Review of: SISALv3: A global speleothem stable isotope and trace element database by Kaushal et al

Kaushal et al present an update of the SISAL database, adding many new records and incorporating new isotope proxies including Mg/Ca, Sr/Ca, Ba/Ca, U/Ca, P/Ca and Sr. These additions necessitated a slight restructuring of the database, which is available as a MySQL database, and csv files. In the paper, they have detailed all of the data fields in the database, as well as highlighting the locations of the data and the time periods they cover. In addition to the speleothem data, they also include land cover and vegetation data, which is useful for interpreting the data.

The editor, Dr. Wang, asked me to provide a more formal review after my previous comment.  This is the third major iteration of this database, so at this point it is quite mature and useful. As mentioned before, it is a great credit to the speleothem community that updates to this database happen frequently, and I look forward to the eventual version 4. I have found it useful for my own research, and I am sure many others do as well. After a more detailed read of the manuscript, I think expanding on the following points would be useful to add to the manuscript to aid those who want to use the database (in addition to the previous point I made of including a table of the new records).

•  With all the new elemental proxies added, it would be useful to include a table describing what climatic variables the proxies are representing. This would strengthen the stated goal of the paper to assessing hydroclimate variability
•  I believe there has been an online interactive version of the database made available since the publication of SISALv2. It might be a good idea to have a short section explaining how that can be used to access the data.
•  The first version of SISAL came out a few years ago. Having a short section showing some example applications of the database would highlight why this database is so important.

Best Regards,

Evan J. Gowan

Kaushal et al. describe the newest version of the SISAL database (SISALv3). I have found previous versions of the SISAL database to be incredibly valuable in my research, and the numerous of publications related to SISAL data (as listed in the introduction) also speaks to this fact. The compilation and standardization of so many records is impressive, and this paper builds on that previous work. 

Overall, the manuscript is well written and provides extensive detail about the changes between SISALv3 compared to SISALv2. The most significant update is the addition of trace element and Sr isotope records. As the authors note, the inclusion of trace element data will help to researchers understand hydroclimate and hydrological processes. New stable isotope data are also provided in the updated database. Minor changes to the database structure and metadata are also discussed. These include the addition of land cover and vegetation metadata fields. The changes represent a significant advance and make SISALv3 a valuable contribution to the field. 

Below is a list of minor comments about the manuscript which I believe will help clarify the text and improve the usability of the SISAL database:

1. Line 69. SISAL does not include climate interpretations. However, both the abstract and introduction justify the need for adding trace element data to the database as “hydroclimate-sensitive geochemical proxies”. As SISAL data is used by a wide-variety of researchers, including those with limited speleothem expertise, expert-defined climate interpretations would be valuable to prevent misinterpretations.
2. Line 81. Does SISAL indicate which data are at seasonal resolution?
3. Lines 144 and 146. Why were the radius of 250 m and 1000m selected? What is the spatial resolution of these datasets?
4. Line 208: It should be clear how records from previous versions have been updated. How does SISAL version control and where is this information listed for users to access?
5. Line 211: I suggest changing the caption from “…datasets in SISALv3” to “…datasets in SISAL” as there does not appear to be any change in this process from previous versions. More generally, the text of section 3 could be streamlined as there appears to be minimal differences from previous versions other than that there is new data to QC.
6. Line 215: This table is useful, but it would be more valuable to know which records have been modified. For example, which 4 time series were altered? I see that there is some documentation in the notes.csv, but this is not easily searchable to find this specific information.
7. Line 222: What is the 60% number based off?
8. Line 230: My understanding of Table 1 is that data with trace element higher resolution values would be identifiable by the “trace_elements_metadatafile” column of the entity table. However, when I open this csv, there are no rows with a “yes” value in this column.
9. Line 250: It is unclear to me why the authors chose to show the increase as percentages rather than as count. A count seems more relevant to describing the actual data.
10. Line 255: A map of trace element records would be valuable to see.
11. Line 268 (Figure 7): It is unclear to me why Figure 5 and Figure 7 are so similar but have different legends. Couldn’t both maps shows SISALv2 data in orange, new data in purple, and other records in black? It would also be interesting to see which sites had alterations (as listed in table 3).
12. Line 278: Do “notes” include any information about past vegetation and land use changes or is everything based on modern?
13. Line 285: The Figure 9 captain is missing a description of panels C & D. The transparency also makes it difficult to see which vegetation type corresponds to each dot.
14. Line 295: It is unclear to me how “using all cover fields together” would be manageable in a programmatic big data analysis. Could the authors provide context to how this may be done?
15. Line 334: Will the Rehfeld et al. (2020) zenodo link be updated to include age model ensembles for new sites?
16. Line 333: To encourage proper credit of data creators, the text “If individual records are extracted” should be revised to clearly state than any analysis of SISAL records should credit the original publications. The current text seems to suggest that data credit is only needed if a single record is extracted. I would also like to see a list of references for new data in SISALv3. 

I also have several additional comments about the data and code archived at https://ora.ox.ac.uk/objects/uuid:ea1c56f7-384f-4df0-8525-6be00ab85870  

• Please clarify if the code is final or if these scripts will be updated before publication. For example, the sisalv3_code is provided at the dataset link. This name implies that the scripts are new, but it is not clear that these scripts are any different from previous SISAL versions. For example, line 61 of the sisal_connect2db.R file is: “dbname='sisalv2',”. 
• Apologies if I missed this, but the README_SISALv3.pdf would be strengthened with an explanation of each zip folder provided. Specifically, it is not immediately clear to me what the sisalv3_standardised_trace_elements_datafiles.zip provides. My assumption is that these are the data measurements on their original depth (before downsampled)
• It is also not clear to me what the purpose of each file within the sisalv3_standardised_trace_elements_datafiles.zip is used for. I opened a random folder (sudwala) and looked at 2 excel sheets (trace_elements_datafile_sudwala_SC1_standardised.xlsx & trace_elements_datafile_sudwala_SC1.xlsx). As far as I can tell, these are the same data. 
• The authors provide several scripts which can be used to analyze the sql data in R/Python/etc. These include three R scripts (which I primary focused on for my review) which demonstrate how to load the data from MySQL, save the data as csv, and load the csv files. These scripts also provide the code for reproducing several figures from earlier SISAL publications. However, given that this is the 3rd version of the SISAL dataset, I am surprised to not see increased documentation about the usability and functionality of the dataset. There are two “low-hanging fruit” ways to improve the code documentation which immediately come to mind:
  
  
  
  
  • The first is to transition into using R Markdown / Jupyter notebooks. I find these resources extremely helpful when referencing code from other projects as I often want to see the output of the code before trying to reproduce the results myself. Although this suggestion may seem trivial, the stated goal of SISAL to “make speleothem data to the wider paleoclimate community” necessities providing detailed documentation for accessing the data. 
  • The publishing of a new SISAL database version seems to provide an opportunity for create a central repository for code associated with analyzing SISAL data. This is especially relevant given the choice of SISAL to store the data in an SQL format. As at least one reviewer on a previous SISAL paper has noted, this is a data format that many (including me) are unaccustomed to (https://doi.org/10.5194/essd-2020-39-RC1). Therefore, providing more resources which demonstrate how to access and use the data would be valuable. Furthermore, there is now a long list of projects using previous versions of the SISAL database (as listed starting on line 90 of the main text) which could be referenced in this central repository. 

The present manuscript has documented the SISALv3 database, which includes new trace element data and more stable isotope records. The inclusion of trace element data adds value to the understanding of hydrological processes in the karst and cave. The authors also documented the revisions to the data structure. The updated database represents a community effort in synthesizing speleothem paleoclimate data with a global scope and thus is of great value to the community. The database can be used to benchmark climate model simulations and to study past changes in climate and hydrology.

The manuscript is very well written and fits nicely with the scope of Earth System Science Data. Below are a few comments that need to be addressed before the publication of this manuscript.

1. Since goal of the database is to allow “comparisons with isotope-enabled climate models and other earth system and hydrological models”, it would be very helpful if the authors could provide example analysis code for this task. Specifically, the Last Glacial Maximum and the mid-Holocene have been the research focus of paleoclimate community for several decades (e.g., the Paleoclimate Modeling Intercomparison Project (PMIP); Kageyama et al., 2017, doi:10.5194/gmd-10-4035-2017; Otto-Bliesner et al., 2017, doi:10.5194/gmd-10-3979-2017). I would suggest the authors provide example codes (preferably using open-source program languages) for fetching stable isotope and/or trace element data for these key time periods and making simple plots. From the paleoclimate molders and proxy experts whom I have talked to, the SISAL database is not easy to use, in particular for model-data comparison. Providing example codes for PMIP users would greatly broaden the use of the database in paleoclimate research.
2. The database has undergone multiple version updates in the past few years. It would be interesting to briefly discuss the future plans from the working group. For example, is there plan for a new release in the future? Moreover, the author mentioned that the current database has included ~60% of the known published data (Line 222), what are the challenges in including more dataset? Do the working group have a plan to cope with these challenges and increase the data inclusion?
3. Line 72: For clarification and comparison, please consider listing either the number of stable isotope and carbon isotope records in SISALv2 or the number of newly added records in SISALv3.

This is an exciting new compilation of global speleothem dataset that adds significant amount of new proxies and records to previous versions of SISAL datasets. The inclusion of published trace metal data in speleothems is particularly important  as it may be one of the most important subjects in future speleothem-based paleoclimate studies. Putting together published δ¹³C data is also important, as it is underutilized in previous speleothem studies in general, and may provide useful information about climate-ecosystem interactions. What makes the compilation even more impressive is the inclusion of background geological settings for the cave sites as well as available maps and speleothem images. With a comprehensive compilation of existing data, this work also points out directions for future speleothem research in terms of regions/time intervals/new proxies. I have looked through the online version of the dataset and believe sufficient instructions are provided for new users to both search the dataset and apply the codes provided. I think it is a high-quality dataset and can be published as is.