General comments

The manuscript describes the UK Met Office regional coupled modelling suite configured for the Singapore region and highlights the performance for the first half of 2019. Two-way coupling improved precipitation and the diurnal cycle over the oceans compared to an atmosphere-only simulation. The focus lies in understanding air-sea interactions and their impact on regional weather, particularly on momentum transfer and large-scale precipitation. The structure of the manuscript could see a few improvements, mainly for readability, clarity, and focus. One major concern is the role of both tuning and evaluation of ocean-only and wave-only components, specifically the wave part for Betamax with a value of 1.75 (Appendix A). It contradicts the primary objective, "the development of a fully coupled system enabling two-way interaction" (line 15-16). The parameter is significantly higher than the Met Office value of 1.39 and is equivalent to that from an ERA5-driven wave hindcast (Alday et al., 2021, https://doi.org/10.1016/j.ocemod.2021.101848). The Miles-Janssen-type wind input is known to significantly increase wave feedback to the atmosphere through the Charnock coefficient in two-way coupled simulations and reduce surface wind speed. In addition, the manuscript could benefit from a clearer description of the coupled system and its components, to support the results of fully coupled and atmosphere-only (uncoupled) simulations presented in section 4. If the coupled system is valid, then surface variables and diagnostics will improve and agree well with observations. On the ocean side, similar concerns as for the waves. "Sensitivity experiments were conducted to optimise configuration choices, including light penetration schemes and bulk forcing formulations, resulting in improved SST representation […]" on lines 156-157 sideline any form of coupling feedback entirely. The total water-side stress is equivalent to the air-side stress only if the wave field is in equilibrium, ie., the sum of wind input, dissipation and nonlinear interaction is zero. Most of the time, the wave field is not in equilibrium, which means that the air-side stress differs from the water-side stress. Atmosphere-wave-ocean coupler should take care of the flux exchange in the wave boundary layer, i.e., similar to Figure 8 in Breivik et al. (2015, https://doi.org/10.1002/2014JC010565).

Specific comments

1. Line 174: "wind field and current interpolation". Interpolation is performed within OASIS-MCT using remap weights. This setting will break the wave model because it does not know the future time step to establish linear regression.
2. Line 185: "The atmosphere, ocean, and wave components […] exchange variables at an hourly coupling frequency". What is the reason for hourly intervals? Convective-scale processes could benefit from a higher coupling frequency.
3. Line 191: "the coupling architecture and exchange variables" appears to double up on stresses received by the ocean. NEMO is subject to stress from both the atmosphere and the waves. It is unclear what wave fields are supposed to mean.
4. Section 4.1. What is the purpose of an ocean-only evaluation? The title of the manuscript is about the development of a fully coupled modelling system.
5. Section 4.2. There appears to be a threshold in the Indian Ocean that the wave model cannot represent swell waves greater than 2 metres. It appears to be a feature coming from ERA5.
6. Line 94-95: "Through evaluation against observational and reanalysis datasets,…." This sentence reads like a summary or highlight and does not belong in the objective statement. Line 192: "The development of this fully coupled system represents the main technical contribution [….]", in the wave model configuration, is the main objective of the paper and should appear at this point.
7. Lines 124-126 are riddled with acronyms. Why list all the applications for data assimilation, ensemble prediction, climate, and coupled configuration here? It is distracting for the reader and has no relevance to the manuscript's objective.
8. The paragraph on lines 315-321 appears better positioned in the introduction and for the justification of the objective.
9. Line 325-326: "Differences [in daily accumulated precipitation] between cSIGV and cSIGV_A are relatively small in this representation, and no clear systematic impact of coupling is evident at this scale." It is quite difficult to read the diagram and tell how different the simulations are from the observations. With a biased distribution, one could explain the drop in fractional skill score in Figure 6.
10. The diurnal cycle of precipitation in Figure 7 lacks units and has inconsistent scales. There appears to be significantly higher precipitation over land at noon, regardless of coupling, which is surprising, since about three-quarters of the domain is ocean and one-quarter land. Is this due to the land surface configuration RAL2-T?
11. Figure 12 shows the mean differences in latent heat flux and their decomposition into components: wind-speed contribution and moisture-gradient contribution. The coupled simulation indicates an increase in the upward flux of about 10-15 W m^-2. The decomposition into the two components highlights regional differences: wind speed confined to the Indian Ocean and moisture in the South China Sea. It would be interesting to the audience whether reduced fluxes arise mainly from weaker surface winds or weaker sea surface temperatures. For example, additional information showing differences in surface winds, Charnock coefficient, and sea surface temperature would help clarify. From a wave-field perspective, on the other hand, it looks like wind-sea contribution in the South China Sea versus swell contribution in the Indian Ocean. Clarification on this would greatly support the authors' claim for improved air-sea interaction in the region.

 cSINGV v1.0: Development and evaluation of a 1.5 km coupled atmosphere-ocean-wave modelling system for the Western Maritime Continent

Rajesh Kumar et al

This manuscript describes a technically substantial achievement: a kilometre-scale, convection-permitting, fully coupled atmosphere–ocean–wave modelling system for the Western Maritime Continent (WMC), built on well-established community models (UM/SINGV, NEMO, WW3) within the Regional Coupled Suite. The paper is well organised, the evaluation is multi-faceted (SST, Hs, precipitation spatial/temporal structure, near-surface variables, case studies, flux decomposition), and the scientific narrative (coupling enhanced LHF via SST-driven humidity gradients that sustained oceanic convection) is coherent and physically plausible. The manuscript can be acceptance in principle, but it would benefit substantially from the clarifications suggested below, several of which bear on the robustness of the central claim that coupling improves precipitation representation.

1. The uncoupled run uses SST prescribed and updated daily from OSTIA, while the coupled run has hourly two-way SST feedback. Since much of the paper's argument rests on diurnal-scale processes (nocturnal offshore propagation, diurnal LHF), how much of the coupling benefit could simply reflect the difference between daily-updated and sub-daily-evolving SST, independent of two-way coupling per se? Authors can add an explicit paragraph in the Discussion or Limitations acknowledging this confound and, using only existing model output, show the actual sub-daily SST variability produced by cSINGV relative to the daily-fixed OSTIA field, so readers can judge the plausible magnitude of this effect?

2. cSINGV_A uses a fixed Charnock parameter of 0.011, while cSINGV derives surface roughness from the WW3 wave state. Authors can diagnose, from the existing coupled output, the effective or implied Charnock parameter or roughness length produced by the wave-dependent formulation over the domain and compare its range with the fixed value of 0.011? This would allow readers to assess whether the two configurations differ sufficiently in roughness to explain part of the flux and wind differences. This caveat should also be stated explicitly in Sect. 4.3 rather than left implicit.

3. Figures 6, 9, and 10 report differences between cSINGV and cSINGV_A qualitatively. Using the existing daily time series already computed for these figures, could the authors apply a simple paired significance test, such as a sign test or day-by-day bootstrap of the RMSE or FSS differences, to determine whether the coupled advantage is statistically distinguishable from zero, particularly for the near-surface variables described as showing "limited sensitivity"?

4. All results derive from single deterministic six-month runs. Since generating an ensemble is not feasible given the computational cost of 1.5 km coupled simulations, the authors should add a clear limitations statement noting that the squall case-study differences (Sect. 4.4) and the regime-dependence conclusion are based on individual realizations of convection. Some of the apparent coupling impact could therefore reflect internal atmospheric variability rather than a systematic coupling effect. Tempering the causal language in this regard would strengthen the scientific credibility of the manuscript.

5. Sect. 3.1 indicates that a wave-only configuration (cSINGV_W) was run for the full January–June 2019 period, yet Sect. 4.2 validates Hs only for January, chosen for Typhoon Pabuk. Since these data already exist, could the authors extend the Fig. 4-type statistics (RMSE, r, SI, KS) to the full six months to confirm that the tuned betamax = 1.75 configuration remains appropriate outside the strong-wind period?

6. Sects. 4.1–4.2 evaluate cSINGV_O and cSINGV_W as standalone components. Since the fully coupled cSINGV run already exists and presumably outputs SST and Hs, could the authors add the equivalent RMSE and bias comparisons against OSTIA, CMEMS, and altimetry from the fully coupled run itself? This will clarify whether two-way coupling degrades or maintains ocean and wave fidelity relative to the standalone components.

7. Both configurations tend to produce more moderate and intense rainfall compared to the observations. Using the existing GPM comparison performed for Fig. 5, could the authors quantify, for example through a simple bias or intensity histogram, whether this wet bias differs systematically in magnitude between cSINGV and cSINGV_A? This would clarify whether coupling exacerbates or mitigates the rainfall bias.

8. The squall case studies rely on four events selected from the existing MSS squall database using a threshold of >50 mm h⁻¹, of which only two show a clear coupling signal. Could the authors report, how many total qualifying events occurred in the study domain and period? This would clarify whether the four selected cases are representative of the broader population or constitute a favourable subsample.

9. Since Sect. 6 frames cSINGV as a step toward an operational system, could the authors report the computational cost, such as core-hours and wall-clock time per forecast day, already logged for the completed coupled and atmosphere-only runs? This is a straightforward reporting that would substantially strengthen the paper's relevance to operational centres.

Minor points

• Figure 3 caption enumerates panels (a), (b), and (c), but the panels are unlabelled in the figure. Please add the corresponding panel labels.

• Please clarify whether the FSS in Figure 6 is computed within the red evaluation box shown in Figure 1 or over the full domain.

• A one-line in-text summary of Tables S2–S4, including the NEMO version changes and the betamax/shallow-water sensitivity results, would aid readability.

Recommendation

Minor revision. The manuscript makes a valuable technical contribution and is scientifically sound in its overall narrative. Addressing the confounding-factors questions and the statistical robustness questions would substantially strengthen the causal claims about coupling benefits, which are currently the paper's central scientific contribution beyond the technical system description.