Please refer to the attached PDF files for the comments.

This study investigates the role of initial conditions in Antarctic sea ice seasonal predictions by comparing two versions of the Australian Bureau of Meteorology’s ACCESS system (ACCESS-S1 with sea ice data assimilation and ACCESS-S2 without). Using over 20 years of coupled simulations, the work reveals seasonally dependent impacts of initial conditions—with January–April initial conditions enhancing forecast skill for up to three months, while winter forecasts of the two systems become statistically indistinguishable after two weeks. The study provides valuable insights for operational sea ice forecasting and model development, but it has notable issues in structure, clarity, and presentation that need addressing. Therefore, this work is suitable for publication after major revisions.

Major Comments:

• The simulation periods (S1: 1990–2012; S2: 1981–2018) and forecast lead times (S1: 212 days; S2: 273 days) of the two systems are inconsistent. Please clarify the rationale for these differences and verify whether the overlapping period (1990–2013) can fully represent the variability of Antarctic sea ice (e.g., covering extreme events and long-term trends). It is recommended to standardize lead times for fairer comparison if feasible.
• Excessive monthly subplots (e.g., Figures 6 and 8) lead to redundant information and lack narrative progression. For results reflecting common characteristics across months, aggregating or simplifying subplots is suggested to avoid distracting readers from core findings.
• The explanation for the anomalous phenomenon (S2’s RMSE decreasing with increasing lead time) is confusing due to disorganized logic and irrelevant information. The proposed causal chain—"systematic thin sea ice in S2 → over-sensitivity to atmospheric/oceanic forcing → overestimated interannual variability at short lead times → convergence to model climatology at long lead times → reduced RMSE"—needs to be clearly structured, with key links (e.g., thin sea ice vs. variability) supported by direct evidence (e.g., spatial correlation between ice thickness and variability). In addition, Figure 9 also presents the RMSE for sea ice concentration anomalies (SICA). Does the same error trend apply to the RMSE of SICA as well?
• Line 253 references a "winter predictability barrier" in the Weddell Sea, but the study’s results show the most prominent barrier in the King Haakon Sea (0°–120°E). Please explain this discrepancy, considering factors such as regional circulation differences or model representation of ocean-sea ice processes.
• The key conclusion "ACCESS-S1 and ACCESS-S2 forecast errors are statistically indistinguishable after just two weeks" is repeatedly mentioned in the abstract and conclusion but lacks explicit textual support. The only relevant evidence is in Figure 9 (parenthetical values), which should be discussed in the main text.
• Figures 11–13 are information-dense but low in informativeness, with cumbersome correspondence between text and subplots. Alternative visualization methods (e.g., spatial aggregation, key region highlighting, or summary statistics) are recommended to convey core messages (e.g., regional skill differences) more directly.

Minor Comments:

• Line 30 has an ambiguity: clarify whether sea ice initial conditions are important in "autumn" (as per Arctic studies) or "spring" (relevant to Antarctic seasons).
• Section titles 3.1 ("Bias") and 3.2 ("RMSE") are overly simplistic. Revise them to reflect core content (e.g., "Bias Characteristics of Sea Ice Extent and Concentration" or "RMSE and Skill Score Analysis of Forecast Anomalies").
• In Figures 3 and 4, flip the color bar to follow conventional standards (blue for underestimation, red for overestimation) and standardize color bars across all relevant figures.
• In Figure 5, align the layout with other figures (place S1 on the left, S2 on the right) and add subfigure labels (e.g., a), b)) for clarity.
• In Figure 6, add subfigure labels and specify the unit of "lead time" (e.g., "months"). Synchronize these revisions across similar figures.
• Consistently use "sea ice" (instead of "sea-ice") throughout the text. Define abbreviations for key terms at their first appearance: sea ice concentration (SIC), sea ice thickness (SIT), and sea ice extent (SIE).
• In Figure 10’s caption, correct "larger than" to "lower than" (consistent with the logic of RMSE exceeding the 95% confidence interval lower bound of persistence forecast RMSE).
• Lines 279–280: Since the persistence forecast is listed separately, present the climatology forecast as a separate reference category for clarity.
• Section 3.3 ("Conclusions") should be moved to an independent Section 4 to serve as the full text’s comprehensive conclusion.