Deep Learning on Airborne Radar Echograms for Tracing Snow Accumulation Layers of the Greenland Ice Sheet

Abstract

Climate change is extensively affecting ice sheets resulting in accelerating mass loss in recent decades. Assessment of this reduction and its causes is required to project future ice mass loss. Annual snow accumulation is an important component of the surface mass balance of ice sheets. While in situ snow accumulation measurements are temporally and spatially limited due to their high cost, airborne radar sounders can achieve ice sheet wide coverage by capturing and tracking annual snow layers in the radar images or echograms. In this paper, we use deep learning to uniquely identify the position of each annual snow layer in the Snow Radar echograms taken across different regions over the Greenland ice sheet. We train with more than 15,000 images generated from radar echograms and estimate the thickness of each snow layer within a mean absolute error of 0.54 to 7.28 pixels, depending on dataset. A highly precise snow layer thickness can help improve weather models and, thus, support glaciological studies. Such a well-trained deep learning model can be used with ever-growing datasets to aid in the accurate assessment of snow accumulation on the dynamically changing ice sheets.

Review article: Feature tracing in radio-echo sounding products of terrestrial ice sheets and planetary bodies

“…In most other featureextraction and image-segmentation applications in glaciology, the features and edges (if multiple are present) are usually not located at such short distances from each other as IRHs could be, for example in glacier grounding lines delineation (Mohajerani et al, 2021). This remains a complex task for any algorithm to detect different features as close as a few pixels from each other and not merge them when there is no merging to be done, but to separately trace them in case of discontinuities, which 965 are another phenomenon that obstructs mapping attempts (Varshney et al, 2021b). They could emerge as a result of tracing algorithm's shortcomings and inefficient tracing capabilities.…”

“…Further work on producing synthetic data with higher quality that could represent real data more realistically might be a suitable solution. Varshney et al (2021b) also used FCN to trace snow layer boundaries. This publication is a combination of Varshney et al (2020) and Rahnemoonfar et al (2021), also employing the same architectures as the latter.…”

“…CNNs have been employed in a variety of fields such as computer vision (Zhao et al, 2024), speech and language processing (Pham et al, 2016), face recognition (Coskun et al, 2017), object detection (Galvez et al, 2018), to name a few. In recent years, CNNs applications have been expanded to the field of glaciology as well, for instance in calving front delineation using synthetic aperture radar (SAR) imagery (Mohajerani et al, 2019;Zhang et al, 2019), grounding line delineation (Mohajerani et al, 2021), and automatic stratigraphy mapping (e.g., Varshney et al, 2021b;Cai et al, 2022;Wang et al, 2020;Donini et al, 2022c).…”

Automatic Extraction of Martian Subsurface Layer from Radargrams Based on PDE Denoising and KL Filter

“…The false detection rate f R and the missed detection rate m R were used for the evaluation [28]. When evaluating the accuracy of subsurface reflector detection algorithms, it is common to use datasets with known labels for validation [26]. However, this method was not applicable due to the lack of standard labeled data for Martian polar subsurface layers.…”

VQA-Aid: Visual Question Answering for Post-Disaster Damage Assessment and Analysis

“…With the advancement of many deep learning techniques, many challenges have been efficiently addressed in recent times regarding polar ice sheet research. Those researches mainly focused on estimating ice-layer thickness [1], ice-layer tracking [2], physics-driven Deep learning simulation for generating ice-imagery [3], etc. However, more different types of information can be extracted from the ice sheet data that are very useful for scientists to have a better understanding.…”