Review of glacier velocity and facies characterization techniques using multi-sensor approach

Glacier velocity estimation and facies characterization in glaciology using remote sensing have been widely used to understand ice dynamics, improve climate modeling, assess hazards, and study the impact of climate change on glaciers. This paper critically reviews the remote sensing techniques, including satellite-based sensors and ground-based measurements, in combination with deep learning algorithms to analyze glacier dynamics and understand the distribution of different facies types within the Himalayan glacier system. A multi-sensor time series data of C-band SAR data (RISAT-1, Sentinel-1) with an object-oriented classification system are used to detect isolated glacier facies such as percolation facies, icefalls, bare ice facies, re-freezing snow, and supraglacial detritus, which is crucial for studying glacier dynamics and assessing their response to climate change. Various deep learning algorithms, such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) and ANN algorithms were reviewed for automated classification and mapping of different facies types present in Himalayan glaciers and to estimate glacier surface velocity. We examined all relevant publications in the field of glacier velocity estimation using remote sensing data of 36,722 glaciers between 1999 and 2018; of these, 32% experienced acceleration, 24.5% experienced deceleration, and 43.5% experienced stability. According to the results of earlier studies of eastern Himalayas, it has been observed an average decline in glacier surface velocity from 15.7 m/yr. in 1994/96 to 12.88 m/yr. in 2018/2020, indicating a decrease of approximately 15% during the study period and the NW Himalayan glacier surface velocity is decreasing from 1994 to 2020, indicating a loss of ice mass in the glacier. In this paper, we review the existing methods of Himalayan glacier surface velocity estimation and the advancements achieved through the integration of multi-sensor data and deep learning algorithms in improving our understanding of glacier dynamics in the Himalayas. However, these studies are focused on 2D study of glacier movement. An attempt has been made to study 3D glacier flow velocity of the Himalayan glacier using SAR interferometry and Offset tracking measurement techniques integrating with MAI (multiple aperture interferometric). The D-InSAR and MAI methodologies can measure displacements and structural deformation in the LOS direction and azimuth orientations, respectively, with an accuracy of a few millimeters to a few centimeters, but are limited by coherence loss, which would be easily affected by the large temporal baseline of SAR data sets or glacier velocity. We considered that the future research trend is to explore the integration of data from multiple sources, such as satellite imagery, field observations, and modeling, to improve the accuracy and resolution of glacier velocity estimation and facies characterization and understand the complex relationships between glacier velocity, facies characteristics, and environmental factors.