The authors present new fluid inclusion isotope data for an exceptionally well dated stalagmite that had been studied for carbonate stable isotope and trace element compositions by Warken et al. (2020). They convincingly separate fluid inclusion δ²H and δ¹⁸O data affected by evaporation and discuss paleotemperatures obtained from samples which are thought to contain primary, unaltered water and calcite oxygen isotope compositions. The paper is well written, concise enough and easy to follow. The new data are valuable and worth to publish.

However, I see two main problems with the data.

1) The evaporation process. The inclusion data clearly reflect evaporation. Evaporation may occur in the atmosphere and on the surface. In this case the precipitating calcite can reflect the shifted water δ¹⁸O value, provided that increasing evaporation is not associated with increasing temperature, whose fractionation effect can counteract the evaporation related isotope shift, leading to constant δ¹⁸O in the calcite. The present study suggests cold and dry conditions, in which case the positive δ¹⁸O change in the water would be enhanced in the calcite due to the elevated calcite-water oxygen isotope fractionation, which is obviously not the case.

Another possibility is evaporation along the karstic water migration route. In this case evaporation would also be associated with enhanced CO₂ degassing and strong PCP. Warken et al. (2020) excludes this possibility as the Sr/Ca ratios are not changing along with the Mg/Ca values. Additionally, the water arriving at the stalagmite surface would already be evaporated and its isotope compositions would be shifted, that should appear in the calcite.

The next possibility is evaporation on the stalagmite surface and this is what the authors suggest as a potential process. However, it is difficult to imagine that calcite is precipitated from an unaffected solution, then the water film is evaporated, and the calcite precipitating from the evaporated water would no show any δ¹⁸O shift. Further, the low H₂O contents should be associated with higher δ¹³C values in the calcite. I tried to plot the water contents on the δ¹³C record presented by Warken et al (2020), and the relationship is not convincing.

I plotted the data and it is apparent that the water content and the fluid inclusion δ²H and δ¹⁸O data are negatively correlated (with r² values around 0.32 for logarithmic fitting). However, it is not clear why the evaporation from the water film on the surface would result in lower amount of fluid inclusions in the calcite. Evaporation should increase the calcite saturation, leading to faster carbonate precipitation and more inclusions trapped.

An alternative process is partial leaking of inclusions. As the sample chips were held in a carrier gas flow, they were most probably heated to remove the adsorptively bound water. Although this is not mentioned exactly, the authors mention 120 °C in the vacuum line to avoid water condensation. Heating speleothems to 120 °C may partially open large inclusions. The changes in climate conditions may change the carbonate fabric, inclusion entrapment, and hence the tendency of inclusion leakage during heating, as well. Partial removal of inclusion-hosted water due to inclusion leaking may lead to lower water content and shifted isotope compositions. This process is also supported by the large variations in water contents and isotopic compositions in the same sample (e.g. sample 24).

Heating the samples at different temperatures (80, 100, 120 °C) in a vacuum system and record the vacuum achieved, or measuring pieces from the strongly evaporated laminae at lower temperature (e.g., 80 °C) would provide means to investigate this potential measurement bias. Additionally, calcite and fluid inclusion petrography is essential in order to see if there is any fabric and textural change related to changes in climate conditions.

Further, the fluid inclusion data should be compared with the averaged δ¹³C, δ¹⁸O and Mg/Ca values in order to see if these evaporation- and degassing-sensitive proxy data are affected. Enhanced cave ventilation is mentioned in line 372 that should be supported by δ¹³C-δ¹⁸O correlation.

2) Paleotemperature uncertainty. The paleotemperature data are valuable, but their scatter is too high to discuss <2 °C differences between different periods. The authors state that the interstadial values are 1.6 °C higher than the MIS3 average. However, the interstadial average contains data from sample 24, which is very inhomogeneous. As the authors mention, some of the data were excluded from the calculations due to the evaporation affect, and the temperatures are taken as maximum values. However, if some periods are affected, and some others are not (due to carbonate texture differences), then the uncertainty becomes much higher that the 1.6 °C difference.

The authors state that the 22.0 ±1.8 °C temperature for the interstadial periods is comparable with the present day condition. However, the former is obtained from inclusion data, the latter is from direct temperature measurement. I know that it is not a nice suggestion to collect more samples from Puerto Rico, but if the authors accidentally have modern samples or at least Holocene stalagmites, only some measurements may strengthen this statement.

Addional comments:

line 155: „Sections with a relatively high water content, e.g., around c. 28 or 34ka...”. The water content at 34 ka is low, as far as I see.

line 194: It would be informative to plot curves of the Johnston et al. (2013) equation for different temperatures in Fig. 4a.

line 295: „Post-depositional enrichment of ¹⁸O in fluid inclusion water due to evaporation”. Post-depositional but pre-entrapment?

line 302: In order to support this sentence, the fluid inclusion data and the proxy records should not only plotted together, but a statistical analysis between these variables (with the high-resolution proxy data averaged for the inclusion sampling) is suggested.

line 332: „a drier state and weak convection”. If the reliable flinc temperatures are plotted together with the Mg/Ca record of Warken et al 2020, there seems to be a positive relationship that supports the assumptions presented here.

line 358: „The pattern largely follows the evolution of the high-resolution calcite stable isotope record (Fig. 2).” Looking at Fig. 2, this is rather risky to state. I made a linear correlation between the inclusion data and the averaged d18Occ values, there is no statistically supported correlation.

This manuscript presents a study of fluid inclusion isotope composition from a speleothem sampled in a cave from Puerto Rico, which today experiences seasonal ventilation variations. The data in this paper are high quality fluid inclusion dD and d18O measurements covering the 46-15 ka period made using state of the art methodology. The excellent team of authors show that the FI data can grouped into two distinct d18O-dD groups. One group falls on the global and local meteoric water lines. This group represents samples FI yield greater than 0.55 microliter/g. The second group, characterized by FI yields less than 0.55 microliter/g shows a strong deviation from the local meteoric water line in an apparent evaporation trend characterized by a dD/d18O slope significantly lower than that of the meteoric water line. The calcite of former group is proposed to form at, or near, temperature d18O equilibrium with the fluid inclusion water. The calcite d18O of the second group is similar in range to the first group, and consequently is in isotopic disequilibrium with the FI waters of the second group. The controls of the proposed evaporation process are climatic-driven ventilation during colder periods. Thus, the first group of FI waters and calcites are considered to reflect high cave humidity warm period conditions, whereas the evaporated samples reflect colder periods characterized be enhanced cave ventilation.

The kinetic mechanism proposed to account for these contrasting modes is very interesting. The speleothem calcite precipitates rapidly in both warm and cold periods in or near temperature isotopic equilibrium with cave drip waters. In the colder periods characterized by enhanced ventilation, evaporation of the dripwater films and the fluid inclusions trapping these evaporated waters occurs, yielding the deviations observed from the meteoric water line. However, the kinetics of the calcite-bicarbonate exchange is so slow that the speleothem calcite does not reset to the evaporated water values: the calcite thus retains the original 'equilibrium precipitation signature' attained prior to the onset of evaporation driven d18O changes. The model is well explained and the findings of the paper are both interesting and innovative.

This is an intriguing experimental data set, well-written, innovative in scope, and the manuscript certainly deserves publication. Nevertheless, there are several points that I would like to see addressed before the manuscript becomes published.

1. What is the advantage of using the Johnson et al (2013) thermometer calibration over the more generally used Tremaine et al (2011) calibration. The Tremaine et al calibration has become something of a standard in cave thermometry.
2. I have some reservation over the explanation of the proposed evaporation trend. The shift of d18O from the meteoric water line value is as much as 12 permil. This is a huge evaporation effect! The authors should use a Rayleigh calculation to estimate fraction of the original water that has evaporated away before leaving the residual water in the fluid inclusions of the evaporated samples. I would guess that the amount of residual water in the fluid inclusion is a small fraction of the calculated original water amount, and that the calculated original water amount is much higher than the water present in the first group samples, which contain about 0.5-2 microliter/g. The authors should explain how their proposed mechanism of fluid inclusion entrapment is compatible with high degrees of evaporation. Is there any alternative mechanism that could account for the apparent evaporation trend e.g, a process during analytical handling?
3. There is no graph of calculated temperature vs age showing the variations with time. I think that such a graph should be added to the paper. This graph could show the disequilibrium temperatures of the second group as well the "good' temperatures of the first group. Also, I am not completely clear how the authors calculated temperatures for the cold period samples. Did they use the water d18O value given by the intersection of the evaporation and meteoric water lines (Fig 3b)? If so, I would point out that one could run several different evaporation lines through the data. As Fig 3b shows, the evaporation line they calculated averages out quite widely disperse data.