The manuscript aims to construct a high-quality Rrs dataset covering coastal waters of both China and the United States. An AC processing framework that integrates standard NIR-based AC and grided aerosol model AC were developed. Moreover, cloud masking constraint was relaxed to get more effective data. While the topic is important as Rrs products in coastal areas is still have relatively high uncertainty as compared to it in open oceans, the high quality promoted by the title does not match the actual results for the following four major reasons.

1. Relaxing cloud masking constraint of course would enlarge effective data portion, but it would degrade data quality or accuracy of the Rrs products especially in clear waters. Therefore, this contradicts the goal of obtaining high-quality Rrs dataset.

2. The representativeness of grided aerosol models in instantaneous satellite images of coastal or shelf waters is questionable. This issue includes two parts. One is the temporal representativeness, since the grided aerosol models are multi-years and monthly averaged data, it can’t represent the instantaneous aerosol conditions. For example, even for some coastal oceans that are often affected by absorbing aerosols, the aerosol type may still be non or weak-absorbing aerosols in a specific day's image. In this case, if the average absorbing aerosol is still used, the accuracy may decrease. Another one is the spatial representativeness, aerosol types are high spatial heterogeneity, the large grid size (5 degree) makes it difficult to ensure that the grided aerosol models can represent the real aerosol types of each pixel, especially for grids with only one AERONET station far from the coastline.

3. There is no scientific basis for using water turbidity to distinguish between grided aerosol models and traditional AF10 aerosol models. The regulation mechanisms of water turbidity and aerosols are different, and there is no necessary connection between the two.

4. The standard NIR AC algorithm has a high degree of uncertainty in turbid water, which has been widely recognized. Therefore, comparing the results with NIR AC algorithm in this study cannot fully demonstrate the superiority of the proposed method. It is recommended to compare it with existing atmospheric correction results for turbid water, such as OC-SMART, POLYMER, NIR-SWIR, etc.

This manuscript describes a dataset of “high-quality” remote sensing reflectance (Rrs) over the coastal waters off China and the United States. Rrs is an important water property; thus, this manuscript is appropriate for this journal. In addition, the manuscript is well written, very easy to follow. However, this reviewer does not think it is ready for publication, at least for the present version.

Main issue:

1. The Rrs product of the global ocean, including coastal waters, is already distributed by NASA (and many other agencies) for decades. Publication and distribution of newly generated products should demonstrate an overall better quality compared to existing ones. However, as presented in Fig. 3, for the 8 wavelengths compared, NASA products show clearly better (in R2, MdAPE, NRMSE) quality for 4 wavelengths (412, 443, 667, and 678 nm), with the quality similar for the other 4 wavelengths (488, 531, 547, and 555 nm). These underperformances are also observed by the authors. In view of these, it would cause confusion to the user community if such a data product were distributed.
2. Fig. 8 shows more data products by the new approach. However, it is not clear if it is due to the new atmospheric correction algorithm, or due to the use of updated masks. If the NASA products were also processed by the updated mask, would the number of valid products be similar?
3. On a minor note. Why just the coast waters of China and the United States? Why not the coastal waters of the global ocean?

This reviewer believes that the above issues should be resolved first before the publication of this paper.

A few minor specifics:

1. Line 19: “(slope: 1.00 ±08)”. Statistically, the slope alone is insufficient to indicate error or uncertainty or accuracy.
2. Line 34, “Ioccg” should be “IOCCG”.
3. Line 148, “located over open ocean … were excluded.” What is the criterion to identify such “open ocean” waters?
4. Fig. 3. It is necessary to include units.