Review of egusphere-2024-115

Tempel et al.

Vegetation response to climatic variability: Implications for root zone storage and streamflow predictions

Review by Andrew J. Guswa, Smith College, aguswa@smith.edu

In this paper, the authors compare the response of long-term, actual evapotranspiration (E_a) to changes in climate (precipitation, P, and/or potential evapotranspiration, E_p) across 286 (mostly humid) catchments.  They do this in two ways:

1) they fit a parametric Budyko curve (with parameter, w) to a decade of data, and then use that Budyko curve to estimate E_a for future decades with new P and E_p.

2) they use the data from future decades to determine E_a directly from the water balance, presuming zero change in storage over that long time.

They quantify the difference between those two values of E_a as delta_E_a. And they also quantify the change in the ratio of E_a to P (I_E) as delta_I_E.

The authors use those two values of long-term E_a (or I_E) to estimate two values of maximum root depth (Sr,max) and the difference between them: delta_Sr,max.

They show that the differences in Sr,max from the two methods are small, as are the subsequent effects on hydrologic modeling when the uncertainty in Sr,max is incorporated into a process-based model.

The study nicely integrates large datasets and modeling to elucidate the minor importance of those differences in Sr,max to hydrologic modeling.  The paper is clear, well-written, and the figures are compelling.  I have a few minor comments (see below), and I also think the authors could discuss further the implications of the simplification they employ to estimate Sr,max.

Regarding the latter, the authors simplify daily actual evapotranspiration (E_a_daily) to be equal to daily potential evapotranspiration scaled by the decadal ratio of actual to potential ET.  They then uses a daily water balance to determine the necessary Sr,max to deliver that daily evapotranspiration.  Using a constant ratio to convert potential ET into actual, however, does not necessarily reflect the behavior of catchment vegetation.  As a counterpoint, one might expect potential ET to be met fully during periods of low E_p (and plentiful water) and actual ET to approach zero during periods of high-demand/drought.  Thus, another way that one could determine the requisite Sr,max – as opposed to equations 4 and 5 – would be to simplify the system such that

1. E_a = E_p if water is available in storage
2. E_a = 0 when the water in storage is depleted
3. Find the value of Sr,max such that the (long-term sum of E_a)/(long-term sum of E_p) equals that desired long-term ratio

Such an approach may better represent vegetation response (albeit a little extreme, along the lines of Milly, 1994), and would be more consistent with the complementary hypothesis for evaporation (see multiple references by Szilagyi)

It may be that the resulting Sr,max does not differ much from that determined from equations 4 and 5, due to the self-limiting process of ET (e.g., whether one removes 5 mm on day one and then zero on day two or 2.5 mm on day 1 and another 2.5 mm on day 2 may not matter).  However, it would be interesting to compare and to see if there is a difference, especially for the monthly/seasonal results, where the differences may have an even larger effect.

I understand this may be beyond the scope of the paper.  Nevertheless, given the significance of equations 4 and 5 on the central message of this paper, I recommend that the authors spend more time discussing those simplifications, alternative simplifications (such as that above), and the potential implications on the results and conclusions.

Relatedly, I think the abstract and discussion would benefit from additional acknowledgment that the catchments used in this study are both snow-free and relatively aseasonal.  Thus, the conclusions may not be extensible to snow-dominated watersheds and/or those with strong seasonality, such as a Mediterranean climate.

Minor comments

Title

Given the nature of the datasets used, I think a more representative title for this work would be “Catchment response to climatic variability: Implications for root zone storage and streamflow predictions.”  The CAMELS datasets are catchment-based, and the authors are not isolating specific vegetation responses.

Nomenclature

I found it somewhat confusing that the meanings of the subscripts modifying evapotranspiration (E) and aridity index (I) were not consistent.  When A was used as a subscript, it meant “actual” when modifying evapotranspiration; however, it meant “aridity” when modifying the index, which – in turn – meant it signified potential (not actual) ET.  Thus, I_A was not the analog to E_A; rather I_E was the analog to E_A.  Perhaps I_A could be used to indicate the evaporative index based on actual ET, whereas I_P could indicate the evaporative index based on potential ET.

Line 52-60                 The authors present their methods of determining Sr,max from a daily water balance (see above).  In essence, the Sr,max is the storage volume needed to ensure that daily ET can be met.  However, that value represents a minimum value for Sr,max, which – of course – could be larger.  It might be worth a comment to that effect, especially since those values of Sr,max are then used in a hydrologic model with a very different mathematics.

Lines 165-172           The numbering scheme used in this paragraph does not exactly match the numbering of the methods sections to which it refers.

Lines 299-307           I particularly appreciate that the authors sought explanatory variables, such as aridity index, for their results.  I expected aridity to be a controlling factor, and it was interesting to learn that it was not.

Line 431                     dangling phrase, “the more equilibrated scenario A”

Equation 5

As written, the equation is circular.  What should be used as the argument of the inequalities on the RHS is the integral from t0 to t of (P_daily - E_A_daily) dt rather than S_D,j,i(t)

References

The reference for Dralle, et al. 2021 is missing from the reference list

Figure 7

I recognize that figure 7 is intended to explain the methodology and not results.  Even so, I recommend that the qualitative character of the distributions for delta_I_E reflect the results of this paper.  That is, the distribution for scenario A should be narrower than that for scenario B; and the mean for scenario B could even be shifted away from zero (compare Figure 7 and Figure 11).  As is, the figure gives the false impression that the uncertainty across all catchments is greater than across the Meuse watershed alone.

The article presents an analysis of the sensitivity of root-zone storage capacity to climate variability. The article is generally clear. My main concern on this article is that I did not clearly see the novel insights provided by this study. The authors say a few times that their results corroborates past findings on the limits of the Budyko framework. Actually, I found that they should more clearly emphasize what is new in their results compared to past studies.

I have also other comments detailed below. I suggest major revision.

1. L67-69: What about groundwater exchanges? Should not they be considered also?
2. (1): Please explain in simple words what low or high values of omega mean in terms of water balance type. Say omega should be positive.
3. L141: At this stage of the article, it is unclear why this specific PE formula was used. Maybe this could be justified in a few words (even if some comments are later provided in the discussion section).
4. L151-154: I do not understand why apparently “leaking” catchments where excluded from the analysis. There are probably many catchments in the remaining dataset, which exhibit groundwater exports even if there are within the limits of the graph. The limit IE>IA is arbitrary and does not really corresponds to underlying processes (groundwater exports or not). Would the analysis be very different if all the catchment had been kept to conduct the analysis? Does this catchment selection partly explain the apparently very consistent behaviour of UK and US catchments?    
5. Section 4: I felt a bit confused in reading the results section. In the first part of the analysis, the Meuse basin seems to be somehow outlier in its behaviour compared to the UK and US datasets, but no clear explanations on this behaviour is found. Therefore the added value of this small catchment sub-set in this part of the analysis is unclear. In the second part where the process-based model is applied, only the Meuse basin is used. Though I understand it is difficult to apply such a model on large datasets, I found it makes the study less clear and the overall conclusions more difficult to draw.  
6. Section 4.1: I found figures 9 to 11 not very useful. Maybe the authors could say what were their results without showing figures (or putting them in SM).
7. Sections 4.3.2 and 4.3.3: I found these parts of the article tedious to read. A lot of detailed information is given, which makes it difficult to draw the overall picture.
8. Conclusion: see main comment above
9. 14: I found this figure not very useful (overall I found there are too many figures in this article and their number could probably be reduced).

Minor comments

• The reference Hulsman et al. (2021) is missing in the list of references.
• UK is used in notations, but GB is used for the CAMELS dataset. Maybe use a single abbreviation to be consistent.  

This study calculates the differences in I_E and evaluates how these differences influence the estimates of root zone storage capacities. It further examines how uncertainties in root zone storage capacities affect streamflow predictions in hydrological models. To some extent, the manuscript is well-structured, detailed, and presents valuable ideas. However, several concerns need to be addressed. Additionally, inconsistent formatting and grammar errors diminish the quality of the paper. Overall quality needs to be enhanced.

1. First concern: Does the paper present climatic variability properly and clearly? Several questions regarding this:
   
   
   1. Using absolute values of the I_A deviation in Figure 8 cannot effectively reflect the change in I_A. Using percentage changes would better illustrate how I_A changes to reflect multi-decadal climatic variability. Your use of percentage changes in Figure 12 (c) for root zone storage capacity changes is a good approach.
   2. Are the values of the aridity index in Figures 1-3 calculated for the entire period? If so, while the values of I_A deviation in Figure 8 are calculated by decades, it might be better to find a consistent way to present I_A and I_A deviation using the same time period (either the entire period or by decades).
   3. If percentage changes in I_A are small for most catchments, climatic variability is small. Then is the conclusion that hydrological responses, in terms of changes in I_E, root zone storage capacities, and streamflow, are generally minor under changing climatic conditions reliable?
2. Second concern, related to 1.c: How many catchments exhibit distinct changing climate conditions? Can percentage changes in I_A and I_E by decades effectively reflect that? If the climate changes are small, their impact on root zone storage capacity changes might be less significant. 

Suggestions for some minor revisions (this is not a comprehensive list):

1. The legends in Figures 1 and 2 should use periods instead of commas, so they should be 0.1 – 0.2, 0.2 – 0.3, etc., not 0,1 – 0,2, 0,2 – 0,3, etc. Additionally, the title of the legends should be "Aridity Index I_A," with the A as a subscript.
2. Figure 11 could be removed; the information is clearly conveyed in the text.
3. Figure 13 (a): Consider adding a secondary axis on the right side of the chart to display the groups of I_E deviations.
4. Units of Figure 13 are incorrect.
5. Lines 332 to 331, do you mean Figure 13?
6. Line 405: The reference Wang et al., 2016 is missing from the list of references. There may be other missing references as well. A comprehensive reference check is recommended.
7. Line 684: two references listed in one line.