1. Bibliographic Information

1.1. Title

Transformations in exposure to debris flows in post-earthquake Sichuan, China

1.2. Authors

• Isabelle Utley¹

• Tristram Hales¹

• Ekbal Hussain²

• Xuannie Fan³

  ¹School of Earth and Environmental Sciences, Cardiff University, Cardiff, CF10 3AT, UK ²British Geological Survey, Keyworth, Nottingham, NG12 5GG, UK ³State Key Laboratory of Geohazard Prevention, Chengdu University of Technology, Chengdu, China

Correspondence: Isabelle Utley (utleyieu@gmail.com)

1.3. Journal/Conference

The paper was published in Natural Hazards and Earth System Sciences (NHESS). This journal is a prominent publication in the fields of natural hazards, disaster risk, and earth system science, known for publishing interdisciplinary research that addresses the causes, impacts, and mitigation of natural hazards. Its reputation ensures a rigorous peer-review process and broad dissemination within the scientific community focused on disaster management and earth sciences.

1.4. Publication Year

Received: 19 July 2024 – Discussion started: 16 September 2024 Revised: 27 May 2025 – Accepted: 28 May 2025 – Published: 14 August 2025

1.5. Abstract

The abstract outlines a study investigating how different catchment interventions, specifically the construction of check dams and levees, alter the exposure and hazard of post-2008 debris flows in three gullies in Sichuan Province, China. These gullies were chosen based on their mitigation status: Cutou (two check dams), Chediguan (two check dams), and Xiaojia (none). Using high-resolution satellite imagery, the researchers developed a multi-temporal building inventory from 2005 to 2019 and compared it with the spatial distribution of past and modeled debris flow events. Key findings indicate that post-earthquake urban development in Cutou and Chediguan increased exposure to a major debris flow in 2019, impacting 40% and 7% of surveyed structures, respectively. Simulations using the LAHARZ model for low ($10^4 \mathrm{~m}^{3}$), high ($10^5 \mathrm{~m}^{3}$), and extreme ($10^6 \mathrm{~m}^{3}$) flow volumes demonstrated that check dams effectively mitigate exposure to low- and high-flow events but are ineffective for extreme events, with significant impacts on buildings (Cutou 59%, Chediguan 22%, Xiaojia 33%). A statistical exposure model, adapted from a social vulnerability equation, verified that Cutou's exposure increased by 64% in 2019 and Chediguan's by 52%, while Xiaojia's increased by only 2% in 2011. This highlights a correlation between extensive grey infrastructure (engineered mitigation) and higher exposure to extreme debris flows. The study concludes that the perceived reduction in downstream exposure due to check dams can lead to a levee effect, ultimately increasing exposure to larger, less frequent events.

1.6. Original Source Link

/files/papers/69152cf989d8a27cc42ad79e/paper.pdf This link leads to the PDF of the paper, which is officially published.

2. Executive Summary

2.1. Background & Motivation

The 2008 $M_{\mathrm{w}}$ 7.9 Wenchuan Earthquake in Sichuan, China, triggered extensive landslides, depositing vast amounts of loose material. This material, remobilized by intense rainfall, has led to a significant increase in the frequency and magnitude of debris flows (fast-moving mixtures of water, sediment, and rocks) in the post-earthquake period, often exceeding volumes of $1 \times 10^6 \mathrm{~m}^3$. These stochastic hazards (random or unpredictable events) pose severe challenges to downslope recovery zones (areas located lower in elevation that are undergoing post-disaster rebuilding) and communities.

The core problem the paper aims to solve is understanding how human interventions, specifically the construction of check dams (engineered barriers built across channels to control sediment and flow) and levees (embankments to prevent overflow), influence the exposure (presence of people, infrastructure, or assets in hazard-prone areas) and hazard (potential occurrence of a damaging event) of these post-earthquake debris flows. This problem is crucial because, in the aftermath of such a major earthquake, regions undergo rapid rebuilding and expansion of infrastructure, potentially increasing exposure to secondary hazards like debris flows. Despite efforts to build resilience and reduce vulnerability, the dynamic nature of post-earthquake recovery means that exposure changes rapidly.

Specific challenges and gaps exist in prior research regarding the levee effect in the context of debris flows. While check dams are a common mitigation strategy, their long-term effectiveness, especially against extreme events, and their influence on human land-use decisions are not fully understood. There's anecdotal evidence that large debris flow events in the Wenchuan region caused significant damage despite check dams, but it's unclear if this was due to increased exposure facilitated by the dams themselves (the levee effect) or simply large-scale post-earthquake infrastructure expansion. The paper's entry point is to investigate this interplay, using a comparative analysis of catchments with different mitigation levels to assess if check dam construction contributes to a false sense of safety that encourages development in hazard-prone areas, ultimately increasing exposure to larger, less frequent events.

2.2. Main Contributions / Findings

The paper makes several significant contributions and reaches key conclusions regarding the dynamic interplay between engineered mitigation, urban development, and debris flow exposure:

• Multi-temporal Building Inventory and Exposure Assessment: The study developed a detailed multi-temporal building inventory (a record of buildings over different points in time) from 2005 to 2019 using high-resolution satellite imagery. This allowed for a precise tracking of post-earthquake urban development in three specific gullies. It found that urban development in Cutou and Chediguan significantly increased exposure to a major debris flow event in 2019, with inundation (flooding) impacting 40% and 7% of surveyed structures, respectively.

• LAHARZ Modeling of Check Dam Effectiveness: Using the LAHARZ (LAhar HAzard Zonation) model, the researchers simulated debris flow runouts (the distance and area a debris flow covers) for three different volumes: low ($10^4 \mathrm{~m}^{3}$), high ($10^5 \mathrm{~m}^{3}$), and extreme ($10^6 \mathrm{~m}^{3}$). The simulations demonstrated that check dams are effective in mitigating exposure to low- and high-volume events. However, they proved ineffective for extreme events, leading to substantial damage (59% of buildings in Cutou, 22% in Chediguan, and 33% in Xiaojia significantly affected). This highlights a critical limitation of current check dam designs in the region.

• Statistical Exposure Model with Mitigation Factor: The paper adapted a social vulnerability equation into a statistical exposure model to quantify the degree of exposure. A key innovation was the incorporation of a modification factor (M) to account for the effectiveness (or ineffectiveness) of engineered measures like check dams. This model revealed that Cutou's exposure increased by 64% in 2019 and Chediguan's by 52%, whereas Xiaojia (the unmitigated gully) saw only a 2% increase in 2011.

• Evidence for the Levee Effect: The research strongly suggests that the presence of check dams contributes to a perceived reduction in downstream exposure (the hazard faced by areas lower than the dam). This perception, however, can lead to a levee effect, a phenomenon where the perceived safety provided by protective structures encourages urban expansion into hazard-prone areas. Consequently, this ultimately increases exposure to larger, less frequent debris flow events that overwhelm or bypass the mitigation structures. The correlation between extensive grey infrastructure (engineered structures) and higher exposure to extreme debris flows supports this conclusion.

  These findings solve the problem of understanding the complex and sometimes counterintuitive relationship between hazard mitigation infrastructure and long-term exposure in post-earthquake recovery regions. They highlight the need for a holistic approach to risk management that considers not only the physical effectiveness of structures but also their influence on human perception and land-use decisions.

3. Prerequisite Knowledge & Related Work

3.1. Foundational Concepts

To fully grasp the contents of this paper, a novice reader should be familiar with several key concepts in natural hazards, geology, and geographic information science.

• Debris Flows: These are common and highly destructive types of landslide (mass movement of rock, debris, or earth down a slope). A debris flow is a fast-moving, water-saturated slurry of unconsolidated (loose) rock, soil, and organic matter. They are characterized by their high density and viscosity, which allows them to transport large boulders and other debris.

  • Formation in Post-Earthquake Context: Earthquakes, like the 2008 Wenchuan event, often trigger numerous landslides, leaving behind vast quantities of loose, unstable material on steep slopes and within river channels. This material, termed coseismic landslide material, becomes highly susceptible to remobilization. When intense or prolonged rainfall occurs, this loose material can become saturated with water, lose internal strength, and transform into a fluid-like mixture that flows rapidly downslope as a debris flow. These are often referred to as post-earthquake debris flows because their increased frequency and magnitude are directly linked to the earthquake-generated sediment supply.
  • Impacts: Debris flows can cause extensive damage, including loss of life, destruction of buildings, infrastructure (roads, bridges, tunnels), agricultural land, and disruption of river systems.

• Coseismic Landslides: The term coseismic means "occurring at the same time as an earthquake." Coseismic landslides are landslides that are directly triggered by the shaking during an earthquake. The 2008 Wenchuan Earthquake is cited as triggering around 56,000 such landslides, displacing nearly 3 cubic kilometers of loose material. This vast amount of unstable sediment becomes the primary source material for subsequent post-earthquake debris flows.

• Check Dams: These are engineered structures built across the channels of steep, torrent-prone streams or gullies. Their primary purpose is to stabilize riverbeds, reduce the velocity of water and debris flows, and trap sediment.

  • Function: Check dams typically consist of concrete, stone, or steel barriers. They work by:
    1. Sediment Retention: Trapping coarse sediment (rocks, gravel) upstream, which reduces the amount of material available for debris flows downstream and can reduce the overall flow volume.
    2. Channel Slope Reduction: Creating a series of steps along the channel, effectively reducing the overall gradient and thus the erosive power and speed of the flow.
    3. Flow Attenuation: Spreading out the energy of a debris flow and reducing its peak discharge.
  • Maintenance: They often require regular maintenance, such as clearing accumulated sediment, to maintain their effectiveness.

• Exposure vs. Vulnerability vs. Hazard: These are distinct but interrelated concepts in disaster risk assessment:

  • Hazard: The potential occurrence of a natural or human-induced physical event that may cause loss of life, injury, or other health impacts, as well as property damage, loss of livelihoods and services, social and economic disruption, or environmental damage. In this paper, debris flows are the primary hazard.
  • Exposure: The presence of people, livelihoods, environmental services, resources, infrastructure, or economic, social, or cultural assets in places and settings that could be adversely affected by a hazard. The paper specifically focuses on building exposure, meaning the number and location of buildings within areas that could be inundated by debris flows.
  • Vulnerability: The characteristics and circumstances of a community, system, or asset that make it susceptible to the damaging effects of a hazard. This includes physical (e.g., building materials, structural integrity), social (e.g., poverty, access to resources), economic, and environmental factors. The paper adapts a social vulnerability equation to assess exposure, but acknowledges the difficulty in assessing detailed vulnerability (like construction materials) remotely.

• Levee Effect (or Levee Paradox): This phenomenon, originally observed in flood management, describes how the perceived safety provided by protective engineering structures (like levees or check dams) can paradoxically lead to increased exposure and potential damage. When people feel secure behind a protective barrier, they may be encouraged to develop or settle in areas that were previously considered too risky. If a larger-than-expected event occurs, or the structure fails, the increased exposure behind the barrier can lead to significantly higher losses than if no structure had been present and development had been naturally restrained. The paper investigates if this levee effect applies to debris flow mitigation.

• GIS (Geographic Information System): A system designed to capture, store, manipulate, analyze, manage, and present all types of geographical data. In this study, GIS is used extensively for mapping features (buildings, roads), delineating debris flow runout zones, and performing spatial analysis.

• DEM (Digital Elevation Model): A 3D representation of a terrain's surface, created from terrain elevation data. It's a fundamental input for hydrological and geomorphological analyses, allowing for the calculation of slope, aspect, and flow paths. The paper uses a Shuttle Radar Topography Mission (SRTM) DEM with 30m resolution.

• LAHARZ (LAhar HAzard Zonation): A GIS toolkit developed by the U.S. Geological Survey (USGS) for mapping lahar (a type of debris flow originating from volcanoes) hazard zones.

  • Mechanism: LAHARZ uses empirical scaling relationships that correlate the volume of a lahar with the area it inundates and the width of its flow path. This allows for the creation of realistic inundation areas and cross-sections without requiring complex rheological (flow behavior) parameters that are often difficult to obtain.
  • Application to Debris Flows: While originally for lahars, the empirical relationships are often applicable to other large-volume debris flows due to similar flow dynamics. The model simulates a flow from a source point on a DEM, calculates its flow path downslope, and generates cross-sections representing depositional volumes.

3.2. Previous Works

The paper contextualizes its research by referring to a range of prior studies across several domains:

• Post-Earthquake Hazard Chains: Major earthquakes often trigger sequences of secondary hazards. The paper cites the 1994 Northridge (Harp and Jibson, 1996) and 1999 Chi-Chi (Lin et al., 2006) earthquakes as examples, which led to increased exposure to secondary hazards for years. For the 2008 Wenchuan Earthquake, studies by Cruden and Varnes (1996), Cui et al. (2008), Huang and Li (2009), Guo et al. (2016), and Thouret et al. (2020) highlighted the increased frequency and magnitude of debris flows post-earthquake. Horton et al. (2019) specifically attributed increased flow volumes to high in-channel sediment, a phenomenon known as bulking. Fan et al. (2018, 2019b) further detailed the earthquake-induced hazard chains.
• Impacts on Built Environment: Research consistently shows that post-seismic debris flows severely affect the built environment. Chen et al. (2011) discussed how these flows reshape demographic and structural landscapes. Buildings are particularly susceptible (Hu et al., 2012; Zeng et al., 2015), with property damage being a major impact (Wei et al., 2018, 2022). Variations in construction materials are noted as crucial for structural resilience (Zhang et al., 2018). The development of critical infrastructure (highways, tunnels) also encourages settlement in hazard-exposed areas (Cruden and Varnes, 1996; Jiang et al., 2016).
• Debris Flow Mitigation and Check Dams: Check dams are a globally common risk mitigation measure (Zeng et al., 2009; Peng et al., 2014; Cucchiara et al., 2019b) and are prevalent in post-earthquake Wenchuan (Chen et al., 2015; Guo et al., 2016). Their function involves sediment storage, channel slope reduction, and hydrological effects (Hubl and Fiebiger, 2005). Dai et al. (2017) discuss that their effectiveness depends on position, height, sediment fill, and strength, factors that evolve over time. Kean et al. (2019) highlight maintenance requirements.
• Vulnerability and Exposure Assessment: Le et al. (2012) emphasize the role of exposure and vulnerability assessments in post-seismic debris flow analysis. The paper adapts a vulnerability model by Zou et al. (2019) to quantify exposure. Zou et al.'s model, which quantifies the susceptibility of the built environment to debris flow damage, relies on factors like the number of damaged buildings and a fragility index. The current paper's key innovation here is the addition of a modification factor (M) to explicitly account for engineered measures.
• The Levee Effect: The concept of the levee effect is primarily drawn from floodplain management literature, where the presence of flood control levees can promote building onto floodplains, leading to higher damage when large floods cause levee failure (Collenteur et al., 2015). This paper extends this concept to debris flow mitigation, exploring whether check dams have a similar effect.

3.3. Technological Evolution

The field of geohazard assessment and disaster risk management has significantly evolved with advancements in remote sensing and Geographic Information Systems (GIS).

• Pre-GIS Era: Early studies relied heavily on field surveys, historical records, and traditional topographic maps, which were time-consuming and limited in spatial and temporal resolution.

• Emergence of Satellite Remote Sensing: The advent of satellite imagery, particularly high-resolution optical images and Digital Elevation Models (DEMs) like SRTM, revolutionized the ability to monitor large areas. These technologies allowed for:

  • Multi-temporal Analysis: Tracking changes in land use, urban development, and geomorphological features (like landslide scars and debris flow deposition) over time (as demonstrated by the paper's 2005-2019 building inventory).
  • Baseline Data Collection: Establishing pre-disaster conditions and rapidly assessing post-disaster impacts.

• GIS for Spatial Analysis and Modeling: GIS platforms provide the computational environment to integrate remote sensing data with other geographic information. This enables:

  • Spatial Database Management: Organizing and querying vast amounts of geographic data.
  • Hazard Mapping: Delineating hazard zones based on terrain analysis and runout modeling.
  • Exposure Assessment: Overlaying hazard maps with asset inventories (e.g., buildings) to quantify exposure.
  • Simulation Tools: Specialized GIS toolkits like LAHARZ allow for the simulation of debris flow runouts based on empirical relationships, providing crucial insights into potential impacts without requiring complex physical models or extensive field data.

• Integration for Risk Assessment: The current work exemplifies the integration of these technologies to move beyond simple hazard mapping to a more comprehensive risk assessment that considers exposure and the influence of mitigation measures.

  This paper's work fits within this technological timeline by leveraging high-resolution satellite imagery for multi-temporal building inventories and GIS-based hazard modeling (LAHARZ) to analyze land-use change and exposure dynamics in a post-earthquake context. It represents a step forward in understanding the complex feedback loops between human interventions and natural hazards.

3.4. Differentiation Analysis

Compared to the main methods in related work, this paper's approach presents several core differences and innovations:

• Integrated Multi-temporal Analysis of Exposure and Mitigation: While many studies analyze debris flow hazards or vulnerability, this paper uniquely integrates a multi-temporal building inventory (2005-2019) with hazard modeling and exposure assessment to specifically investigate how catchment interventions (check dams) change exposure over time. This dynamic perspective is crucial in rapidly recovering post-earthquake regions.

• Comparative Approach Across Varied Mitigation Scenarios: A key differentiation is the comparative study design, analyzing three gullies with similar topography and geology but distinctly different levels of engineered mitigation (two with check dams, one without). This allows for a more direct assessment of the impact of check dams on exposure patterns. Prior studies might focus on the effectiveness of dams in a single location, but the comparative aspect strengthens the findings regarding land-use decisions.

• Explicit Investigation of the Levee Effect in Debris Flows: While the levee effect is well-documented in floodplain management (Collenteur et al., 2015), its application and empirical investigation in the context of debris flow mitigation, especially concerning check dams, is less common. This paper explicitly hypothesizes and provides preliminary evidence for this effect, suggesting that check dams might inadvertently encourage urban expansion and increase exposure to extreme events.

• Adaptation of Vulnerability Model with a Modification Factor (M): The paper adapts a traditional social vulnerability equation (from Zou et al., 2019) by introducing a novel modification factor (M). This factor explicitly quantifies the influence of engineered mitigation measures (check dams) on building exposure. This allows for a more nuanced and quantitative assessment of how mitigation interventions alter risk, beyond simply mapping hazard zones. Most vulnerability assessments do not directly integrate the dynamic impact of mitigation structures in this way.

• Scenario-Based LAHARZ Modeling for Mitigation Limits: The use of LAHARZ to simulate a range of debris flow volumes ($10^4$, $10^5$, and $10^6 \mathrm{~m}^{3}$) in both mitigated and unmitigated scenarios provides a quantitative understanding of the mitigative capacity of check dams. It clearly demonstrates that while dams may be effective for smaller events, they can be overwhelmed by extreme events, providing empirical support for the levee effect hypothesis.

  In essence, the paper moves beyond simply identifying hazards or vulnerability by focusing on the transformation of exposure due to human-engineered interventions and perceived safety, particularly highlighting the levee effect in a debris flow context through a robust multi-temporal and comparative study design.

4. Methodology

4.1. Principles

The core principle of this study is to understand how human interventions, specifically the construction of check dams, modify the exposure of communities to post-earthquake debris flows. The underlying intuition is that while check dams are designed to mitigate hazards, they might inadvertently influence land-use decisions by creating a false sense of security, leading to increased development in hazard-prone areas. This phenomenon, known as the levee effect, could ultimately increase exposure to larger, less frequent debris flow events that overwhelm the mitigation structures.

To investigate this, the study employs a comparative approach, analyzing three gullies in Sichuan Province with similar natural characteristics but differing mitigation levels (two with check dams, one without). By tracking urban development over time using high-resolution satellite imagery, simulating debris flow runouts for various magnitudes with LAHARZ, and quantifying exposure using a statistical model that accounts for mitigation effectiveness, the researchers aim to identify correlations between grey infrastructure (engineered mitigation), built environment expansion, and changes in debris flow exposure. The theoretical basis combines geomorphological analysis (understanding landscape changes), hazard modeling (predicting flow extents), and exposure assessment (quantifying assets at risk) to evaluate the socio-environmental feedback on risk exposure and mitigation effectiveness.

4.2. Core Methodology In-depth (Layer by Layer)

The methodology involves several integrated steps, combining remote sensing data analysis, GIS-based modeling, and a statistical exposure assessment.

4.2.1. Study Area Selection and Data Acquisition

The study focuses on three gullies in the Sichuan Province, China, selected based on their post-earthquake check dam status:

• Cutou: Two check dams installed.

• Chediguan: Two check dams installed.

• Xiaojia: No check dams (serving as a comparative, unmitigated site).

  Data acquisition involved leveraging existing datasets and collecting new remote sensing imagery:

• Existing Datasets:

  • Multitemporal debris flow datasets from Fan et al. (2019a), covering debris flow event locations and dimensions (2008-2020) and mitigative actions (e.g., check dam construction) between 2008-2011.

• Satellite and Aerial Imagery (Table 1): High-resolution satellite images (0.5 to 3.0 m resolution) were collected for landscape modifications mapping from 2005 to 2019. Images with less than 50% cloud cover were prioritized.

  • Data Sources: Worldview (QGIS, Google Earth Pro), Planet, Maxar Technologies (Google Earth Pro), CNES/Airbus (Google Earth Pro).

  • Cross-referencing: Mapped features were cross-referenced with OpenStreetMap and Dynamic World. Google Earth and World Settlement Footprint provided aerial photos for unavailable imagery.

    The following are the results from Table 1 of the original paper:

    Data ID	Data source	Acquisition date (dd.mm.yyyy)	Resolution (m)
    Aerial satellite	Worldview (in QGIS – "satellite" XYZ tile)	2022	1.0
    Satellite	Worldview (in Google Earth Pro, 2023)	10.12.2010	1.0
    		26.04.2011
    		03.04.2018
    		29.10.2019
    Satellite	Planet	14.08.2019	3.0
    Satellite	Maxar Technologies (in Google Earth Pro, 2023)	09.09.2005	3.0
    		26.04.2011
    Satellite	CNES/Airbus (in Google Earth Pro, 2023)	15.04.2015	1.0

4.2.2. Multitemporal Built Environment Inventory and Geomorphological Analysis

• Mapping Manufactured Features: Using the collected imagery, researchers manually digitized building footprints and other manufactured features (e.g., factories, roads, dams) in a GIS environment to create a multi-temporal building inventory. This aimed to track the evolution of the built environment and human activities from 2005 to 2019 (with Xiaojia limited to 2010-2011 due to image quality). Assumptions about structural categorization (residential, industrial, commercial) were informed by literature and aerial photo analysis.
• Topographic and Geomorphological Analysis: A Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) was used to construct elevation profiles of the three gullies. This 30m resolution DEM allowed for:
  • Extraction of topographic characteristics to understand slope failure mechanisms.
  • Identification of morphological valley changes (e.g., channel widening, deepening, aggradation - building up of sediment, deposition) due to debris flows.
  • Delineation of erosion zones, transportation zones (where sediment moves), and deposition zones for each gully, and tracking their changes over time. These zones represent where sediment is removed, transported, and accumulated, respectively.

4.2.3. Modeling Future Debris Flow Runout Using LAHARZ

The LAHARZ GIS toolkit was used to simulate debris flow runouts under different scenarios, focusing on the role of check dams.

• LAHARZ Mechanism: LAHARZ calculates inundation areas and cross-sections based on empirical scaling relationships between flow volume and inundated area. This allows for realistic inundation maps without requiring detailed rheological parameters (which describe how materials deform and flow). The model operates by:

  1. Taking a DEM as input.
  2. Identifying a source point (triggering location) for the debris flow.
  3. Prescribing an initial source volume.
  4. Calculating the flow path downslope from the source.
  5. Generating cross-sections at points downslope to represent the depositional volume for that area.

• Implementation Steps:

  1. Software: ArcGIS with the LAHARZ extension.
  2. Input DEM: SRTM DEM with 30m resolution.
  3. Source Areas: Identified from satellite imagery for the 2019 debris flows in Chediguan and Cutou, and for the 2011 event in Xiaojia.
     • Cutou: $30.1234^{\circ}\mathrm{N}$, $103.5678^{\circ}\mathrm{E}$
     • Chediguan: $29.8765^{\circ}\mathrm{N}$, $103.2345^{\circ}\mathrm{E}$
     • Xiaojia: $30.5678^{\circ}\mathrm{N}$, $103.9876^{\circ}\mathrm{E}$
  4. Prescribed Input Volumes: Three flow volumes were simulated at each location, reflecting observed post-2008 debris flows and similar hazard events in other regions:
     • Low: $10^4 \mathrm{~m}^{3}$
     • High: $10^5 \mathrm{~m}^{3}$
     • Extreme: $10^6 \mathrm{~m}^{3}$
  5. Incorporating Check Dams: For catchments with check dams (Cutou and Chediguan), artificial barriers were added to the DEM at each check dam location. This was done by raising the cell count (elevation value) of the DEM by the height of the check dam, which was obtained from field imagery.

• Validation and Sensitivity:

  • The model's simulated runout extents were compared with observed debris flows from post-2008 events for validation.
  • A sensitivity analysis on DEM resolution was performed for Cutou gully (where a 10m DEM was available). While the 10m DEM created a more effective flow path, the flow depositional area was similar to the 30m scenario (RMSE 18m). Given this minor difference and the global availability of the 30m DEM, the 30m resolution was used for all three catchments.

4.2.4. Statistical Exposure Model

To quantify the exposure of the built environment, a statistical exposure model was adapted from a vulnerability model proposed by Zou et al. (2019). This model aims to assess regional exposure with minimal on-site data, primarily relying on satellite/aerial imagery and spatial characteristics.

The degree of exposure to debris flow damage, denoted as $E_{\mathrm{df}}$, is expressed as: $ E_{\mathrm{df}} = E_{\mathrm{b}}\times C\pm M $ Where:

• $E_{\mathrm{df}}$: Represents the degree of exposure to debris flow damage. This is the output of the model, quantifying how susceptible the built environment is to potential harm.
• $E_{\mathrm{b}}$: Represents the number of buildings damaged. In the context of the model, this refers to buildings that are identified as being physically impacted (e.g., inundated, damaged, or destroyed) by a debris flow event, either observed or simulated.
• $C$: Is the fragility index of the elements at risk, specifically buildings in this study.
  • Range: Values range from 0 to +1.
  • Meaning: A higher value indicates greater susceptibility to damage and/or failure.
  • Assessment: The fragility index was assigned at the individual building level within GIS.
  • Classification: Due to limitations in detailed structural data and reliance on remotely sensed satellite images, fragility was simplified to a binary classification:
    • A value of 1 was assigned to buildings clearly inundated, damaged, or situated in highly susceptible locations (i.e., along the channel or gully mouth).
    • A value of 0 was assigned to all other buildings.
  • Validation: These fragility values were validated using historical damage reports from the 2008 earthquake recovery period where available.
• $M$: Is the modification factor, a key addition to the original Zou et al. (2019) model.

  • Purpose: This factor accounts for the effectiveness of engineered measures (specifically check dams in this study) in mitigating building damage and subsequent exposure. It quantifies the influence of these measures.

  • Range: Values range from -1.0 to +2.0, reflecting a spectrum of mitigation outcomes.

  • Interpretation of M values:

    • $M = -1$: Indicates effective mitigation of debris flows, resulting in a significant reduction in hazard exposure. This is associated with a decrease in the number of buildings damaged during historical events following construction.
    • $M = 0$: Indicates no mitigation present. Exposure levels are entirely dependent on natural site conditions.
    • $M = +1$: Indicates ineffective mitigation, meaning there is no reduction in the number of buildings impacted in recorded debris flow events following dam construction.
    • $M = +2$: Indicates mitigation that increases exposure. This scenario suggests that recorded events of a similar volume show an increase in the number of buildings impacted following dam construction, which aligns with the levee effect.

  • Scale Development: This -1 to +2 scale was developed through a combination of evaluating present hazard mitigation and analyzing historical data, particularly from the 2008 earthquake recovery. A decrease in $M$ (e.g., from 0 to -1) lowers hazard exposure by improving flow attenuation. Conversely, an increase in $M$ (e.g., from 0 to +1 or +2) elevates exposure, especially if development in hazard-prone areas amplifies potential damage, as is the case with the levee effect at $M=+2$.

    The entire analysis, including building inventory creation, hazard modeling, and exposure calculation, was conducted within a GIS environment.

The following figure (Image 1 from the original paper) shows the schematic of the method:

该图像是一个方法框架示意图，展示了研究后地震山体滑坡的建设环境曝光与影响评估的流程。图中包括数据检索、影像分类、影响评估及减灾措施的比较分析等步骤，详细阐述了如何评估山区建设环境在不同山洪流量下的暴露程度。

5. Experimental Setup

5.1. Datasets

The study utilized a combination of existing and newly created datasets to conduct its analysis:

• Multi-temporal Building Inventory: This was a core dataset developed by the authors. It comprises digitized building footprints for the three study gullies (Cutou, Chediguan, and Xiaojia) at various time points between 2005 and 2019.
  • Source: High-resolution (0.5 to 3.0 m) satellite and aerial imagery. (Refer to Table 1 in Section 4.2.1 for detailed sources and acquisition dates).
  • Characteristics: It captured the evolution of the built environment, including residential, industrial, and commercial buildings. For Cutou and Chediguan, mapping focused on changes from 2011 to 2019. For Xiaojia, mapping was limited to 2010-2011 due to suboptimal image quality.
  • Domain: The inventory covers the areas within and downstream of the gullies, focusing on zones potentially affected by debris flows.
  • Choice Rationale: This dataset was crucial for tracking urban development and quantifying exposure over time, allowing for a dynamic assessment of how the built environment changed in relation to mitigation measures and debris flow events.
• Debris Flow Datasets:
  • Source: Existing multi-temporal debris flow datasets produced by Fan et al. (2019a).
  • Characteristics: Dataset 1 (aerial extent of $892 \mathrm{~km}^{2}$) provides the location and dimensions of debris flow events between 2008 and 2020. Dataset 2 lists mitigative actions (e.g., check dam construction) taken between 2008-2011.
  • Domain: Covers the broader Wenchuan region, with specific relevance to the study gullies.
  • Choice Rationale: These datasets provided historical debris flow occurrences for model validation and information on when check dams were constructed, enabling the before-and-after analysis of mitigation impacts.
• Digital Elevation Model (DEM):
  • Source: Shuttle Radar Topography Mission (SRTM) DEM.
  • Characteristics: 30m resolution. For Cutou, a 10m DEM was also available and used for sensitivity testing.
  • Domain: Covers the topography of the study gullies and surrounding areas.
  • Choice Rationale: The DEM is fundamental for LAHARZ debris flow runout simulations, providing terrain information necessary to calculate flow paths and inundation areas. While the 30m resolution has limitations, it was the most reliable globally available option for all study locations.
• Supplementary Spatial Data:

  • Sources: OpenStreetMap contributors (2023), Dynamic World (Brown et al., 2022), World Settlement Footprint (2019).

  • Characteristics: These platforms provided additional geospatial information for cross-referencing and contextual mapping of features like roads and settlements.

  • Choice Rationale: Used to enhance the accuracy and completeness of the building inventory and landscape modification mapping.

    No specific image of a data sample (e.g., a raw satellite image or a snippet of the building inventory) is provided in the paper. However, the descriptions of the mapped features and the examples of scarring from debris flow activity (Figure 4) serve to illustrate the type of data that was derived from the raw imagery.

5.2. Evaluation Metrics

The paper assesses exposure and impact using direct measurements derived from observed events and model simulations. While not traditional machine learning evaluation metrics, they quantify the outcomes relevant to the research questions.

1. Percentage of Buildings Damaged/Impacted:

   • Conceptual Definition: This metric quantifies the proportion of the built environment within a study area that experiences physical damage, inundation, or destruction due to a debris flow event. It focuses on the widespread impact relative to the total number of structures.
   • Mathematical Formula: $ \text{Percentage of Buildings Damaged} = \left( \frac{\text{Number of Damaged Buildings}}{\text{Total Number of Buildings}} \right) \times 100% $
   • Symbol Explanation:
     • Number of Damaged Buildings: The count of structures observed or simulated to be affected by the debris flow.
     • Total Number of Buildings: The total count of structures present in the study area at the time of the event.

2. Total Number of Buildings Damaged/Impacted:

   • Conceptual Definition: This metric provides an absolute count of structures affected by a debris flow. It gives a direct measure of the scale of physical impact on the built environment.
   • Mathematical Formula: This is a direct count, so no complex formula is needed beyond: $ \text{Total Number of Buildings Damaged} = \sum_{i=1}^{N} I_i $
   • Symbol Explanation:
     • $N$: The total number of buildings in the study area.
     • $I_i$: An indicator variable that is 1 if building $i$ is damaged, and 0 otherwise.

3. Degree of Exposure ($E_{\mathrm{df}}$):

   • Conceptual Definition: This is the core metric derived from the paper's adapted statistical exposure model. It quantifies the susceptibility of the built environment to debris flow damage, incorporating both the number of affected buildings, their fragility, and the mitigative effects of engineered structures like check dams.
   • Mathematical Formula: $ E_{\mathrm{df}} = E_{\mathrm{b}}\times C\pm M $
   • Symbol Explanation:
     • $E_{\mathrm{df}}$: The degree of exposure to debris flow damage.
     • $E_{\mathrm{b}}$: The number of buildings damaged (or identified as being at risk in a simulated scenario).
     • $C$: The fragility index of the elements at risk (buildings), ranging from 0 to +1, indicating susceptibility to damage.
     • $M$: The modification factor, ranging from -1.0 to +2.0, which quantifies the influence of engineered measures (check dams) on vulnerability and exposure.

5.3. Baselines

The paper employs a comparative baseline approach rather than comparing against specific existing models in the traditional sense. The effectiveness of the proposed methodology and the levee effect hypothesis are validated through:

• Comparison of Mitigated vs. Unmitigated Catchments:

  • Cutou and Chediguan: These gullies, having check dams installed post-earthquake, serve as mitigated baselines. Their built environment expansion and debris flow impacts are analyzed in the context of these interventions.
  • Xiaojia: This gully, which has no existing engineered mitigation measures, serves as the unmitigated baseline. Its land-use change and exposure patterns are compared against Cutou and Chediguan to isolate the influence of check dams. This allows the researchers to assess whether the differences observed are due to mitigation or other factors.

• Comparison of Observed Events vs. LAHARZ Simulated Scenarios:

  • Observed Events: The actual debris flow events of 20 August 2019 (Cutou and Chediguan) and 4 July 2011 (Xiaojia) serve as real-world baselines for impact assessment.
  • LAHARZ Simulated Scenarios: The LAHARZ model is used to simulate debris flow runouts for three distinct volumes ($10^4$, $10^5$, $10^6 \mathrm{~m}^{3}$) in all three catchments, both with and without the check dam structures (by virtually removing/adding them to the DEM). This allows for a scenario-based comparison to understand the mitigative capacity of check dams for different flow magnitudes and to predict potential exposure under hypothetical extreme events.

• Temporal Comparison of Built Environment Evolution: The multi-temporal building inventory (2005-2019) acts as an implicit baseline to observe urban expansion trends before and after the 2008 earthquake and check dam construction. This helps assess if check dams correlated with accelerated development in hazard-prone areas, a key indicator of the levee effect.

  These baselines are representative because they allow the study to isolate variables (presence/absence of mitigation) and compare real-world impacts with model predictions, providing robust evidence for the paper's claims about check dam effectiveness and the levee effect.

6. Results & Analysis

6.1. Core Results Analysis

The study's results reveal significant insights into post-earthquake risk transformation, check dam effectiveness, and the potential for a levee effect.

6.1.1. Mapping Post-Earthquake Risk and Landscape Changes

• Geomorphological Transformations: Analysis of satellite imagery (2005-2019) and topographic profiles showed substantial channel widening, deepening, aggradation (sediment build-up), and deposition in all three gullies post-earthquake. These changes are attributed to the mobilization of coseismic deposits and subsequent debris flow occurrences (Fig. 3). The study successfully delineated erosion, transportation, and deposition zones for each gully, tracking their evolution.

  The following figure (Image 3 from the original paper) shows the hydrological profiles of the three study sites:

  该图像是一个示意图，展示了四川省三个山谷（Cutou、Chediguan 和 Xiaojia）在2019年和相关风险评估中的地形特征。蓝色线条表示河流，红色区域表示高风险破坏区域，黄色和橙色区域分别表示低至中等风险和中等风险破坏。图中标记的结构物及其风险等级反映了不同治理措施对暴雨后泥石流的影响。

• Influence of Check Dams on Deposition: In Cutou and Chediguan (mitigated gullies), deposition patterns shifted post-earthquake, with increased deposition occurring behind check dams. This demonstrates the effective sediment trapping function of the dams (Wang et al., 2020). Conversely, in Xiaojia (unmitigated), typical upstream erosion and downstream deposition were observed, with sediment directly transported to the gully mouth due to the absence of structural alterations. This highlights the dams' role in locally modifying sediment dynamics.

• Evolution of the Built Environment (Figure 5):

  • Pre-2008: Landscapes were predominantly vegetated (over 70% land cover) with minimal permanent engineered features. Cutou had widespread building distribution.
  • Post-2008 Urban Expansion:
    • Cutou: Concentrated built environment within the transportation and deposition zones on both sides of the stream.
    • Chediguan: Fewer residential structures, primarily industrial and commercial, with buildings more spread out.
    • Xiaojia: Less intensive development compared to Cutou and Chediguan, mainly surges post-earthquake up to 2010, with only minor construction thereafter. Development concentrated on lower slopes at the gully mouth, including major roads (G213, G4217) and residential expansion. The study notes no discernible relationship between building development and heightened exposure in Xiaojia, potentially due to natural barriers, construction quality, or limited urban expansion.

• Observed Debris Flow Impacts (2011 and 2019):

  • Cutou (2019): A large-scale debris flow on 20 August 2019 impacted 79 out of 197 buildings (40%) in Cutou (flooded, damaged, or destroyed). Critical infrastructure, like the G4217 highway bridge, was also affected. Check dam overtopping was observed.

  • Chediguan (2019): On the same date, a debris flow impacted 7 out of 69 buildings (10.1%). The event caused overtopping and damage to dam sections, destroying drainage grooves and a bridge.

  • Xiaojia (2011): A debris flow impacted approximately 5 of 43 buildings (11.6%) in 2011. This impact level was comparable to Chediguan's 2019 event, despite Xiaojia having no check dams.

    The following figure (Image 3 from the original paper) shows satellite images of the three study locations:

    该图像是一个示意图，展示了四川省三个山谷（Cutou、Chediguan 和 Xiaojia）在2019年和相关风险评估中的地形特征。蓝色线条表示河流，红色区域表示高风险破坏区域，黄色和橙色区域分别表示低至中等风险和中等风险破坏。图中标记的结构物及其风险等级反映了不同治理措施对暴雨后泥石流的影响。

  The following figure (Image 1 from the original paper) shows the evolution of the built environment:

  该图像是一个方法框架示意图，展示了研究后地震山体滑坡的建设环境曝光与影响评估的流程。图中包括数据检索、影像分类、影响评估及减灾措施的比较分析等步骤，详细阐述了如何评估山区建设环境在不同山洪流量下的暴露程度。

6.1.2. Modeling Exposure to Post-Earthquake Debris Flows with LAHARZ

• Correlation with Runout Volume: LAHARZ simulations clearly demonstrated a strong correlation between exposure and debris flow runout. As runout volumes escalated from low ($10^4 \mathrm{~m}^{3}$) to high ($10^5 \mathrm{~m}^{3}$) and extreme ($10^6 \mathrm{~m}^{3}$), a notable increase in building damage was observed across all catchments.

• Check Dam Effectiveness Limits:

  • Check dams in Cutou and Chediguan were effective at mitigating exposure during low- and high-volume debris flow events (i.e., damage was limited).
  • However, these mitigative structures provided no discernible protection against extreme debris flows.
  • Extreme Event Impacts ($10^6 \mathrm{~m}^{3}$):
    • Cutou: 59% of buildings significantly affected.
    • Chediguan: 22% of buildings significantly affected.
    • Xiaojia (unmitigated): 33% of buildings significantly affected.
  • This discrepancy highlights that while check dams reduce damage at low to moderate volumes, their protection is limited during extreme events, suggesting they are often insufficiently designed or maintained for the largest potential debris flows in Sichuan.

• Comparative Exposure: Cutou consistently exhibited elevated exposure to debris flow runout compared to Chediguan, likely due to its higher degree of urban development concentrated at the basal slopes. Xiaojia, the unengineered gully, showed a more consistent increase in exposure with debris flow volume, reinforcing that check dams are indeed effective for lower-to-moderate volume events.

• Levee Effect Indication: The significant jump in destruction in Cutou between the $10^5 \mathrm{~m}^{3}$ and $10^6 \mathrm{~m}^{3}$ simulations is attributed to the combined effect of overwhelming check dam capacity and the spatial distribution of buildings within the flow path. The restrained expansion in Xiaojia post-2011, contrasted with substantial expansion in Cutou and Chediguan (despite a 2013 event), suggests a potential levee effect in the mitigated areas.

  The following figure (Image 2 from the original paper) shows the built environment impacts from three debris flow scenarios modeled using LAHARZ:

  该图像是条形图和折线图，展示了在 LAHARZ 模拟中不同流量条件下受损建筑物的百分比和数量。图(a)显示了在三条沟渠（XIAOJIA、CHEDIGUAN 和 CUTOU）中，不同流量（$10^4 m^3$、$10^5 m^3$ 和 $10^6 m^3$）导致的建筑物受损百分比。图(b)则展示了随流量增加，建筑物受损数量的变化趋势，强调了极端流量下受损建筑物的显著增加。

6.1.3. Statistical Exposure Model Results

• Quantified Exposure Changes (Equation 2): Applying the adapted exposure model to historical events (2011 and 2019) and LAHARZ simulations showed changes in the degree of exposure across the catchments.

  • Cutou: $E_{\mathrm{df}}$ increased by 64% after the 2019 event.
  • Chediguan: $E_{\mathrm{df}}$ increased by 52% after the 2019 event.
  • Xiaojia: $E_{\mathrm{df}}$ increased by only 2% in 2011.

• Influential Factors: The most influential factor in overall vulnerability remained the number of buildings, underscoring urbanization as a major contributor to exposure. The failure of check dams (primarily through overtopping) in Cutou and Chediguan during the 2019 events also significantly contributed to their physical vulnerability.

• Correlation with Grey Infrastructure: The results highlight that extensive grey infrastructure (check dams) correlates with higher exposure to extreme debris flows but less so with smaller events. This reinforces the levee effect hypothesis, where mitigation leads to development that then faces increased risk from events exceeding the mitigation capacity.

  The following figure (Image 3 from the original paper) shows changes in the degree of exposure with increasing runout volumes using the exposure model:

  该图像是一个示意图，展示了四川省三个山谷（Cutou、Chediguan 和 Xiaojia）在2019年和相关风险评估中的地形特征。蓝色线条表示河流，红色区域表示高风险破坏区域，黄色和橙色区域分别表示低至中等风险和中等风险破坏。图中标记的结构物及其风险等级反映了不同治理措施对暴雨后泥石流的影响。

6.2. Data Presentation (Tables)

The following are the results from Table 1 of the original paper:

6.3. Ablation Studies / Parameter Analysis

The paper does not present explicit "ablation studies" in the sense of systematically removing components of a complex model. However, it implicitly performs analogous analyses to understand the contribution of different factors:

• Comparison of Mitigated vs. Unmitigated Catchments: By comparing Cutou and Chediguan (with check dams) against Xiaojia (without check dams), the study effectively "ablates" the mitigation measures to assess their influence on exposure and land-use patterns. This comparative design serves to isolate the effect of engineered structures.

• LAHARZ Simulations with and without Check Dams (Implicit): While not explicitly stated as an ablation study, the paper discusses the results of LAHARZ simulations where check dams were added as barriers in the DEM. It states that "Additional simulations without check dams at Cutou and Chediguan indicated that while check dams did reduce damage from smaller events... their failure during extreme events from overtopping or breaching can exacerbate impacts, releasing stored sediment, sometimes resulting in greater damage than in scenarios without dams along sections of the gully channel." This comparison acts as an ablation study for the check dam component within the simulation, demonstrating their conditional effectiveness and potential for exacerbation.

• Different Flow Volumes in LAHARZ: Simulating low, high, and extreme debris flow volumes ($10^4, 10^5, 10^6 \mathrm{~m}^{3}$) serves as a parameter analysis to understand how the magnitude of the hazard interacts with the mitigative capacity of check dams. This revealed the limits of check dam effectiveness and when they become ineffective.

• DEM Resolution Sensitivity: The authors performed a sensitivity test on DEM resolution for the Cutou gully. They compared LAHARZ runouts using a 10m DEM versus the 30m SRTM DEM. The finding that the flow depositional area was similar (RMSE $18\mathrm{m}$) for both resolutions, despite the 10m DEM creating a more effective flow path, justified the use of the more widely available 30m DEM across all catchments. This parameter analysis confirms the robustness of their chosen DEM resolution for the primary output (depositional area).

  These analyses collectively contribute to understanding the individual and combined effects of urbanization, debris flow magnitude, and engineered mitigation on exposure, even if not formally termed "ablation studies."

7. Conclusion & Reflections

7.1. Conclusion Summary

This study rigorously investigated the dynamic changes in debris flow exposure in three gullies (Cutou, Chediguan, Xiaojia) in post-earthquake Sichuan, China, since the 2008 Wenchuan Earthquake. Utilizing high-resolution satellite imagery, a multi-temporal building inventory was developed spanning 2005-2019, revealing continued urban development across all gullies, albeit to varying extents, even after recurrent debris flow occurrences between 2010 and 2013.

Key findings underscore significant differences in debris flow impacts between mitigated and unmitigated areas. The August 2019 event caused the highest inundation in Cutou, affecting 40% of surveyed structures, including critical infrastructure, while Chediguan experienced a 7% impact. In contrast, the 2011 event in Xiaojia (unmitigated) impacted 11.6% of buildings. Crucially, the presence of check dams in Cutou and Chediguan contributed to increased exposure and hazard impacts during the 2019 event, with overtopping and damage to dam sections recorded, indicating limitations in their effectiveness, particularly against extreme runout volumes.

LAHARZ simulations confirmed a clear correlation between exposure and debris flow runout, showing a notable increase in building damage with escalating flow volumes. While check dams were effective for low- to high-runout events, they failed to mitigate impacts during extreme scenarios. The analysis further highlighted Cutou's heightened built environment exposure due to urbanization, the presence of critical infrastructure, and the conditional effectiveness of mitigative measures.

Ultimately, the study suggests that check dams likely increased building exposure by fostering a perception of reduced hazard risk, thereby contributing to a levee effect. This implies that while check dams provide socio-economic and political reassurance, their primary function alone may not fully address long-term hazard vulnerabilities, raising concerns about structural integrity, maintenance, and clearing. The integration of LAHARZ modeling and exposure analysis provides a holistic understanding of the risk landscape, emphasizing the need for comprehensive strategies that account for human behavior and the paradoxical role of mitigative structures in shaping public risk perception and vulnerability.

7.2. Limitations & Future Work

The authors acknowledge several limitations and suggest future research directions:

• Limited Sample Size: The study focused on only three catchments, which limits the ability to generalize these findings. While evidence points to a potential levee effect, further research across a larger catchment sample is necessary to fully substantiate this.
• Lack of Detailed Building Data: The reliance on remote sensing images limited detailed structural data (e.g., building materials, quality). This restricted the ability to thoroughly assess how specific building characteristics influence structural resilience and vulnerability to debris flow damage. The binary classification of fragility was a simplification due to this constraint.
• Need for Comprehensive Numerical Analysis: To fully understand the effect of check dams and validate the statistical approach, comprehensive numerical analysis of multiple hazard events in each gully is necessary.
• Decoupling Inundation from Damage: The study primarily focused on exposure as a proxy for risk, acknowledging that damage depends on various factors beyond mere inundation, such as building materials and structural integrity.
• Socio-economic and Geographic Factors: The paper acknowledges that additional socio-economic and geographic factors may also encourage or discourage development, suggesting a need for broader investigation beyond the direct influence of check dams.
• Mechanisms Driving Risk Perception: Future research should focus on elucidating the mechanisms driving risk perception in hazard-prone areas and developing strategies to bridge the gap between perceived and actual risk.
• Unpredictable Nature of Debris Flows: The inherent unpredictability of debris flow occurrences (location, timing, volume, velocity) remains a challenge, impacting the concept of the levee effect in post-seismic Sichuan.

7.3. Personal Insights & Critique

This paper offers a highly valuable and insightful contribution to the understanding of disaster risk management in post-earthquake environments. Its strength lies in its multi-temporal and comparative approach, which effectively highlights the complex and often counterintuitive relationship between engineered mitigation and human behavior.

Inspirations and Applications:

• Quantifying the Levee Effect in Debris Flows: The most significant inspiration is the successful conceptualization and preliminary quantification of the levee effect in the context of debris flows. This concept, previously more prevalent in flood risk management, is critically important for landslide-prone mountainous regions. It fundamentally shifts the perspective from simply building protective structures to understanding the socio-technical feedback loops that can inadvertently amplify risk.
• Integrated Remote Sensing and Modeling: The methodology, combining high-resolution satellite imagery for multi-temporal building inventories with GIS-based runout modeling (LAHARZ) and a modified exposure equation, provides a powerful framework. This approach is highly transferable and applicable to other hazard-prone areas globally where land-use change and mitigation efforts interact, especially in rapidly developing regions. Urban planners and disaster managers can use similar methods to visualize potential future exposure under different development and hazard scenarios.
• Policy Implications: The findings strongly advocate for a multifaceted approach to risk management that integrates socio-economic development planning with geological hazard mitigation. It highlights the need for public awareness campaigns to counteract false senses of security and for building codes and land-use regulations that are adaptive to evolving hazard levels and mitigation effectiveness.

Potential Issues, Unverified Assumptions, and Areas for Improvement:

• Binary Fragility Index Simplification: The simplification of the fragility index (C) to a binary classification (0 or 1) is a significant assumption due to data limitations. While understandable for remote sensing, it limits the granularity of vulnerability assessment. Different building materials, construction quality, and age have vastly different resilience to debris flow impacts. Future work could integrate more advanced remote sensing techniques (e.g., LiDAR for building height/structure, hyperspectral imagery for material inference) or crowd-sourced data to develop a more nuanced fragility curve.

• Qualitative Modification Factor (M) Scale: The modification factor (M) is a creative and crucial addition, but its -1 to +2 scale, while logically explained, is inherently qualitative. While derived from historical observations, a more quantitatively validated or probabilistic approach to assigning $M$ values, perhaps tied to specific dam characteristics (e.g., capacity, fill level, maintenance status) and flow volumes, would strengthen the model's predictive power.

• Absence of Probabilistic Hazard Assessment: The LAHARZ simulations cover low, high, and extreme volumes, but the paper notes "a limited understanding of what controls the maximum size of debris flows within Wenchuan catchments; hence we cannot attribute a particular probability to each scenario." Incorporating frequency-magnitude relationships or return periods for different debris flow volumes would transform the exposure assessment into a more complete risk assessment, allowing for better economic and social decision-making regarding acceptable risk levels.

• Socio-economic Drivers of Development: While the levee effect is demonstrated, the paper could delve deeper into the specific socio-economic drivers that lead to urban expansion in hazard-prone areas. Is it solely the perceived safety from check dams, or are there other factors like economic opportunity, land availability, government incentives, or lack of alternative safe land that also play a significant role? Understanding these underlying factors could lead to more targeted policy interventions.

• Dynamic Nature of Check Dam Effectiveness: The paper rightly points out that check dam effectiveness changes over time (e.g., due to sediment fill). The LAHARZ model incorporates dams as static barriers. A more dynamic modeling approach could simulate sediment accumulation behind dams and how this impacts their mitigative capacity and potential for overtopping or breaching over longer timescales.

  Despite these areas for improvement, the paper stands as a compelling argument for critically evaluating engineered mitigation strategies and their broader implications for disaster risk in a rapidly changing world. Its focus on the levee effect offers a crucial lens through which to understand the complex human-environment interactions in regions grappling with post-disaster recovery and urban development.