the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Jun 2022
Quantification of meteorological conditions for rockfall triggers in Germany
Katrin M. Nissen
Stefan Rupp
Thomas M. Kreuzer
Björn Guse
Bodo Damm
Uwe Ulbrich
Download
- Final revised paper (published on 23 Jun 2022)
- Supplement to the final revised paper
- Preprint (discussion started on 17 Aug 2021)
Interactive discussion
Status: closed
-
RC1: 'Comment on nhess-2021-243', Anonymous Referee #1, 27 Aug 2021
General comments
The manuscript of Nissen et al. deals with the set up of a logistic regression model to derive the probability of occurrence of rockfalls in Germany, based on meteorological and hydrological variables. The paper is interesting and the writing is fluent and clear. I enjoyed reading it. Regarding results, the authors are able to quantify the impact of increasing rainfall and increasing subsurface water (i.e., pore water, water in fractures) in terms of variations of probability of occurrence of rockfalls. Despite its general good quality and interesting findings, I believe there are certain aspects that need improvement.
- While presenting the results of both the selection of predictors procedure and the logistic regression model, often some apsects are not shown. Since the manuscript is very concise and there isn’t an excessive number of figures, I suggest to show some additional detail.
- Results are clearly presented, expressed in quantitative terms and well-linked to the performed analysis. However, I believe that presenting them with an additional operative perspective could give a lot of added value to the manuscript. In which conditions did most of the rockfalls occurred? Above which percentile of precipitation, subsurface water content? Would it be possible to translate the probabilities of occurrence in a matrix of no-, low-, mid- and high-hazard? See also specific comments #23, #24, #27.
- In terms of presentation, to favor readability, I suggest to consider a more rigorous and classic structure with Intro – Study Area and Data – Methods – Results – Discussions – Conclusions. Now, each step is presented with methods and results together. I got lost just a couple of times, not compromising the general understanding of the work, but I would find it easier to follow with the suggested modification. In addition, dealing the paper with rockfalls, I suggest a thourough check of the use of terms dealing with subsurface water (e.g., pore water, water in fractures, soil moisture, subsurface moisture etc.), since it is not always clear to what exactly the authors are reffering to.
- Discussions need to be extended with at least an additional paragraph comparing the results with previous literature on the topic (see comment #25)
Comments related to contents
- L9-10: Precipitation minus potential evaporation corresponds to what was defined as simulated soil moisture or parameterised pore water? Also, it was stated that moisture observations are not available so which was the assumption made to evaluate the performance of the proxy?
- L18-20: I believe that the described conditions can be most common between the end of the Winter season and the beginning of Spring (late March-April) buti it could be an Alpine bias of mine.
- L27-28: Consequently […] site specific. The sentence probably needs a couple of example references.
- L34-35: in this sentence, is promote used to say that the weathering mechanisms are preparatory-predisposing factors?
- L36: wetting and drying of porous rocks (esp. argillaceous) à I am pointing out this sentence as an example but I think it should be clarified throughout the manuscript how the term porosity (pore) is used (other example in the abstract, L4). Is it matrix porosity only or does it include discontinuities (joints, fractures, etc)? I suggest to specify it in the text. In this specific case, if talking of matrix porosity, maybe it is better to use sandstones as example rather than argillaceous rocks.
- L40-44: earthquakes could be added to the list, although probably not relevant in Germany.
- L52: water in rock cracks can act as a weathering agent through both physical and chemical processes. I would expect not only water presence but also wet-dry cycles causing repeated high water (over-)pressures in fractures to reduce the rock mass strength (i.e., weather the rock mass). I suggest to add some additional explanation to the sentence.
- L58-72: Rockfalls have the year of occurrence but the model is based on hourly, daily and weekly data. How was it possible to fit the model? How could you relate specific values of the meteorological and hydrological variables to the occurrence of the different rockfalls?
- L94-96: I would move this last paragraph towards the end of section 2.1 and introduce Fig. 2 there.
- L112-119: Please add some details regarding the model. In particular, it is mentioned that it is calibrated using gauge measurments; are these gauge measurements soil moisture sensors or else? Which is the time resolution of the model? Also, the model allows simulations through the entire column from the surface to a depth of approximately 1.8 m; within this depth, does it allow to distinguish between actual soil and rock? Is the water infiltration process modelled in the same way for both materials?
- L120: what is meant for operationally available? If I think of a model used for operational purposes I would refer to numerical weather predictions (short or mid-range, i.e. few days) rather than climate models (decades).
- L133-134: Please specify the accumulation periods that were tested. How was the performance of the different accumalation periods evaluated? The fact that the weekly period behaved best is a result. As stated in the general comment, I suggest to consider the possibility of re-organizing the manuscript with a more rigid and classic structure (Intro- Study area – Methods – Results – Discussion – Conclusion).
- L135-138: it is true that with this approach trivial areas (e.g., flat terrain, no rock) were excluded but it could be that potential unstable areas were excluded too. It is a reasonable approach to set-up the model but if predictions are necessary the areas should be filtered based on other terrain and land-use data. I suggest to motivate it explicitly, including the part on the exclusion of the grid boxes with events occurred in periods not covered by meteorological/hydrological data.
- L145: How are the range of values selected?
- L149-164: Mostly results, therefore same suggestion as comment #12. Also, The manuscript does not have an excessive amount of Figures, so why not showing part of the results?
- L179-180: is resolution intended as spatial, temporal or both? In any case I miss the direct link with model comparison. I would expect two models to be comparable if the input (training) data correspond, while in this section it is said that they might change according to the data used (daily precipitation, soil moisture, hourly precipitation).
- L209: it is not clear to me how the cluster predictor works. Can you please explain with additional detail the rationale of including it in the logistic regression model? From L199-201 I understood that the cluster number was just used to subsample the available data and split them in training and test sets.
- L212-213: Why considering all the interaction terms? Physically, what do you expect the product of two terms can explain that their addition does not?
- L218-219: The introduction of AIC belongs to methods.
- L225: Similarly to comment L212-213, what could be the physical meaning of the interaction term between the local percentile of daily precipitation and soil moisture? Why was it decided to include it in the model?
- L227-228: Where is it possible to see that the customisation of the model for regions does not improve the performance?
- L242: Where is it possible to see it?
- L258-259: Among the 237 events used to fit Model 16, given D at its median value or below, how many events occurred for conditions of precipitation below the median and how many for conditions below the 90th percentile? Is it the same in all the three study areas?
- L262-263: Same as above but given D at its 95th percentile or below.
- L267-298: In the discussion section, at least a paragraph should be dedicated to the comparison of results with analyses of previous studies. Perhaps the subject is not identical, but in the introduction several studies are cited that discuss causes and relationships of climate and hydrological variables with rockfalls (e.g., Bajni et al., 2021; D’Amato et al., 2016; Macciotta et al., 2017; Saas and Oberlechner, 2012). I think it could be interesting to know how your results, or the general indications given by your results, compare to those of these studies and maybe other similar ones.
- L275: Data regarding other regions of Central Europe are not presented. I would phrase it a bit more carefully saying that given the similar climatological, hydrological, geological and topographical characteristics the model could be applied in other low mountain areas of Central Europe with success. An evalutaion of its performance would still be needed. The same comment applies to the conclusions (L303) and the title.
- L277-289: in this paragraph, false alarms and prediction errors are discussed. However, in the results there isn’t a real attempt to set probability thresholds and quantify these values. I know how difficult it is but based on the derived models and the recorded occurrences, could you suggest combinations of values to define no (low) hazard, medium hazard, high hazard? This comment is linked to comments #23 and #24 too.
- L312: frost days will decrease, but freeze-thaw cycles could increase at specific elevations. I’ll suggest to phrase it more carefully.
Minor editorial comments
- L5: both for the day of the event and the days leading up to it.
- L36: dissolution in carbonatic rocks
- L49: all climatic factors that promote
- L53: a statistical model that
- L58: rockfall data that. Please check throughout the manuscript the use of which/that.
- L61: database, which […]. The database mainly covers the last 200 years
- L62: Information on 670 rockfall events are included in the landslide dataset.
- L68: while the remaining
- With the majority of them (n=621) recorded from
- L105: 1 km x 1 km or 1 x 1 km2
- L258: Less/more precipitation leads to a below/above
-
AC1: 'Reply on RC1', Katrin Nissen, 17 Sep 2021
Dear Reviewer,
thank you very much for your kind and constructive comments.
This reply is not a paper revision. Here, we would like to describe how we intend to address your main suggestions and to answer the most important questions.Operative perspective:
The advantage of the logistic regression model is that it can be used to determine/predict the probability for a rockfall event if the local meteorological and hydrological conditions are known/forecasted. (This procedure was used in order to construct Figure 4). Which probability is regarded as a low, medium or high risk can be defined by the operator using the model. With a predefined matrix this flexibility would be lost.However, we would like to note that an operational warning system only based on the meteorological and hydrological conditions is not practical in our study region. This becomes clear when looking at Figs. 4a and 4c. The probability for a rockfall event is climatological (i.e. number of events divided by length of the time series) if subsurface water (D) and daily precipitation are of median values. The rockfall probability becomes above average if D and/or precipitation are further increased. Thus, a mid-hazard warning would have to be issued in almost 50 percent of all days. As the climatological probability in our study region is so low, on most of these days this would be a false alarm.
We will add extra information in the revised version of the paper on the number/percentage of events that occurred under different meteorological and hydrological conditions to address your points #23 and #24.
Structure:
We were hoping that deviating from the classic structure (with a separate methods section) and including the description of the methods in the sections they are actually used in, would help the flow of the paper. Our approach has the advantage that we can explain the method using the actual application as an example. We will rethink this decision. Hopefully, our second reviewer will also comment on that issue.Terminology
We will check and clarify our use of the terms dealing with subsurface water and porosity.Discussion:
We agree that an additional paragraph comparing the results with previous literature on the topic is needed.Answer to specific questions:
1. The performance of D refers to its ability to improve the statistical model.4. Yes, “promote” is used to say that the weathering mechanisms are preparatory-predisposing factors.
8. For most events the exact date is known. These are the events that were used in the analysis. The hour is not needed as we evaluated the daily maximum of hourly precipitation.
13. When the model is applied to climate simulations a terrain filter will indeed be necessary.
14. The number of bins was set to 6. The restriction that each bin has to contain the same number of events, determines the range included in a bin.
16. The sentence refers to the spatial resolution.
18. (and 20) This might become clear if you look at Fig. 4. Without the interaction term the relationship shown in Fig. 4a would be the same as in 4b and that in 4c would be the same as in 4d. The approach is able to reflect that precipitation is more effective if the preconditions in terms of sub-surface water are favourable
-
AC3: 'Reply on RC1', Katrin Nissen, 20 Dec 2021
Addition to AC1
This addition addresses the remaining points we haven’t covered in our first reply to RC1
General:While presenting the results of both the selection of predictors procedure and the logistic regression model, often some aspects are not shown. Since the manuscript is very concise and there isn’t an excessive number of figures, I suggest to show some additional detail.
In order to keep the paper short and concise we would like to refrain from adding additional figures. All figures needed to understand the methodology are already included. Rather than expanding the main paper we will add a supplement showing additional material e.g. the result of the sensitivity analysis performed for the WOE analysis and the effect of different predictor combinations in the logistic regression model on the prediction of the rockfall probability.Comments related to specific contents:
1. Please see AC1.
2. L18-20: The seasonal cycle does indeed show a peak in winter and spring.
3. L27-28: We will extend the introduction and add more references.
4. Please see AC1.
5. L36: Both reviewers pointed out a number of expressions that need to be defined more detailed (porosity, pore-water, promote, soil moisture). We will address this by defining the terms and homogenizing their usage in the revised version of the manuscript.
6. L40-44: Earthquakes are indeed not relevant for our study region.
7. L52: We will extend the introduction adding additional information.
8. Please see AC1.
9. L94-96: We will restructure the manuscript adapting a more classical outline (introduction, data, methods, results, discussion and conclusions).
10. L112-119: We will extend our description of the soil moisture model.
11. L120: We will replace the term “operationally available” by “stored”.
12. L133-134: For D accumulation periods between 14 and 6 days were tested in terms of their ability to improve the logarithmic skill score of the logistic regression model. The skill increased with decreasing duration and reached its peak at 7 days.
13. Please see AC1.
14. Please see AC1.
15. L149-164: We will restructure the manuscript adapting a more classical outline (introduction, data, methods, results, discussion and conclusions). We will also add a supplement showing additional figures.
16. Please see AC1.
17. L209: Using the clusters as an additional predictor has the effect of fitting 4 different models, one for each cluster. We will rephrase L227 where this is explained.
18. L212-213: The best result in terms of the logarithmic skill score is achieved by using all possible combinations. The relationship between D and precipitation is linear if only the sum is used. A linear relationship does not reflect the fact that precipitation becomes more efficient if D is high. The product reflects this fact but seems to overestimate the probabilities for high precipitation or D percentiles. Including the sum in addition to the product has a dampening effect on the high probabilities. We will extend the discussion and include this information. A figure will be added as supplementary material.
19. L218-219: We will add a methods section and move the introduction of the AIC.
20. Please see our reply to 18.
21. L227-228: Models 9, 10 and 12 don’t improve the logarithmic skill score.
22. L242: This is not shown. We can add a figure to the supplementary material. We think that adding figures for all the sensitivity studies we performed to the paper makes it more difficult to follow the main story.
23. L258-259: We will add extra information in the revised version of the paper on the number/percentage of events that occurred under different meteorological and hydrological conditions.
24. L262-263: See answer to 24.
25. L267-298: We agree that an additional paragraph comparing the results with previous literature on the topic is needed.
26. L275: Both reviewers suggest to be more restrictive when naming the region in which the statistical model can be applied. The intention was to express that the applicability of the model does not stop at the state borders. We will change the title of the manuscript and discuss in more detail the possibility to apply the model in other Central European low mountain regions with similar climatological, hydrological, geological and topographical characteristics.
27. Please see AC1.
28. L312: We will rephrase this sentence.
-
RC2: 'Comment on nhess-2021-243', Anonymous Referee #2, 29 Nov 2021
General comments
The manuscript proposes a statistically-based approach to quantitatively assess the impact of a set of meteorological and hydrological variables on rockfall occurrence in some selected low-mountain regions in Germany. These variables are thus considered as potential triggering factors and have been analysed for the event day and for the days before. The authors conclude that the logistic regression-based model used is able to detect changes in the probability of rockfall occurrence in the study area. Precipitation at daily scale turn to be the main triggering factor and a 5-parameters model based on the interaction of daily precipitation, freezing-thaw cycles and increase in sub-surface water is the most appropriated in terms of skill.
The manuscript represents a valuable and innovative contribution to the understanding of climate variables-related impact on the hydrogeological risk. Outcomes of the research are very interesting and the paper is in general well-written and fluent. I’m not a mother-tongue but I think that the English is good. Nevertheless, I personally think that some major changes are needed and would improve the overall quality of the paper.
First, I think that he paper should be restructured in a more standardized and classic way to improve readability, in particular the data and methods sections. I suggest to state clearly each part e.g., Data, Methodology, Results etc. The manuscript as is mixes together data, methods and results and I must admit that I encountered some difficulties in following the text.
The description of data and methods (e.g., methods to simulate soil moisture and pore water proxy) is not sufficiently complete. Further specification and clarification should be included throughout the text. Adding some figures related to not shown analysis could be of help in this (see Specific comments).
The Discussion as is lacks of comparisons with similar studies on the same topic (i.e., relations among climate, hydrological variables and rockfall occurrence) to assess how they differ or agree.
Other specific comments are listed hereinafter. Moreover, I think it is a bit ambitious to say that the model is valid for the entire Central Europe as stated in the title as well.
Specific comments
Title
I suggest to revise the title. The considered dataset is not representative of Central Europe actually.
Abstract
L 10: Maybe it should be clarified that precipitation minus evapotranspiration is a pore water proxy (as an alternative to simulated soil moisture).
Introduction
L 33: Is the term “promote” related to predisposing condition for rockfall occurrence? Check the meaning throughout the text.
L 45-55: Actually, studies focusing on statistically-based approach to assess the linkage between climate forcing and rockfall/landslide occurrence are increasing worldwide and especially in mountain-areas like e.g., the one investigated in this work. I suggest to discuss further this point and mention some relevant works in this context, highlighting how this study adds value and contributes to shade light on climate and hydrological-related trigger mechanisms.
Meteorological and hydrological variables
L 112- 119: I would add some further details on the methodology adopted. What’s the resolution of the model? Can the model distinguish between different types of lithologies across the entire column?
L 114: In general, the use of terms related to soil moisture ad sub-surface water throughout the text looks a bit misleading to me, since it is not always clear to what the authors refer to.
Figure 1: Please add the meaning of the cluster in the caption.
L 120: It is not clear if the authors refer exactly to climate scenarios or to a shorter-term prediction. Before, there was no reference on the use of climate change scenarios for this analysis.
L 120: Not clear how the authors validated the pore water proxy if soil moisture information are available only for some sites and simulations. In general, I think that the authors should provide more details on the methodology and procedures used to calculate the soil moisture and pore water proxy.
L 128-129: Please rephrase.
L134: Please specify what are the accumulation periods considered.
L 145: How the authors select the range of values for each variable?
L 149: These are part of the results.
L 155-164: This is a mix of results and methods. Please see the general comment.
L 156: Considering time spans before the rockfall events is useful not only for thawing process, but for the antecedent moisture condition as well. It looks like that the time span including the days leading-up to the failure have been considered only for thawing process while, as can be derived from L 134, different accumulation periods prior to the failure have been investigated also for pore water. Please clarify better this point.
L 157: What time spans did the authors consider?
L 161: It is not very clear to me how the authors could compare the importance of such variables if the number of rockfalls and grid change depending on the spatiotemporal resolution of the respective datasets. Maybe it is worth to include the results of the consistency test here or as Supplementary Material.
L 165: This part seem to be more appropriate for a discussion.
L169: Please add some reference.
L 178: It is not very clear to me how the WOE considers the fact that more than one landslide occurs in a grid box.
L 179: I guess that the resolution is intended as spatial?
L 213: This point has to be clarified. Please discuss further why using all possible combinations of predictors instead of simply the sum or product could add value to the analysis.
Results
L 218: AIC has not been introduced before. Please add further details in the methodology section.
L 238 The authors should introduce the term “across-site percentile” and clarify it.
L 246 This can be due also to the DEM resolution, that could not be sufficiently higher to detect the exact location of a rockfall.
L 254 AIC considers both how well the model reproduces the data and the number of variables (maximum likelihood estimates of the model) used to build the model. The lower the AIC value, the better the model fits as the authors rightly state at L 220. If I rightly understood, AIC is much lower in model 15 compared to the selected one (16). Is the choice due to the number of rockfall events involved? The authors partially explained the final selection; thus, I would suggest to discuss further this point.
Discussion
L 270: The authors rightly state that there is no guarantee that the three sampling locations are representative for the entire Germany but at L 275 they say that the approach can be reasonably extended in Central Europe regardless of the geographical and local settings. I think that additional analyses are needed to state this actually.
L 284 I would say that also the inclusion of more climate and hydrological variables could be of help in decreasing the number of missed alarms.
L 291 As I understood correctly, time series of different lengths have been used depending on the considered variable. To determine the percentile of the variable in the lead-up of the event with relation to the previous period, the authors use both a local and across-site percentile. Across-site percentiles could be misleading in this context. Does “across-percentile” mean that the same variable is compared across records of different weather stations with different lengths while the local percentile is referred to the local time-series?
L 303 Be careful about saying that the model is representative for low-mountain regions in Central Europe. I would say in Germany, at most.
Technical corrections
L 6: and instead of “as well as”
L 49: “that” instead of “which”
L 105: Be consistent throughout the text: 1 km x 1 km 1 km2 (L 102).
L 258: “a below” instead of “an below”
-
AC2: 'Reply on RC2', Katrin Nissen, 17 Dec 2021
Dear Reviewer,
thank you very much for your constructive comments.
In the following we would like to describe how we intend to address your suggestions and to answer the most important questions.General comments:
- Structure: Both reviews suggest a more classical outline (data, methods, results). We will therefore restructure the paper.
- Terminology: Both reviewers pointed out that we should clarify the use of certain terms. In detail these are:
-sub-surface water
-pore water
-soil moisture
-promote
We will address this by defining the terms and homogenizing their usage in the revised version of the manuscript. - Discussion: We agree that an additional paragraph comparing the results with previous literature on the topic is needed. This was pointed out by both reviewers.
- Region of applicability: Both reviewers suggest to be more restrictive when naming the region in which the statistical model can be applied. The intention was to express that the applicability of the model does not stop at the state borders. We will change the title of the manuscript and discuss in more detail the possibility to apply the model in other Central European low mountain regions with similar climatological, hydrological, geological and topographical characteristics.
Reply to specific comments:
Title: Please see our comment “Region of applicability”.
Abstract
L 10: We will state that precipitation minus evapotranspiration is a pore water proxy.
Introduction
L 33: Please see our comment “Terminology”.
L 45-55: We will extend the literature review on the linkage between climate forcing and rockfall/landslide occurrence in the introduction.
Meteorological and hydrological variables
L 112- 119: We will extend the description of the mHM model in the revised version of the manuscript. In this model setup, the soil component in mHM consists of six soil horizons (vertical levels) up to a depth of 2m. The horizontal simulation resolution is 5x5 km (grid cell size). All processes, states and fluxes are calculated for each cell. This includes among others land cover, slope, aspect, soil texture, clay percentage, sand percentage and mineral bulk density.
L 114: Please see our comment “Terminology.
Figure 1: We will add information on the clusters to the figure caption.
L 120: The model will be used to analyse climate scenario simulations. This will be clarified in a revised version of the manuscript.
L 120: D and the soil moisture simulations are not validated against observations. Instead their ability to improve the statistical model in terms of the logarithmic skill score is compared. We will clarify that in the revised manuscript.
L 128-129: Will be rephrased.
L 134: For D accumulation periods between 14 and 6 days were tested in terms of their ability to improve the logarithmic skill score of the logistic regression model. The skill increased with decreasing duration and reached its peak at 7 days.Selection of potential predictors
L 145: In order to allow comparison of the WOE for different variables the same number of bins needs to be selected for each variable. Here it was set to 6. Each bin has to contain the same number of observations. This restriction determines the range included in a bin.
L 149: Please see our comment “Structure”.
L 155-164: Please see our comment “Structure”.
L 156: For the two variables moisture precondition and freeze-thawing cycles the days before the rockfall events were considered. We will clarify this in a revised version of the manuscript.
L 157: Time spans between 2 and 14 days were tested.
L 161: We think that the analysis of the relationship between rockfall and the individual potential predictors is most robust if all available data is used. We therefore show this result in Fig. 3 of the manuscript. In order to make sure that using periods of different length for the different predictors does not distort the information of the IV value on the importance of the predictors we conducted a sensitivity analysis using only the period available for all variables. We can add this figure as supplementary material. It is too similar to Fig 3 to warrant an extra figure in the main paper.
L 165: We will restructure the paper.
L 169: This sentence refers to the relationships described in the introduction. We will add an explanation here.Construction of a statistical model
L 178: Input for a WOE analysis are two equally long vectors. The first contains the concatenated time series of a meteorological observation extracted from the grid boxes closest to the positions of the rockfall events. The second one contains the information if a rockfall event occurred at this time step and location (yes/no). If more than one landslide occurs in a grid box the time series associated with this grid box is included only once in the WOE analysis.
This procedure of removing duplicated time series is not practical for the logistic regression were the time series of the different meteorological variables are analysed together. Two rockfall events may be located in the same grid box of the coarser observational data set but in different grid boxes of the higher resolution data set. The number of sites used for model fitting would vary for each predictor combination. This would make a comparison of the statistical models impossible. It is important to ensure that the model selection process is done by comparing statistical models fitted using an equal number of time series and time series of equal length. Comparing models 14 and 11 instead of 14 and 15 demonstrates how important this is.
L 179: We will add “spatial” to the text.
L 213: The best result in terms of the logarithmic skill score is achieved by using all possible combinations. The relationship between D and precipitation is linear if only the sum is used. A linear relationship does not reflect the fact that precipitation becomes more efficient if D is high. The product reflects this fact but seems to overestimate the probabilities for high precipitation or D percentiles. Including the sum in addition to the product has a dampening effect on the high probabilities. We will extend the discussion and include this information
Results
L 218: AIC will be introduced in the new methodology section.
L 238: “across-site percentile” is the percentile over all grid boxes which include the location of an event. This will be added to the text.
L 246: We will add this to the discussion
L 254: The low AIC must be seen in perspective, as model 15 was fitted using fewer and shorter observational time series. The AIC of model 15 shouldn’t be directly compared to model 16. Please see also our comment to L178. We will clarify that in a revised version.
Discussion
L 270: Please see our comment “Region of applicability”
L 284: From the literature review we conclude that we have included the most important climate and hydrological variables. Including too many variables can result in overfitting. As we see by looking at model 10 this can already happen with 16 parameters.
L 291: As we did not use data from stations directly but a gridded product (based on station observations) the length of the time series is the same for all locations. Local and across-site percentile calculations are based on the same length of time series.
L 303: Please see our comment “Region of applicability”
-
AC2: 'Reply on RC2', Katrin Nissen, 17 Dec 2021
Peer review completion
