Measurement of Arctic sea ice from satellite altimetry: the potential and limitations of CryoSat-2 SARIn mode

Alessandro Di Bella

    Research output: Book/ReportPh.D. thesis

    96 Downloads (Pure)


    Sea ice plays a fundamental role in the global climate system, influencing directly the albedo of our planet and regulating the exchange of heat between the atmosphere and the ocean. Observations from satellites and submarine data have shown a rapid reduction of the ice-covered area and a general thinning of Arctic sea ice in the last three decades. Satellite altimetry can be used to infer sea ice thickness from the direct measurement of the sea ice freeboard, i.e. the height of the ice surface above the local sea level. However, in the freeboard-to-thickness conversion the freeboard and the associated errors are typically multiplied by a factor of ∼10, thus, it is fundamental to both improve the accuracy of freeboard estimates as well as to minimise their uncertainty.

    The largest source of freeboard uncertainty, after the contribution due to the lack of knowledge of the Arctic snow cover, originates from the poor knowledge of the sea surface height (SSH) in icecovered regions. CryoSat-2’s (CS) interferometric mode (SARIn) enables to process waveforms whose power echo is dominated by the strong reflection from off-nadir leads, referred to as "snagged" waveforms, which are usually discarded in common SAR altimetry data processing. In fact, the available phase information can be used to correct for the associated range error and to retrieve a larger number of valid SSH measurements which, ultimately, increases the accuracy as well as reduces the uncertainty of the area-averaged SSH. The SARIn phase information  is currently not used by the scientific community in the estimation of sea ice freeboard and
    thickness, probably because of the scarce SARIn coverage of the Arctic Ocean. However, despite changes in the SARIn acquisition mask throughout the years, CS still operates in SARIn mode along the entire coastline of the Arctic Ocean. In this work, an assessment of the potential and limitations of the CS SARIn mode with respect to the estimation of the sea ice freeboard and thickness in the Arctic is performed. Besides being of interest to the sea ice community, such an assessment could inform the proposal and design of future SARIn-only satellite  altimetry missions.

    The first part of this project investigates how the phase information provided by the CS SARIn acquisition mode affects Arctic sea ice freeboard and thickness retrievals as well as their corresponding random uncertainties. An Arctic sea ice processor (ASIP) is developed at DTU to process CS SAR and SARIn L1b waveforms according to both a regular SAR processing scheme, where only the power echoes are used, as well as by using the SARIn phase information. It is shown that CS SARIn mode can accurately detect off-nadir leads up to 2 km from the satellite nadir and correct for the associated range overestimation using the interferometric information. The comparison of along-track sea ice freeboard estimates from ASIP with airborne measurements from the ESA CryoVEx and NASA Operation IceBridge (OIB) campaigns indicates that, by using the phase information, the average random freeboard uncertainty of single CS freeboard estimates can potentially be reduced without introducing a bias on the average freeboard heights. Differences between SAR and SARIn freeboard heights at the boundaries of the SARIn mask,  due to the larger noise of SARIn waveforms compared to SAR, are analysed and it is shown that continuity between SAR and SARIn regions can be achieved. By correcting for the  overestimated range due to the snagging effect, it is possible to process more waveforms than in a regular SAR
    processing scheme. The larger amount of both the processed waveforms and the detected leads is then increased significantly. In SARIn areas, this results in a 14% and 13% average reduction of the gridded random uncertainty of freeboard and thickness estimates, respectively, when compared to the results from the regular SAR reference case. During the analysis performed in this part of the study, an issue in the ESA Baseline C IPF1, the processor responsible of computing SAR and SARIn L1b waveforms, is detected. This issue, causing inaccurate values of phase difference to be computed for some SARIn waveforms at the boundaries of the SARIn
    mask, is solved in cooperation with ESA and Dr. Scagliola from ARESYS S.r.l.. The upcoming CS Baseline D L1b products will include this improvement which will benefit any application that exploits the phase information from SARIn L1b products near the boundaries of the SARIn mask.

    The second part of the project describes the work carried out at the NASA Jet Propulsion Laboratory (JPL) in cooperation with Dr. Kwok and Dr. Armitage. Here, a second sea ice processor (MPASIP) is developed to investigate if the number of valid sea ice as well as lead measurements could be increased by processing multiple peaks of single CS SARIn waveforms, using the associated phase information, to further reduce the sea ice freeboard and thickness uncertainties. From Sentinel-1 SAR images, it is shown that the contributions from sea ice reflections close to the satellite nadir and specular returns from off-nadir leads can potentially be separated for some SARIn waveforms. Additionally, using the SARIn phase information enables to discard echoes originating on land and to retrieve freeboard heights closer to the coast compared to other available sea ice products.

    Radar freeboard retrievals from MPASIP show a general good agreement with estimates from the Alfred Wagner Institute (AWI) and a general overestimation over multi-year ice (MYI) compared to the JPL processor. The comparison of sea ice freeboard and thickness estimates with OIB airborne data shows a good correlation (0.7) and indicates that MPASIP and AWI overestimations in MYI regions are retracker-dependent. The analysis in SARIn areas reveals a larger correlation of MPASIP sea ice freeboard and thickness with OIB estimates compared to AWI. This is possibly due to a better sampling of the local sea surface of MPASIP, achieved by processing additional parts of the SARIn waveforms representing returns from off-nadir leads. Comparing MPASIP and ASIP, the increased amount of both lead (5 times) and sea ice (2.5 times) measurements, obtained by  processing multiple peaks of single SARIn waveforms, results in an average reduction of the gridded random freeboard and thickness uncertainties of 54% and 52%, respectively. This corresponds to a reduction in the total sea ice thickness uncertainty of 30%, when the systematic contributions from the snow depth and density are taken into account.
    Original languageEnglish
    Place of PublicationKgs. Lyngby
    PublisherTechnical University of Denmark
    Number of pages135
    Publication statusPublished - 2019


    Dive into the research topics of 'Measurement of Arctic sea ice from satellite altimetry: the potential and limitations of CryoSat-2 SARIn mode'. Together they form a unique fingerprint.

    Cite this