Please see the attached pdf file for review, thanks.

The manuscript entitled “Regime-dependence when constraining a sea ice model with observations: lessons from a single-column perspective” investigates the impact of several plausible observation types of sea ice on the sea ice variables in three different sea ice regimes: PACK ICE, SEASONAL ICE, and FIRST-YEAR ICE. The authors conducted a series of perfect model experiments using the linked DART and Icepack and analyzed the results using several metrics comprehensively. The authors find that the efficacy of sea ice DA varies significantly across observation types and sea ice regimes. While the former finding is not new, the latter finding is novel. This study suggests that DA strategies need to be tailored to observation types and regimes, which provides insights for the sea ice DA community. The experiments are well-designed, and the manuscript is well written. I very much enjoyed reading it. Albeit, I have several comments for the authors to address before publication, which I consider minor. 

1. The authors presented a lot of results in the paper, yet do not provide enough speculations or discussions. It’s helpful to know which observation types perform better than others, or which observations work better in certain regimes, but are they model specific? How translatable are the findings? Digging deeper into the mechanisms that contribute to their different performance might be more valuable. 

2. The format of the paper needs to be cleared up. Figures are referred to as Fig. XX in some places and Figure. XX elsewhere. Please be consistent throughout the paper. Also, there are no sub-labels in the plots, some of which are fine, but not when they are referred to in the paper, e.g., Figure 4. The citation also needs to be edited, e.g., Petty et al (2023a) is referred to in the text, but it’s not clearly labeled in the reference list section 

More detailed comments are listed below

1. The definition of the three regimes is a bit sporadic in the paper. Since it’s an important concept in the paper, I’d suggest clearly defining them in the experimental setup, maybe in a separate subsection.

2. There’s a lot of information in Figure 2 that’s not discussed in the paper. For example, although the aggregate SIC spread is small throughout the year except in the summer, the spreads of individual categories are decent. What does this suggest, and how will this impact DA results? I believe these discussions are valuable. Another thing that catches eyes is the huge spread of SIC in the FIRST-YEAR ICE. Basically it ranges from 0 to close to 1 in winter. Is it representative of the model uncertainty? Figure 4 is meant to evaluate if the ensemble in the single column grid points are representative of their regimes, but it only shows the time-averaged (co)variances, which does not consider the large seasonal variation, especially in the FIRST-YEAR ICE regime.  

3. The panels in Figure 4 are not labeled but referred to in the text. Please add sublabels. 

4. In Figure 4, the right panel doesn’t provide additional information to the left panel, at least the authors didn’t elaborate on it. I’d suggest re-arrange the panels in Figure 4 to add the seasonal variations of spread and trim the panels that are not discussed in the text. 

5. Block 240: why does FBR differ so much from FBL in FY ICE? The authors discussed their different performances in snow depth estimates, but didn’t mention their differences in ice volume. The possible contributing factors are also not mentioned. Since the two observation types are two key derived products, it’s worth investigating why one offers more value than the other, and the reasons behind it.