This is the review for the paper titled “An in-situ daily dataset for benchmarking temporal variability of groundwater recharge” by Malakar et al. The authors estimated groundwater recharge per unit specific yield (RpSy) at the 485 groundwater monitoring wells across the United States. They adopted the water table fluctuation method on the daily groundwater table time-series. Here are my major comments:

- I like the way the authors represented groundwater monitoring well data in this manuscript. They performed analysis using daily time-scale data.

- The authors mentioned that the RpSy concept was introduced to reduce errors associated with the recharge estimation arising from the uncertainties in the specific yield (Sy) data. RpSy is nothing but the dh – the change in water table. I think this may not provide the representative values of change in recharge rates. For example, groundwater recharge signature in alluvium will differ a lot from the hard rock areas with a similar change in water table. I think the authors can reduce emphasizing RpSy as a central point of this manuscript, rather, focus more on creating the unique database. 

- Section 3.1 and Figure 3: Correlation between RpSy and USGS-based groundwater recharge show less than 0.5 values across the majority of the mid-western, dryland areas. I understand USGS-based recharge estimates using the water budget approach, can the mismatch show bias in precipitation or any other data? The magnitude mismatch is understandable, the patterns should match unless uncertainties present in the data. Authors may consider using other recharge data for comparison in those areas.

- Figure 4a and 4c look similar to me and they are not reflecting the patterns observed in Figure 4b and 4d. Please revisit the figures.

This paper provides a useful dataset of estimated daily recharge per unit specific yield (RpSy) across 485 well locations in the US derived using the water table fluctuation method on daily groundwater table time series. Overall the paper seems to be a useful contribution, but I suggest that the following aspects are further considered before publication:

• To understand interannual variations in recharge (Fig. 5) would it be useful to not just consider a timeseries plot, but also make scatter plots of drivers (PPpt, or Ppt-ET) and responses (recharge). This may help to better understand and quantify their linkages.
• In figure 4, panel a and d look completely identical but should be different. Check if an presentation error is made here. 
• the quality of all figures rather limited. For example, Figure 1: make the timeseries somewhat more readable (using a highjer resolution figure output and a slight change of line styles may help here. Please check all figure to potentially up the standard. 
• For the comparison of center of mass, please make more direct comparison of these datapoints than just maps. Also note that a “centroid” does not match “the date on which the cumulative value is half of the yearly total in a water year”. A centroid represents the mean of a cumulative (recharge) distribution whereas the textual description would represent the date of that matches the median.

This is a review of “An in-situ daily dataset for benchmarking temporal variability of groundwater recharge” by P. Malakar et al. This paper describes the development of a benchmark dataset of groundwater recharge per unit specific yield (RpSy). The authors apply an established Water Table Fluctuation / Master Recession Curve method and QA/QC measures to produce daily groundwater level variation time series data for 485 sites.

As the authors note, there are currently no daily timescale data sets in the literature with which one can compare estimated/modeled results for groundwater recharge. Due to usefulness of the results, I recommend acceptance of the paper after minor revisions.

My main request is that more analysis and discussion be given to the RpSyu data, which is the version of the data that includes time varying specific yield, and which is provided in this data set alongside the RpSy data. The authors state that because the correlation between these two data sets is greater than 0.8, they will not include RpSyu data in the comparisons with USGS data, Ppt, ET, etc. It is not surprising that RpSy and RpSyu would generally correlate with each other, but without a comparison between RpSyu data and the USGS data, a user cannot determine whether the additional complexity of the RpSyu calculation adds any value. Some analysis here would help the user decide whether to use the provided RpSy or RpSyu. At a minimum, the map of R2 between the USGS data and the RpSy data should have an equivalent map for RpSyu, and there should be some summary statistics that help readers understand whether RpSy or RpSyu more closely matches the temporal variation in recharge.

Other comments:

• The interannual variation data in Figure 5 should be presented in table form, showing the R2 between the data sets (and including columns for the RpSyu data).
• Can you comment more in the discussion on the appropriate way to use these data to evaluate models? You make some mention already, but more clear statements on this point would be useful. Such as, temporal variation but not magnitude between these data and recharge estimates, what to do if you have some specific yield numbers to apply for a given area, etc.
• Figure 4a and 4c are identical – must be some error in the figure production.

The paper has a few minor/grammatical errors and could use a proofreading. A few examples:

• L25: strike “the rate of”
• L36: “in East Africa is” to “in East Africa are”
• L36: “ratio” to “fraction”