TL;DR
This paper explores how future CRISTAL satellite altimetry data will estimate snow depth on Arctic sea ice. By using current CryoSat-2 (Ku-band) and SARAL AltiKa (Ka-band) missions, the study reveals that while both can track snow depth, their radar signals interact differently with snow and ice, leading to biases that need careful consideration for CRISTAL's success.
What Problem Does This Paper Address?
The upcoming CRISTAL mission aims to provide crucial data on Arctic sea ice thickness and snow depth. However, a fundamental challenge lies in understanding how Ku-band and Ka-band radar signals interact with the complex snow and sea ice layers. This interaction dictates where the radar "sees" the surface, influencing the accuracy of derived snow depth. Previous studies have often attributed discrepancies between radar altimetry and independent measurements to radar penetration biases. This paper addresses this by using existing missions to mimic CRISTAL's dual-frequency approach, investigating these biases and their dependence on surface properties to better anticipate CRISTAL's performance.
Methods
The authors employ a sophisticated approach involving:
- Physical Waveform Modeling (LARM): They use a Lognormal Altimeter Retracker Model (LARM) to simulate radar altimeter echoes from sea ice and snow. This model accounts for surface roughness and scattering mechanisms.
- Dual-Frequency Altimetry: They leverage data from CryoSat-2 (Ku-band SAR mode) and SARAL AltiKa (Ka-band LRM mode) to mimic CRISTAL's dual-frequency sensing.
- Comparison with Laser Altimetry: ICESat-2 laser altimetry data, considered a more direct measure of the air-snow interface, is used as a reference for freeboard comparisons.
- Snow Depth Estimation: Two methods are used to estimate snow depth:
- KuKa: Difference between CryoSat-2 and AltiKa freeboards.
- KuLa: Difference between CryoSat-2 and ICESat-2 freeboards.
- Waveform Simulations for CRISTAL: They simulate CRISTAL's expected performance by modeling radar echoes based on observed snow and ice properties, investigating the impact of surface roughness and snow properties on the retrieved heights.
Datasets
The study utilizes data from three key satellite missions:
- CryoSat-2 (CS2): Ku-band radar altimeter data in SAR and SARIn modes (October 2018 - April 2023).
- SARAL AltiKa (AK): Ka-band radar altimeter data in LRM mode (October 2018 - April 2023).
- ICESat-2 (IS2): Laser altimeter data, specifically the ATL20 daily and monthly gridded sea ice freeboard product (October 2018 - April 2023).
Additionally, the study references and compares with:
- Airborne and In Situ Datasets: Including data from the AWI IceBird Winter 2019 campaign, MOSAIC campaign transects, and CRESIS/University of Kansas airborne radar data.
- Reference Sea Surface Models: Such as the DTU21 solution.
Key Insights and Ideas
- Frequency-Dependent Penetration: The core insight is that Ku-band radar tends to penetrate deeper into the snowpack, potentially reflecting off the snow-ice interface, while Ka-band radar is more sensitive to the air-snow interface and upper snow layers. This difference is crucial for estimating snow depth.
- Surface Roughness Dominance: Waveform simulations suggest that for CRISTAL, the height of the detected snow and ice interfaces is more sensitive to multi-scale surface roughness than to snow properties themselves. This highlights the importance of accurately characterizing surface topography.
- Retracking Thresholds Matter: The way radar waveforms are "retracked" (interpreted to determine surface height) significantly impacts freeboard estimates. Different thresholds are needed for different surface conditions (e.g., rough multi-year ice vs. smooth first-year ice) to accurately represent the air-snow or snow-ice interface.
- Seasonal Evolution of Biases: Ka-band freeboards from AltiKa show a slower thickening rate over winter compared to ICESat-2, suggesting potential low biases due to uncertain penetration over first-year ice in spring.
- CRISTAL's Potential: Despite current limitations, the dual-frequency nature of CRISTAL holds promise for reducing snow depth uncertainties, potentially by providing more direct measurements of both the air-snow and snow-ice interfaces.
Key Results
- Freeboard Differences: CryoSat-2 radar freeboards are consistently thinner than AltiKa and ICESat-2 freeboards. AltiKa and ICESat-2 freeboards show similar patterns but differ in magnitude.
- Snow Depth Estimation: The KuLa (CryoSat-2 minus ICESat-2) snow depth product shows better agreement with independent observations (correlation of 0.79, RMSE of 7 cm) compared to the KuKa (CryoSat-2 minus AltiKa) product (correlation of 0.70, RMSE of 9 cm).
- Seasonal Accumulation Rates: KuLa estimates a mean seasonal snow accumulation of 1.6 cm/month over multi-year ice (MYI) and 1.9 cm/month over first-year ice (FYI). KuKa estimates are significantly lower, around one-third of KuLa rates.
- Model Simulations: Simulations suggest that CRISTAL Ku-band retrievals will track a median elevation 3% above the snow-ice interface, while Ka-band retrievals will track a median elevation 10% below the air-snow interface. However, these can be off by over 20% in some cases.
Limitations and Future Work
- Retracking Algorithm Sensitivity: The study highlights that the choice of retracking algorithm and assumptions about surface scattering significantly influence freeboard estimates, making it challenging to disentangle geophysical biases from processing artifacts.
- Surface Roughness Impact: While simulations show surface roughness is a key factor, accurately characterizing its multi-scale variations remains a challenge.
- Snow Properties Complexity: The simplified physical model used for simulations may underestimate the complexity of layered snowpacks and volume scattering effects.
- Future Work: The authors suggest further research into:
- The covariance of snow and sea ice surface topography at various scales.
- Bounds on millimeter-centimeter radar-scale roughness and controlling factors.
- The evolution of radar returns from the same snow/ice over time to observe changes in snow properties.
- Utilizing auxiliary remote sensing data or model outputs to support CRISTAL waveform interpretation.
- Investigating the impact of different snow properties and surface roughness on CRISTAL's dual-frequency SAR mode.
Summary updated