the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 May 2026
Two decades of GNSS-IR observations reveal an asymmetric decline in coastal sea ice phenology of the Beaufort Sea
Abstract. Arctic coastal sea ice phenology is a critical climate indicator, yet sub-kilometer long-term observations remain scarce, limiting our understanding of complex local freeze-thaw dynamics. This study investigates the long-term evolution and thermodynamic drivers of coastal sea ice using a continuous, 20-year (2003–2023) high-resolution dataset derived from ground-based GNSS Interferometric Reflectometry (GNSS-IR) at Tuktoyaktuk in the Beaufort Sea. By employing a physics-based Amplitude Integration Factor (AIF) method, we successfully bridged decadal hardware discrepancies to extract an uninterrupted, thermodynamically consistent climatological record. Trend analysis of the 20-year record reveals a statistically significant shortening of the continuous ice season by 4.63 days per decade (p = 0.04). This climatological decline is profoundly asymmetric, driven primarily by a substantially delayed autumn freeze-up (+3.40 days per decade) rather than an advanced spring breakup (−1.42 days per decade), underscoring the dominant influence of enhanced summer oceanic heat uptake and thermal memory. The physical reliability of these localized observations is corroborated by their strong coupling with accumulated Freezing Degree-Days (R² = 0.74). Crucially, cross-scale comparisons demonstrate that GNSS-IR detects autumn freeze-up onset 5.5 ± 3.7 days earlier than 4-km gridded satellite products (IMS). This systemic lead time confirms the unique capability of GNSS-IR to resolve initial nearshore frazil ice formation – a critical sub-grid thermodynamic process typically diluted in coarse-resolution remote sensing. Ultimately, this work provides an essential high-resolution baseline for validating regional climate models.
- Preprint
(38556 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 22 Aug 2026)
-
RC1: 'Comment on egusphere-2026-1160', Anonymous Referee #1, 03 Aug 2026
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1160/egusphere-2026-1160-RC1-supplement.pdfReplyCitation: https://doi.org/
10.5194/egusphere-2026-1160-RC1 -
RC2: 'Comment on egusphere-2026-1160', Anonymous Referee #2, 11 Aug 2026
reply
This manuscript uses the Amplitude Integration Factor (AIF) to analyze a 20-year GNSS-IR record of coastal sea ice phenology at the TUKT station and examines long-term changes in ice-out, freeze-up, and ice-season duration. The long observation period, the focus on the poorly observed nearshore environment, and the attempt to homogenize GNSS observations across receiver changes are potentially valuable. The comparison with IMS, ERA5, in situ meteorological observations, and Sentinel-2 imagery also provides useful multi-scale context.
I have also reviewed the authors’ previous work describing the development and validation of the AIF indicator. Accordingly, the methodological novelty of AIF itself is not a central issue in my evaluation of this manuscript. The key question here is whether this indicator can be reliably extended to a 20-year climate-scale record, and whether the resulting phenological dates and long-term trends are supported by sufficiently rigorous calibration, uncertainty assessment, and independent validation.
At present, several central conclusions appear stronger than the analyses can support. In particular, the long-term climatological conclusion depends strongly on the treatment of an individual year, while the hardware homogenization, transition-date detection, and physical interpretation of AIF require stronger validation. I therefore recommend major revision before the main conclusions can be considered sufficiently robust.
Major comments
1. Section 2.2: GNSS-IR reflection footprint
Section 2.2 states that the reflection footprint ranges from approximately 8 to 40 m in radius and covers 50–400 m2. These values appear geometrically inconsistent. Moreover, the First Fresnel Zone is elliptical rather than circular, so describing it by a single radius may be misleading. The authors should verify these values and provide a consistent description of the spatial footprint. If the exact dimensions are not essential to the subsequent analysis, I suggest removing the specific radius and area values rather than introducing potentially confusing spatial-scale estimates.
2. Section 2.3.2: Definition of the IMS comparison
The IMS processing requires further clarification. Section 2.3.2 defines regional ice presence using an 11×11 pixel window, corresponding to approximately 44×44 km, with an "ice probability" threshold of 0.5. However, later sections discuss the comparison in terms of a single 4-km IMS grid cell or a 16-km2 pixel. These represent substantially different spatial supports.
The authors should clearly specify whether the phenological dates are derived from the central IMS pixel, an 11×11 pixel window, or a spatially aggregated ice fraction, and explain how the 0.5 threshold is calculated. This distinction is essential for interpreting the subsequent comparison with the much smaller GNSS-IR reflection footprint.
3. Sections 3.1–3.2: Role of AIF and observational validation of its physical basis
AIF was previously developed and validated by the authors in Song et al. (DOI: 10.1109/TGRS.2022.3155051). Although AIF remains central to the present analysis, an extensive re-derivation of the indicator does not appear necessary here. Citing the previous study and briefly describing the principle and implementation used in the present work should be sufficient. If the AIF calculation differs from the previous study, these differences should be clearly identified and justified. I therefore do not regard the development of AIF itself as a primary methodological contribution of the present manuscript.
Instead, the manuscript would benefit more from showing the observed elevation-dependent amplitude curves, P (Eq. 4), under different sea-surface states. In particular, do the observed dSNR amplitudes show the same elevation-dependent behavior as the theoretical or simulated results in Figure 3? For example, is the observed amplitude over sea ice consistently higher than that over open water throughout the 5°–15° elevation range? Such an observational comparison would provide more direct support for the physical basis of the AIF analysis than simply presenting integrated AIF values for ice and water.
4. Section 3.3.1: Inter-calibration across receiver changes
The manuscript reports an approximately 27% reduction in AIF magnitude following the 2015 receiver change and subsequently applies satellite-specific subtraction of the mean AIF during DOY 200–250. However, this normalization removes an additive offset, whereas the hardware effect is described as a proportional reduction in signal magnitude.
The authors should clarify whether the receiver-induced change is additive, multiplicative, or more complex, and demonstrate that the adopted normalization adequately removes its effect on both the open-water baseline and the seasonal ice–water contrast. In addition, the manuscript states that several hardware changes occurred during the 20-year period, while only the 2015 change is discussed in detail. A complete equipment history and evidence that other hardware changes do not introduce detectable discontinuities would strengthen the reliability of the long-term record.
5. Section 3.3.2: Phenological transition criteria and the added value of AIF
The transition-date criteria require further clarification. With σ_ref=0.19, the nominal detection threshold is 3σ_ref=0.57 in calibrated AIF units. However, ice-out additionally requires a “sustained low value,” for which no duration is specified, whereas freeze-up requires AIF to remain above the threshold for at least five days. These asymmetric persistence criteria introduce some subjectivity and should be defined more explicitly.
They also suggest that the method is primarily intended for retrospective phenological analysis. If so, the authors should clarify the specific advantage of AIF over conventional spectral-amplitude metrics. Although spectral amplitude may provide weaker instantaneous ice–water separation, under the same smoothing, thresholding, and persistence-based post-processing it may potentially provide comparable transition dates. A direct comparison of AIF and conventional spectral amplitude in terms of state separability and transition-date stability would better demonstrate the added value of AIF for long-term phenological analysis.
6. Section 4.1: Physical interpretation of the 2019 seasonal phases
The physical interpretation of the five AIF phases appears more specific than the independent observations can support. Several processes, including snow densification, destructive metamorphism, exposure of bare ice, melt-pond formation, and changes in dielectric loss, are inferred without direct field observations.
Based on the temperature evolution and satellite imagery presented in the manuscript, Phase II and Phase III appear more straightforwardly interpretable as snow-melt-dominated and ice-melt-dominated periods, respectively. I suggest that the authors distinguish more clearly between surface states directly supported by independent observations and physical processes inferred from the AIF behavior. The phase-specific interpretation should be moderated unless additional independent evidence is available.
7. Section 4.2.2: Treatment of the 2018 anomaly and significance of the long-term trend
The exclusion of 2018 currently appears to be motivated primarily by its strong influence on the fitted regression and its characterization as a meteorological anomaly. When the complete record is retained, the corrected ice-duration trend is −4.19 days per decade with p=0.06, whereas the statistically significant value of −4.63 days per decade with p=0.04 is obtained only after excluding 2018.
An extreme year in a 19-year climate record is not necessarily an invalid observation. It may represent genuine coastal climate variability and should not be removed solely because its inclusion reduces statistical significance. The full-record result should therefore remain the primary estimate, while the result excluding 2018 can be presented as a sensitivity analysis unless an independent criterion demonstrates that the 2018 observation is invalid. The corresponding statements in the Abstract, Results, and Conclusions should be revised accordingly.
8. Section 4.3: Interpretation of the thermodynamic mechanism
The manuscript attributes the delayed freeze-up primarily to enhanced summer ocean heat uptake and a "thermal memory" effect. However, the study does not directly analyze ocean heat content or autumn ocean thermal anomalies sufficient to quantify this mechanism. Moreover, the freeze-up trend itself is not statistically significant either with or without 2018.
The observed association with air temperature and FDD supports a thermodynamic influence on freeze-up timing, but it does not by itself demonstrate that enhanced summer ocean heat storage is the dominant mechanism responsible for the long-term trend. The authors should distinguish more clearly between mechanisms directly demonstrated by the present analysis and physically plausible interpretations based on previous literature.
9. Section 4.4: Interpretation of the 5.5-day GNSS-IR–IMS offset
The reported mean freeze-up offset of 5.5 ± 3.7 days is potentially interesting, but its physical interpretation depends strongly on the IMS spatial aggregation described in Section 2.3.2. If the IMS phenological date is actually derived from an approximately 44×44 km window, it is not appropriate to explain the offset simply as a comparison between the GNSS-IR footprint and a single 4-km or 16-km2 IMS pixel.
Therefore, until the exact IMS processing and spatial support are clarified, the 5.5-day lead should not be interpreted as definitive evidence of a sensor-resolution effect. The distinction between spatial-resolution effects, different spatial sampling domains, and genuine nearshore–offshore differences should be made more carefully.
Minor comments
- Lines 55 and 57:The manuscript contains several formatting and language issues, including duplicated punctuation and inappropriate spacing. A careful proofreading of the manuscript is recommended.
- Line 63:Unless the dielectric permittivity values are representative of conditions at the study site or their applicability to the TUKT environment can be justified, I suggest avoiding overly specific numerical values here.
- Figure 1:The station symbols in the legends do not appear to be consistent. In addition, Figure 1b appears to show that some of the plotted First Fresnel Zones, particularly those associated with the 5° and 10° elevation angles, overlap the land surface. Please verify the reflection-zone geometry and the statement that the selected footprints are entirely located over water.
- Table 1:Describing Sentinel-2 as having a temporal resolution of 5 days may be misleading for this application. Although the nominal revisit interval may be approximately 5 days, the number of usable cloud-free images over the study area is much lower. Please distinguish nominal revisit time from actual image availability.
- Figure 3:Figure 3 should be placed after its first citation in the main text.
- Figures 4 and 7:Please define the meaning of the ±1σ uncertainty envelope. It is unclear whether it represents inter-satellite variability, uncertainty in the filtered estimate, a confidence interval, or another quantity.
- Figure 6:The Sentinel-2 images shown in Figure 6 appear not to match the imagery available for the corresponding dates and location in Copernicus Data Space. The authors are requested to verify the acquisition dates and image products.
- Tables 3 and 4, Figure 8, and associated text:The manuscript uses several different temporal ranges, including 2004–2022, 2004–2023, a 19-year record, and a 20-year record. Please make these descriptions consistent and state the exact number of observations used in each regression.
Citation: https://doi.org/10.5194/egusphere-2026-1160-RC2
Data sets
Two Decades of Arctic Coastal Sea Ice Phenology: A GNSS-IR Record from the Beaufort Sea (2003–2023) Minfeng Song et al. https://doi.org/10.5281/zenodo.18452514
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 187 | 74 | 19 | 280 | 18 | 19 |
- HTML: 187
- PDF: 74
- XML: 19
- Total: 280
- BibTeX: 18
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1