General comments

The manuscript “A new snow module improves predictions of isotope-enabled MAIDENiso forest growth model” describes a new approach to simulate a soil moisture taking into account off-growing-season precipitation (snow, ice) and snow melting. The authors have made a considerable effort to improve the soil water block of the MAIDENiso by introducing an one of the principal water source (snow/ice) for trees growing up in boreal zone. The paper is well structured and written. The introduction provides a comprehensive overview of the background information and the pertinent literature, and demonstrated the need for the current study. They involved a wide range of statistical and modeling techniques for checking their hypothesis and confirming their results. The new block of the soil water procedure improves the understanding some biological and ecological processes, particularly the influence of winter precipitation on delta18O.

But there is a serious issue which should be considered in the MS. Calibration of any multidimensional model can be considered as a fitting observed values of some process by the model optimization. But independent validation (or verification) of the model is using to estimate a forecast ability of it. In the MS I did not find how the new block was validated independently. Based on my own experience we can fit even a stochastic process by multidimensional model using advanced optimization procedure very well. But we are not able to forecast that process (we cannot fit stochastic values by the calibrated model on independent time interval). To be sure that the model is able to forecast I recommend the authors to calibrate the model on some time interval and then check how the calibrated model works on independent time interval.

I would suggest to publish the MS after major revision.

Specific comments

Section 50: The authors wrote: “Among the previously mentioned models, the only exception is the Biome3 model (Rathgeber et al., 2003)...” It is not correct! As an example, the VS-model has a sub-block of snow melting (see Vaganov et al., 2006; Shishov et al., 2016). Could the authors introduce additional references there?

Paragraph 2.1.1: I would suggest to make the Paragraph 2.1.1 shorter, and simplify it by moving some formulas and description into appendix section. The well-known MAIDENiso is not a topic of the MS discussion.

Paragraph 2.1.2: How many new parameters involved in the soil moisture block? I would suggest to introduce a table there with short description of new parameters (included their dimensions).

Section 170: Is the new code of the MAIDENnIso available in some public depositories?

Paragraph 2.2 Along with calibration I strongly recommend to describe a validation procedure of the model.

Paragraph 3.1 I would recommend to calibrate and validate the new model on two independent time intervals.

Section 275. The authors wrote: “The values obtained for both sites were quite different”. What does "quite different" mean? Are they statistically different?

Section 275: This part “This potentially indicates that the physical factors that are neglected when assuming a constant ra (such as wind speed) are quite different between the two sites. The higher values at Tungsten indicates that snow sublimates at a slower rate, likely because the average wind speed during winter at this site is smaller than at Caniapiscau” is an interpretation which can be moved to discussion section

Sections 280-285 can be moved to Discussion section.

Section 295. It seems to me the simulation GPP results with(without) snow are statistically the same (or even identical). Could the authors check it and indicate this in the MS?

Section 315 can be moved to Discussion section.

Section 330: Could the authors statistically check a difference between two simulations (Fig. 8c,

Figure 5: Where is the red curve on the Figure 5? Possibly, both versions (with/without snow) produce the same GPP results.

General

This paper is about the modification of the MAIDENiso model to incorporate a new physical based snow module, including a pool of ice in each soil layer. The paper is well structured and written, considers earlier work quite well, and clearly shows improvements of hydrological simulations and stable oxygen isotopes in tree-ring cellulose.

The topic is interesting for the journal audience, however, the results and thus conclusions are solely based on a calibration strategy without including a validation of the calibrated model (also pointed out by Anonymous Referee #1). I also recommend, to include a validation period to check the calibrated model (the considered time series lengths are long enough to support this kind of analysis).

Specific comments:

• Page 4, lines 91-94: what type of mixing of soil water is incorporated in the model (for example between mobile and immobile soil water domains)?
• Page 8, line 226 and 229: the GPP data from stations seems quite far from the study sites (1023 km and 650 km), with the assumption that obtained parameters were similar at the studied sites. Could you please elaborate a bit more why this assumption is (probably) valid?
• Page 8, line 245: besides NSE, the Kling-Gupta efficiency (KGE), that is increasingly used as an alternative metric in hydrology, could provide useful additional information.
• Page 12, paragraph 4.1: for the Discussion, but also for the results section, a nice addition is to provide delta18O for the hydrological outputs of the model. Many studies have reported that groundwater or streamwater does not show an evaporation fractionation signal compared to (top) soil or trees. Adding this to the analysis would make this study even more attractive to the more hydrological oriented readers of this journal.
• Page 12, paragraph 4.4: The current study could also be extended to sites that are snow dominated in the winter, but do have a Mediterranean climate, where trees are more dependent on water derived from snow melt, as this was not the case for the study sites as part of this study.