Through a dynamic downscaling approach, the authors provided a high-resolution (9km) climate dataset for CA covering 10 commonly used meteorological elements. The manuscript is well organized while requiring minor revisions.

• Title: In fact, this study does not show ecological or hydrological applications of this high-resolution dataset. Therefore, the phrase “for ecological and hydrological applications” should be removed from the title.
• Introduction: Why this method is used for downscaling of climate projections. A criticism-featured literature review on downscaling methods is needed.
• Model and experiments- Bias-correction technique: why did you choose these three GCMs (MPI-ESM-MR, CCSM4 and HadGEM2-ES)? The authors need to add an explanation in the manuscript.

Thank the authors and editor for bringing me to this work, the bias-correction method based dynamical downscaling is a totally new technique for me and very interesting. Base on this method, the authors created high resolution historical and projected gridded climate datasets in Central Asia (CA), which are very important and useful to the target region. The manuscript is generally well written and I have only some minor points as following.

1, I suggest to also provide the key geostatic variables from the WRF downscaling for data archive, e.g. topography, soil type, land cover type.

2, Addressing the data quality is very important, especially for a data journal. As a potential user, I expect to see the spacial distribution of biases (those versus CRU, ERA5 and stations), so I can have a basic impression on the accuracy of the data when applied to any sub-region or basin for ecological and hydrological studies. Thus, I recommend to add related figures, but you do not need to deliver in-depth scientific discussions. I also suggest to provide similar figures as figure 8 for other variables instead of only give precipitation as an example. So the readers can directly know the quality of each variable when compared with station observation.

You can add such figures as main content or even as an appendix is acceptable.

Additionally, I suggest to consider topographic elevation difference between WRF simulation and observation (CRU, ERA5, station) during comparison/evaluation, for these three variables: pressure, temperature, relative humidity. Because base on my own experience, this is very important over mountainous/complex-terrain regions.

3, Please clarify the origin of the station data during evaluation.

It is my pleasure to review the paper entitled “HCPD-CA High-resolution climate projection 1 dataset in Central Asia for ecological and hydrological applications” by Qiu et al. The authors produced a high-resolution (9km) climate projection dataset over Central Asia based on the dynamically downscaled results being combined with multiple bias-corrected global climate models. This dataset is assumed to serve as a scientific basis for assessing the impacts of climate change over Central Asia on many sectors.

Given the importance of this work and its potential impact, I would like to providing the following comments for improving the manuscript.

• The time range in 2.2 (1981-2005/2026-2050) is different from that in 2.3 (1985-2005/2060-2050). Could you please explain or make them consistent for readable?
• Fig. 2 should be explained in detail, especially for the right side of the figure. There are some sub-questions are occurred from this figure:

-How to achieve the action named “WRF with the optional combination of physical schemes” with the input from Bias-corrected GCMs and Observation data?

-The term “WRF with the optional combination of physical schemes” seems to run WRF model again, however, WRF model has been executed in previous step. This term can be renamed properly.

-Is the spatial resolution of Bias-corrected GCMs same to the one of the WRF model output? If not, how to match them?

• In Fig.3, the observational data for temperature is CRU data, the ones for precipitation and other variables are ERA data. Could you please give more reasonable explanation?
• It is recommended to do more comparisons between the simulation from control experiment and the one being combing Bias-correction GCMs, and to show the improvement by introducing the Bias-correction GCMs.
• Please use the data doi linkage instead of the uuid linkage for accessing the dataset in Lines 22-23 and Lines 187-188.

This manuscript describes a high-resolution climate projection dataset in central Asia, which could be potentially useful for hydrological or ecological applications. More proofs of its reliability and confidence levels are necessary to present before it could be used in the consequent applications.

The horizontal resolution of 9-km was used in this study. My question is why 9-km? 9-km is an awaked resolution not large as precedent quarter degree, not small as convection-permitting modeling. At this resolution, whether the cumulus parametrization is used or not is still an open question. 

The bias-corrected GCMs were used as forcing in this study. What does the simulation performance look like if the bias correction was not used? The previous study also claimed that similar performances were obtained in the Tibetan Plateau using either reanalysis or GCMs as forcing in the historical period. Major differences only occur in the temporal changes or linear trends. Almost the same historical simulations were also presented in figures 3-6 using three different forcings GCMs in this study. Therefore, it would be good to present the differences between using bias-corrected GCM or not. 

The future changes in 2031-2050 compared to 1986-2005 are studied. It is neither the end of this century 2100 nor the target year of zero carbon emission 2060. It would be better to describe the importance of this period in the future in central Asia. In addition, all the future changes are based on the model. Suggest adding historical changes or linear trends to solid the credibility or reliability of future changes.