the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Apr 2026
How precipitation lapse rates shape runoff simulations and flood frequency estimates in mountainous regions
Abstract. Precipitation lapse rates (PLRs) play a key role in hydrological simulations of mountainous catchments. However, they are often poorly represented in the precipitation estimates and are typically simplified as constant and positive values in the hydrological models. In this study, we combine a stochastic weather generator with a hydrological model to investigate how PLRs affect runoff simulations for several mountainous catchments in Switzerland. In the weather generator, the PLR adjusts precipitation from station elevation to mean catchment elevation, while in the hydrological model it redistributes precipitation among elevation zones. By systematically varying the PLRs in both the weather generator and the hydrological model between 0 % and 10 %, we found effects on mean seasonal and annual runoff, as well as on extreme floods, depending on catchment characteristics and precipitation network properties. Specifically, higher-elevation catchments were less sensitive compared to lower-elevation catchments. Increasing PLRs tended to increase summer floods, while decreasing PLRs tended to increase winter floods. In addition, the seasonality of frequent floods was more sensitive to changes in PLRs than that of rare floods. Moreover, flood seasonality was primarily controlled by the PLR in the hydrological model, while flood magnitude was mainly driven by the PLR in the weather generator through its effect on precipitation amounts. These findings highlight the need for a more comprehensive investigation of the assumption of a constant lapse rate in hydrological models, particularly in mountainous regions where precipitation gradients are strong and observations are limited.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Hydrology and Earth System Sciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(2245 KB) - Metadata XML
-
Supplement
(206 KB) - BibTeX
- EndNote
Status: open (until 13 Jun 2026)
- RC1: 'Comment on egusphere-2026-2338', Anonymous Referee #1, 24 May 2026 reply
-
RC2: 'Comment on egusphere-2026-2338', Anonymous Referee #2, 29 May 2026
reply
Overall assessment:
This manuscript pursues a worthy goal: to quantify the effect of often arbitrarily chosen precipitation lapse rates on hydrological models. However there are important problems of structure and the results are overly justified in ways that go beyond what the data show, often discussing and interpreting what might very well be in fact modelling artifacts. I believe the paper could be made much more concise and readable by trimming down the physical interpretation of the lapse rates that are actually not physics, but properties of the models used.
General comments:
- What I see in the results is that larger PLRs result in more Q, and more extremes. However this can be explained by the fact that most rain gauges are below the average catchment altitude, therefore positive lapse rates will result in more water than measured at the rain gauge (especially when the nearest-neighbor Thiessen method is used). In a way, in this context the positive lapse rates are "creating" water, especially since the model calibration is kept constant. This alone could explain the correlation of Q with PLR and also the correlation of maximum flows with PLRs. As such, I would need to be convinced that the correlations found are not artifacts of the experimental setting.
The effect mentioned above can particularly affect summer floods, therefore it may be a possibility to focus the discussion on the reverse effect observed with winter floods. - I feel that the discussions and interpretations are overreaching. In p.16 (results), or in the discussion section, the manuscript discusses processes to a large extent. However, it is only based on model runs, therefore the processes are only those that have been implemented in HBV, which is imperfect as all models.
The manuscript should focusing more on the uncertainties of the hydrological model, since it is the main target of the research. The way it is structured and the way things are explained though make it seem that the general effect of lapse rates in floods is analyzed based on physical understanding. This deviates the main target.While section 5.5.3 acknowledges these limitations, the fact remains that it makes many of the discussions less relevant.
-
The same comment extends to the stochastic rainfall generator model, which also has limits (even though GWEX is very well established). For instance, a lack of seasonal shift in the very rare floods (1000 years) might be a feature of the precipitation generator rather than a question of lapse rates. What would really help here is a discussion and analysis of the results of the GWEX model, because currently the reader does not see the output rainfall amounts, their seasonality, etc.
To push the argument further, the sensitivity analysis does not enforce physical consistency when it allows PLR_WG and PLR_HBV to be different. While this is fine as these are “statistical” parameters that represent modelling practices, they cannot be discussed in physical terms. -
The design of the figures should be revisited. Many figures are too sketchy. There are important inconsistencies in the size of plots and axes.
-
There are shortcomings in the structure of the paper. Notably
- some of the data that are presented in the results sections
- some methods are presented in the results section
- Some discussion elements are provided in the results section
Specific comments:
l.24: one could argue that convection is exactly the uplift of air masses or the reason behind the uplift rather than the result, so maybe use more exact language
l.94-95: what is the reason of including this paragraph since in the current approach the PLR is also linear, positive and constant
l.99-100: In some catchments, the station elevation might be close to the average elevation, while in other catchments not. Does this create a bias as some catchments are corrected more than others?
l.102: The unit should be specified in % per 100m. Also, what about negative lapse rates?
l.133-134: maybe it would be beneficial to include a reasoning on why these numbers and intervals were chosen, since only 5% is explained below. Also, Why is this range chosen? A reference would be needed to justify it.
Section 2 is very short. It could be beefed up by moving here some of the material in figure 3b and the associated text, which is largely about data description. Also, don't see 27 catchments on figure 1.
Section 3.1: the text is not very clear in this section. Maybe provide a schematic representation of the workflow and different sensitivities strategies.
l.143: some more background on these papers could be useful since they seem to be a prequel to this work
l.180-186: since there is a large number of parameters with possibly complex multivariate interactions, it is hard to assess the effect of using reducing this to only 3 representative parameters sets (min, med, max). Can a sensitivity study be shown, maybe on a single catchment, to reassure me that these 3 parameters sets are representative of the entire ensemble?
The logic of the simulations carried out is much better explained in the first sentences of section 3.4 than in section 3.1. I think it should be explained only once, and with a schematic, as mentioned above.
l.214: What is the rationale of this metric for flood seasonality, as others could be envisioned.
l.222: Here I understand the term elevation zones as what is commonly called elevation bands. It would be useful to specify that these terms are equivalent.
l.240: this description is a bit redundant and can be clarified
There should be a notation for the different lapse rates in eq. 6
Figure 2 is useful, but I have the following comments:
- It should come at the beginning of section 3 rather than at the end
- It should provide more details of the sensitivity workflow strategy with both PLRs
- The picture on the left, while funny, is not very informative and could be removed to save space
- Overall, the graphical design could be improved
Figure 3 shows that rain gauges are largely located in the lower portion of catchments, and that lapse rates are mostly extrapolated to higher altitudes. This should be explicitly discussed.
Figure 3: The labels PLR000 etc are not very clear. Either rename or explain in legend
l.357: Where does this change in mean annual maximum precipitation comes from? It seems that there is even a change of 80% in some cases, so it might be beneficial to explain how is this change is imposed in the model
Section 4.1 mixes results and data description, some of it could be moved to the data description section.
Section 4.2: At this stage, it is not clear for the reader why the threshold elevation is introduced and its relation with the PLR. This should have been explained beforehand, when the entire methodology is explained.
Figure 4 is graphically imbalanced with some very small and some very large plots. In c), it is unclear why the individual samples needed in the whiskers plot. Axes limits are inconsistent.
Section 4.3: This sensitivity analysis is presented as results, but was not described in the methods section. Therefore, the reader is confused because there seems to be a discrepancy between what has been described in the methods and the content of the results. It is unclear how the sensitivity coefficient epsilon is calculated. Frustratingly, I could not find it in any of the equations given. I am unfortunately confused and not able to follow the discussions on positively or negatively correlated sensitivity, since sensitivity is not clearly defined.
The discussion of figure 8 should revolve on the effect of varying lapse rates, however these are not clearly discussed. Instead, the discussion is about the processes underlying floods. To understand, the reader has to go back to the text describing equation 6. Graphically, the colors do not convey a meaning.
I find figure 9 redundant with the second column of figure 5
l.415 and below: The explaination of the rationale for RI should come when RI is explained, in the methods section, rather than in the results
In figure 11, I don't see a clean and interpretable dependence of the relationship with PLR
l. 456-458: I disagree with this statement: usually higher-elevation catchments tend to have thinner (or no) soils, less alluvial aquifers, and therefore can store less groundwater than lower-altitude catchments.
l.491: the insulating affect of snow on glaciers is highly nonlinear and unlikely to be fully represented in HBV
l. 514: I believe that this pattern is not convincingly explained. Since with higher lapse rates we get more snow in most catchements, since the mean elevation of most of them is above the station altitude. Is it an albedo or local circulation/temperature effect? Or could it be that with lower lapse rates we get more precipitation in elevations below the station, which is mostly rain rather than snow, and that contributes to direct floods caused by the rainfall amount rather than the snowmelt? (This would contrast Figure 9 though)
l.532: I am not sure I get the explanation. Is it implied that in that case it is easier for the snow to start melting in higher elevations already in the winter? Is such a claim that intuitive or self-justified?
Minor comments/typos:
l.249: word events written twice
l.266: add a reference to figure 3a
l.283: in -> on
caption of figure 4: s -> is
l.550-551: rephrase
l.655: rephrase
Citation: https://doi.org/10.5194/egusphere-2026-2338-RC2 - What I see in the results is that larger PLRs result in more Q, and more extremes. However this can be explained by the fact that most rain gauges are below the average catchment altitude, therefore positive lapse rates will result in more water than measured at the rain gauge (especially when the nearest-neighbor Thiessen method is used). In a way, in this context the positive lapse rates are "creating" water, especially since the model calibration is kept constant. This alone could explain the correlation of Q with PLR and also the correlation of maximum flows with PLRs. As such, I would need to be convinced that the correlations found are not artifacts of the experimental setting.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 239 | 88 | 24 | 351 | 26 | 17 | 19 |
- HTML: 239
- PDF: 88
- XML: 24
- Total: 351
- Supplement: 26
- BibTeX: 17
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript is about how precipitation lapse rates (PLRs) affect runoff simulation and flood frequency in Swiss mountain catchments, as authors used weather generator outputs to drive HBV hydrological model(s) with differential PLRs from 0 to 10 % per 100m in both components and also isolated the effect of hydrological model PLRs itself by using specific 5% PLRs in WG. The topic is interesting and methodologically is conceptually sound. However, some sections raise questions that should be answered during the revision.
General Comments:
GC1: The PLRs range from 0 to 10 % is presented as a reasonable choice, yet 10% seems a bit on the extreme side. The introduction itself (later also discussion) mentioned that short duration extremes can have negative lapse rates (reverse orographic effect). As the WG produces hourly data, the question is why not also explore the reverse orographic effect and/or if both effects could be somehow combined in WG (different based on seasonality? winter is more prone to daily extremes, summer to hourly). This could be a bit further discussed I believe (maybe as suggestion for future work?), no matter that it is already mentioned in the limitations.
GC2: The baseline of 5% PLR is justified by referencing previous work of the same group (Line 144), while in the introduction they introduced some work with PLR 5%. This raises a question: is the 5% typical across Switzerland only or also used somewhere else? What is the mean value in Switzerland? Might be nice to support the baseline at Line 144 with references out of group too.
GC3: The WG is parameterized on 120 stations (daily, precipitation) across Switzerland and output is later computed via Thiessen polygons to catchment scale. Fig. 3b tries to explain representativeness of stations per basin, which is good, but the manuscript is missing the map showing these stations. Fig. 1 also makes it a bit confusing. -> What does "fully modeled with HBV" mean and why such basins are highlighted? Does it mean that such basins were modeled as semi distributed HBV and the rest as lumped? Lastly, the Sarine and Maggia basins are repeatedly used as example catchments; it would be nice to show them on the map and maybe provide some basic characteristics of these two catchments.
GC4: The daily to hourly disaggregation by analogue method is referenced extensively but is not adequately described in the manuscript. Authors included five references at Line 161 for details, yet only Viviroli et al. 2022 actually documents this step (Evin et al., 2019, 2018 and also GWEX R code include only 3d to daily). More targeted citations, together with a brief description in the manuscript, would help the reader follow the method. Now, I have these questions in mind: Can the method extrapolate beyond the envelope of observed days? Is the method validated and sound to maintain the daily to hourly extremes? Is there randomness in the selection or just RMSE (This might influence the sampling of extremes as a single extreme day can be selected more times, especially on the tail). Last, did you actually check if the WG produces close enough extreme statistics, for example GEV daily and hourly scale for some return period (20y?) in comparison to observation station data?
GC5: I somehow expected a clearer cross-analysis comparing whether PLRs matter more in the WG or in the Hydrological Model. The dual experiment presented is methodologically sound but the narrative in the manuscript requires the reader to connect the dots themselves. Starting section 4.1, the sensitivity of mean annual precipitation to PLR_WG is essentially an expected outcome. Later sections, when authors focus on mixed signals of PLR_WG and PLR_Hydro, make it harder to disentangle which one actually drives the final results, especially when the hydrological model is calibrated with 5% PLR (Line 180). After that, section 4.7 shows that seasonality of floods is highly controlled by PLR_HBV and magnitude by PLR_WG. This is again a bit of an expected outcome, but overall I feel that the manuscript downplays the WG contribution. I would suggest reorder the results section because now it could raise a question of why they even bothered to do PLR_WG if the main factor is PLR_HBV (as it sounds like that in 4.7 after excessive analysis in 4.3, 4.4, 4.5), but the final decision is of course on the authors' shoulders.
GC6: To me, the flood frequency analysis is not clear. The authors didn't mention if AMFs are extracted from daily or hourly discharge series. This matters because each can have different behavior (seasonality too). All analysis is later performed as empirical and on medians I assume. But it raises the question, how certain are you with 30 scenarios for a 1000y return period? Does it follow some expected distribution? Do you have larger uncertainty CI intervals for higher return periods? This should be acknowledged when discussing the uncertainty results.
GC7: The temperature lapse rates are calculated as daily climatology — does it include some uncertainty bounds and year by year fluctuations, or is it assuming that each day has this particular lapse rate based on mean? It would be nice to include a brief explanation in the manuscript. Additionally, it would be beneficial to mention that the temperature lapse rate can vary due to the diurnal cycle, although I understand this was outside the scope of your study.
Specific Comments:
L34: Missing comma after citations in parentheses and missing full stop (dot) at the end of the sentence.
L263: "ΔElevation is illustrated in Fig. 3b" — based on the figure, this should refer to Fig. 3a.
L327–328: "While small but positive sensitivity coefficients occur in winter and spring, they are close to zero in summer for almost all catchments. That suggests that precipitation change does not drive summer snowmelt." This interpretation needs revision. Winter sensitivity coefficients range from roughly 0 to 0.75 and spring from 0 to 0.5; these are not "small." For summer, the trend appears to be pulled higher by a single point on the left side; without it, the signal would look similar to spring. The current wording risks overstating the contrast between seasons.
Fig. 7e: The figure is somewhat confusing. Each catchment is represented by four points, colored by elevation, and the conclusion is drawn from the full sample of 108 points. Does the relationship hold for each catchment individually, or is it an artifact of aggregation? A supplementary version of the same figure with each catchment colored by PLR would help.
Fig. 10: The title of subplot (b) should be moved slightly higher; it currently overlaps with the panel.
L417: "range ts compared" — is "ts" meant to stand for something, or is it a typo?
L519: "This suggests that floods occurring in winter and early spring result from a combination of liquid precipitation and potential earlier snowmelt." The mechanism behind "earlier snowmelt" could be explained more explicitly. I guess lower PLRs lead to more snow falling at lower elevations, where it melts earlier; alternatively, wetter snow and a thinner snowpack at higher elevations could degrade more quickly. A short clarification of which mechanism the authors invoke would help.
L545: "PLR_GWEX" should be "PLR_WG" for consistency with the rest of the manuscript.
L616: Typo: "Czeck Republic" → "Czech Republic".
Bibliography: Bertoncini and Pomeroy (2025) appears twice as two identical entries (2025a and 2025b). Please further check if this is the only one.
Note that there might be more typos, so please double check those as well.