General comments:

The paper addresses research questions which are both important and relevant to the scope of HESS. Through comparing several modelling approaches with different types of data, this study presents novel ideas and data to demonstrate the importance of including lateral groundwater flow in models representing soil moisture, particularly when simulating saturated peatland conditions and lowland forest. These results are significant given many models do not currently include lateral groundwater flow, and the importance of accurate simulation of soil moisture. This is of particular importance for peatlands, given their role in mitigation of climate change impacts.

The paper is well-structured and written clearly and concisely. The abstract summarises the research well and the introduction reviews the current gaps in the literature, explaining the need for this research. Methods and assumptions are clearly defined, and the results support the conclusions with well-presented figures and analysis. Limitations to the study are well-described and supplementary materials clear.

I have a few specific comments, but otherwise an excellent paper. 

Specific comments:

Abstract - Could highlight why this research is needed within the first couple of sentences.

Line 104: “at range” –> across/over a range

Line 116: plant Latin names in italics normally?

Figure 1. It’s not very clear how a) fits into b) and the scale of b) compared to c). It would also be nice in figure 1a) to see which is the forest site and which is the mire site (perhaps clarify in the figure legend in brackets). Perhaps also add the river network over the top so this is clear.  

Line 144: I assume the above-ground fluxes and state variables computed in the canopy submodel are the same for all 3 versions of SpaFHy? I would just add a sentence to make sure this is clear.  

Line 227: Did you do any sensitivity tests to confirm that depth-to-bedrock was not important? Or was 5m decided based on any other studies/references?

Line 242/Figure 1: Not very clear which soil moisture sampling locations are those bi-weekly vs the 56 additional locations in the figure. I can also see there are locations labelled with m and l, but you don’t explain what these two stand for.

Line 269: “of midday” -> at or for midday

Table 2: Figure legend above table rather than below

Figure 3: Please add in the figure legend for A) as observed (obs) and simulated (mod). In addition, there is no SWE in the figure. For B) and C) figure legend I think as you have specified what each parameter is in A, it makes sense to do so here too i.e. Air temperature (T)

Line 313: “groundwater rechange” -> groundwater recharge

Figure 5. figure legend – add “(obs)” for “in-situ measured”  

Figure 6: as previous plots have low=dark blue, high = yellow, this should be consistent for canopy fraction too

Line 384: “follow mostly” -> mostly follow

Line 508: I assume there was no groundwater level data to compare with model simulations.

Review of HESS-2024-81

I read this manuscript describing a modelling investigation of lateral groundwater flow in a subarctic catchment with interest. The authors use multiple model parameterizations to quantify hydrologic fluxes and states in a sub-Arctic catchment, with a focus on soil moisture. The results point to the importance of model inclusion of lateral groundwater to improve simulation of soil moisture, with the SpaFHy-2D set-up generating a greater range in soil moisture conditions than the other two parameterizations, most notably increasing moisture levels for peatland soils, which are expected to have higher saturation.

Overall I think the work has merit, is suited to HESS, and can be of interest to the readership of the journal with revisions. I found the paper to generally well written, especially the introduction, which provides the reader with a generally clear picture of the research needs and direction. I do however highlight below a number of areas concern, including a caution in how the model results have been evaluated. The model, despite the potential afforded by the new parameterization approach described here, is nonetheless not demonstrating particularly strong ability to simulate soil moisture, and a more rounded presentation of model performance in analyzing the results is merited. The authors point to the potential for enhanced process representation as being potentially needed, which I would agree with as cold-regions processes, including snowmelt dynamics, soil freezing and infiltration inhibition, and more are not yet considered in the model. Here I summarize some of the overarching areas for improvement that I suggest for this work.

1. In the methods, additional detail on what processes are captured and how they are represented in the model are warranted. How are precipitation and snowmelt represented to characterize hydrological dynamics of upper layer? Does water percolating through the soil bucket immediately reach the catchment outlet once it drains below this layer, or does the model account for transit time to the outlet, and if so how is flow routing handled? How are snowmelt dynamics modelled? Do frozen soils prevent infiltration?
2. Section 2.5 While there are a wealth of soil moisture measurements observations described here, it is not clear how these correspond to the modelling framework. Which depths represent the organic layer, and which depths represent the rooting zone? Section 2.4 makes effort to descript the horizontal resolution of the modelling, but the vertical resolution has not been described. This detail is important in linking observational data to the modelling approach, and a fuller explanation here would be most helpful.
3. In the results, there is a pattern of the results paragraphs starting with pseudo figure captions rather than topic sentences. As a result, it is not easy for the reader to easily parse out key themes from the analysis conducted. These sometimes appear later in the paragraphs as well. Removing these would improve conciseness and provide a more direct overview of the findings emerging from this work.
4. In the results, there is need for a more systematic approach to presenting evaluation metrics. While overall metrics are presented for some of the model fits, the results describe permutations of these relationships that are not quantified in the text. Importantly, some of the description of the model performance is not well supported by the analyses shown (see detailed examples below). That the model, even with 2D representation, does not perform particularly well should be emphasized, as there is lots of direction provided in the discussion about how this work could be improved in the future. It is important that the model performance, and its limited ability to capture soil moisture dynamics be described with appropriate supporting quantitative metrics.
5. While I understand the intent behind including the SAR data, my assessment is that they are being relied on too heavily in this work. Given that the SAR data do not capture very well other observations of soil moisture, e.g. due to spatial differences in representivity, there is limited potential in using them to assess model performance. This is an area where the paper can be streamlined, perhaps by earmarking some of this for the SI rather than the main text.

Line comments:

Introduction

L23: “region, climate change”

L27: “affecting tree health, mortality”

L33: Here it might be helpful to expand on the C aspects a bit. This are of research is important for understanding GHG exchange from terrestrial environments, but also lateral export of DOC (and nutrients, as noted) to waterbodies.

L41: I wonder if undulation is the best term to use here. I tend to think about undulating surfaces as occurring over space, but for groundwater it seems implied here that this is temporal variability in groundwater height at a given location, rather than an undulating groundwater surface that rises up and down repeatedly along a linear plane.

L58: “, and extend point-scale…”

L93: perhaps use “shallow soil moisture” here. Also, I think you can simplify this question to one of ‘where’, since, as worded, it seeks to look at temporal variability. In my mind, this makes the ‘when’ part of this question redundant.

L96: perhaps it is worth adding here an investigation of the accuracy of the SAR measurements using point scale observations. This seems a prerequisite to using soil moisture estimates to evaluate model predictions.

Methods

L107: can you include the proportion of precipitation falling as snow? This would be helpful.

L109: use elevation instead of altitude (which generally refers to height above the ground surface).

L111: the image and text below suggest roads and ditches as potential human disturbances. Perhaps the extent of human disturbance (while still small) could be described in more detail here.

L118: Above (L116) the scientific name is used, but here the common name appears. Suggest defining first and using standardized naming convention for all plant species. Perhaps the journal has a convention for this.

L129: What soil depth is used/modelled here to represent the rooting zone?

L147: It seems strange that precipitation and snowmelt are not also represented here to characterize hydrological dynamics of this upper layer? Does water percolating through the soil bucket immediately reach the catchment outlet once it drains below this layer, or does the model account for transit time to the outlet? How are snowmelt dynamics modelled?

L181: Can you provide more detail on constant h in streams and ditches, as this seems to be a strange assumption to make as these would be dynamic in time and space. Perhaps these are not spatially explicit in the model, but the reader would benefit from a fuller description here, as it is not clear how catchment discharge is captured if streams have constant h.

L242: During what time period were bi-weekly observations made, did this span the full calendar year?

L249: Porosity provided here (0.88) does not match that in table S2 (0.89)?

L278: Is ‘matching’ meant here instead of ‘non matching’, the context of this description seems to be characterizing the data available to use, rather than the inverse(?).

Results

L307: Why are different evaluation metrics being used for Q and ET?

L325: It would be helpful to explain here why the morning flyover predicts drier soil conditions. It seems that this pattern could depend on the time of year, and whether snowmelt is occurring during the day.

L335: Please explain what is meant by “especially in terms of ranking the positions”. It seems clear from the results that some locations are captured well, and others poorly.

L337: Yes, but there are also extended periods of strong underestimation that should not be overlooked.

L345: Some metrics should be provided on all of these evaluations.

L350: I don’t believe that this statement is supported by the data. R2 values are low. There are commonly large absolute errors approaching 0.5 m³ m^–3 at the upper end of this range. Large relative errors at lower observed soil moisture levels and a tendency to overpredict is clear. Perhaps use of NMAE would offer a better assessment here, but importantly, the model abilities should be described with greater rigour.

L353: Again here, a more objective description of the model performance is required. There are few predictions at higher moisture with the 2D model that lie close to the 1 to 1 line. If 0.55 is used as a threshold for evaluating performance, it is recommended that metrics for observed moisture levels above and below this level, as well as overall be presented. Likewise, it would be helpful to provide metrics for the different landcover or canopy closures discussed in the text. Many of the examples in the following paragraph relate to landcover, so this seems a better factor to use in Figure 6 than is canopy closure.

L370: It would be helpful to cite Figure 6 here in addition to mentioning this occurs in peatlands.

L374: Again here, there remain large deviations with this model, and care should be taken to not oversell what the model is capable of.

L380: Perhaps I have misunderstood something, but I am having trouble understanding the value in comparing the model to SAR measurements for a 5 cm depth with modelled data, given that those SAR measurements haven’t been validated as being in strong agreement with observed data (Figure 6D). I appreciate that the SAR data provide an opportunity to compare model results between observations, but this would only seem useful if those SAR data are effectively capturing the hydrological state, and this has not been shown clearly. For this reason, I am unconvinced that section 3.4 belongs in the manuscript. The discussion beginning at lines 469 seems to support this notion.

L402: Does this statement about differences being highest in wet conditions hold if the panels for homog.canopy at q = 0.5, 0.9 are blank, and negligible as stated at L409?

Discussion

L414-416: It has been shown that the model parameterization shapes this (with improved but not strong performance in 2D), but observational data as shown/analyzed do not demonstrate this directly.

L417: It remains hard to see from the model performance illustrated that the models are ‘reliably’ predicting soil moisture. That they predict moisture variability is besides the point if the predictions are not also accurate.

L425/26: The figure cited to support this statement (Figure S7) doesn’t show a comparison of the 2D model with HydroGeoSphere.

L434: See earlier recommendation to consider these classes instead of vegetation.

L437: shallow soil moisture

L487: Yes, this is important as snowmelt is radiation rather than temperature driven, so this suggests that the process might be arriving at the right answer for the wrong reason.

Conclusions

L515: “shaping model simulations of soil moisture dynamics”

Given the unreliability of the SAR observations, it would be beneficial to touch on the large errors in soil moisture simulation in this section, and focus less on the SAR data. Certainly, the model progression has led to the ability to simulate a wider range of moisture conditions, but given the performance demonstrated, the model predictions are probably not robust enough to see applied use yet. Given this, a strong argument should be made in the conclusion for continued model performance to improve on that shown here.

Figures and Tables

Figure 1. “and its hydrological measurement stations”

Tables in supplementary information should be labelled with S, to distinguish from the manuscript.

Table S1 and S2 should read “Soil type–specific”

Table 2: This is not a complete list. At a minimum a more descriptive caption is needed here that directs the reader to additional model parameters provided in the SI.

Figure 3. These panels are too small to be legible at print scale. It seems that panel B and C should be presented first, as this summarizes raw data, while the other two panels are results oriented. What period is captured by panels B and C? On panel A, why is snow presented as a line, rather than having Psnow and Pliquid as stacked bar plot to give total P. This doesn’t allow for easy interpretation. Are snowpack observations available to evaluate model predictions?

Figure 7. While described as qualitative, this figure isn’t particularly easy to interpret, as it isn’t always easy to distinguish between points and the underlying land use. As this largely conveys the same information as Figure 6, but less effectively, this seems a good candidate to move to the Supporting information.

Figure 8: “Spatial patterns”. This plot is hard to evaluate. Please include performance metrics to allow diagnosis of model performance.

Figure 10. An improved caption is needed here. This isn’t demonstrating lateral groundwater flow, but rather model parameterization that includes this process. It isn’t immediately clear why panels E and F are blank.

General comments

There are places where hyphens are used where negative symbols are needed.

Notation style should be harmonized, e.g. there are instances of unit/unit² and unit unit⁻²

It would be helpful to have the Figures in the SI appear in the same order in which they are cited in the text.

This is my first review of the paper "Multi-scale soil moisture data and process-based modeling reveal the importance of lateral groundwater flow in a subarctic catchment" by Jari-Pekka Nousu et al.

The paper is well-written and presents a valuable contribution to the literature. As models and data improve in resolution, many processes become scale-dependent, making what is overlooked at a coarse scale crucial at high resolution. This manuscript addresses this issue by comparing different model parameterizations and SAR-based soil moisture with a robust experimental dataset.

I do not have major comments on the study, but I suggest some moderate revisions:

Update the Bibliography: The bibliography is outdated. Please revise it to include more recent works. I have provided some suggestions in the annotated PDF.

Enhance Section 3.4: Section 3.4 is overly qualitative and could be improved significantly. Consider incorporating metrics to quantitatively demonstrate the differences between SAR data, various model parameterizations, and in situ data. Temporal stability analysis, as discussed in Dari et al. 2019 (https://www.sciencedirect.com/science/article/abs/pii/S0022169419300575), could be particularly useful. Comparing different statistical spatial measures from various soil moisture spatiotemporal dynamics would be highly relevant.

Clarify SAR Estimates: While there is already a paper on SAR estimates, more detailed information about the retrievals should be included in this manuscript to provide better context.

Based on these points, my recommendation is moderate revisions. I have also attached the annotated PDF with additional comments for further guidance.