Global assessment of how averaging over spatial heterogeneity in precipitation and potential evapotranspiration affects modeled evapotranspiration rates

Rouholahnejad Freund, Elham; Fan, Ying; Kirchner, James W.

doi:https://doi.org/10.5194/hess-24-1927-2020

Articles | Volume 24, issue 4

https://doi.org/10.5194/hess-24-1927-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-24-1927-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 24, issue 4

Research article

|

16 Apr 2020

Research article |

| 16 Apr 2020

Global assessment of how averaging over spatial heterogeneity in precipitation and potential evapotranspiration affects modeled evapotranspiration rates

Elham Rouholahnejad Freund, Ying Fan, and James W. Kirchner

Download

Final revised paper (published on 16 Apr 2020)
Supplement to the final revised paper
Preprint (discussion started on 15 Mar 2019)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review comments on the paper by Rouholahnejad Freund et al.', Anonymous Referee #1, 16 Apr 2019
- AC1: 'Response to Reviewer #1 comments', Elham Rouholahnejad Freund, 03 May 2019
RC2: 'Review', Anonymous Referee #2, 03 Jun 2019
- AC2: 'Response to reviewer#2', Elham Rouholahnejad Freund, 17 Jun 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (further review by editor and referees) (04 Jul 2019) by Miriam Coenders-Gerrits

The authors present a study where they investigated the effects of spatial averaging on modelled evaporation (ET) estimates at the global scale. They make use of the Budyko framework to model ET and use this same framework to spatially 'average' ET and to determine the heterogeniety bias. This method was already presented in a previous paper (Rouholahneja Freund & Kirchner, 2017), but is now applied at the global scale. Only applying a method to the global scale without adding new insights (or very limited) is not enough for a new publication. As also mentioned by reviewer #1, this should be improved. Clearly show the added value of this study above the previous one. In the abstract (L40-41) it is promised to proved insights in the underlying mechanisms, but these can not (or limited) be found in the manuscript.

Furthermore, I have some doubts on your methodology. You frame your study in such way, that it will help to quantify errors in ET due to spatial averaging for ESMs. To investigate this, you don't use a ESM, but you choose for the Budyko framework for simplicity reasons. However, the Budyko framework is a first order estimate, meant for large catchments under steady-state (see also comment reviewer #1). I wonder if the gridcells can be considered as mini-catchments under steady state. What is the effect of lateral flow, irrigation, advected energy (dry gridcell next to wet gridcell) etc.? This should be more discussed in detail. Furthermore, I wonder if the study shows the real ET-heterogeniety bias (as far as you can at all), or that I am looking at uncertainties in rainfall products? Because looking at figure 1b, the bias is largest once P/PET deviates most between 2 locations. This is the case when either P or PET differs most between locations. Often PET differs less than P, so the bias is dominated by differences in P (as shown in figure 4). Hence I am not surprised to see that the bias is related to topography, because it is well known that P changes significantly with altitude (and also becomes more uncertain). In relation to this, I wonder if figure 5b would not look similar once you plot climate zone against the standard deviation in P?

Based on these concerns, plus the 2 critical recommendations by the reviewers I advise to do a major revision of your manuscript and emphasize what we can learn from this study in addition to the previous study (what are the new insights). Hereafter, I will send out the manuscript for a new review round. Please also have a careful look at the comments of reviewer #1. They will help to improve the quality of the paper.

Specific comments:
- P1L23: what do you mean by "landscapes where P is inversely related with PET"?
- P3L78: in my view ET and LH are synomyms.
- P5L151-153: the hypotheses miss link with previous text. Please make the connection clearer.
- fig 1b: Please change this figure to the common way Budyko curves are drawn, i.e.: aridity on x-asis (aridity=PET/P) and not the wetness index
- P7L208: P/PET is not aridity, become once P/PET becomes larger the index becomes wetter instead of dryer
- P11L262-264: this line somehow suggest that you prefer prismP, because you then have larger biases? Why?
- Fig 4: I would swop a) and b)
- Fig 5b: would be nice to also make this graph for standard deviation of P and PET. Likely it's similar.
- P15L314: add comma after e.g.

Hide

AR by Anna Mirena Feist-Polner on behalf of the Authors (21 Aug 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (27 Aug 2019) by Miriam Coenders-Gerrits

RR by Anonymous Referee #3 (01 Sep 2019)

Suggestions for revision or reasons for rejection

Review of “Global assessment of how averaging over spatial heterogeneity in precipitation and potential evapotranspiration affects modeled evapotranspiration rates” by Elham Rouholahnejad Freund et al.

This is my first review of this work. The manuscript by Elham Rouholahnejad Freund et al. addresses the interesting issue of scaling of water and energy exchange at the land surface. While the manuscript focusses on a novel issue that could provide an interesting new addition to decades of literature on scaling, I do not find the manuscript in its current version to be convincing. This has to do with: a) a poor link between the main methodology and motivation as outlined in the Introduction, and b) a complete neglect of surface heterogeneity and scale-dependency in the Budyko n-parameter. In my view, the work could be a valuable contribution to HESS only if these deficiencies are addressed.

The work is motivated by potential scaling issues in ESMs. In my view, these relate mainly to effects of land surface heterogeneity (land use, soil type, groundwater tables, soil moisture, all of which can show large variability on small scales). To the degree these depend on forcing, this will to a large extent be caused by spatio-temporal variability of rainfall (i.e. convective storms leading to temporary wetting of part of a water-limited region only) and not just spatial variability. This is a big simplification where most of the scaling problem already is solved, and which is inherent to the choice for Budyko. A possible solution would be to solely focus on scaling issues within the Budyko framework due to P and PET. This would be a fairly novel approach, and it would avoid (artificially) linking too much to ESMs.

My second concern deals with the choice for a single Budyko n-parameter. Effectively, the authors show that at larger scales due to forcing heterogeneity, the Budyko curve tends to become more linear. But what is the motivation for the baseline choice of n? Where is it shown that this value corresponds better to observations (i.e. is a more valid model) at finer scales (1 km) than at courser scales (1 degree)? Would the results not strongly depend on the choice of n? In reality, n will also very strongly depend on land use (see for example Fig. 1 in https://www.hydrol-earth-syst-sci-discuss.net/hess-2018-634/#discussion or any of the many other studies on this subject). I think any analysis of scaling should focus on the main nonlinearities to avoid becoming a purely academic exercise (nothing wrong with the latter, but then it should be presented as such). As a minimum, I would expect a sensitivity analysis on how the results depend on the value of n, accompanied by a discussion on how n might vary locally and with scale. Ideally, I would see the manuscript being restructured towards a more theoretical analysis of effects of forcing heterogeneity on the Budyko model, which would result in a clearly testable hypothesis that Budyko curves should become more linear with increasing scale as a result of heterogeneity, and based on the maps regions can be identified where this effect should be largest and best observable. It should also be noted that the value of n used in the analysis is not reported, at least I could not find it.

Hide

RR by Anonymous Referee #1 (01 Oct 2019)

Suggestions for revision or reasons for rejection

I now read the revised manuscript version and rebuttal letter of the manuscript entitled “Global assessment of how averaging over spatial heterogeneity in precipitation and potential evapotranspiration affects modeled evapotranspiration rates”. I appreciate authors effort in improving the manuscript and clarifying several aspects of their work. Having said that, I must also admit that I am still not fully convinced by some of the arguments the authors used in the reply letter to present (and defend) their work. Therefore, I highlight below those important points that need to be addressed:

- Authors treat P and PET as model input data defined at the same scale (see lines 292-294). However, this does not reflect how typically ESMs dynamical cores are designed and/or have been evolving. Forcing terms (e.g., precipitation, temperature, humidity) are defined at the atmospheric model grid (usually coarser) while PET is calculated at the PFT-level using land surface features (e.g., LAI, aerodynamic resistance). In light of this, variability in modelled P and PET occurs at different spatial scales. This makes, in my opinion, the calculation of a correction factor for ET less straightforward. Could authors elaborate on this point?

- Authors affirm that their work highlights under which climate conditions averaging in P and PET has an effect on ET estimates. Without giving clear explanations I suspect that the largest differences found for Cs and Ds climate zones are still mainly driven by topography. Note also that this analysis is limited to the CONUS domain where there are not many sampling points for certain climate zones (e.g., Ds). This information should be provided in the plots and the statistical significance of the differences should be tested. An analysis at the global scale would be certainly more convincing.

- The grid-scale dependence is tested for Switzerland and I don`t think we can “extrapolate” this exponential relationship everywhere around the globe. If you want to convince the reader you need to repeat this assessment for all regions (identified for instance in Fig. 3) where averaging effects are not negligible. Juxtaposing the different curves will (or will not) support the existance of a general “scaling” relationship with the grid resolution.

- In the first iteration I asked authors to provide more information about the value of “n” parameter. This information is still missing in the manuscript. Can the authors provide some concrete numbers on the sensitivity of their global estimates with respect to different “n” values?

Other comments:
- Please do not include any discussion in the captions of the figures. See Figure 3-5-6.
- Lines 36-37 in the abstract. I do not see how the results of this paper can be used for guiding a more detailed mechanistic modelling. Note also that averaging- or grid-scale effects have been largely reported also when using mechanistic models. So please remove this sentence.
- You do not want to quantify the absolute magnitude of the averaging effects and at the same you claim that your methodology is potentially a way for correcting such bias. The second statement imply a sort of quantification, in my opinion.
- I found an imbalance between the emphasis you put on the introduction and the actual findings of the manuscript. As I said in the first iteration, this is an applied study of a previously described methodology that does not contain general insights.
- Lines 317-319 (“most likely due to the sharp spatial gradient…”). This is a quite generic statement.

Hide

ED: Reconsider after major revisions (further review by editor and referees) (15 Oct 2019) by Miriam Coenders-Gerrits

AR by Elham Rouholahnejad Freund on behalf of the Authors (26 Nov 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (05 Dec 2019) by Miriam Coenders-Gerrits

RR by Anonymous Referee #3 (03 Jan 2020)

RR by Anonymous Referee #1 (07 Jan 2020)

ED: Publish subject to minor revisions (review by editor) (23 Jan 2020) by Miriam Coenders-Gerrits

AR by Anna Mirena Feist-Polner on behalf of the Authors (27 Jan 2020) Author's response Manuscript

ED: Publish as is (10 Feb 2020) by Miriam Coenders-Gerrits

Short summary

Evapotranspiration (ET) rates and properties that regulate them are spatially heterogeneous. Averaging over spatial heterogeneity in precipitation (P) and potential evapotranspiration (PET) as the main drivers of ET may lead to biased estimates of energy and water fluxes from the land to the atmosphere. We show that this bias is largest in mountainous terrains, in regions with temperate climates and dry summers, and in landscapes where spatial variations in P and PET are inversely correlated.