A Novel Distributed Model for Predicting Runoff-Induced Multi-Instability Risk Along Highway Corridors Under Heavy Rainfall

Comprehensive analysis of rainfall-induced landslides using a multi-field coupling discrete element model

An Unsaturated Soil Material Point Method for Simulating Rainfall-Induced Landslides

Rainfall infiltration frequently triggers slope failures by elevating pore water pressure and compromising the shear strength of unsaturated soil layers. Modeling these severe geological events remains computationally challenging. Standard grid-based techniques, such as the Finite Element Method (FEM), typically fail due to severe mesh distortion under large deformations, whereas the Discrete Element Method (DEM) demands excessive computational resources. Addressing this gap, we develop an advanced theoretical framework utilizing the Material Point Method (MPM) integrated with a liquid-solid-gas three-phase mechanics model. We first verify the algorithmic accuracy using a 1D unsaturated soil column test. Subsequently, the framework is deployed to capture the dynamic displacement and mechanical responses of a 2D rainfall-induced landslide. Benchmarking against FEM data confirms that our multiphase MPM accurately models the infiltration process and subsequent structural collapse. Ultimately, this approach offers a highly robust computational strategy for analyzing large-scale landslide deformations and improving predictive assessments.

Rapid Assessment of Waterlogging Risk Induced by Typhoon Rainfall: A Case Study of Typhoon Haikui (2311) in South China

Typhoon rainfall can trigger urban waterlogging within hours, but hydrodynamic modelling is impractical for rapid screening. For Typhoon Haikui (2311) across Guangdong, Guangxi, and Hainan, we integrated half-hourly Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG) Final Run V07 precipitation, LandScan population exposure (E), and an impervious-surface factor (I) derived from the China Land Cover Dataset into a Typhoon Rainfall Risk Index (TRRI). Rainfall hazard (H) combines event-total rainfall, rainfall duration, and maximum 3 h rainfall. Hazard clustered along the Guangdong coast, where maximum event-total and 24 h rainfall reached 529.7 and 272.8 mm, respectively. TRRI high-risk zones occupied 12.45% of the three-province land area. Defined by the largest weighted component, H-, E-, and I-dominated cells accounted for 60.5%, 33.4%, and 6.1% of the high-risk class. Of 22 reported waterlogging locations in the event-focused domain, 19 fell in high-risk zones; the 86.36% hit rate versus a 26.09% background area yielded an enrichment ratio of 3.31 (p < 0.001). In a descriptive full-domain check, all 49 distinct locations were medium or high. Relative to H alone, TRRI reclassified 20.59% of land upward and 20.64% downward, while 58.77% remained unchanged. TRRI supports rapid event-scale screening and risk prioritization for targeted disaster-prevention decision-making.

Flood Hazard Assessment using Rainfall Data (1995-2025) with a Hydrograph Approach and an Eco-Geomorphic Overview in Padang City, Indonesia.

Padang City experiences recurrent flooding because intense rainfall, rapid runoff from eastern catchments, low-lying floodplains, and coastal-estuarine backwater effects interact across a developed watershed continuum. Existing assessments often examine rainfall, runoff response, spatial hazard, and ecological functions separately, limiting integrated flood-risk reduction and land-use planning. This study aims to 1) analyze areal rainfall frequency for 1995-2025 using the Annual Maximum Series of 24-hour Rainfall (R24/AMS), the Log-Pearson Type III (LP-III) distribution, and the Kolmogorov-Smirnov (K-S) goodness-of-fit test; 2) generate a preliminary regional-scale runoff hydrograph using the Nakayasu Synthetic Unit Hydrograph (SUH) to characterize event magnitude and response timing under design rainfall; and 3) map flood-hazard zones using GIS-based weighted overlay scoring that integrates hydrological, geomorphological, and ecological parameters. Rainfall records indicate interannual variability and repeated ≥150 mm/24 h events, including 447 mm on 07/03/2024 and 261 mm on 25/11/2025. Design 24-hour rainfall increases from 160.87 to 334.14 mm across 2-50-year return periods, producing 4-hour intensities of 22.13-45.97 mm/h. The preliminary Nakayasu scenario yields peak discharges of approximately 135-250 m³/s with short times to peak and should be interpreted as screening evidence rather than basin-specific engineering estimates. The flood-hazard map classifies 537.97 km² (77.41%) as Safe, 96.74 km² (13.92%) as Low, 36.76 km² (5.29%) as Moderate, and 23.49 km² (3.38%) as High. Moderate- and high-hazard areas cluster in fluvial-marine lowlands such as floodplains, backswamps, coastal alluvium, river corridors, and estuaries. These findings provide a basis for Ecosystem-based Disaster Risk Reduction (Eco-DRR) through retention-space protection, river setbacks, wetland restoration, upland recharge conservation, and erosion-sediment control.

A Coupled Hydrodynamic–Optimization Framework for Flood Prediction and Green Gray Stormwater Management: Multi Objective Trade Offs in a Tropical Catchment

Tropical coastal cities in Southeast Asia are highly vulnerable to pluvial and fluvial flooding, which is exacerbated by rapid upland land use change and a reduction in natural water retention. This study develops and applies a novel decision support tool that couples a JAX accelerated diffusive wave solver for shallow water equation (SWE) with a multi objective optimization framework inspired by the Non dominated Sorting Genetic Algorithm II (NSGA II). The model is designed to identify the optimal location and sizing of upland retention basins, balancing two competing objectives: minimizing peak flood depth and maximizing groundwater recharge. Applied to a synthetically realistic representation of Roxas City, Philippines, the framework reveals a clear “mitigation ceiling” under extreme storm events. For the 150 mm/h 1 in 50 year event, the peak flood depth stabilizes at 0.589 m, irrespective of the specific Pareto optimal configuration, indicating that the current basin network has reached its volumetric limit for peak reduction. Nevertheless, the “Max Recharge” strategy achieves the same flood safety as the “Min Depth” strategy while capturing an additional ~1,000 m³ of water (~20% more), demonstrating that it is the superior all round solution for both flood mitigation and water security. The proposed methodology known as SHOMS has been piloted as a project known as Hydrointelligence System (HIS) in Roxas City; where a transferable, physics based, multi objective planning tool that supports local government units in the Philippines in designing nature based stormwater management interventions is being implemented and such can be replicated across the world.

Multi-hazard modelling of hydro-geomorphological events for early warning systems. Application to the Val d’Aran (Central Pyrenees, Spain)

(English) Hydro-geomorphological hazards such as floods and rainfall-induced landslides frequently co-occur in mountainous regions, driven by shared meteorological forcing but governed by distinct hydrological and geotechnical mechanisms. Improving their anticipation requires modelling frameworks capable of consistently representing subsurface hydrological conditioning and its interaction with transient triggering processes. This thesis develops and evaluates a physically based, forecast-driven multi-hazard modelling framework for early warning applications in the Upper Garonne River Basin (Val d’Aran, Central Pyrenees, Spain). The research progresses through three interconnected stages. First, an event-based slope stability model is enhanced through a novel multi-objective calibration strategy that explicitly represents both antecedent and final slope stability conditions. Results demonstrate that inadequate representation of subsurface hydrological states leads to physically inconsistent pre-event instability, highlighting the sensitivity of slope response to antecedent moisture conditions. Building on this finding, a continuous distributed hydrological–geotechnical modelling system (SCLAM) is developed to dynamically simulate snowmelt, rainfall–runoff processes, groundwater redistribution, and slope stability within a unified framework. By replacing externally imposed antecedent conditions with internally simulated hydrological states, the model improves physical consistency and identifies Baseflow Excess as a main indicator of destabilizing subsurface conditions. The coupled framework is subsequently extended into a prototype multi-hazard early warning system integrating flood and landslide forecasting under bias-corrected ensemble meteorological forcings. The system operates at 30 m spatial resolution over more than one million grid cells and completes a full forecast-driven multi-hazard simulation cycle in approximately one minute, demonstrating its operational feasibility under real-time constraints. Results show that flood hazard responds primarily to short-term precipitation intensity, whereas landslide hazard is governed by cumulative subsurface conditioning. The asymmetric yet conditionally coupled behaviour of both hazards underscores the necessity of a unified, state-based modelling approach. Overall, this research demonstrates that physically consistent and computationally efficient multi-hazard modelling is feasible for operational early warning in mountainous regions, providing a transferable framework that links improved subsurface representation, hydro-geotechnical coupling, and forecast-driven implementation. (Català) Els perills hidrogeomorfològics, com les inundacions i les esllavissades desencadenades per la pluja, ocorren sovint en regions muntanyoses, i estan impulsats per un mateix esdeveniment meteorològic però es regeixen per mecanismes hidrològics i geotècnics diferents. La seva anticipació requereix marcs de modelització capaços de representar de manera consistent el condicionament hidrològic del subsol i la seva interacció amb processos d’activació transitoris. Aquesta tesi desenvolupa i avalua un marc de modelització multiperill, basat en models físics i orientat a la predicció, per a la seva aplicació en un sistema d’alerta a la capçalera de la conca del riu Garona (Val d’Aran, Pirineus Centrals, Espanya). La recerca s’estructura en tres etapes. En primer lloc, es millora el model d’estabilitat de vessants basat en esdeveniments mitjançant una estratègia de calibració multiobjectiu que representa explícitament les condicions prèvies i finals d’estabilitat. Els resultats mostren que una representació inadequada dels estats hidrològics del subsol condueix a inestabilitats que són físicament inconsistents abans de l’episodi de pluja desencadenant, posant en evidència la sensibilitat a la humitat antecedent. A partir d’aquest resultat, es desenvolupa un sistema continu i distribuït de modelització hidrològica–geotècnica (SCLAM) que simula el desgel, l’escorrentiu, la redistribució de les aigües subterrànies i l’estabilitat de vessants en un marc unificat. En substituir les condicions antecedents imposades per estats hidrològics simulats, el model millora la coherència física i identifica l'excés de cabal de base com a indicador integrador de les condicions desestabilitzants del subsòl. El marc s’amplia a un prototip de sistema d’alerta primerenca multiperill que integra la predicció d’inundacions i esllavissaments sota esdeveniments meteorològics en conjunt corregits per biaix. El sistema opera a 30 m de resolució sobre més d’un milió de cel·les i completa un cicle de simulació en aproximadament un minut, demostrant la seva viabilitat operativa. Els resultats indiquen que el perill d’inundació respon principalment a la intensitat de la precipitació a curt termini, mentre que el d’esllavissades depèn del condicionant acumulat del subsòl. Aquest comportament asimètric, però acoblat, reforça la necessitat d’un enfocament unificat i basat en estats hidrològics subsuperficials. En conjunt, la tesi demostra que la modelització multiperill és consistent amb la física, computacionalment eficient i és viable per a la seva implementació en sistemes d’alerta de manera operativa en regions muntanyoses, oferint un marc transferible que integra una representació millorada de l’estat hidrològic subsuperficial, l’acoblament hidrogeotècnic i la incorporació de forçaments de predicció meteorològica. (Español) Los peligros hidrogeomorfológicos, como las inundaciones y los deslizamientos inducidos por la lluvia, coocurren con frecuencia en regiones montañosas, impulsados por un forzamiento meteorológico común pero regidos por mecanismos hidrológicos y geotécnicos distintos. Su anticipación requiere marcos de modelización capaces de representar de forma consistente el estado hidrológico del subsuelo y su interacción con los procesos transitorios de su desencadenamiento. Esta tesis desarrolla y evalúa un marco de modelización multi-peligro, físicamente basado y orientado a pronóstico, para aplicaciones de alerta temprana en la cuenca alta del río Garona (Val d’Aran, Pirineos Centrales, España). La investigación se estructura en tres etapas. En primer lugar, se mejora un modelo de estabilidad de laderas basado en eventos mediante una estrategia de calibración multiobjetivo que representa explícitamente las condiciones previas y finales de estabilidad. Los resultados muestran que una representación inadecuada de los estados hidrológicos del subsuelo conduce a la ocurrencia de inestabilidades físicamente inconsistentes antes del evento, evidenciando la sensibilidad a la humedad antecedente. A partir de este resultado, se desarrolla un sistema continuo y distribuido de modelización hidrológico–geotécnica (SCLAM) que simula el deshielo, la escorrentía, la redistribución de aguas subterráneas y la estabilidad de laderas en un marco unificado. Al sustituir las condiciones antecedentes impuestas por estados hidrológicos simulados, el modelo mejora la coherencia física e identifica el exceso de flujo base como principal indicador de condiciones desestabilizadoras del subsuelo. El marco se amplía a un prototipo de sistema de alerta temprana multi-peligro que integra la predicción de inundaciones y deslizamientos a partir de forzamientos meteorológicos corregidos por sesgo. El sistema opera a 30 m de resolución sobre más de un millón de celdas y completa un ciclo de simulación en aproximadamente un minuto, demostrando su viabilidad operativa. Los resultados indican que el peligro de inundación responde principalmente a la intensidad de la precipitación a corto plazo, mientras que el de deslizamientos depende de los estados acumulados antecedente del subsuelo. Este comportamiento asimétrico, pero acoplado, refuerza la necesidad de un enfoque unificado y basado en los estados subsuperficiales. En conjunto, la tesis demuestra que la modelización multi-peligro físicamente consistente y computacionalmente eficiente es viable para la alerta temprana operativa en regiones montañosas, ofreciendo un marco transferible que integra la representación mejorada del estado hidrológico subsuperficial, el acoplamiento hidro-geotécnico y la implementación de forzantes de pronóstico meteorológico.