The dynamics of peak head responses at Dutch canal dikes and the impact of climate change

Strijker, Bart; Kok, Matthijs

doi:https://doi.org/10.5194/egusphere-2024-1495

Preprints

https://doi.org/10.5194/egusphere-2024-1495

Preprints

20 Sep 2024

| 20 Sep 2024

The dynamics of peak head responses at Dutch canal dikes and the impact of climate change

Bart Strijker and Matthijs Kok

Abstract. The management of water and flood risk levels in low-lying polder regions depends on the performance of canal dikes. Heavy rainfall can lead to peak hydraulic heads within the dikes affecting their stability, which can induce dike breaches. Variations in head responses and head statistics are both relevant for regional flood risk analysis of canal dike systems. This study examined the dynamics of peak heads in canal dikes on a national scale using time series models calibrated on observed heads. Various model structures are evaluated and a nonlinear model performed the best. These models were used to simulate long-term head time series. Subsequently, dike clusters were identified based on the coincidence of peak heads, allowing for the identification of dikes where peaks are caused by (dis)similar types of rainfall events. The differences and similarities in peak head response were related to physical dike characteristics. While no single significant relationship emerged, the soil type combined with the width of the dike appears to be important factors influencing the variation in head responses. However, the presence of the same soil type and dike widths in multiple clusters indicates that these characteristics do not yield a definitive outcome for the head response. Moreover, peak head statistics across various dikes were derived and indicated that extreme and yearly load conditions are relatively close to each other, with a median decimation height of only 15 centimeters. The head statistics are affected by climate change, characterized by increasing winter precipitation and summer evaporation. By 2100, extreme peak heads are expected to occur between 3 times less and 8 times more frequently, depending on the climate scenario and the type of canal dike.

Received: 30 May 2024 – Discussion started: 20 Sep 2024

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.

Download & links

Preprint (PDF, 3154 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (3154 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

12 Sep 2025

The dynamics of peak head responses at Dutch canal dikes and the impact of climate change

Bart Strijker and Matthijs Kok

Nat. Hazards Earth Syst. Sci., 25, 3355–3379, https://doi.org/10.5194/nhess-25-3355-2025,https://doi.org/10.5194/nhess-25-3355-2025, 2025

Short summary

Bart Strijker and Matthijs Kok

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-1495', Anonymous Referee #1, 02 Dec 2024

The paper reports on a comprehensive analysis of dike strength and failure for a large number of polder dikes across the Netherlands. Using a model for dike 'ground'water head calculation and time series of precipitation and evapotranspiration, the model produces time series of water heads in dikes. Results are statistically analysed, including grouping of dike types across the area.
The subject fits well to the scope of NHESS, and addresses a field that has received increasing attention, both in water management and (applied) research into dike stability and flood risk.
Still the manuscript requires revision before final publication. My main points are:
- I miss a spatial component in the discussions of clusters and 'correlation': how would a map indicating cluster member of dike sections look - is there any spatial correlation, or not?
- The meaning of the clusters determined from the statistical analyses remains a unclear - what do they tell us, and how do these contribute to the overall objective, i.e. predicting dike stability over large areas, under conditions of extreme precipitation. I got the impression that a different type of clustering would have added more insight.
- The discussion section should me restructured and complemented with thoughts about climate change impacts, and implications for the dikes, and their management.
- The conclusions are too much a summary: rewrite these in 'conclusive' style, not a story of what you did.
- At several points the text is wordy, or unclear - see specific comments.
In the attached pdf I have put my detailed comments.

Citation: https://doi.org/10.5194/egusphere-2024-1495-RC1
- AC1: 'Reply on RC1', Bart Strijker, 07 Feb 2025
  
  We appreciate the Reviewer’s helpful comments and suggestions and have addressed them individually in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2024-1495-AC1
RC2:
'Comment on egusphere-2024-1495', Anonymous Referee #2, 20 Jan 2025
This manuscript outlines an application of time series models using impulse response functions to model the hydraulic heads observed with dike systems in the Netherlands. Different model structures are tested to simulate the heads, with a nonlinear-threshold model (TARSO) found to perform the best. This is an interesting result, that could teach us something about how the heads in dike systems respond to precipitation and potential evaporation. The study attempts to relate model characteristics to various physical characteristics of the dike systems, with moderate success. I generally found the manuscript well written and the figure quality appropriate. The topic fits the scope of the journal. I have a couple of major comments that should be addressed, and some minor technical comments at the bottom.

One thing I was missing in the manuscript is an explanation and interpretation of why the threshold nonlinear model structure (TARSO) works best for the hydraulic heads in Dikes in the Netherlands. This is a surprising outcome to me, that deserves more thought and might be informative for future attempts to model the heads in dikes. This model was designed for a different type of system (groundwater levels in polders, influenced by ditches falling dry and being activated). Perhaps there is topping-off of the heads in dikes. These types of models are commonly used to gain understanding of how groundwater systems function, and why. A discussion of this type is currently missing from the manuscript but would be a welcome addition.

Looking at Figure 6, I was very surprised by the simulated behavior of the FlexModel given that all models share the same input and not be that far off from each other. I made a quick script to model the data using the Pastas default values to better understand the result but got an average of R²=0.68 for the FlexModel, much higher than the reported value of 0.32. I suspect some suboptimal choices were made for that model. Perhaps the Authors can revisit the scripts and double-check this, or explain this result in more detail.

I assessed the manuscript for its reproducibility. I appreciate the authors providing the original head, precipitation, and evaporation data is provided. This data provides a unique dataset on head measurement in dike systems, which might be worth highlighting in the manuscript. I note here that none of the data underlying the results shown in the figures and tables are shared, nor are the scripts that lead to the results. This makes it difficult to verify the results and/or build upon this work. I would recommend the authors to share the scripts and output data on a FAIR repository to improve the reproducibility of this study.

Minor technical suggestions:
L84: Potential evaporation

L195: Figure 4 and 5 appear to be the same. I am not sure if another figure is meant to be here. Otherwise Figure 5 can be removed.

L192: How well does GeoTop actually work for dikes? I can imagine that these are not related at all, given that dikes are built by humans with specific materials. This may influence the results later on relating outcomes to the soil types, i.e., would the relationship with the soil type improve if the GeoTop data is left out? Some consideration about this would be good here.

L238: I think Sm is substituted by R, not the other way around.

L276: simulate “the heads”.

L288: How was this threshold of 0.7 determined? How sensitive are the result to changing this threshold.

L302: Analyses

L303: every “…”?

L327: r2 was previously referred to as R2, check throughout

L329: The FlexModel is a nonlinear recharge model, the other three models are not. I think the TARSO model is meant here, which still computes recharge using a linear equation.

L342: replace by “in the summer of 2019”

L343: disturbances

L345: scatter plot in Figure XX.

L366: I don’t understand what “peak block” is, please clarify.

L496: Apparently, the head time series were filtered using some reliability criteria in this study. This should be mentioned in the section describing the data. What reliability criteria were used?

L501: Remove “explicitly”. Uncertainty was not considered, as I understood from the manuscript.
Citation: https://doi.org/10.5194/egusphere-2024-1495-RC2
- AC2: 'Reply on RC2', Bart Strijker, 07 Feb 2025
  
  We appreciate the Reviewer’s helpful comments and suggestions and have addressed them individually in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2024-1495-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-1495', Anonymous Referee #1, 02 Dec 2024

The paper reports on a comprehensive analysis of dike strength and failure for a large number of polder dikes across the Netherlands. Using a model for dike 'ground'water head calculation and time series of precipitation and evapotranspiration, the model produces time series of water heads in dikes. Results are statistically analysed, including grouping of dike types across the area.
The subject fits well to the scope of NHESS, and addresses a field that has received increasing attention, both in water management and (applied) research into dike stability and flood risk.
Still the manuscript requires revision before final publication. My main points are:
- I miss a spatial component in the discussions of clusters and 'correlation': how would a map indicating cluster member of dike sections look - is there any spatial correlation, or not?
- The meaning of the clusters determined from the statistical analyses remains a unclear - what do they tell us, and how do these contribute to the overall objective, i.e. predicting dike stability over large areas, under conditions of extreme precipitation. I got the impression that a different type of clustering would have added more insight.
- The discussion section should me restructured and complemented with thoughts about climate change impacts, and implications for the dikes, and their management.
- The conclusions are too much a summary: rewrite these in 'conclusive' style, not a story of what you did.
- At several points the text is wordy, or unclear - see specific comments.
In the attached pdf I have put my detailed comments.

Citation: https://doi.org/10.5194/egusphere-2024-1495-RC1
- AC1: 'Reply on RC1', Bart Strijker, 07 Feb 2025
  
  We appreciate the Reviewer’s helpful comments and suggestions and have addressed them individually in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2024-1495-AC1
RC2:
'Comment on egusphere-2024-1495', Anonymous Referee #2, 20 Jan 2025
This manuscript outlines an application of time series models using impulse response functions to model the hydraulic heads observed with dike systems in the Netherlands. Different model structures are tested to simulate the heads, with a nonlinear-threshold model (TARSO) found to perform the best. This is an interesting result, that could teach us something about how the heads in dike systems respond to precipitation and potential evaporation. The study attempts to relate model characteristics to various physical characteristics of the dike systems, with moderate success. I generally found the manuscript well written and the figure quality appropriate. The topic fits the scope of the journal. I have a couple of major comments that should be addressed, and some minor technical comments at the bottom.

One thing I was missing in the manuscript is an explanation and interpretation of why the threshold nonlinear model structure (TARSO) works best for the hydraulic heads in Dikes in the Netherlands. This is a surprising outcome to me, that deserves more thought and might be informative for future attempts to model the heads in dikes. This model was designed for a different type of system (groundwater levels in polders, influenced by ditches falling dry and being activated). Perhaps there is topping-off of the heads in dikes. These types of models are commonly used to gain understanding of how groundwater systems function, and why. A discussion of this type is currently missing from the manuscript but would be a welcome addition.

Looking at Figure 6, I was very surprised by the simulated behavior of the FlexModel given that all models share the same input and not be that far off from each other. I made a quick script to model the data using the Pastas default values to better understand the result but got an average of R²=0.68 for the FlexModel, much higher than the reported value of 0.32. I suspect some suboptimal choices were made for that model. Perhaps the Authors can revisit the scripts and double-check this, or explain this result in more detail.

I assessed the manuscript for its reproducibility. I appreciate the authors providing the original head, precipitation, and evaporation data is provided. This data provides a unique dataset on head measurement in dike systems, which might be worth highlighting in the manuscript. I note here that none of the data underlying the results shown in the figures and tables are shared, nor are the scripts that lead to the results. This makes it difficult to verify the results and/or build upon this work. I would recommend the authors to share the scripts and output data on a FAIR repository to improve the reproducibility of this study.

Minor technical suggestions:
L84: Potential evaporation

L195: Figure 4 and 5 appear to be the same. I am not sure if another figure is meant to be here. Otherwise Figure 5 can be removed.

L192: How well does GeoTop actually work for dikes? I can imagine that these are not related at all, given that dikes are built by humans with specific materials. This may influence the results later on relating outcomes to the soil types, i.e., would the relationship with the soil type improve if the GeoTop data is left out? Some consideration about this would be good here.

L238: I think Sm is substituted by R, not the other way around.

L276: simulate “the heads”.

L288: How was this threshold of 0.7 determined? How sensitive are the result to changing this threshold.

L302: Analyses

L303: every “…”?

L327: r2 was previously referred to as R2, check throughout

L329: The FlexModel is a nonlinear recharge model, the other three models are not. I think the TARSO model is meant here, which still computes recharge using a linear equation.

L342: replace by “in the summer of 2019”

L343: disturbances

L345: scatter plot in Figure XX.

L366: I don’t understand what “peak block” is, please clarify.

L496: Apparently, the head time series were filtered using some reliability criteria in this study. This should be mentioned in the section describing the data. What reliability criteria were used?

L501: Remove “explicitly”. Uncertainty was not considered, as I understood from the manuscript.
Citation: https://doi.org/10.5194/egusphere-2024-1495-RC2
- AC2: 'Reply on RC2', Bart Strijker, 07 Feb 2025
  
  We appreciate the Reviewer’s helpful comments and suggestions and have addressed them individually in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2024-1495-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (18 Feb 2025) by Paolo Tarolli

AR by Bart Strijker on behalf of the Authors (25 Mar 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (25 Apr 2025) by Paolo Tarolli

RR by Anonymous Referee #1 (27 May 2025)

RR by Anonymous Referee #2 (06 Jun 2025)

RR by Anonymous Referee #3 (09 Jun 2025)

Suggestions for revision or reasons for rejection

This study presents a compilation of hydraulic head measurements from polder regions across the Netherlands and conducts a data-driven time series analysis of current and future trends in head. The authors find that a non-linear threshold model best reproduces the observed data, and observation data may be clustered into 4 groups based on coincident observations of peak head in space and time. Under four different climate scenarios, changes to the return interval of extreme head values may be negative or positive because future outcomes depend on the change in future hydroclimatic variables. This study is likely to be of interest to the scientific community, particularly to those involved in hazard assessments of low-lying canal dikes. I look forward to seeing it in print.

The manuscript would benefit from a round of minor revisions. Suggested edits to improve the manuscript primarily relate to manuscript organization. There are multiple places where content in the Introduction, Methods, and Results would fit better in a different section. In addition, the results of the climate change analysis are interesting and will be of interest to the community, but they merit further attention in the Discussion section. Lastly, there are some grammatical errors to resolve. Specific recommendations on each of these themes (as well as a few minor ones) are presented by line number below.

Line 12-13: I recommend rephrasing this sentence to: “These models were used to simulate 30-year time series of head under current and plausible future climate scenarios.”

Line 56: remove comma after “heavy rainfall event”

Lines 84-98, and Figure 2: This paragraph and Figure 2 are useful to understanding the methodological framework of the study, but they feel out of place in their current location at the end of the Introduction section. This paragraph and figure summarize the Methods used in the study; for this reason, I suggest moving both the paragraph and Figure 2 to the very start of the 3. Method(s) section, before section 3.1. This will fit more naturally within the structure of the manuscript and will give readers a helpful summarization of methodology right before diving into all of the details. Note that this change will also require renumbering Figures 3 & 4 and their respective references in the text. The final sentences of the previous paragraph (i.e., lines 78-82, starting with “This study aims to…”) seem like a more natural end to the Introduction section, as they provide a general roadmap of the study. The authors may also choose to expand on the roadmap with an additional statement or two on the novelty of their approach and the research gap that they address.

Line 85: the acronym “IRF” is defined in three separate places in the text (here, line 193, and the caption of Table 1), but “IRF” is only used as a column header in Table 1. I would recommend removing this acronym from the text since it is not used, or alternatively, replacing the following instances of “impulse response function” in the text with “IRF”.

Line 89: remove comma after “resulting in a set of models”

Line 110: fix spelling of “focusses”

Figure 4: I find this figure to be very helpful. One recommended change: it is not clear on first glance what is meant by the terms “90th percentile profile” and “10th percentile profile" – my first interpretation was that they referred to the phreatic surface profiles. These terms are eventually defined as points drawn from elevation profiles on line 309. Please make this clearer in either the figure itself or in the figure caption.

Line 197: the index of summation under the summation symbol is a capital “M”, the same as the upper bound of the summation. Should this be a lower-case “m”?

Line 205: I understand that the authors use more than one type of response function (e.g. exponential, gamma) that are not defined in the manuscript. This seems fine, but can the authors provide a citation here to point interested readers to a resource with fuller descriptions of the response functions?

Line 215: this water budget does not include surface runoff (follow-up: as described in the Discussion section 5.1); is surface runoff unimportant/less relevant to polder canal hydrology? This merits a brief mention here, too.

Table 1: please replace all instances of “lineair” with “linear”

Lines 246-249 and Figure 5: These results belong in the Results section.

Line 296-298: though not entirely clear, this sentence seems to imply that only the exponential distribution was used throughout the study. If this was the case, then for clarity, I would recommend removing the discussion and definition of the general form of the generalized Pareto distribution and instead define just the exponential distribution. If not, then please clarify where/how the Type I, II, and III distributions were utilized in the text.

Line 310: as with “IRF”, the target levels acronym “TL” is not used in the text (other than in Figure 4) and should be removed. Its definition in the caption of Figure 4 should remain.

Line 317: “clusters of dikes” is mentioned here, but no methodological details are provided. They are instead described later in the Results section 4.2.1 (lines 394-411). Most of the paragraph on lines 394-411 belongs in the Methods, not the Results.

Section 4. Results: there are several places where interpretation of results (and attribution of causes behind the results) are included in the Results section but likely belong in the Discussion section. Moving them would help focus the reader’s attention on the results of this study while helping to bolster the discussion of study results within a wider scientific context, which is largely missing from the Discussion section. In general, I encourage the authors to more fully differentiate presentation of results in the Results section from interpretation of results and their connection to past work in the Discussion. The lines below are suggested places to consider moving content from the Results to the Discussion:
• Lines 379-386: beginning with “The difference in response time can be…”
• Lines 473-479: all sentences on these lines
• Parts of lines 498-503 that infer explanations for the results

Line 330: recommend using “Therefore,” in place of “So,”

Figure 6 and lines 350-351: what is “A” in the block response plot? It seems to be the peak response value. Additionally, lines 350-351 state “The peak of the block response represents the maximum increase in head level that would occur,” implying that the block response has dimensions of length (e.g., units of meters), but the y-axis label in the block response plot is “Response [-]”, implying it is dimensionless. Can the authors please clarify?

Line 425-426: This first sentence is unnecessary; it can be removed without loss of detail.

Lines 464-466: The definition of decimate height and how it is used would be more appropriate in the Methods section, probably section 3.2.

Line 485: little information is provided in the manuscript on the climate change analysis portion of the study. Greater description of the datasets presented in section 2.2.2 would be helpful to the reader, i.e., fuller descriptions of SSP5-8.5, SSP1-2.6, and the wet- and dry-trending scenarios that were introduced on line 176. Sentences from this section that describe methods should be moved to an appropriate section in the Methods (e.g., lines 485-488, the definition of the probability factor on line 494-495). The authors might consider a new section at the end of the Methods to bring the descriptions of climate change data and methods together in one place.

Line 502: the authors state that the “precipitation increases” in the climate scenarios with drying trends; this seems like a contradiction, although it may not be. Perhaps additional details on the methods behind the climate-change analysis (as suggested above) would alleviate this confusion.

Lines 503-504: The authors state “For climate scenarios in 2100 with a wetting trend, the frequency of occurrence of extreme load levels is increasing,” but the right plot in Figure 11 seems to indicate that the low-emission, wet regional response scenario (Lw) results in little-to-no change in probability factor. Please clarify.

Line 516: change “Frist” to “First”

Lines 629-631: One important aspect of the Conclusions and the Abstract is that extreme peak heads “…are expected to occur between 3 times less and 8 times more frequently by 2100, depending on the climate scenario and the type of canal dike,” but this statement is not expounded upon in the Discussion. In addition to the above-mentioned modifications of the Results/Discussion sections, I would recommend a few sentences in the Discussion that explicitly tease this conclusion apart. For instance, under which specific scenarios (and timeframes) do the extreme peak heads tend to become less frequent, and under which do they become more frequent? The right plot in Figure 11 also shows some interesting results that are not discussed: the high emission + drying scenario (Hd) shows large declines or no change to the probability factor in 2050 but generally shows increases in 2100. Likewise, the high emission + wetting scenario (Hw) shows a slight increase in 2050 but a much larger increase across the board in 2100. Can the authors infer any causes behind these temporal trends? Finally, a discussion of precipitation seasonality is mentioned in the Abstract, Introduction, Conclusions, and the Appendix, but is almost entirely absent from the Discussion. Trends in seasonality and the importance of the type of canal dike when considering future climate conditions should be explicitly commented upon in the Discussion.

Hide

ED: Publish subject to minor revisions (review by editor) (13 Jun 2025) by Paolo Tarolli

AR by Bart Strijker on behalf of the Authors (23 Jun 2025) Author's response Author's tracked changes Manuscript

ED: Publish as is (04 Jul 2025) by Paolo Tarolli

AR by Bart Strijker on behalf of the Authors (09 Jul 2025)

Journal article(s) based on this preprint

12 Sep 2025

The dynamics of peak head responses at Dutch canal dikes and the impact of climate change

Bart Strijker and Matthijs Kok

Nat. Hazards Earth Syst. Sci., 25, 3355–3379, https://doi.org/10.5194/nhess-25-3355-2025,https://doi.org/10.5194/nhess-25-3355-2025, 2025

Short summary

Bart Strijker and Matthijs Kok

Viewed

Total article views: 928 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
595	80	253	928	32	45

HTML: 595
PDF: 80
XML: 253
Total: 928
BibTeX: 32
EndNote: 45

Views and downloads (calculated since 20 Sep 2024)

Month	HTML	PDF	XML	Total
Sep 2024	96	7	25	128
Oct 2024	45	6	71	122
Nov 2024	44	2	49	95
Dec 2024	27	9	54	90
Jan 2025	48	4	40	92
Feb 2025	42	8	3	53
Mar 2025	12	4	0	16
Apr 2025	9	6	1	16
May 2025	22	7	1	30
Jun 2025	23	8	7	38
Jul 2025	27	2	1	30
Aug 2025	91	16	1	108
Sep 2025	109	1	0	110

Cumulative views and downloads (calculated since 20 Sep 2024)

Month	HTML	PDF	XML	Total
Sep 2024	96	7	25	128
Oct 2024	45	6	71	122
Nov 2024	44	2	49	95
Dec 2024	27	9	54	90
Jan 2025	48	4	40	92
Feb 2025	42	8	3	53
Mar 2025	12	4	0	16
Apr 2025	9	6	1	16
May 2025	22	7	1	30
Jun 2025	23	8	7	38
Jul 2025	27	2	1	30
Aug 2025	91	16	1	108
Sep 2025	109	1	0	110

Viewed (geographical distribution)

Total article views: 875 (including HTML, PDF, and XML) Thereof 875 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 12 Sep 2025

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (3154 KB)
Metadata XML

Short summary

This study examines how hydraulic head levels in canal dikes respond to heavy rainfall, potentially causing instabilities and flooding. Using time series models and simulating long-term head levels, we identified clusters of dikes where head peaks are driven by similar rainfall events. Statistical analyses show that extreme and yearly conditions are close. However, extreme conditions are expected to become more frequent due to climate change, though some dikes will be less affected than others.


Total:	0
HTML:	0
PDF:	0
XML:	0