; )
We coded three algorithms for debris cover extent mapping based on the Google Earth Engine (GEE) geospatial big data analysis platform, and we tested them with data on glaciers in the Hunza Valley using Landsat satellite imagery
The main conclusions are i) The random forest (RF) has the best accuracy among the classification algorithms, with a Kappa coefficient of 0.94, an overall accuracy of 96.02%, and determination coefficient (R2) of the linear fit between the manual digitisation of 0.98
Summary
Known as ‘alpine solid reservoirs’, are a promising natural freshwater resource and a sensitive indicator of global climate change (Kaab et al, 2012; Yang, 1995; Zemp et al, 2019; ). Large-scale glacier inventory initiatives mainly include Global Land Ice Measurements from Space (GLMS) (Raup et al, 2007), Randolph Glacier Inventory (RGI) (Pfeffer et al, 2014), Glacier Area Mapping for Discharge from the Asian Mountains (GAMDAM) (Nuimura et al, 2015), and the Second Glacier Inventory of China (Liu et al, 2015). All of these were aimed at generating a global dataset of land glaciers. Due to the reflectance of supraglacial debris is similar to that of non-glacier slopes (Paul et al, 2004) and because there is a lack of continuous, large-scale, high-quality optical images that are not affected by cloud and terrain shadows, mapping the supraglacial debris cover based on remote sensing is challenging compared to mapping clean ice or snow
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call