A Snow Properties-Aware Deep Learning Framework for Large-Scale Estimation of Penetration Bias in TanDEM-X DEMs

The accurate measurement of snow and ice dynamics is crucial for understanding glaciological processes and the impacts of climate change. TanDEM-X, an X-band spaceborne interferometric synthetic aperture radar (InSAR) mission, offers global, high-resolution digital elevation models (DEMs) that are invaluable for these studies; however, the inherent variability in radar wave penetration into snow and ice poses challenges in accurately estimating surface elevation changes, snow mass balance and snow depth. Variations in the acquisition geometry and snow properties can affect the estimation of the radar mean phase center, leading to penetration bias and an underestimation of the surface topographic height. In this work, we propose a novel deep learning framework for estimating the penetration bias of TanDEM-X DEMs over ice sheets, by combining the knowledge of the physical properties of snow and of the InSAR system for the development of a robust regression framework. Due to the lack of extended reference data, which jeopardizes the use of fully-supervised data-driven approaches, we propose a deep learning approach based on two intrinsically connected tasks: a first unsupervised snow facies segmentation model designed to capture the overall properties of the snowpack independently of the single-pass InSAR acquisition geometries; and a subsequent downstream penetration bias regression model. To ensure that the robustness against the InSAR geometries of the first model is preserved, we propose the use of multitask learning for the downstream task by jointly training both the segmentation and regression models. This approach ensures that the snow-related feature representations identified during the pretext unsupervised segmentation task are consistently leveraged to improve the final regression performance. Experimental results over the Greenland Ice Sheet, using IceBridge laser altimeter measurements as reference data, demonstrate that our method estimates the penetration bias with a coefficient of determination of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}=90 \%$</tex> and an RMSE of only 0.65 m, independently of the InSAR acquisition geometry and snow properties. Furthermore, we show that our proposed approach can be transferred to alpine glaciers - characterized by high-relief topography - with minimal fine-tuning, being able to aid glaciological applications such as snow depth estimation by reducing the estimation error with TanDEM-X DEMs by about 66% on average. The work performed here is the first in the literature to develop models that can be used for large-scale TanDEM-X penetration bias correction, designing deep learning architectures that are robust against varying single-pass InSAR acquisition geometries and validated for both ice sheets and alpine glaciers. The results showcased in this paper are crucial for enhancing the accuracy of TanDEM-X DEMs over snow and ice-covered regions, thereby improving our understanding of glaciological processes and their response to climate change, and can be highly relevant for the design of future single-pass InSAR missions such as the ESA 10th Earth Explorer Harmony mission.