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.
We present a new digital elevation model (DEM) of the bed, with a 1 km gridding, of the Weddell Sea (WS) sector of the West Antarctic Ice Sheet (WAIS). The DEM has a total area of ∼ 125 000 km2 covering the Institute, Möller and Foundation ice streams, as well as the Bungenstock ice rise. In comparison with the Bedmap2 product, our DEM includes new aerogeophysical datasets acquired by the Center for Remote Sensing of Ice Sheets (CReSIS) through the NASA Operation IceBridge (OIB) program in 2012, 2014 and 2016. We also improve bed elevation information from the single largest existing dataset in the region, collected by the British Antarctic Survey (BAS) Polarimetric radar Airborne Science Instrument (PASIN) in 2010–2011, from the relatively crude measurements determined in the field for quality control purposes used in Bedmap2. While the gross form of the new DEM is similar to Bedmap2, there are some notable differences. For example, the position and size of a deep subglacial trough (∼ 2 km below sea level) between the ice-sheet interior and the grounding line of the Foundation Ice Stream have been redefined. From the revised DEM, we are able to better derive the expected routing of basal water and, by comparison with that calculated using Bedmap2, we are able to assess regions where hydraulic flow is sensitive to change. Given the potential vulnerability of this sector to ocean-induced melting at the grounding line, especially in light of the improved definition of the Foundation Ice Stream trough, our revised DEM will be of value to ice-sheet modelling in efforts to quantify future glaciological changes in the region and, from this, the potential impact on global sea level. The new 1 km bed elevation product of the WS sector can be found at https://doi.org/10.5281/zenodo.1035488.
19 November 2021 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Monitoring properties of ice sheets in polar regions is one of the main challenges in glaciology. There is a large amount of heterogeneous radar data from the polar regions that have been gathered through expensive missions. However, retrieving meaningful information from this large volume of data is still a great challenge. With the advancement of machine learning techniques in recent years, many scientists are eager to take advantage of these algorithms and techniques to explore and mine Arctic and Antarctic data. These advancements, however, have happened mainly in the area of supervised learning where the models are data hungry and require large amounts of annotated data. Generating simulated data can be an effective and inexpensive approach to provide large labeled datasets for training machine learning models. In this work, we explore two approaches to simulate arctic snow radar echogram images, namely a radar scattering physics based approach combined with some statistical measures and a purely data-driven approach based on a conditional generative adversarial network. Using several image comparison metrics, we analyze the utility of both methods for the purpose of simulating echograms. Our results show that the physics simulator generates images with good structural similarities, while the purely data-driven approach achieves better textural similarities for simulated image. Finally, we also show that by augmenting our real dataset by the simulated echograms, we can improve our deep learning model for tracking internal layers of snow.
24 June 2021 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
We developed a portable ultra-wideband radar system capable of reconfigurable operation in multiple frequency bands (separate or simultaneous) spanning from microwaves through millimeter waves. The instrument provides a compact solution for fine-resolution measurements of elevation changes and superficial snow/firn thickness from low-altitude, mid-sized airborne platforms. In this article, we provide an overview of the radar system design and its performance during laboratory testing. We demonstrate its application in aerial surveys of snow layer thickness at S/C bands, dual-band airborne altimetry at Ku-/Ka-bands, and present first-order comparisons with coincident airborne lidar data.
During the concluding phase of the NASA Operation IceBridge (OIB), we successfully completed two airborne measurement campaigns (in 2018 and 2021, respectively) using a compact S and C band radar installed on a Single Otter aircraft and collected data over Alaskan mountains, ice fields, and glaciers. This paper reports seasonal snow depths derived from radar data. We found large variations in seasonal radar-inferred depths with multi-modal distributions assuming a constant relative permittivity for snow equal to 1.89. About 34 % of the snow depths observed in 2018 were between 3.2 and 4.2 m, and close to 30 % of the snow depths observed in 2021 were between 2.5 and 3.5 m. We observed snow strata in ice facies, combined percolation and wet-snow facies, and dry-snow facies from radar data and identified the transition areas from wet-snow facies to ice facies for multiple glaciers based on the snow strata and radar backscattering characteristics. Our analysis focuses on the measured strata of multiple years at the caldera of Mount Wrangell (K'elt'aeni) to estimate the local snow accumulation rate. We developed a method for using our radar readings of multi-year strata to constrain the uncertain parameters of interpretation models with the assumption that most of the snow layers detected by the radar at the caldera are annual accumulation layers. At a 2004 ice core and 2005 temperature sensor tower site, the locally estimated average snow accumulation rate is ∼2.89 m w.e. a−1 between the years 2003 and 2021. Our estimate of the snow accumulation rate between 2005 and 2006 is 2.82 m w.e. a−1, which matches closely to the 2.75 m w.e. a−1 inferred from independent ground-truth measurements made the same year. The snow accumulation rate between the years 2003 and 2021 also showed a linear increasing trend of 0.011 m w.e. a−2. This trend is corroborated by comparisons with the surface mass balance (SMB) derived for the same period from the regional atmospheric climate model MAR (Modèle Atmosphérique Régional). According to MAR data, which show an increase of 0.86 ∘C in this area for the period of 2003–2021, the linear upward trend is associated with the increase in snowfall and rainfall events, which may be attributed to elevated global temperatures. The findings of this study confirmed the viability of our methodology, as well as its underlying assumptions and interpretation models.
Ice properties inferred from multi-polarization measurements, such as birefringence and crystal orientation fabric (COF), can provide insight into ice strain, viscosity, and ice flow. In 2008, the Center for Remote Sensing of Ice Sheets (CReSIS) used a ground-based VHF (very high frequency) radar to take multi-channel and multi-polarization measurements around the NEEM (North Greenland Eemian Ice Drilling) site. The system operated with 30 MHz bandwidth at a center frequency of 150 MHz. This paper describes the radar system, antenna configurations, data collection, and processing and analysis of this data set. Within the framework derived from uniaxial ice crystal model, we found that ice birefringence dominates the power variation patterns of co-polarization and cross-polarization measurements in the area of 100 km2 around the ice core site. The phase shift between ordinary and extraordinary waves increases nonlinearly with depth. The ice optic axis lies in planes that are close to the vertical plane and perpendicular or parallel to the ice divide depending on depth. The ice optic axis has an average tilt angle of about 11.6° vertically, and its plane may rotate either clockwise or counterclockwise by about 10° across the 100 km2 area, and at a specific location the plane may rotate slightly counterclockwise as depth increases. Comparisons between the radar observations, simulations, and ice core fabric data are in very good agreement. We calculated the effective colatitude at different depths by using azimuth and colatitude measurements of the c axis of ice crystals. We obtained an average effective c axis tilt angle of 9.6° from the vertical axis, very comparable to the average optic axis tilt angle estimated from radar polarization measurements. The comparisons give us confidence in applying this polarimetric radio echo sounding technique to infer profiles of ice fabric in locations where there are no ice core measurements.