This study produces high spatial resolution climate datasets including 15 variables as complement to already existing temperature and precipitation datasets over the recent past period. For the future period, the delta-change method was used to obtain the future climatic anomalies of precipitation and temperature. Such dataset is indeed really important as it offers us an opportunity to understand the climate dominance on other processes at higher spatial resolution and it can also facilitate the high-resolution process-based model simulation as forcing datasets. However, I find the data description mainly focused on northern hemisphere, and some sentences are just common sense, which should be revised a bit. I have a few concerns about the dataset validation and its performance in terms of the comparison with other existing available products. The validation part should be extended to include the comparison at different time scales: seasonal cycle, inter-annual variability, etc as the monthly dataset is provided.

Major comments:

1 Please describe the method you used to create the dataset in abstract as well (in one sentence) about interpolation.

2 The author spent a lot of words on the performance of each climate variable, but not on the process to create high spatial resolution dataset. Does the interpolation really introduce new information during the downscaling process?      

3 Please use a table to summarize the dataset you used including the external data, specifying the spatial and temporal resolution, time period, and the source.

4 Can you provide other evidence to justify the process of wind speed data?  

5 The validation part should be extended, including comparison with other climate datasets not only the station-based measurement.

6 I see that some figures show the difference between climatological means of 2071-2100 and 1981-2010, and others show the range of monthly means for the period 1981-2010. The former one is for variables that use temperature and precipitation only, right? You can also show range of monthly means for the common time period.  

Minor comments:

Line 33 live -> lives

Line 35 key to what

Line 51 please add references for the usage in macroecology

Line 59 please use unit mm yr^-1

Line 99 why the period 1981-2010 was chosen rather than 1980-2018?

Line 150 to amount of

Line 189 Please also mention the ERA5 spatial resolution here.

Line 231 Why is clt downscaled to 1.5 arcmin here?

Line 481 is it really an east-west belt from Ethiopia to Sierra Leone?

Figure 2 why did you only focus on the northern hemisphere?

Figure 4 Have you checked whether the wind data also shows stilling reversal around 2010 like Zeng et al (2018)?

Line 560 is this potential NPP based on Miami model really meaningful? What are the benefits of this NPP dataset relative to CMIP5/6 outputs?

Line 716 please also specify the comparison time period.

Can you use other related (not exactly the same) variables to validate the performance of fourth-order and fifth-order climatic variables?

Line 743-744, can you provide references more recently?

Reference 

Zeng, Z., Ziegler, A.D., Searchinger, T. et al. A reversal in global terrestrial stilling and its implications for wind energy production. Nat. Clim. Chang. 9, 979–985 (2019). https://doi.org/10.1038/s41558-019-0622-6

It is indeed very useful to develop a number of climatic variables at the interface between ecology and climate science. Such efforts can act as a link, connecting the two disciplines and helping ecologists include climatic variables into their modelling efforts, at a very high spatial resolution. Moreover, projects like this can encourage the collaboration between climate scientists and ecologists, further developing the field of "climate change ecology". Even though this study is ambitious and could potentially assist multiple ecological studies, it is lacking in terms of validation.

Major comments:

1. The fact that the data from the five climate models, and from the three scenarios (SSPs) are included independently is positive. It allows users to quantify the uncertainty introduced by these two sources (model spread, scenario uncertainty). More models (perhaps in future releases) would be even better.

2. The validation presented in this paper is not convincing enough. Fifteen pages of the results section are used for describing and visualizing the variables, while only one page is devoted to validation of the data. A simple description/visualization does not add much value, and instead a comparison to the station data would have been more useful here. For example, Karger et al. (2017), compared their datasets with station data and other products using maps (spatial) and time series (temporal). This helps in gaining confidence to the CHELSA data by seing the biases over space (maps) and time (time series). For example, can two ecological studies, one in Africa and one in the Arctic, use CHELSA data with the same level of confidence? Can one study use CHELSA data for spring and for autumn with the same level of confidence? 

3. It is understandable that only a number of variables (8) that are more easily measurable were validated (mostly 1st and 2nd order). It is these eight variables, after all, that are combined in order to create the other seven variables (3rd and 4th order). However, validating more than eight variables would give more confidence to the user. For example, in the case of npp, could remote sensing products be used? https://neo.gsfc.nasa.gov/view.php?datasetId=MOD17A2_M_PSN

4. Only a global value for r, RMSE, MAE and bias is provided. What about different regions or different seasons? Do all regions/seasons show similar spatial and temporal patterns?

5. Throughout the paper, the words "high spatiotemporal resolution" are used. The spatial resolution is definitely high. Is the temporal resolution of this dataset (monthly) considered high for ecological applications?

Minor comments:

Line 33: live -> life

Line 35: This dataset's highest temporal resolution is monthly, is this enough for detecting extremes? Please provide examples and/or references

Line 71: Same comment as in Line 35

Line 80: Please provide references

Line 447: Please explain how r, MAE, RMSE and bias are calculated. Did you use time series from each station, and then averaged them globally to obtain a single number (the number shown in Table 1)?

Line 732: Please provide references for "Moreover, they are comparably robust to gaps in the network of meteorological field stations and therefore provide more reliable estimates in remote areas, compared to alternatives from station-based interpolations"

Line 762: "... realistic patterns in our exemplary remote-area, high-resolution maps in ...". Realistic compared to what? Please define "realistic" here.

Reference:

Karger, D. N., & Zimmermann, N. E. (2019). Climatologies at High resolution for the Earth Land Surface Areas CHELSA V1 . 2: Technical specification. Scientific Data, 4(April), 170122. https://doi.org/10.1594/WDCC/CHELSA