General comment

This manuscript describes an attempt to improve the methodology behind the Copernicus OMI indicator regarding SST extremes. The authors run an assessment of two different approaches regarding the use of different initialisation data (interim) or a multi-product ensemble approach that integrates regional and global model outputs with satellite-based observations. The main results show the benefits of these two approaches with respect to the actual procedure.

My only concern is that the analysis has only been run for 2024, which causes some uncertainty about the robustness of the new methodology. Indeed, the authors indicate in the conclusions section that a statistical analysis of the new methodology over decadal time scales would be necessary for a statistical characterisation of error distributions. This analysis would improve the consistency of the new approaches before operational applications.

As this seems to be the next step for the authors, my recommendation is to accept the manuscript for publication in its current form.

Minor comments

Which is the period referring to “the corresponding climatological 99th percentile”? Is it “(1993–2023)”? Why is this period used? Have different reference periods been tested?

The period should be indicated and justified in the data section, or different reference periods should be tested in the discussion sections.

Figure 3 caption does not describe panel (c)

Manuscript revision

This study assesses two methodological improvements to the SST extremes Ocean Monitoring Indicator (OMI) in the IBI region: the replacement of the NRT dataset with the Interim product and the adoption of a multi-product ensemble approach. The Interim dataset improves internal consistency, whereas the ensemble method enhances robustness and allows the estimation of uncertainty. Overall, the work is relevant and well structured, but several major revisions are needed before publication.

1. Choice of the 99th-percentile anomaly as a metric

I have concerns regarding the use of the anomaly of the 99th percentile as the main metric for characterizing SST extremes, including the annual maps. The 99th percentile reflects only the very upper tail of the distribution and is therefore highly sensitive to short-lived spikes, noise, or inconsistencies across datasets—especially in coastal regions, where observational and modelling limitations are more pronounced. As a consequence, part of the signal may reflect artefacts rather than physically meaningful extreme events.

Although the 99th percentile captures the most intense warm anomalies, it may filter out persistent but moderate events that are ecologically relevant (e.g., typical marine heatwaves). Lower percentiles (e.g., 95th) or multi-metric approaches (e.g., combining intensity and duration) are commonly used in the literature and may offer more physically interpretable results.

Furthermore, because the current OMI is defined as the anomaly of the 99th percentile, the indicator becomes doubly sensitive—both the annual value and the baseline depend on extreme-tail behaviour. This amplifies uncertainty and complicates interpretation, especially when comparing the performance of different products or analysing trends.

1. Methodological clarity on the annual anomaly maps

The manuscript does not clearly explain how the annual 99th-percentile anomaly maps are generated. While it is clear that the computation is performed on a per-pixel basis, the precise steps are not described. It remains unclear whether the annual 99th percentile is derived from daily SST values for the entire year, from a subset of data, or through some form of temporal aggregation or filtering. The construction of the climatological reference is also not explicitly documented—whether it is based on yearly percentiles averaged across years, or on other approaches.

Clarifying these aspects is crucial for transparency, reproducibility, and the physical interpretation of the resulting anomaly fields. As it stands, it is difficult to fully understand how Figures 1–2 were produced and how to be interpreted.

1. Use of the terms “deterministic” and “probabilistic”

The manuscript frequently refers to “deterministic” and “probabilistic” methodologies. However, all three OMI formulations are based on statistical thresholds (percentiles), and the use of these adjectives may give the impression of fundamentally different methodological classes. I recommend removing these terms, unless a clearer conceptual distinction is provided and justified.

Section-specific comments

Abstract

• Lines 17–19: These lines repeat information already provided in lines 12–14; one of the two should be removed.
• Line 14: The expression “to support decision-making” is vague. Consider replacing it with “to characterize the reliability of this OMI” or something similar.

1. Introduction

• Line 31: The reference to “decision-makers and other stakeholders” appears unnecessary.
• Line 33: The citation “von Schuckman et al., 2026” appears incorrect; please verify (also at line 49).
• Line 41: Specify here the climatological period of the reanalysis product.
• Line 42: The argument about “ensuring timely operational delivery” is unclear, since the OMI is updated annually (line 56).
• Lines 57–59: This sentence is overly self-congratulatory; rephrase or support the claims with references.

1. Data and Methods

• Lines 91–92: This repeats the idea introduced in line 46; consider removing.
• Lines 122–127: The construction of MULTI-OMI should be explained more thoroughly. Specify whether the ensemble anomaly is computed as a simple average, and confirm whether each product maintains its native spatial resolution (as suggested by lines 250–251). Also explain how non-significant pixels are defined.
• Line 129: Consider replacing “common reference point” with “common reference dataset”.
• Lines 135–146: This paragraph repeats earlier content and should be removed or rephrased.

1. Analysis

• Line 194: The expression “rather than a formal estimate of uncertainty” is confusing, since the ensemble is explicitly introduced to estimate uncertainty. Reword this sentence.
• Lines 195–196: This sentence is too vague; rewrite or remove.
• Figure 2b: The presence of STD values over non-significant areas (as defined in Fig. 2a) is confusing. Clarify how non-significant pixels are handled and be consistent with the definition (see also next comment).

1. Assessment

• Line 238 / Fig. 3a: Fig. 3a corresponds to Fig. 2a but without masking non-significant areas. It is not clear how the sign of anomalies is chosen in grey regions. In practice, it appears that in-situ observations are being overlaid onto areas classified as non-significant. Then, a consistent treatment of non-significant pixels is needed. Non-significant regions should be represented consistently (e.g., grey shading), otherwise interpretation becomes difficult. Considering the overall manuscript, I would suggest to adjust Fig. 3a to Fig. 2a accordingly.
• Lines 245–246: The term “simplified metric” is unclear. If the metric cannot produce robust confidence intervals, its use should be justified.
• Figure 3b: The meaning of the red and blue bars for OMI-OBS is unclear. Do they represent uncertainty? One would expect a single value, as for OMI-NRT and MULTI-OMI. Please revise. Moreover, OMI-INT is missing, and the caption lacks the description of Fig. 3c.
• Lines 254–255: Provide a clearer explanation of why spatial variability is “more accurate”.
• Line 263: The sentence suggests that the decrease in STD for MULTI-OMI is problematic; clarify the intended meaning.
• Line 265: The RMSE how it is derived? From the Taylor diagram?
• Table 2: OMI-INT is missing and should be included.
• Lines 294–296: MULTI-OMI does not reduce misclassification per se; rather, it provides the option to classify pixels as non-significant. Consider rephrasing accordingly.

1. Conclusions

• Lines 306–332: This section repeats material already presented several times; a more concise synthesis is recommended.
• Line 323: “Deterministic error metrics” is not appropriate; mean bias and RMSE are standard statistical metrics.
• Line 329: The claim that OMI-INT is more consistent because it derives from the same climate data record is not demonstrated. Provide justification or temper the statement.
• Line 380: The term “substantial improvements” appears too strong given the results; consider softening the language. This applies elsewhere in the manuscript as well.