Based on the intelligent dasymetric mapping (IDM) method, the authors generated a set of high resolution population density product by using various datasets. This is interesting and significant since population change impacts almost every aspect of global change. However, it is noteworthy that the method described in this paper is not clear. To make it more persuasive, some suggestions includes:1) To better present how the data was processed and used, a flow chart is recommended; 2) It’s better to list the spatial resolution for the raster data, and the feature types of vector data, such as point, line or polygon.

The authors describe the application of the IDM method to further improve the existing EPA EnviroAtlas Dasymetric Population Map for conterminous United States, mostly through the further expansion of uninhabitable areas. The resulting data product shows decreased estimation error in an assessment using census tract data as input and blocks as test units. The fine resolution of the gridded population density estimates will be very useful for various applications as described in the introduction of the manuscript.

While the value of such an improved dasymetric map is unquestionable given the resolution and the spatial extent (CONUS), there are some reservations that the authors are encouraged to address:

• The result of defining more extensive and spatially expanded uninhabitable areas and using them for dasymetric refinement is very intuitive and expected. This strategy is about using robust constraining variables to reach estimates of higher accuracy.
• This makes this study a rather incremental step to revise and improve an existing data product by including additional data layers that allow to identify more uninhabitable areas.
• IDM is a slightly dated (though still effective) method for population refinement. Given the nature of the data and the statistical frameworks available it is surprising that the authors have not tested different methods (e.g., machine learning) to cross-compare performance and resulting data quality. The given problem at hand seems to be one that could be treated effectively with learning frameworks.
• Estimating the average population density of target zones across a whole state is not effective and represents a well-known limitation of IDM. The authors mention the further refinement of their estimation procedure for urban and rural classes as future work. However, it seems that this would have been an important step already in this data product revision. The estimation of population densities for different urban classes (intensity of development) will result in much more accurate data layers. This would allow the model to proceed at the scale at which variation is observed across counties and states.

General: The different data layers and processing steps need more precise descriptions. The authors are also encouraged to provide a broader background on dasymetric modeling including uncertainty concerns and typical limitations reported in the literature. Also, there are other fine-resolution population datasets, and it is recommended that the authors make sure their data layers are compared to those existing products.