the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Aug 2026
Why has flood risk increased in a millennia-old mountainous town? Engineering-induced transformation of compound flood pathways in Baifusi, China
Abstract. Baifusi, a long-inhabited mountain-river town in China, coexisted with floods, yet damaging inundation has become concentrated in the recent engineering era. This study asks why flood risk has risen markedly over the last decade. We combine hydrological monitoring, field floodmarks, historical satellite images, drainage calculations, reservoir-operation rule learning, operation logs, surveyed terrain reconstruction, one-dimensional–two-dimensional coupled hydrodynamic scenario modelling, threshold analysis and interpretable surrogate-model evidence. The evidence shows that rainfall magnitude alone is insufficient: the 2016 event had a rare 24 h rainfall return period but a moderate flood-peak return period, while the 2023 event combined high Najitan release, Qiedao tributary inflow, Jinlongtan downstream backwater and local drainage blockage. Matched scenarios show a persistent 1.56–2.08 m tributary increment, whereas the downstream-backwater increment decreases from 1.67 m in the 5-year scenario to 0.03 m in the 100-year scenario. Relative to the 357.01 m hospital-entrance threshold, effective protection capacity decreases from about 42 years under the mainstream-only boundary to about 17 years with tributary inflow and 10 years under the full compound boundary. Archival images indicate a shift from sparse riverfront occupation in 2009 to a dense riverfront building belt by 2023. Pump-station calculations show that internal drainage capacity is 2.62–3.00 m³ s⁻¹, so ponding and backflow become critical when river stage is high. Flood-risk management in engineered mountain towns should move beyond a single return-period defence standard towards a process-based system that jointly manages upstream releases, tributary timing, downstream stage, drainage outlets and exposure control.
- Preprint
(2861 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-4327', Anonymous Referee #1, 08 Sep 2026
-
AC1: 'Reply on RC1', Y. Liu, 02 Oct 2026
We thank the referee for the positive assessment and constructive comments. We accept the recommendations and will strengthen the quantitative results, supplementary event cases, validation and presentation as detailed below.
Major comment 1: Emphasize quantitative and process-based findings in the abstract.
Response: We agree. We will shorten the background and add the principal quantitative findings on local flood levels, tributary contribution, downstream backwater and threshold exceedance. The findings will be updated using the supplementary calculations and validation results, with their applicable conditions stated.
Major comment 2: Define the research gap and connect it to the objectives.
Response: We agree. We will revise the Introduction to explain the limitations of single-peak, single-return-period and rainfall-only assessments for historic mountain towns. The objectives will be stated explicitly: identifying compound flood pathways, evaluating local impact thresholds, and developing management implications for towns facing similar constraints.
Major comment 3: Provide details of model configuration and scenario construction.
Response: We agree. We will add a configuration summary covering the computational domain, mesh and cross-sections, roughness zoning, boundary locations and sources, downstream-stage treatment, initial conditions, time step, wetting and drying, and convergence checks. We will document the vertical reference and harmonization of terrain, surveyed sections, floodmarks and gauge levels, together with the representation of bridges, drainage and storage. A scenario table will list the inputs and assumptions for each case.
Major comment 4: Clarify the meaning, synchronization assumptions and limitations of S1–S3.
Response: We agree. We will specify the tributary reference condition, downstream-stage boundary and timing assumptions for S1–S3, and define their scope as comparisons under present-day conditions. We will add the tailwater-only combination to complete a two-by-two tributary/tailwater comparison and evaluate conditional effects and interaction. Event replay and prescribed timing tests will be identified separately.
Major comment 5: Strengthen validation and document floodmarks and event boundaries.
Response: We agree. We will supplement the event calculations and add an observed–simulated comparison table and a floodmark location map. The table will report the number of usable marks, event year, coordinates, vertical datum, measurement uncertainty, simulated stage, and signed and absolute errors in metres. We will document the 2023 calibration and 2024 evaluation boundaries, parameter settings and use of observations during model development. Validation evidence for the configuration used in this study will be presented separately from the supporting results of Xiao et al.
Major comment 6: Distinguish the mechanisms of the 2016, 2021, 2023 and 2024 events.
Response: We agree. We will add an event-comparison table covering local rainfall, mainstream discharge, tributary inflow, downstream stage, peak timing and observed inundation. The Results and Discussion will identify the dominant mechanisms supported by these records and state the attribution limits for events with incomplete observations.
Major comment 7: Add semi-quantitative evidence of exposure expansion.
Response: We agree. Where comparable dated imagery and spatial records are available, we will add indicators of building-footprint change, riverfront-buffer occupation and development in consistently defined low-lying areas. The comparison period, coverage and classification method will be stated. Periods with insufficient comparable evidence will be described qualitatively, with the limitations specified.
Major comment 8: Develop the implications for return-period-based flood protection.
Response: We agree. We will expand the Discussion with comparisons of local peak levels, exceedance duration and impacts under different boundary combinations at the same nominal design class. We will revise the protection terminology to conditional scenario-based threshold-crossing indices, explain the probability limitations and interpolation assumptions, and evaluate the local thresholds at their corresponding locations.
Major comment 9: Focus the conclusions on the research objectives.
Response: We agree. We will organize the Conclusions around compound flood mechanisms, local protection thresholds and management implications. Repeated methodological and numerical details will be shortened. The recommendations will focus on coordinated monitoring and warning, drainage under high downstream stages, and protection of heritage buildings and essential facilities, with the scope of application stated.
Minor comment 1: Standardize units.
Response: We agree. We will standardize units in the text, equations, tables and figures, including m³ s⁻¹, m and events yr⁻¹, and check the definitions and observation periods of event-rate indicators.
Minor comment 2: Shorten figure captions.
Response: We agree. We will shorten captions to the information needed to read each figure and move extended methodological explanations and interpretation to the main text.
Minor comment 3: Improve table layout.
Response: We agree. We will adjust column widths and line breaks, standardize labels and precision, and place units in the column headings.
Minor comment 4: Check equations and notation.
Response: We agree. We will check equation numbering and cross-references, define variables and units at first use, and state the time intervals and error denominators explicitly.
Minor comment 5: Document imagery sources, dates and authorization.
Response: We agree. We will add sources, acquisition dates and permissions or licences for maps, remote-sensing images and UAV images, together with the spatial reference and relevant processing information.
Minor comment 6: Improve sentence readability.
Response: We agree. We will divide long sentences, remove repetition and standardize terminology throughout the manuscript.
Citation: https://doi.org/10.5194/egusphere-2026-4327-AC1
-
AC1: 'Reply on RC1', Y. Liu, 02 Oct 2026
-
RC2: 'Comment on egusphere-2026-4327', Anonymous Referee #2, 30 Sep 2026
This manuscript addresses an important and practically relevant problem: how interacting reservoir releases, tributary inflow, downstream backwater and local conditions affect flooding in a historic mountain town. I particularly appreciate the integration of field evidence with hydraulic scenarios and impact thresholds. The study is well suited to NHESS and has the potential to make a useful contribution. My overall assessment is positive. However, revision is needed to establish the strength of the historical attribution, clarify the probabilistic interpretation of the results, and demonstrate the reliability of the model evidence supporting the central conclusions.
Major comments
1. Distinguish present-day hydraulic mechanisms from historical engineering attribution
The S1–S3 experiments diagnose the present-day effects of adding tributary inflow and downstream backwater, but they do not reconstruct pre-engineering conditions. In particular, S1 excludes the naturally occurring tributary while retaining the current geometry and model configuration. Therefore, the 42–17–10-year sequence supports a diagnosis of current compound-flood mechanisms, but does not by itself demonstrate a historical deterioration caused by engineering. Please distinguish clearly between these two claims. If historical geometry and operation data are insufficient for a matched past-versus-present comparison, the causal wording in the title, abstract, and conclusions should be moderated accordingly. The argument that moderate local rainfall should historically have been non-damaging should also be qualified, because upstream runoff, antecedent conditions, and event timing remain important.
2. Clarify the probability meaning of the reported “effective protection capacity”
The reported values of approximately 42, 17, and 10 years are central to the paper, but their probabilistic interpretation is unclear. Combining mainstream, tributary, and downstream boundaries within the same nominal return-period class does not establish the annual exceedance probability of the compound boundary state or of town inundation. Please explain how the design hydrographs and Jinlongtan stages were derived, including the underlying frequency records, distributions, and timing assumptions. Unless the joint probability of the drivers is explicitly quantified, these values should be presented as conditional, scenario-based threshold-crossing indices rather than actual protection return periods. The interpolation procedure should also be justified, and corresponding results for the other local thresholds would help demonstrate robustness.
3. Demonstrate that the hydraulic attribution is robust to model and boundary uncertainty
The main conclusions depend on relatively small scenario differences and threshold exceedances, yet the uncertainty of the S1–S3 model is not sufficiently characterized. Please distinguish validation of the HydroMPM configuration used here from supporting results reported in Xiao et al. (2026), and provide observed versus simulated values for the available floodmarks, together with locations and measurement uncertainty. A focused sensitivity analysis should assess the robustness of the main scenario ordering and threshold crossings to plausible uncertainty in inflows, downstream stage, roughness, terrain, and peak timing. This is particularly important where exceedances or backwater increments are only a few centimetres. The unresolved 2023 Najitan release discrepancy should also be reconciled or propagated as a boundary range. Finally, because S2−S1 and S3−S2 depend on the order in which drivers are introduced, these increments should be described as conditional marginal effects unless interactions are explicitly tested.
Additional minor comments
Reservoir-operation learning and routing (Section 3.3; Fig. 5c): specify which hydraulic boundaries use measured releases and which use LightGBM predictions. Report the selected predictors, time lags and flood-event test performance, with feature selection and tuning confined to training data. Explain the optimized-routing assumptions, initial storage and operational constraints so the reported improvement can be assessed.
Figures and terminology: Fig. 6d labels the result as “first town overtopping”, although the calculation uses the hospital threshold and lower street-entry thresholds exist. Revise the label accordingly. Define the normalization in Fig. 5a and the error denominators in Fig. 3b. Keep hazard, exposure, threshold exceedance and flood risk distinct throughout the manuscript.
Presentation and reproducibility: shorten the repetitive introduction and methodological framing, and make the new contribution relative to Xiao et al. (2026) explicit. A compact supplement containing the scenario inputs, validation table, processed results and interpolation script would considerably improve transparency while respecting the stated restrictions on raw operational data. Check figure references, units and significant figures throughout.
Citation: https://doi.org/10.5194/egusphere-2026-4327-RC2 -
AC2: 'Reply on RC2', Y. Liu, 08 Oct 2026
We sincerely thank the reviewer for the positive assessment and careful, constructive comments. We appreciate the recommendations on attribution, probability interpretation and model reliability. We accept the main recommendations and have clarified the manuscript accordingly. Our responses and planned supplementary analyses are outlined below.
Major comment 1: Present-day mechanisms and historical attribution
We appreciate this important distinction and agree that the scope of the scenario evidence should be explicit. The revised manuscript clarifies that S1–S3 represent the present-day hydraulic system, with constant reference tributary inflow in S1. These comparisons identify conditional hydraulic responses; quantitative attribution of long-term change requires comparable historical geometry, operation and event records. The abstract, discussion and conclusions now acknowledge this limitation. The interpretation of local rainfall also accounts for upstream runoff, antecedent conditions and event timing. We respectfully retain the original title to express the central research question, while qualifying the historical interpretation in the text.
Major comment 2: Probability interpretation and threshold crossing
We thank the reviewer for identifying this ambiguity. We agree that assigning the same nominal return-period class to several drivers does not establish their joint annual probability. The approximately 42-, 17- and 10-year values are now defined as conditional, scenario-based threshold-crossing indices for the 357.01 m hospital reference. They indicate the loading class at which this elevation is reached within each prescribed boundary combination. Boundary synchronization is stated explicitly. Appendix A compares interpolation in T, log(T) and 1/T for four local elevations and includes an illustrative threshold-sensitivity check. We will supplement the frequency records, fitted distributions, design-hydrograph construction and downstream-stage derivation, including their timing assumptions.
Major comment 3: Validation, uncertainty and conditional effects
We appreciate the reviewer’s emphasis on the reliability of small scenario differences and agree that validation must apply to the configuration used in this study. Claims relying on validation of another configuration have been removed. We will supplement event- and location-matched floodmark comparisons, consistent vertical datums and measurement uncertainty, together with focused sensitivity cases for inflow, downstream stage, roughness, terrain and peak timing. These checks will assess scenario ordering, threshold crossings and the 0.03 m increment.
Figure 3 presents the two 2023 release records separately; their discrepancy will be reconciled or propagated as a boundary range. Section 4.4 and Appendix A also clarify that S2−S1 and S3−S2 are conditional marginal effects and explain their dependence on the order of boundary changes.
Minor comment 1: Reservoir-operation learning and routing
We agree that boundary sources and routing assumptions need to be transparent. Section 3.3 clarifies that observed releases define the event boundary and prescribed hydrographs define the design scenarios. Unsupported learning-model performance and optimization claims have been removed. The routing method now specifies the storage balance, 368.00 m initial level, five-gate full-opening outlet relation and 369.00 m constraint, distinguished from the 370.00 m normal pool level. Appendix A provides the interval-mean inflow reconstruction, with its temporal resolution clearly identified. We will verify the comparison baseline of the reported 18.50 % peak reduction so that its meaning can be assessed consistently.
Minor comment 2: Figures and terminology
We thank the reviewer for these specific corrections. Figure 6d now identifies the 357.01 m hospital-elevation threshold; Section 4.3 distinguishes it from lower street-entry thresholds. Figure 5a defines normalization by each indicator’s maximum across the three events. The former error panel in Figure 3 has been replaced by a sourced hourly monitoring comparison. The text distinguishes project-summary values from hourly maxima and consistently separates hazard, exposure, threshold exceedance and flood risk. Stage and inundation-area responses are also evaluated separately.
Minor comment 3: Presentation and reproducibility
We appreciate this suggestion and agree that a clearer contribution and accessible supporting material will strengthen the manuscript. The contribution is now stated as an interpretable assessment linking compound boundary configurations, stage–area responses, conditional factor contributions and local thresholds. The adopted terrain method is cited to previously published work, while the present analysis is described explicitly. We will further shorten repetitive framing and provide the prepared processed results and analysis scripts with the scenario inputs and validation table in a compact supplement, respecting restrictions on raw operational data. Figure references, units and significant figures have been checked.
We thank the reviewer again for the recommendations, which have helped clarify the interpretation and strengthen the planned verification.
Citation: https://doi.org/10.5194/egusphere-2026-4327-AC2
-
AC2: 'Reply on RC2', Y. Liu, 08 Oct 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 251 | 117 | 55 | 423 | 40 | 39 |
- HTML: 251
- PDF: 117
- XML: 55
- Total: 423
- BibTeX: 40
- EndNote: 39
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Overall Evaluation
This manuscript addresses the recent increase in flood risk in Baifusi, a long-inhabited mountainous river town. The study develops a process-oriented framework that links reservoir operation, tributary inflow, downstream backwater, terrain constraints, limited drainage capacity, and expansion of exposure in low-lying areas. The topic is well aligned with the scope of NHESS, particularly its focus on natural hazard processes, risk mechanisms, and disaster-risk management. Overall, the manuscript presents a relevant case, a clear scientific question, and useful implications for compound flood-risk governance. I recommend acceptance after revision.
Major Comments
1. The abstract should emphasize clearer quantitative and process-based insights. The authors should reduce generic background statements and highlight the main hydrological findings for the study area.
2. The introduction should define the research gap more explicitly. In particular, the limitation of single return-period, single-peak, or rainfall-only flood assessment should be clearly linked to the specific objectives of the paper.
3. The methods section should provide more detail on model configuration, including mesh setting, roughness zoning, boundary conditions, downstream-stage treatment, vertical datum harmonization, wetting-drying treatment, and scenario construction.
4. The S1, S2, and S3 scenarios are central to the manuscript. Their physical meaning, boundary-synchronization assumptions, and limitations should be explained more clearly.
5. Model validation should be strengthened. The authors should report the number and spatial distribution of floodmarks, the error-calculation method, and the boundary conditions used for the 2023 calibration and 2024 validation events.
6. The comparison of typical flood events could be improved. The manuscript should more clearly distinguish the dominant mechanisms of the 2016, 2021, 2023, and 2024 events, such as rainfall dominance, tributary contribution, backwater amplification, and external-mainstream flooding.
7. The evidence for exposure expansion in low-lying areas remains mainly qualitative. If possible, the authors should add semi-quantitative indicators, such as building expansion, riverfront-buffer occupation, low-lying development intensity, or representative cross-section changes.
8. The discussion should further develop the implications for return-period-based flood protection. A key message of the paper is that the same nominal return period may lead to very different impacts under different engineering-boundary combinations.
9. The conclusions should be more focused. They should be written as clear concluding statements aligned with the research objectives, rather than repeating detailed results and discussion.
Language and Formatting Issues
1. Units should be consistent throughout the manuscript, such as m³ s⁻¹ and events yr⁻¹.
2. Some figure captions are too long and should be shortened.
3. Table layout, column width, and line breaks need further adjustment.
4. Equation numbering, variable definitions, and symbol formatting should be checked carefully.
5. Maps, remote-sensing images, and UAV images should clearly indicate data source, date, and authorization.
6. Some long sentences should be divided to improve readability.
Recommendation
The manuscript addresses a relevant hydrological and natural-hazard problem, fits the journal scope, and provides useful insights for compound flood-risk management in mountainous towns. I recommend acceptance after minor revision.