Incorporating Pipe Age and Sizes into Pipe Roughness Coefficient Estimation for Urban Flood Modeling: A Scenario-Based Roughness Approach

Abstract

With climate change, the frequency and severity of localized heavy rainfalls are increasing. Thus, for urban drainage networks (UDNs), particularly those in aging cities such as Seoul, Republic of Korea, flood risk management challenges are mounting. Conventional design standards typically apply uniform roughness coefficients based on new pipe conditions, neglecting the ongoing performance degradation from physical influences. This study introduces a methodology that systematically incorporates pipe age and size into roughness coefficient scenarios for higher-accuracy 1D–2D rainfall–runoff hydrologic–hydraulic simulations. Eleven roughness scenarios (a baseline and ten aging-based scenarios) are applied across seven UDNs using historical rainfall data. The most representative scenario (S3) is identified using a Euclidean distance metric combining the peak water-level error and root mean square error. For two rainfall events, S3 yields substantial increases in the simulated mean flood volumes (75.02% and 76.45%) compared with the baseline, while spatial analysis reveals significantly expanded inundation areas and increased flood depths. These findings underscore the critical impact of pipe deterioration on hydraulic capacity and demonstrate the importance of incorporating aging infrastructure into flood modeling and UDN design. This approach offers empirical support for updating UDN design standards for more resilient flood management.

1. Introduction

Owing to climate change, the frequency of localized heavy rainfall has increased, thereby exposing the significant performance limitations of existing urban drainage networks (UDNs). In Seoul, where much of the land surface is impervious, stormwater runoff rapidly escalates during rainfall events, frequently overwhelming the hydraulic capacity of UDNs. A notable event occurred in September 2011, when 381.5 mm of cumulative rainfall was recorded within 3 h, resulting in eight fatalities and more than KRW 130 billion in property damage [1]. The damage was primarily attributed to the reduced hydraulic capacity of aging pipes and insufficient conveyance, with comprehensive replacement hindered by limited funding and spatial constraints [2]. These challenges highlight the necessity of quantitatively evaluating existing pipe performance and improving hydraulic analysis and design standards that explicitly consider pipe aging and its associated roughness changes.

However, detailed pipe-level data—such as installation year, size, and network topology—are often incomplete, outdated, or unavailable in practice. This limitation is well documented in urban drainage system studies, where data scarcity is reported to constrain robust hydraulic modeling and infrastructure planning [3,4]. To address this, recent studies have introduced surrogate modeling and data-driven approaches that compensate for uncertain or missing pipe attributes [5,6]. Despite these advancements, the Republic of Korea’s “Sewer Design Standards” still recommend roughness coefficients of 0.013 for circular pipes and 0.015 for rectangular pipes [7], values that correspond to newly installed pipes. In reality, roughness increases over time due to pipe age, sedimentation, biofilm growth, and structural deterioration [8,9,10]. In Seoul, approximately 59% of the drainage pipes were installed more than 30 years ago, and many are part of combined networks. Persistent maintenance challenges and sediment infiltration can elevate roughness coefficients above 0.020, markedly diminishing hydraulic performance [11,12,13].

Numerous studies have been performed in the urban drainage system domain [3,4,5,6,14,15,16]. Among these, it was demonstrated that the implications of such roughness increases are substantial. Elevated roughness values can distort rainfall–runoff modeling, leading to underestimation of overflow volumes and flood risk under standard design rainfall conditions [17,18,19]. Previous studies have addressed these issues in three primary ways: (1) assessing drainage pipe deterioration, (2) quantifying runoff changes due to increased roughness, and (3) defining upper bounds for roughness coefficients. For example, Tran [20] applied a Markov chain model to predict deterioration of sewer pipes, while Song et al. [21] simulated sediment accumulation and structural degradation, confirming that such processes directly elevate roughness values. Other studies have demonstrated that the effects of roughness increase on flood inundation. Kim et al. [22] showed, using EPA-SWMM, that raising n from 0.011 (new pipes) to 0.017 (deteriorated pipes) significantly affected runoff volumes and inundation risks. Won et al. [23] found that incremental increases amplified downstream peak flows, while Park et al. [24] reported that roughness changes from 0.015 to 0.120 could double peak water levels depending on flow conditions. Empirical studies [25,26,27] further showed that actual roughness values can range widely (0.042–0.220), proposing practical benchmarks such as 0.013 for new pipes, 0.015 for long-term operation, and 0.032 when sediment or obstructions are present [27].

Although these contributions are valuable for UDN maintenance and post-construction evaluation, they share a critical limitation: they do not provide systematic guidelines for incorporating pipe age and physical attributes into roughness coefficient determination during the design stage. Similarly, urban drainage calibration studies generally apply uniform roughness across the entire network or sample values randomly within a specified range, thereby neglecting the variability associated with pipe age or sizes [28,29,30,31,32]. Yet, among the parameters influencing hydraulic performance (e.g., size, slope, and alignment), roughness has been shown to exert the most significant effect on conveyance capacity [22,25]. If roughness values for aged pipes are not systematically adjusted, models underestimate flood volumes by ignoring increased hydraulic resistance. Conversely, overly conservative values obscure spatial variability and overstate system deterioration. Both outcomes risk flawed flood risk assessments and unreliable design.

Accordingly, precise identification of roughness coefficients is essential for UDN design and flood risk management. To address these limitations, the present study proposes a practical framework that assigns roughness coefficients based on pipe age, size, and other physical attributes. Specifically, 11 candidate scenarios are defined, each representing different combinations of pipe attributes. Seven UDNs in Seoul are analyzed under 37 rainfall events using a coupled 1D–2D rainfall–runoff hydrologic–hydraulic simulation. The observed water levels at multiple monitoring points are systematically compared with simulated values under each scenario. To robustly evaluate performance, both the peak water-level error (PWLE) and root mean square error (RMSE) are synthesized into a Euclidean distance metric, enabling integrated assessment. The scenario most frequently yielding the lowest errors across monitoring locations is selected as the representative guideline.

Finally, by contrasting the highest-ranked scenario with a baseline assuming new-pipe roughness throughout the network, this study demonstrates the improved accuracy and applicability of the proposed approach. Our study contributes by quantifying the practical consequences of neglecting pipe age and size in flood simulations and systematically demonstrating how roughness variations can lead to under- or overestimation of inundation risks. By integrating service life and physical attributes into UDN modeling, we highlight the critical need for refined design guidelines and provide the results directly relevant to engineering practice.

2. Study Materials

2.1. Study Areas

Seoul experienced severe urban flooding during a catastrophic rainfall event on 8 August 2022, which caused significant harm to humans and property damage across multiple districts, including Dongjak-gu, Gangnam-gu, and Guro-gu [1]. Low public awareness of sewer infrastructure, inadequate maintenance, and reduced drainage capacity due to sediment and waste accumulation within pipes were identified as the primary causes of this damage [33]. With the expansion of urban development in Seoul into upstream areas, the runoff volume in urban areas has increased. However, downstream pipes remain connected to the system, which exacerbates capacity limitations. In addition, in low-lying zones with gentle slopes, inefficient stormwater discharge contributes to recurrent flood events. Under these conditions, even small increases in pipe roughness owing to aging, sediment accumulation, or structural deterioration can dramatically decrease the flow velocity and drainage capacity as the gravitational driving forces are already weak. Therefore, accurate estimation of the roughness coefficient is critical for these regions. Inappropriate or underestimated values can cause severe underestimation of the flood risk and hinder the design of effective flood mitigation measures. Therefore, precise roughness coefficient determination is essential for diagnosis and management of flood vulnerability in low-slope, low-lying urban catchments.

The climate of Seoul is characterized by distinct seasonal variations. The city receives average annual precipitation of 1417.9 mm, with the lowest monthly average occurring in January (16.8 mm) and the highest in July (414.4 mm). The summer months (June, July, and August) contribute significantly to the total, with 892.1 mm or approximately 63% of the annual rainfall accumulating during this period; thus, a strong seasonal pattern is apparent. The average annual relative humidity is 61.8%, with the lowest levels (54.6%) being observed in February and March and the highest (76.2%) during summer. Geologically, the area is predominantly underlain by granitic rocks, and its topography features steeply sloped ridge lines.

In this study, seven UDNs in Seoul for which observed water-level data were available were selected, and a 1D–2D rainfall–runoff hydrologic–hydraulic simulation model was used to investigate the effects of pipe aging and sizes on roughness coefficient variation. The UDNs were labeled AB, AH, BA, BF, DC, FD, and FG, as shown in Figure 1, and were selected based on comprehensive considerations (i.e., their total areas, total lengths, and water level locations). All the selected UDNs had a history of flooding or were located in areas where the stormwater-pipe conditions critically affect urban drainage safety.

Figure 1. Configuration for seven UDNs used in this study. The black lines represent the pipe layouts, while the red dashed lines indicate the boundaries of the drainage catchments.

The seven UDNs considered in this study exhibited varying topological complexities. BF had the largest network, comprising 2064 nodes and 2291 links, followed by AB, with 1811 nodes and 1936 links, and DC, which contained 1836 nodes and 1982 links. AH and BA were moderate-scale networks with 1428 nodes and 1488 links and 1222 nodes and 1349 links, respectively. Finally, FG comprised 870 nodes and 913 links, whereas FD had the smallest network configuration, with 214 nodes and 219 links. Detailed information on the physical characteristics of the selected UDNs is presented in Table 1. In addition, the statistical distribution of pipe ages across all the UDNs is presented in Table S1 of the Supplementary Materials (Section S3).

Table 1. Detailed physical characteristics of selected UDNs.

	AB	AH	BA	BF	DC	FG	FD
Maximum pipe size (m)	9.00	3.10	2.50	4.00	3.10	3.50	3.40
Minimum pipe size (m)	0.15	0.30	0.15	0.25	0.30	0.30	0.45
Total pipe length (km)	100.55	79.47	68.30	132.14	99.35	45.36	10.23
Total drainage area (km2)	10.71	7.25	5.26	7.39	8.58	2.98	2.92

2.2. Data

Two measurement datasets obtained from rainfall and water-level gauge stations in Seoul, Republic of Korea, and managed by the Korea Meteorological Administration were employed in this study. These datasets were employed for the 1D–2D rainfall–runoff hydrologic–hydraulic simulation model. A total of 39 data points (e.g., rainfall and water levels) were collected, which were derived from a historical urban flooding event in the Republic of Korea. This significant flood provided essential observational data (i.e., rainfall and water levels) for urban flood risk computation [see Supplementary Figures S1 (rainfall) and S2 (water level)].

In addition, Figure 2 illustrates the spatial measurements necessary to construct the 1D–2D simulation model (XPSWMM 2018.2.1 used in this study). Therefore, the most recent and internally consistent UDN dataset available was employed in this study; it was assumed that this field-verified dataset was both complete and accurate for computational processes. Additionally, the detailed rainfall–runoff model parameters for all the UDNs are presented in Table S2 of the Supplementary Materials (Section S4).

Figure 2. Illustrations of spatial measurement data required for 1D–2D rainfall–runoff hydrologic–hydraulic simulations: (a) building footprints used to represent impervious surfaces and (b) digital elevation model (DEM) applied for terrain representation.

3. Methodology

3.1. Overview

This study proposes a novel four-step methodology devised to identify the best pipe roughness scenario among all the considered scenarios and to investigate system performance with respect to flood risk reduction. The workflow is shown in Figure 3. First, multiple pipe roughness coefficient scenarios (i.e., Scenarios 0–10 or S0–10) were defined based on the pipe ages and sizes. Second, rainfall–runoff hydrological–hydraulic simulations were individually conducted for each scenario under predefined rainfall events in the selected UDNs. In the third step, the best pipe roughness scenario was determined based on the Euclidean distance approach for two different error metrics (PWLE and RMSE), reflecting the difference between the simulated and observed water levels. Finally, the effects of flood risk reduction were investigated using the proposed model, and its system performance was compared with that of the baseline scenario (i.e., uniform roughness values of 0.013–0.015 based on new pipes). This section describes the proposed methodology in detail.

Figure 3. Overall workflow of the proposed methodology, showing the sequential steps from scenario definition to flood risk quantification.

3.2. Definition of Pipe Roughness Scenarios (Step 1)

Over time, various types of debris, such as leaves, sediment, and waste, gradually accumulate inside urban drainage pipes. This build-up increases the pipe roughness coefficient, which is a major factor in flow capacity reduction [33]. However, the current drainage pipe design standard of the Republic of Korea neglects pipe ages or physical characteristics. Instead, a uniform roughness value based on new pipes is applied, which fails to reflect the hydraulic resistance occurring under real operating conditions. This study aims to improve the accuracy of runoff modeling for aged pipes by quantifying changes in roughness according to pipe age and size and incorporating these adjusted values into scenario-based simulations [34].

The rationale for defining roughness scenarios based on pipe service life is well established, as numerous field surveys and CCTV inspections have confirmed that Manning’s n increases with pipe aging [35,36]. Furthermore, several studies have demonstrated that sediment accumulation, biofilm growth, and structural deterioration can cause Manning’s n to rise to as high as 0.032 as pipes deteriorate [37]. Therefore, the roughness coefficient scenarios defined in this study—namely, the assumption of a gradual increase in roughness with service life—are grounded in empirical evidence, thereby reinforcing their validity.

In the proposed approach, the roughness coefficient scenarios are categorized into six groups based on the year of installation of the pipe (new: <3 years, then in 10-year increments up to >40 years) and four groups based on the pipe sizes and reflecting differences in hydraulic sensitivity: ≤400 mm, 500–700 mm, 800–1000 mm, and >1000 mm [38]. Based on these classifications, ten scenarios (S1–10) were developed in this study, each with incrementally increasing roughness values [39]. S1 reflected commonly assumed conditions with roughness values of 0.013–0.020, whereas S10 corresponded to severely aged conditions, with higher roughness values of 0.026–0.030 [40]. For all the scenarios, new pipes (less than 3 years old) were assigned fixed roughness values: 0.013 and 0.015 for circular and rectangular pipes, respectively. The values increased in stages according to the pipe age [7,41]. Additionally, a baseline scenario (S0) was defined, in which new pipe roughness values (circular: 0.013; rectangular: 0.015) were uniformly applied to all pipes, regardless of their age or size.

These ten scenarios were applied to 10,178 stormwater pipes across the seven selected drainage networks in Seoul. Thus, each pipe was automatically assigned a roughness value based on its year of installation and size. In cases where installation data were missing, interpolation from neighboring pipes was used. Notably, this scenario-based approach has strong practical applicability and enables hydraulic performance assessments that consider various aging conditions even in unmonitored basins [18]. By comparing the simulated and observed water levels for each scenario, the most appropriate roughness patterns for the target areas were identified. Through this structured scenario design, the influence of roughness assumptions on hydrologic and hydraulic simulation outcomes was quantitatively assessed; the findings contributed to the development of a practical guideline for avoiding under- or overestimation of flow capacity. The roughness values for each scenario are provided in Table S3 of the Supplementary Materials (Section S5).

Notably, the selected UDNs included pipes installed between 1946 and 2017, allowing pipe age classification into intervals for periods ranging from 3 to more than 40 years. Based on these criteria, roughness coefficient scenarios were developed (Section 3.4). The observed water-level data were organized and refined by pipe age, enabling a comparative analysis with simulated rainfall–runoff conditions. This preprocessing enabled a quantitative evaluation of the runoff response sensitivity and applicability of the roughness coefficients under varying pipe age conditions.

3.3. Rainfall–Runoff Hydrologic–Hydraulic Simulation (Step 2)

Two modeling approaches were employed to simulate the hydrologic and hydraulic behaviors in urban areas: the 1D (SWMM) model developed by the U.S. EPA and the 1D–2D coupled model available through XPSWMM provided by Computational Hydraulics International. The XPSWMM is a quasi-distributed model capable of simulating runoff generation and hydrologic–hydraulic flow in urban areas under single-event or continuous rainfall conditions. This model simultaneously considers rainfall–runoff processes in subcatchments and routes through UDNs [42,43]. In this study, the 1D SWMM was first used to compare the errors between the observed and simulated water levels under the different pipe roughness scenarios. Then, the 1D–2D coupled model was used to analyze the flood behavior characteristics in relation to the hydraulic resolution and spatial data accuracy. Specifically, the 1D model was applied considering the computational efficiency and flexibility, whereas the coupled model was employed to enhance the spatial accuracy of the surface runoff and flood propagation results.

From the hydrologic perspective, the XPSWMM performed rainfall–runoff simulations for each subcatchment. That is, the catchment was discretized into subcatchments with an average area of 2428 m² using a 0.3 m resolution digital surface model. The rainfall was used as the primary hydrologic input; the surface runoff, infiltration, and evaporation losses were the outputs. The initial runoff was calculated based on the depression storage, which was set at 1.25 and 2.5 mm for impervious and pervious areas, respectively, following the recommendations of the American Society of Civil Engineers [44]. The infiltration loss was calculated using the Green–Ampt model implemented in SWMM, with parameters derived from soil types [45]. For the hydraulic component, dynamic wave routing was used to simulate unsteady flow, and the outputs included the water level, flow rate, and flood duration at each pipe and node [46]. The model parameters were extracted from high-resolution infrastructure data provided by the Seoul Metropolitan Government, along with a digital elevation model and soil maps. The time step was set to 1 s to ensure numerical stability.

The 1D–2D integrated modeling was performed in the XPSWMM environment. A grid-based domain comprising nonuniform square meshes was applied to simulate a 2D free-surface flow. Slope and roughness values were assigned to each 2D cell based on the ground elevation, and the inlet nodes were connected between the 1D and 2D domains through orifices. This allows water to spread into the 2D domain under flood conditions and re-enter UDNs under certain conditions [47,48]. Depending on the inclusion of inertial terms, either diffusive wave or fully dynamic wave simulations were possible, and the model structure followed the same hydrologic and hydraulic configurations as those defined in the 1D SWMM. Using this approach, the relationship between the data completeness, hydraulic resolution, and flood reproducibility was quantitatively analyzed, and a foundation for reducing spatial uncertainty in urban flood simulations was established.

3.4. Evaluation Metrics (Step 3)

To evaluate the appropriateness of a given roughness coefficient in terms of the pipe age and sizes, two different evaluation metrics were employed. First, the accuracy of the peak water levels obtained from the 1D SWMM simulation model was evaluated based on the PWLE through comparison with the observed peak levels. This comparison revealed whether the error rate remained within 5%. Second, the overall accuracy of the simulated time series was assessed using the RMSE, which measured the deviation between the simulated and observed water levels. Finally, to determine the best roughness scenario, a prioritization calculation based on the Euclidean distance method was conducted by integrating the PWLE and RMSE.

The PWLE assessed whether the simulated peak water level converged to within a 5% error of the observed peak level. This metric was particularly useful because the peak water level is directly associated with the urban flood risk and is, thus, a critical reference for assessing the appropriateness of the applied roughness coefficient. The expression is as follows:

$$PWLE={\displaystyle \frac{\left|{WL}_{obs,i}-{WL}_{sim,i}\right|}{{WL}_{obs,i}}}\times 100$$

(1)

where WL_obs,i denotes the water level observed at time step i and WL_sim,i represents the corresponding simulated water level at time step i obtained from the 1D SWMM.

The RMSE numerically quantified the similarity between the simulated and observed water-level time series. The RMSE was calculated as the square root of the average of the squared residuals between the observed and simulated water levels at each time step:

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{\left({WL}_{obs,i}-{WL}_{sim,i}\right)}^2}$$

(2)

where n denotes the number of data points in the time series. Because RMSE reflected the overall goodness-of-fit of the simulation, it provided a comprehensive measure of accuracy across all time steps.

However, relying on either PWLE or RMSE alone could introduce bias, as the former emphasizes peak conditions while the latter focuses on overall time-series fidelity. To reconcile these potentially conflicting tendencies, the Euclidean distance method was applied to integrate both metrics into a single composite index. In this approach, the rankings of each scenario according to PWLE and RMSE were treated as coordinates in a two-dimensional space, and the Euclidean distance from the origin (0,0) was calculated as follows:

$$d=\sqrt{{\left(p_1-q_1\right)}^2+{\left(p_2-q_2\right)}^2}$$

(3)

where p₁ and p₂, and q₁ and q₂ represent the values of each scenario under two different metrics (i.e., PWLE and RMSE). The scenario with the shortest Euclidean distance to the origin was considered the highest-ranked roughness coefficient scenario (Figure 4). This composite framework ensured a balanced and robust evaluation, avoiding the pitfalls of single-metric bias and enabling practically meaningful assessments of roughness scenarios across diverse rainfall conditions.

Figure 4. Conceptual diagram of the Euclidean distance method, illustrating how PWLE and RMSE are integrated into a single performance index.

3.5. Flood Risk Quantification Regarding Underestimation (Step 4)

As discussed above, application of a uniform roughness coefficient based on new pipes in baseline scenarios can cause underestimation of flood risks [34,49]. To evaluate this limitation quantitatively, the highest-ranked scenario, as identified through the proposed approach, was compared with the baseline scenario, in which all the pipes were assumed to have new pipe roughness values. The comparison was conducted using the 1D–2D SWMM simulation results, specifically focusing on the flood inundation characteristics. For each scenario, flood maps were generated and evaluated using three metrics: flood depth, volume, and extent. Note that the flood depth refers to the maximum water level observed in each 2D mesh cell during peak flooding, the flood volume denotes the cumulative amount of water transferred from the 1D drainage network to the 2D surface grid through orifices over the simulation period, and the flood extent represents the total area of the 2D mesh cells experiencing water depths greater than 0 mm at the peak moment of flooding.

4. Results

4.1. Prioritization Rank for Pipe Roughness Coefficient Scenarios

To investigate the accuracy of each scenario, heatmaps were prepared, as shown in Figure 5. The results for the overall prioritization ranking for each scenario (S0–S10), including the baseline and predefined scenarios, are shown in terms of (a) PWLE, (b) RMSE, and (c) Euclidean distance. Note that a deeper blue color indicates a lower error and better performance, whereas a deeper red color represents a higher error and poorer performance.

Figure 5. Performance results for all pipe roughness coefficient scenarios across rainfall events and all UDNs based on three evaluation metrics: (a) PWLE, (b) RMSE, and (c) Euclidean distance. Blue areas indicate lower errors (good performance), whereas red areas indicate higher errors (poor performance). Key scenario differences are highlighted across UDNs and rainfall events.

As shown in Figure 5a, predominantly red patterns were observed for the PWLE results of S0 (baseline), indicating poor performance across most rainfall events. In contrast, specific scenarios S3–S5 exhibited stronger blue patterns, reflecting superior accuracy in simulating peak water levels. Among these, S3 showed the most consistent distribution across all the UDNs, highlighting its stable performance. These results suggest that the uniform roughness values used in S0, based on new pipes, fail to capture the hydraulic effects of pipe aging and size. However, for RMSE (Figure 5b), the trend was reversed. S0 generally showed lower errors (blue), while S3–S5 displayed higher errors, particularly in smaller events. This discrepancy reflects the difference between PWLE, which emphasizes peak water levels, and RMSE, which evaluates the overall conformity of the hydrograph. Scenarios beyond S6 exhibited further degradation, as excessive roughness values reduced simulation fidelity.

To reconcile these contrasting tendencies, the Euclidean distance metric was applied (Figure 5c). This composite index confirmed that S3 consistently achieved the best overall performance across UDNs and rainfall events, balancing both peak-level accuracy and overall hydrograph fidelity. While S4 and S5 occasionally performed well, their results were less stable. Conversely, S0 persistently underperformed, indicating that reliance on new-pipe roughness values underestimates the true hydraulic resistance of aged networks. In summary, these findings demonstrate that S3 offers the most practically robust scenario. By systematically reflecting pipe age and size, S3 provides an available standard design for urban flood risk management.

Finally, the pipe roughness coefficients proposed in S3 were represented by pipe age and sizes, as listed in Table 2. The highest coefficient value, 0.024, corresponded to rectangular pipes with sizes of less than 400 mm, whereas the lowest value, 0.013, applied to circular pipes with the same size range. Notably, these values exceed the current design standard range (0.011–0.020), indicating a more accurate representation of in-field conditions such as pipe aging, sediment accumulation, and increased surface roughness.

Table 2. Pipe roughness coefficients correspond to pipe ages and sizes for the highest-ranked scenarios (S3).

Pipe Sizes (mm)	400		700		1000		>1000
Pipe age (years)	Circular	Rectangular	Circular	Rectangular	Circular	Rectangular	Circular	Rectangular
3	0.013	0.015	0.013	0.015	0.013	0.015	0.013	0.015
10	0.017	0.019	0.016	0.018	0.015	0.017	0.014	0.016
20	0.019	0.021	0.018	0.020	0.017	0.019	0.016	0.018
30	0.020	0.022	0.019	0.021	0.018	0.020	0.017	0.019
40	0.021	0.023	0.020	0.022	0.019	0.021	0.018	0.020
>40	0.022	0.024	0.021	0.023	0.020	0.022	0.019	0.021

4.2. Flood Risk Quantification Between Baseline and Highest-Ranked Scenarios

In the preceding subsection, the proposed methodology was used to prioritize the various pipe roughness scenarios, and the highest-ranked scenario (S3) was identified using systematically defined roughness coefficients, as summarized in Table 2. This subsection presents a quantitative comparison between the highest-ranked scenario (S3) and S0, in which all the pipes were assumed to have the roughness of new pipes to enable assessment of the differences in the hydrologic–hydraulic simulation results. Specifically, the analysis focused on the changes in the total flooding volume, flood depth, and inundation extent across all the UDNs using a 1D–2D simulation model. The comparison was conducted using two real-world extreme rainfall events that occurred in the Republic of Korea in September 2010 and July 2011, rather than synthetic or design storms. These events, which were characterized by high total rainfall and prolonged duration, were selected to enhance the flood reproduction performance and evaluate the robustness of the scenarios under severe urban flooding conditions.

Table 3 compares the flooding volumes across the seven UDNs (AB, AH, BA, BF, DC, FG, and FD) under the two extreme rainfall events of 2010 and 2011 using S0 and S3. While S0 incorporated uniform roughness coefficients based on new pipes, S3 reflected pipe ages and sizes, thereby improving the hydraulic realism of the simulation. The results showed a consistent increase in system flooding volume under S3 across all the UDNs, indicating that S0 substantially underestimated flood risk owing to its oversimplified representation of hydraulic resistance in aging pipes.

Table 3. Comparison of flooding volumes across seven different UDNs under two different extreme rainfall events from 2010 and 2011.

2010 rainfall event
UDN	AB	AH *	BA	BF	DC	FG *	FD
S0	228,789.62 (–)	213,340.12 (–)	186,966.26 (–)	231,501.91 (–)	126,902.10 (–)	27,366.80 (–)	230,806.06 (–)
S3	324,701.78 (41.92%)	484,604.01 (127.15%)	299,971.21 (60.44%)	378,963.59 (63.70%)	218,500.15 (72.18%)	58,442.47 (113.55%)	337,419.58 (46.19%)
2011 rainfall event
UDN	AB	AH	BA	BF	DC	FG	FD
S0	390,761.02 (–)	248,929.08 (–)	255,490.35 (–)	189,448.63 (–)	148,666.77 (–)	28,335.25 (–)	174,242.80 (–)
S3	482,079.84 (23.37%)	488,130.70 (96.09%)	380,983.83 (49.12%)	337,135.98 (77.96%)	293,835.29 (97.65%)	54,700.23 (93.05%)	344,879.08 (97.93%)

* it represented that results showed the largest changes in flooding between S0 and S3.

In 2010, the highest increases (indicated in bold font in Table 3) were observed in AH (127.15%) and FG (113.55%), followed by BF (63.70%). These UDNs contained relatively high proportions of pipes with elevated roughness coefficients, as confirmed by the distribution of the observed values across the network. This increase in pipe roughness, caused by aging and sediment accumulation, contributed to higher hydraulic losses and greater sensitivity to rainfall extremes. Although smaller increases were obtained for AB (41.92%) and FD (46.19%), the growth remained substantial. In 2011, the increase rates remained high or intensified, with marked differences appearing for FD (97.93%), DC (97.65%), FG (93.05%), and BF (77.96%). Importantly, these increases were not artifacts of the model but stemmed from the more accurate representation of pipe roughness, which yielded higher internal friction losses and reduced outflow capacity. AH and FG, in particular, highlight cases where degradation-induced hydraulic inefficiencies were accurately captured by the 1D–2D simulation model.

Figure 6 and Figure 7 illustrate the spatial distributions of flood inundation across the AH and FG drainage networks under the extreme rainfall events of 2010 and 2011. In AH (Figure 6), the application of S3 produced a broader spread of shallow flooding (0.3–0.5 m), while the areas experiencing deep flooding (>1.0 m) were relatively reduced compared with S0. This indicates that deterioration-induced increases in roughness lowered the overall discharge efficiency, dispersing floodwaters more widely but at lower depths. A comparable pattern was observed in FG (Figure 7), where shallow flooding expanded, yet the extent of deep flooding showed only a modest increase. These spatial patterns suggest that pipe deterioration shifts flood risks from localized severe inundation to more widespread but shallower waterlogging.

Figure 6. Flood-inundation maps of the AH drainage network for the extreme rainfall events of (a) 2010 and (b) 2011, comparing results between baseline (S0) and highest-ranked scenario (S3). Black lines indicate the layout of pipes within the UDN. Color shading indicates flood depth categories, where red represents higher flood depths (i.e., higher flood risk level) and blue represents lower flood depths (i.e., lower flood risk level).

Figure 7. Flood-inundation maps of the FG drainage network for the extreme rainfall events of (a) 2010 and (b) 2011, comparing the results between baseline (S0) and highest-ranked scenario (S3). Black lines indicate the layout of pipes within the UDN. Color shading indicates flood depth categories, where red represents higher flood depths (i.e., higher flood risk level) and blue represents lower flood depths (i.e., lower flood risk level).

Beyond total flood volumes, the adoption of S3 revealed critical differences in flood depth and inundation extent that provide direct insights for urban planning and drainage retrofitting. Under S3, the average flood depth across inundated cells tended to decrease, whereas the inundation extent slightly increased. This pattern indicates that incorporating age- and size-dependent roughness shifts flood risk dynamics: the number of shallowly inundated areas increases, while the occurrence of deeper, high-risk inundation is reduced. In other words, although more areas are exposed to minor flooding, the probability of life-threatening deep floods is mitigated, lowering the overall hazard severity. Conversely, S0—by assuming uniformly low roughness—reproduces smaller total inundation areas but concentrates flooding in fewer locations with substantially greater depths. Such conditions are more dangerous for residents and urban assets, suggesting that conventional assumptions may give decision-makers a misleading sense of safety.

Overall, these findings emphasize that applying a uniform roughness coefficient (S0) underestimates the spatial and hydraulic complexity of aging urban drainage networks. By contrast, the proposed framework (S3) provides a more nuanced risk profile: while the inundation footprint increases slightly, the associated hazard level is reduced by limiting the occurrence of deeper floods. This distinction offers actionable insights for urban planners and engineers, underscoring the need to integrate pipe age and diameter information into future retrofitting and adaptation strategies.

5. Discussion

This study demonstrated the validity of applying pipe roughness coefficients that explicitly reflect pipe age and size in seven urban drainage networks (UDNs). Unlike uniform values traditionally used in design standards, the proposed approach highlights that pipe roughness is not a simple calibration parameter but a critical variable representing physical deterioration in drainage systems. Among the evaluated cases, the highest-ranked scenario (S3) was identified, integrating pipe shape, size, and age. Its impacts on urban flooding were assessed using a 1D–2D rainfall–runoff hydrologic–hydraulic model, and flood damages were quantified through inundation depth distributions.

The S3 scenario defines roughness coefficients based on pipe characteristics—circular or rectangular sections, size ranges from ≤400 to >1000 mm, and ages spanning 3 to more than 40 years. These values are consistent with empirical field measurements reported by Coelho and Azevedo [27]. In contrast, conventional standards that apply fixed values (e.g., 0.013 or 0.015) risk overestimating hydraulic capacity, as observed in the uniform baseline scenario (S0). Our results showed that incorporating deterioration-sensitive roughness values substantially altered simulation outputs: average flooding volumes across the UDNs increased by 75.02% and 76.45% for the 2010 and 2011 extreme rainfall events, respectively. Particularly in the AH system, which contains a high proportion of aged pipes, the increase reached 127.15% and 96.09%. These outcomes demonstrate that pipe deterioration intensifies flow resistance and contributes directly to surface flooding, especially at manholes and other hydraulic junctions.

The flood depth distribution analysis further revealed that S3 produced a distinct spatial pattern: shallow flooding (0.3–0.5 m) expanded across broader areas, while deep flooding (>1.0 m) became more localized compared with S0. This shift indicates that aging-induced roughness reduces peak discharge efficiency, dispersing floodwaters more widely but at lower depths. Such spatial characteristics highlight the importance of identifying zones where severe inundation persists, as targeted rehabilitation in these critical segments could yield substantial reductions in high-risk flooding and associated secondary impacts, including structural damage and traffic disruption. Despite the demonstrated value of the proposed approach, its applicability depends on the availability of detailed pipe-level data such as installation year, diameter, and network configuration. In practice, these data are often incomplete or outdated, limiting direct transferability. To address this challenge, recent studies have introduced surrogate data modeling and machine learning–based inference techniques that estimate missing attributes and improve simulation robustness [5,6]. Integrating such methods with the present framework could broaden its relevance, particularly in cities with insufficient infrastructure records [3,4,50].

The practical implication is clear: pipe ages and sizes dependent roughness coefficients should be systematically considered in design standards for urban flood risk modeling. For example, in the Republic of Korea, the current “sewer design standards” fixed values such as 0.013 for concrete circular pipes and 0.015 for box culverts without considering deterioration. This study demonstrates that such uniform assumptions underestimate flood volumes and inundation extents by neglecting the progressive increase in hydraulic resistance due to aging and sediment deposition. By contrast, the proposed framework provides empirically grounded ranges—for instance, increasing from 0.013 in pipes less than 10 years old to over 0.020 in pipes older than 40 years, with further adjustments for geometry. Amending design manuals to include these differentiated coefficients would enable engineers to better account for deterioration effects during both new system design and rehabilitation planning.

6. Summary and Conclusions

This study proposed a novel methodology for incorporating pipe aging effects into pipe roughness settings for UDS simulations. By categorizing pipe roughness scenarios (S1–S10) based on pipe age and sizes, and applying those scenarios across seven UDNs in Seoul, this study introduced a data-driven calibration strategy that captures the hydraulic degradation of aging sewer infrastructures. Using dual evaluation metrics, the PWLE and RMSE, the best-performing scenario (S3) was identified via a Euclidean distance-based prioritization method. S3 was then applied to 1D–2D rainfall–runoff hydrologic–hydraulic simulations to quantify its influence on flood inundation across different rainfall scenarios.

The results revealed that S0 (the baseline scenario), for which uniform roughness coefficients based on new pipes were assumed, significantly underestimated urban flood risks. For two extreme rainfall events occurring in 2010 and 2011, the total flood volumes increased on average by 75.02% and 76.45%, respectively, when S3 was applied. For the AH and FG drainage networks, in which the aging infrastructure was concentrated, the increases exceeded 120%, with new flood inundation zones emerging across critical flood vulnerability zones such as underpasses and major roads. Furthermore, the number of flood-inundated 10 m × 10 m grids exceeding 0.2 m in depth expanded by hundreds to thousands in most UDNs. These results empirically confirm that pipe roughness coefficients should not be treated as mere modeling constants but are, rather, essential proxies for physical deterioration that directly shape the hydrologic–hydraulic performance of a given system.

This study had several limitations that should be addressed in future research. First, the roughness coefficients were estimated only from pipe installation year and size without direct inspection data such as CCTV surveys, condition ratings, or field tests. To overcome this limitation, future studies should integrate multi-source data fusion approaches that combine CCTV inspections, sensor networks, and maintenance records to refine the empirical basis of deterioration-informed roughness estimation. Second, the analysis relied on two historical rainfall events, which restricted the robustness of the conclusions. Expanding the evaluation to probabilistic rainfall generation and climate projection scenarios would enable more comprehensive assessments under diverse hydrologic conditions. Finally, the roughness coefficients in this study were assumed to change stepwise with pipe age and size, but they did not dynamically evolve over time. Future research should, therefore, develop and test dynamic roughness degradation models that continuously adjust coefficients, better capturing temporal variations in pipe conditions. By addressing these limitations, future research can improve the proposed framework into a more adaptive tool for urban drainage planning and design, enhancing its integration into design standards and long-term planning for urban flood risk management.