Paper summary: In response to a growing body of research indicating that plants routinely use water from bedrock, these authors asked the question: what happens to modeled plant transpiration if, instead of relying exclusively on soil water, they are allowed to access a deeper bedrock bucket? They found that having access to more water improved the accuracy of  transpiration in a widely used land surface model (when compared to actual sap flow data) in places with pronounced dry seasons. The authors suggest that this provides additional motivation for the better inclusion of plant-available bedrock water in land surface models. The manuscript is well written and easy to read. 

I am supportive of the goals of this manuscript and would like to see it published, but I would also appreciate the authors considering how they could address what I perceive are two shortcomings:

1.  The study illustrates transpiration dynamics using field data for sapflow at four sites, but no actual local field information about the storage dynamics of the bedrock underlying soil, local rooting profiles, etc. is provided. So, there is little meaningful context regarding the subsurface properties at the sites (properties that are the primary focus of the paper). This means the study essentially looked at the effect of varying a model parameter (water storage bucket size) on T and found that the default model configuration could be improved upon. Other default model parameters could have also been varied (the PFT properties, for example), and modeled transpiration might have been improved as well. So, while the authors have shown that changing a model parameter from the default can improve model performance (larger storage buckets can improve T representation [and I don’t doubt that this is the likely reason]), without any actual data showing that plants use deeper water from bedrock at these sites it has not been demonstrated that this is mechanistically why T has improved for these particular sites. Is any of this context available at the four study sites, and could it be added to the paper? Based on the findings of the paper, what should be done by the modeling community? Should the water storage bucket just be freely calibrated instead of prescribed? What exactly is the goal of changing this parameter? To improve accuracy of historically observed T, or to better predict T under non-stationary climate, etc?
2. Other studies have already shown that increasing the size of the storage bucket accessible to plants can improve modeled T patterns in seasonally dry (e.g., Mediterranean) climates (e.g., Ichii, K., Wang, W., Hashimoto, H., Yang, F., Votava, P., Michaelis, A. R., & Nemani, R. R. [2009]. Refinement of rooting depths using satellite-based evapotranspiration seasonality for ecosystem modeling in California. Agricultural and Forest Meteorology, 149(11), 1907-1918.). Yes, these studies do this by changing rooting depths, or adding deeper soil (rather than calling it bedrock), but isn’t the fundamental result the same: more stored water accessible to plants? What exactly is the novel finding in this study in relation to what these other studies have done (which is to change a model parameter that ultimately allows for more water storage for plants, thereby resulting in a better T or ET estimate)?

Other items:

Table 1: Is the p50 correct for the Russian site? I am surprised it would be such a low water potential in such a cold climate. 

• I understand the goal of Figure 4: compare modeled to actual sapflow patterns by time (note that nowhere in the figure or caption is this stated, however). This figure is extremely difficult to comprehend, even after quite a few minutes of study. It is also worth noting that a continuous variable is reported as an area (circle area) rather than a length, leading to potential interpretation ambiguities. Can these not be plotted as regular time series points, whose values vary along a continuous rather than categorical y-axis?

Figures 5  and A2-A5 are not legible when printed on standard paper and need to be reformatted so that they can be read. 

Line 165: It is reported that in order to mimic the hydraulic behaviour of fractured bedrock, it is modeled as a pile of sand (90% sand, 10% clay). This model choice is not supported by any reference to literature on bedrock hydraulic properties, and surprised me as it is not how I would conceive of bedrock hydraulic properties.

Summary:

The authors manipulate existing CLM 5 soil texture and depth to bedrock (DTB) parameters to account for plant-accessible moisture stored in fractured bedrock beneath soils. They compare simulated to actual transpiration across four sites in Europe with different climate conditions and find that the two simulated bedrock scenarios (one with 1.5 m of simulated “bedrock” and another with additional “fracturing”) better match observed transpiration during the summer for the sites with a pronounced dry season. 

This work motivates further exploration of how bedrock water is accessed by plants and how this process is represented in hydrologic and ecophysiological models. The manuscript is exceptionally well motivated and contextualized, and was easy to follow. The conclusions are well reasoned and of interest to a number of communities engaged in biogeosciences research. Comments are shared to increase clarity. 

Comments and questions for the authors:

Why are the rooting profiles illustrated as linear but described as exponential? 

Is there an expectation that model agreement should be improved during energy limited periods or in energy limited sites (e.g. line 303, 405). Limitations outside of water availability could be better quantified and described in the results and discussion. 

Are the default parameters reported in line 145 site specific and if so, are they reported? 

A statement on the rationale behind the specific three model configurations would benefit the reader. Why these three and not other possibilities? Additionally, in line 167, how does the 90% sand and 10% clay mimic fractured bedrock? Justification is needed here. 

The Pelletier et al 2016 dataset is a model output and not reflective of local site conditions per se. As far as I understand, it is only validated for depth to bedrock in the US using groundwater well data (which is rarely available in uplands areas like the sites in this study.)  The language around use of the dataset should be couched to reflect that the dataset does not provide DTB at the four sites. 

In line 80, what does “fully developed” mean in this context? 

In line 75, an additional possibility is that belowground biomass distributions may change over time in response to water stress (e.g. Liu et al, 2019). 

Is there site specific subsurface information (from e.g. the papers cited in the site descriptions) that could be added to contextualize the DTB increase needed to improve model performance? 

The overprediction of transpiration during spring and rapid drying of the root zone is a very interesting result that models representing deeper water stores will have to grapple with. The discussion of plant hydraulics in L420-440 is thorough and very well done, but are there perhaps other additional factors that could be considered as well? For example, dynamic belowground biomass, fungi, or the role of multi-porosity systems (e.g. Schwinning, 2020).  

Is it necessary to have well developed soil to access groundwater (Line 355)? 

The definition of bedrock within the paper is a bit inconsistent, specifically in the caption of Figure 1. For example, bedrock in CLM5 is considered impermeable but bedrock is represented as a combination of sand and clay.  Clarification is needed here. 

Some comments about figures: 

Figure A1: Is this a boxplot? It seems like a timeseries. A description of the points vs.

lines is needed in the caption. 

Figure 2: Is there a legend label missing (corresponding to pink or orange)?

Figure 3: This is the most impactful figure but it is very difficult to tell the different model configurations apart. The caption says boxes but there don’t seem to be boxes in the figure. 

Figure 4: Some of the concepts in this figure could potentially be better represented by scatterplots for specific times or model configurations that are most significant to the results. Including a simple illustration of how model-data agreement is improved with bedrock water storage under specific conditions could make the paper potentially more impactful and approachable to non-CLM experts. 

References cited: 

Li, H., Si, B., Wu, P. and McDonnell, J.J., 2019. Water mining from the deep critical zone by apple trees growing on loess. Hydrological Processes, 33(2), pp.320-327.