pith. sign in

arxiv: 2604.22809 · v1 · submitted 2026-04-14 · 💻 cs.CE · cs.SI· cs.SY· eess.SY

The Elusive Nature of Roughness: Linking Hydraulics and Graph Theory for Water Distribution Networks Model Calibration

Karol Dykiert , Mateusz Stolarski , Micha{\l} Czuba , Wojciech Cie\.zak , Piotr Br\'odka This is my paper

Pith reviewed 2026-05-10 14:38 UTC · model grok-4.3

classification 💻 cs.CE cs.SIcs.SYeess.SY
keywords water distribution networkspipe roughness calibrationnetwork partitioninggraph theoryclusteringhydraulic modelingoptimizationtopology
0
0 comments X

The pith

Grouping pipes using hydraulic and graph attributes produces stable roughness calibration results comparable to manual methods.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tests whether partitioning water distribution networks into groups based on hydraulic properties and graph theory metrics can improve the calibration of pipe roughness coefficients. Traditional calibration relies on costly field work or manual heuristics that may not be repeatable. By using a high-fidelity model as benchmark, the study compares clustering approaches and finds that attribute-based groups lead to optimization results that are stable and match manual calibration quality for important pipes. Including graph data helps stabilize the process while hydraulic attributes define clearer clusters.

Core claim

Attribute-based grouping of pipes, leveraging both hydraulic and graph-derived attributes, yields stable and repeatable roughness estimates through optimization that are comparable to manual calibration for hydraulically significant pipes. Hydraulic attributes produce more distinct clusters, graph information improves robustness, and density-based clustering achieves similar accuracy to k-means with lower computational effort in certain setups.

What carries the argument

Attribute-based grouping via density-based clustering and topology-driven strategies that combine hydraulic parameters with graph metrics to partition the network for efficient roughness calibration.

If this is right

  • Calibration becomes more repeatable and less dependent on individual expert choices.
  • Graph-based attributes can be added to hydraulic data to enhance optimization stability.
  • Density-based clustering offers a way to maintain accuracy with reduced computation compared to k-means.
  • The method provides a systematic alternative to manual heuristics for large networks.
  • Network topology is shown to be important for reliable parameter estimation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar grouping strategies could extend to calibrating other parameters like demand or valve settings in the same networks.
  • Applying this to networks without a high-fidelity model would require validation against limited field data.
  • Further research might explore how different graph metrics affect cluster quality and calibration outcomes.
  • The approach may generalize to other infrastructure networks modeled as graphs, such as power grids.

Load-bearing premise

The selected hydraulic and graph attributes capture the main factors that cause errors in roughness estimation, and the high-fidelity model is an accurate proxy for real network behavior.

What would settle it

Running the calibration on the same network but with actual field pressure and flow measurements instead of the high-fidelity model outputs would show if the grouped results deviate substantially from manual calibration accuracy.

read the original abstract

Accurate pipe roughness estimation in large-scale water distribution networks is often hindered by the high cost of traditional field methods. This study investigates whether network partitioning, by utilizing hydraulic and graph-derived attributes, can enhance the calibration of these parameters. Using a high-fidelity model of a real network as a benchmark, we evaluate density-based clustering, and topology-driven grouping strategies. Optimization experiments demonstrate that attribute-based grouping yields stable, repeatable results comparable to manual calibration for hydraulically significant pipes. While hydraulic attributes generate more distinct cluster structures, the inclusion of graph-based data improves calibration robustness by stabilizing the optimization process. Notably, density-based clustering achieves similar accuracy to k-means while reducing computational effort in specific configurations. Although the method does not eliminate all sources of uncertainty, results suggest that topology-informed grouping provides a systematic, reproducible, and computationally efficient alternative to manual heuristics, highlighting the critical role of network structure in reliable parameter estimation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that partitioning water distribution networks using a combination of hydraulic attributes and graph-derived topology features enables more stable and reproducible roughness coefficient calibration than manual heuristics. Optimization experiments on a high-fidelity model of a real network are reported to show that attribute-based grouping (including density-based clustering) produces results comparable to manual calibration for hydraulically significant pipes, with graph attributes improving robustness and density-based methods sometimes matching k-means accuracy at lower computational cost.

Significance. If the benchmark is independently validated and the quantitative results are fully reported, the work would offer a systematic, topology-informed alternative to ad-hoc roughness calibration that could reduce subjectivity and field costs in large-scale WDN modeling. The explicit linkage of graph-theoretic attributes to hydraulic parameter estimation is a potentially useful contribution, though its practical impact depends on demonstrating improvement over existing automated calibration techniques.

major comments (2)
  1. [§4] §4 (Case study / benchmark description): The high-fidelity model is used as ground-truth benchmark for all optimization experiments, yet the manuscript provides no description of its construction, parameter sources, boundary conditions, or quantitative agreement with independent field pressure and flow measurements. This is load-bearing for the central claim because the reported stability and comparability to manual calibration are only meaningful if the benchmark accurately reproduces real-network hydraulics rather than sharing structural assumptions with the tested methods.
  2. [§5] §5 (Optimization experiments / Results): The abstract and results claim that attribute-based grouping yields 'stable, repeatable results comparable to manual calibration,' but no specific error metrics (e.g., RMSE or MAE on roughness values or simulated heads/flows), number of optimization runs, statistical significance tests, or tabulated comparisons with manual calibration are presented. Without these, the strength of the repeatability and comparability assertions cannot be evaluated.
minor comments (2)
  1. [Abstract] Abstract: The statement that 'density-based clustering achieves similar accuracy to k-means while reducing computational effort in specific configurations' is not accompanied by the accuracy measure used or the exact configurations, making the claim difficult to interpret.
  2. [§3] §3 (Methodology): The definitions and normalization of the hydraulic plus graph attributes used for clustering are not fully specified, nor is the rationale for the chosen distance metric or clustering hyperparameters.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and will revise the manuscript to incorporate the requested information and quantitative details.

read point-by-point responses

  1. Referee: [§4] §4 (Case study / benchmark description): The high-fidelity model is used as ground-truth benchmark for all optimization experiments, yet the manuscript provides no description of its construction, parameter sources, boundary conditions, or quantitative agreement with independent field pressure and flow measurements. This is load-bearing for the central claim because the reported stability and comparability to manual calibration are only meaningful if the benchmark accurately reproduces real-network hydraulics rather than sharing structural assumptions with the tested methods.

    Authors: We agree that the current manuscript lacks sufficient detail on the high-fidelity model. In the revised version we will expand §4 with a dedicated subsection that describes: (i) the model's construction from GIS, as-built drawings and SCADA data; (ii) sources of all pipe and node parameters; (iii) boundary conditions (reservoir heads, demand patterns, control settings); and (iv) quantitative validation against independent field pressure and flow measurements, including RMSE, MAE and correlation coefficients. These additions will demonstrate that the benchmark reproduces observed hydraulics independently of the partitioning methods under test. revision: yes

  2. Referee: [§5] §5 (Optimization experiments / Results): The abstract and results claim that attribute-based grouping yields 'stable, repeatable results comparable to manual calibration,' but no specific error metrics (e.g., RMSE or MAE on roughness values or simulated heads/flows), number of optimization runs, statistical significance tests, or tabulated comparisons with manual calibration are presented. Without these, the strength of the repeatability and comparability assertions cannot be evaluated.

    Authors: We acknowledge that the results section currently presents only qualitative statements. We will revise §5 to include: (i) tabulated RMSE and MAE values for both roughness coefficients and simulated heads/flows across all methods; (ii) the exact number of independent optimization runs performed for each configuration; (iii) results of statistical significance tests (e.g., paired t-tests or ANOVA) comparing attribute-based, manual and baseline approaches; and (iv) direct side-by-side tables contrasting the new methods with the manual calibration outcomes. These quantitative elements will allow readers to evaluate the claimed stability and comparability. revision: yes

Circularity Check

0 steps flagged

No circularity: results from empirical optimization on external benchmark

full rationale

The paper's claimed results arise from optimization experiments that compare attribute-based grouping strategies against manual calibration, using a high-fidelity model of a real network as an external benchmark. No equations, parameters, or predictions are shown to reduce by construction to their own inputs, fitted values, or self-citations; the evaluation relies on independent simulation runs and clustering performance metrics rather than self-definitional loops or renamed known results. The derivation chain remains self-contained through direct empirical testing against the benchmark.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No explicit free parameters, axioms, or invented entities are detailed in the abstract; the approach assumes standard clustering and optimization techniques apply directly to the roughness problem.

pith-pipeline@v0.9.0 · 5487 in / 969 out tokens · 23307 ms · 2026-05-10T14:38:08.986698+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages

  1. [1]

    overfitting

    Introduction 1.1. The problem of roughness formulation Water resources have become increasingly critical under the pressures of climate change, which exacerbates hydrological extremes such as prolonged droughts, increased evaporation, and altered precipitation patterns [1]. These phenomena place additional stress on drinking water supply systems, intensif...

    work page

  2. [2]

    Materials and Methods The analysed system is a high-pressure zone (HPZ) of the WDN (Figure 2), consisting of six District Meter Areas (DMAs). The dataset, associated preprocessing procedures and the initial model calibration were described in detail in our previous publication [19]; concise summaries are provided in Appendices A and B for reference. Figur...

    work page

  3. [3]

    The results are reported in terms of clustering quality, optimisation performance, and the consistency of calibrated roughness values

    Results This section presents the outcomes of the clustering and calibration experiments. The results are reported in terms of clustering quality, optimisation performance, and the consistency of calibrated roughness values. 3.1. K-means and the elbow method To determine the optimal number of clusters for the k-means algorithm, we employed the elbow metho...

    work page

  4. [4]

    The results are examined in the context of network grouping strategies and their implications for roughness estimation in water distribution networks

    Discussion This section interprets the observed patterns in clustering structure and calibration performance. The results are examined in the context of network grouping strategies and their implications for roughness estimation in water distribution networks. 4.1. Determining the number of clusters in k-means The elbow method behaved largely as expected;...

    work page

  5. [5]

    Increasing the number of clusters generally improved repeatability and, consequently, solution reliability, although the smallest tested k achieved comparable performance

    Conclusions The k-means algorithm proved to be the most suitable method for clustering the WDN in the context of roughness calibration. Increasing the number of clusters generally improved repeatability and, consequently, solution reliability, although the smallest tested k achieved comparable performance. HDBSCAN demonstrated effectiveness similar to k -...

    work page

  6. [6]

    Acknowledgment This research was partially supported by the Ministry of Science and Higher Education, industrial PhD programme, DWD/6/0543/2022 and the Academia Profesorum Iuniorum, funded by the Wrocław University of Science and Technology

    work page 2022

  7. [7]

    The data can only be accessed by requesting the data owner

    Data availability The research was carried out using data collected by the Municipal Water and Sewerage Company in Wrocław and was obtained under an agreement on providing data for scientific purposes. The data can only be accessed by requesting the data owner

    work page

  8. [8]

    CRediT • Conceptualization: Karol Dykiert (KD), Mateusz Stolarski (MS), Michał Czuba (MC), Wojciech Cieżak (WC), Piotr Bródka (PB), • Data curation: KD, • Formal analysis: KD, • Funding acquisition: PB, • Investigation: KD, MS, MC, • Methodology: KD, MS, MC, WC, PB, • Project administration: KD, • Software: KD, MS, MC, • Supervision: PB, WC, • Validation:...

    work page

  9. [9]

    Climate Change and Water Resources

    Frederick KD, Major DC. Climate Change and Water Resources. Clim Change. 1997;37(1):7-23. doi:10.1023/A:1005336924908

    work page doi:10.1023/a:1005336924908 1997

  10. [10]

    Preliminary analysis of the preparation of Polish water utilities to implement mandatory risk management in accordance with the Drinking Water Directive 2020/2184

    Ramm K. Preliminary analysis of the preparation of Polish water utilities to implement mandatory risk management in accordance with the Drinking Water Directive 2020/2184. Appl Water Sci. 2022;12(8):186. doi:10.1007/s13201-022-01710-7

    work page doi:10.1007/s13201-022-01710-7 2020

  11. [11]

    Water quality for citizen confidence: The implementation process of 2020 EU Drinking Water Directive in Nordic countries

    Bayona-Valderrama Á, Gunnarsdóttir MJ, Rossi PM, et al. Water quality for citizen confidence: The implementation process of 2020 EU Drinking Water Directive in Nordic countries. Water Policy. 2024;26(8):793-816. doi:10.2166/wp.2024.013

    work page doi:10.2166/wp.2024.013 2020

  12. [12]

    Making waves: Creating water sensitive cities in Australia

    Fogarty J, van Bueren M, Iftekhar MS. Making waves: Creating water sensitive cities in Australia. Water Res. 2021;202:117456. doi:https://doi.org/10.1016/j.watres.2021.117456

    work page doi:10.1016/j.watres.2021.117456 2021

  13. [13]

    Walski M. T, V. Chase D, A. Savic D, Grayman Walter, Beckwith Stephen, Koelle E. Advanced Water Distribution Modeling and Management. 2003

    work page 2003

  14. [14]

    Water Supply and Distribution Systems (2nd Edition)

    Savic DA, Banyard JK. Water Supply and Distribution Systems (2nd Edition). ICE Publishing. https://app.knovel.com/hotlink/toc/id:kpWSDS000B/water-supply- distribution/water-supply-distribution

    work page

  15. [15]

    Decoupling elevation errors from pipe roughness calibration in hydraulic network models

    Du K, Yu J, Zheng F, Kapelan Z, Savic D. Decoupling elevation errors from pipe roughness calibration in hydraulic network models. Water Res. 2026;290. doi:10.1016/j.watres.2025.125058

    work page doi:10.1016/j.watres.2025.125058 2026

  16. [16]

    An all-purpose method for optimal pressure sensor placement in water distribution networks based on graph signal analysis

    Zhou X, Wan X, Liu S, Su K, Wang W , Farmani R. An all-purpose method for optimal pressure sensor placement in water distribution networks based on graph signal analysis. Water Res. 2024;266:122354. doi:https://doi.org/10.1016/j.watres.2024.122354

    work page doi:10.1016/j.watres.2024.122354 2024

  17. [17]

    Parameter estimation in water distribution networks

    Kumar SM, Narasimhan S, Bhallamudi SM. Parameter estimation in water distribution networks. Water Resources Management. 2010;24(6):1251-1272. doi:10.1007/s11269-009-9495-1

    work page doi:10.1007/s11269-009-9495-1 2010

  18. [18]

    Pipe roughness calibration approach for water distribution network models using a nonlinear state observer

    Torres L, Jiménez-Cabas J, Ponsart JC, Theilliol D, Jiménez-Magaña MR, Guzmán JE V. Pipe roughness calibration approach for water distribution network models using a nonlinear state observer. Results in Engineering. 2024;23:102713. doi:https://doi.org/10.1016/j.rineng.2024.102713

    work page doi:10.1016/j.rineng.2024.102713 2024

  19. [19]

    Simpler Is Better–Calibration of Pipe Roughness in Water Distribution Systems

    Zhao Q, Wu W , Simpson AR, Willis A. Simpler Is Better–Calibration of Pipe Roughness in Water Distribution Systems. Water (Switzerland). 2022;14(20). doi:10.3390/w14203276

    work page doi:10.3390/w14203276 2022

  20. [20]

    Network Science

    Barabási AL, Pósfai M. Network Science. Cambridge University Press; 2017

    work page 2017

  21. [21]

    Introduction to Graph Theory

    Wilson RJ. Introduction to Graph Theory. Longman; 1996

    work page 1996

  22. [22]

    GC4NC: A Benchmark Framework for Graph Condensation on Node Classification with New Insights

    Gong S, Ni J, Sachdeva N, Yang C, Jin W. GC4NC: A Benchmark Framework for Graph Condensation on Node Classification with New Insights. Published online November 9, 2025

    work page 2025

  23. [23]

    Fortunato, D

    Fortunato S, Hric D. Community detection in networks: A user guide. Phys Rep. 2016;659:1-44. doi:10.1016/j.physrep.2016.09.002

    work page doi:10.1016/j.physrep.2016.09.002 2016

  24. [24]

    A faster algorithm for betweenness centrality*

    Brandes U. A faster algorithm for betweenness centrality*. J Math Sociol. 2001;25(2):163-177. doi:10.1080/0022250X.2001.9990249

    work page doi:10.1080/0022250x.2001.9990249 2001

  25. [25]

    Implementing Network Science To Enhance Water Distribution Network Pipe Roughness Calibration

    Dykiert K, Stolarski M, Czuba M, Cieżak W , Bródka P . Implementing Network Science To Enhance Water Distribution Network Pipe Roughness Calibration. In: The University of Sheffield; 2025. doi:10.15131/shef.data.29920964.v1

    work page doi:10.15131/shef.data.29920964.v1 2025

  26. [26]

    Shuffled complex evolution approach for effective and efficient global minimization

    Duan QY , Gupta VK, Sorooshian S. Shuffled complex evolution approach for effective and efficient global minimization. J Optim Theory Appl. 1993;76(3):501-521. doi:10.1007/BF00939380

    work page doi:10.1007/bf00939380 1993

  27. [27]

    Implementing data science to enhance water distribution system modelling

    Dykiert K, Cieżak W , Bródka P . Implementing data science to enhance water distribution system modelling. Instal. Published online October 2025:40-45. doi:10.36119/15.2025.10.6

    work page doi:10.36119/15.2025.10.6 2025

  28. [28]

    Hagberg, Daniel A

    Hagberg AA, Schult DA, Swart PJ. Exploring Network Structure, Dynamics, and Function using NetworkX. In: 2008:11-15. doi:10.25080/TCWV9851

    work page doi:10.25080/tcwv9851 2008

  29. [29]

    k-means++: the advantages of careful seeding

    Arthur D, Vassilvitskii S. k-means++: the advantages of careful seeding. In: Proceedings of the Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms. SODA ’07. Society for Industrial and Applied Mathematics; 2007:1027-1035

    work page 2007

  30. [30]

    Finding Groups in Data: An Introduction to Cluster Analysis

    Gentle JE, Kaufman L, Rousseuw PJ. Finding Groups in Data: An Introduction to Cluster Analysis. Biometrics. 1991;47(2):788. doi:10.2307/2532178

    work page doi:10.2307/2532178 1991

  31. [31]

    Accelerated hierarchical density clustering, in: IEEE International Conference on Data Mining Workshops (ICDMW), pp

    McInnes L, Healy J. Accelerated Hierarchical Density Based Clustering. In: 2017 IEEE International Conference on Data Mining Workshops (ICDMW). 2017:33-42. doi:10.1109/ICDMW.2017.12

    work page doi:10.1109/icdmw.2017.12 2017

  32. [32]

    Scikit-learn: Machine Learning in Python

    Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research. 2011;12:2825-2830

    work page 2011

  33. [33]

    Scaling Up Graph Neural Networks Via Graph Coarsening

    Zengfeng H, Zhang S, Xi C, Liu T, Zhou M. Scaling Up Graph Neural Networks Via Graph Coarsening. 2021. doi:10.48550/arXiv.2106.05150

    work page doi:10.48550/arxiv.2106.05150 2021

  34. [34]

    Aditya and Jin, Wei , title =

    Hashemi M, Gong S, Ni J, Fan W , Prakash B, Jin W. A Comprehensive Survey on Graph Reduction: Sparsification, Coarsening, and Condensation. 2024. doi:10.24963/ijcai.2024/891

    work page doi:10.24963/ijcai.2024/891 2024

  35. [35]

    Physics Reports 486, 75–174

    Fortunato S. Community Detection in Graphs. Phys Rep. 2009;486. doi:10.1016/j.physrep.2009.11.002

    work page doi:10.1016/j.physrep.2009.11.002 2009

  36. [36]

    Blondel, Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefebvre

    Blondel V, Guillaume JL, Lambiotte R, Lefebvre E. Fast Unfolding of Communities in Large Networks. Journal of Statistical Mechanics Theory and Experiment. 2008;2008. doi:10.1088/1742-5468/2008/10/P10008

    work page doi:10.1088/1742-5468/2008/10/p10008 2008

  37. [37]

    Directed Louvain : Maximizing Modularity in Directed Networks

    Dugué N, Perez A. Directed Louvain : Maximizing Modularity in Directed Networks

    work page

  38. [38]

    doi:10.13140/RG.2.1.4497.0328 Appendix A. Dataset description The analysed system is a high -pressure zone (HPZ) of the Wrocław water distribution network, consisting of five District Meter Areas (DMAs) supplied by a pumping station that regulates pressure. While the zone includes several physical interconnections with neighbouring DMAs, these links remai...

    work page doi:10.13140/rg.2.1.4497.0328 2023

  39. [39]

    topology correction – assigning elevations to nodes based on a Digital Elevation Model and merging similar pipes

    work page

  40. [40]

    water demand processing – creating water demand patterns and assigning them to nodes

    work page

  41. [41]

    pressure data smoothing – filtering the reference pressure measurements

    work page

  42. [42]

    In summary, the dataset represents a simplified (skeletonised) HPZ model that preserves the original network structure, according to MPWiK requirements

    data screening – removing nighttime measurements from all time -series data. In summary, the dataset represents a simplified (skeletonised) HPZ model that preserves the original network structure, according to MPWiK requirements. Nodal demands are described in high detail due to the large number of individual water demand patterns. The model supports exte...

    work page 2025