| 12 Jan 2026
Uncertainties in 45+ years of Sea-Ice Area and Sea-Ice Area trend observations
Abstract. In this study, we assess and quantify uncertainties associated with observation-based estimates of Arctic and Antarctic Sea-Ice Area (SIA). In particular, we examine uncertainties inherent in SIA estimates from a single product, based on the EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF) sea ice concentration record, and compare these with the uncertainty in SIA as estimated from the spread across several satellite products. We provide results of a refined uncertainty estimation method that propagates local sea-ice concentration (SIC) uncertainties to hemispheric SIA estimates, accounting for spatial and temporal error correlations. While SIA itself shows notable seasonal and long-term variability, the uncertainties associated with these measurements have remained relatively stable over time. The seasonal cycle of the uncertainties is not directly following the SIA seasonal cycle, but is linked largely to the distribution of the ice. In the growing seasons, the SIC fields are more compact with a rather sharp boundary between high and low sea ice concentrations compared to the more diffuse boundary in the melting season. This seasonal evolution of the so called Marginal Ice Zone (MIZ) leads to relatively large SIA uncertainties in the melting season and smaller uncertainties in the growing season.
We also show that the differences between SIA estimates from different SIC products is characterized by seasonally varying biases. After accounting for these biases, the remaining differences are consistent with our new single-product SIA uncertainty. The presented uncertainty product complements the inter-product approach by providing dynamic daily and monthly estimates which do not rely on the selection and availability of other products. It represents the non-systematic (bias-free) component of the uncertainties, which, among other things, determines the significance of new SIA extremes.
We estimate that non-systematic uncertainties in SIA trend estimates are 11 · 103 km2/dec. (Arctic) and 14 · 103 km2/dec. (Antarctic). We further infer the presence of additional systematic uncertainties in these trend estimates, which indicates that a longer time-series will not be sufficient to remove trend uncertainties. These systematic uncertainties are related to inherent differences between the processing chains of the different products, which reveals an unfortunate influence of methodological choices in the SIC product development on SIA trends.
This manuscript presents a method for propagating local sea-ice concentration uncertainties to hemispheric sea-ice area estimates over the passive microwave record. The study makes three clear contributions that advance our understanding of observational uncertainty in sea ice.
1. The work provides a dynamic, daily and monthly uncertainty product that is self-contained within a single satellite record (OSI SAF) and does not depend on the availability or selection of other products. This fills a gap in the community's toolkit for assessing the significance of new SIA extremes and for climate model evaluation.
2. Secondly, the authors identify the Marginal Ice Zone as the primary physical driver of the seasonality of SIA uncertainty.
3. Finally, the comparison between the single-product stochastic uncertainty and the inter-product spread provides compelling evidence that the large differences between existing SIA trend estimates are dominated by systematic methodological choices in the processing chains, rather than noise. The implication that extending the time series alone will not resolve trend uncertainties is an important message for both the observational and modelling communities.
The study is both timely and important. Reliable uncertainty estimates for SIA are needed for climate assessments and model evaluation, but little is known about these uncertainties over the full satellite record. By separating the stochastic and systematic components, the authors provide appropriate uncertainty estimates for relative comparisons. The Monte Carlo generates a 50-member ensemble that preserves the spatial and temporal error correlations. Three refinements to a previously published method improve consistency with the operational sea ice index. The study acknowledges limitations: it relies entirely on the OSI SAF SIC uncertainties without independent verification and only partially captures systematic errors from melt ponds and interpolation. However, this first step constitutes an important contribution and, following consideration of my points outlined below, would make an interesting contribution to The Cryosphere
The manuscript is generally well-structured and explains complex statistical concepts clearly, though some sections rely heavily on prior work. The statistical framework is rigorous, the validation strategy of comparing the ensemble spread against bias-corrected inter-product differences is well conceived, and the paper is concise and logically structured.
Major concerns:
1. Definition of the MIZ. The finding that SIA uncertainty scales with the square root of MIZ length (the 50% SIC contour) is important. But the width of the MIZ varies regionally and seasonally due to wave exposure, ocean heat and ice dynamics. This is particularly relevant for Antarctica, where a substantial MIZ is almost always present. Would a 2D metric (e.g. MIZ area, or area of grid cells within 15-80% concentration or similar) provide a stronger or more physically complete predictor? If MIZ length already captures most of the variability, demonstrating this explicitly would strengthen the physical interpretation and help assess robustness under future conditions where MIZ character may change.
2. The paper relies heavily on Wernecke et al. (2024), and while the three refinements introduced here are clearly described, readers unfamiliar with the earlier work may struggle to follow key steps. I found it necessary to consult the 2024 paper to fully understand the processing chain. A brief step-by-step summary of the workflow (even in schematic or bullet-point form) would substantially improve accessibility without requiring full re-derivation of the method.
3. The authors are transparent that their method is fully reliant on the SIC uncertainty component from OSI SAF without independent verification. The agreement with bias-free inter-product spread provides indirect validation, but a discussion on whether this could be partly coincidental if both approaches share common limitations would be helpful/ A brief sensitivity analysis, or at a minimum adding a short discussion of how results might change if input SIC uncertainties were scaled by some factor, would help readers assess the robustness of the quantitative estimates.
Minor points
Title: The current title repeats "Sea-Ice Area" twice, which reads awkwardly. You might consider rewording this.
Abstract: The finding about systematic uncertainties and their implications for trend analysis is arguably the paper's most consequential message, but is somewhat buried in the final sentences. You might consider elevating this. Several sentences are long and could be broken for clarity. Hedging phrases such as "so called" and "among other things" could be trimmed without loss of meaning.
Figure 1 caption: the abbreviation STD is not defined (elsewhere in the manuscript your write out "standard deviation").
Line 120: (or at the most appropriate location) The choice of 60 Monte Carlo ensemble members is stated but not justified. A brief note on why 50 is sufficient or a reference to convergence testing in the 2024 paper would be helpful.
Line 151: "ocen" should read "ocean"