The manuscript “Modeling Spatial Fields of Extreme Precipitation – A Hierarchical Bayesian Approach” by Rahill-Marier et al. introduces a hierarchical Bayesian model for modeling spatial rainfall for extreme events of a specified duration which can be considered in regional hydrologic models to perform a regional hydrologic risk analysis. The spatiotemporal dependence is modeled through multivariate normal with partial pooling for the marginal parameter (mu). The proposed model is used to model the spatial field of rainfall at all 9 stations in New York City. The proposed framework and its application to New York City are interesting and well presented. I have some (minor) comments and technical corrections, especially concerning the ability of the model to capture the spatial dependence structure.

1. I would like to see further analysis about why the empirical quantiles for Staten Island (SI) are consistently underestimated. For example, authors could analyze whether the correlation between SI and other stations is overestimated or not. That could explain the overestimation of the magnitude. In Figure 1, SI shows lighter colors pattern for different durations, which is the opposite of the Central Park pattern.
2. I suspect that it is the first one but would be good if the authors mention which model (partial pooling or no-pooling model). was used to generate figures 3 and 4
3. L129: Event?
4. L155: The authors say, “Fifty-four models (one for each duration and each site)”, but they did not mention which specific duration we consider. Please, specify this.
5. L158-159: mu and Sigma are leftover.
6. L165-166: Authors already mentioned it in L 160-161
7. L193: overestimated?

Comments on: Technical Note: Modeling Spatial Fields of Extreme Precipitation – A Hierarchical Bayesian Approach by Rahill-Marier et al.

The technical note presented a Hierarchical Bayesian Approach for modelling the spatial fields of extreme precipitation, aiming at simulate rainfall field when extreme rainfall occurring in one or more stations. Hierarchical Bayesian framework is applied for shrinking the rainfall in the space. In general, the technical note is well-written and mathematically rigorous. I am supportive of the publication of the manuscript if the following comments can be addressed.

1. The contribution needs to be highlighted in the introduction. To my understanding, the key difference between a tradition regional rainfall frequency analysis and this spatial fields analysis is the input data, while the Hierarchical Bayesian model is not very different. Thus, it is important to clarify the contribution of this work.
2. For a T-year return period, the return level at a gauge is usually calculated based on the quantile of the annual maxima, which should be independent of the model. In another word, even if calculated with different models or different subsets of data, the return level should be somehow consistent. On line 99, the logged-mean is pooled in the space, where annual maximum (from the target gauge) and non-annual maximum (from other gauges) are pooled. Since the non-annual maximum will usually be smaller than the annual maximum, will the T-year event be systematically under-estimated with the approach developed in this technical note? If so, there is a way of overcoming this shortage?
3. Line 30, the definition of R_djki is hard to understand.
4. Line 80. The simultaneous fraction analysis showed how many events are concurrent, while it does not provide any information on the spatial dependence of the intensity of extreme rainfall. Although considering spatial structure is important anyway, authors need to make a better justification for this sentence.

The present paper explains the break in Clausius-Clapeyron scaling rate in India through use of observation and a surface energy balance approach balance by thermodynamic. The reasoning and the basis of the work is fine, there are few comments I’d like the authors to address before publication (minor revision), as follows:

1. I ‘d like to know if the authors have made cross-validation when implementing the method.
2. Other than the present assumed distribution of Log-normal, are the GEV or GPD method also tested for comparison?
3. Line 10 “gaging site” -> “gauging site”
4. Line 27 “gage”-> is it gauge?
5. Line 30 the annotation is quite hard to understand, I have hard time transforming from R to A. I would expect a simpler annotation used than the one here.
6. Line 83, “across the gauge sites was developed”-> “that was developed”