A Tsunami in Lisbon Where to run?
Daniel André Silva Conde
Dissertação para obtenção do Grau de Mestre em Engenharia Civil
Júri
Presidente: Prof. Doutor António Jorge Silva Guerreiro Monteiro Orientador: Prof. Doutor Rui Miguel Lage Ferreira Co-Orientador: Prof. Doutor Carlos Alberto Ferreira de Sousa Oliveira Vogal: Prof. Doutora Maria Ana Viana Baptista Vogal: Engenheira Maria João Martins Telhado
Outubro 2012
To my parents, Armando e Guilhermina. Thank you. Aos meus pais, Armando e Guilhermina. Obrigado.
Acknowledgments
This dissertation was developed at Instituto Superior Técnico - Technical University of Lisbon, Portugal, under the guidance of Professors Rui M. L. Ferrira and C. Sousa Oliveira. During this time, part of the work was supported by CEHIDRO - Centro de Estudos de Hidrossistemas - with two research initiation scholarships, also under the guidance of Professor Rui M. L. Ferreira.
For the great opportunities that were given to me with no hesitation and for all the support, knowledge and motivation transmitted throughout this year, my sincere gratefulness to Professor Rui M. L. Ferreira. And I really do mean sincere.
My many thanks to Ricardo for all the support and valuable help with STAV-2D. And of course to Edgar, for keeping the morale high when odds weren’t.
Also, to Professora Maria Ana Baptista, my sincere thanks for all the kindness and availability. Without her contribution this work wouldn’t be possible. To Professor C. Sousa Oliveira for the thoughtful insights about various topics on seismicity and tsunamis.
To my dear colleagues, specially André Marques and Pedro Pinotes, for making those hardworking hours as enjoyable as they could have been. Also to my fellow colleagues and friends at the Residência Engo Duarte Pacheco for providing the most hilarious 5 years that one could have had.
My non-translatable-to-words gratitude to my parents. To my mother, Mira, and my father, Armando, for the constant support, guidance and love throughout my entire existence.
To my grandfather, Joaquim, for the sensitive human being that he is and inspires to be. To my grandmother, Celeste, for being the craziest grandmother on this earth. And I mean this as the truest compliment. To my grandmother, Alcínia, for all the support and welcoming affections every single time she laid eyes on me.
To my godparents, Fátima and Sérgio, for all the support, affection and enjoyable moments that we have had together so far. Also, my affectionate gratitude to my older "godfather", Sérgio, whose memory will last forever.
A very special place for the younger family members. To Zi and Lino for having the most inspiring relation that can be seen between two brothers. Everyone feels better if you’re around. Also, to the rising star in the game (football!), Miguel for the sincere feelings and affections. To Catarina for the special
i connection we share. No matter how long or how far I am away from all of you, I will never cease to have four brothers.
And that’s about it. Oh wait... (ahah!) my loving gratitude to my girlfriend, Dora, for making the last 5 years probably the best ones that I have had so far. Thank you for fulfilling a long lasting void.
It would take forever to thank everyone I owe some part of myself or my life. To all those I have not mentioned, my apologies and sincere thanks.
ii Resumo
Uma revisão recente do histórico de tsunamis em Portugal mostrou que o estuário do Tejo foi afetado por vários tsunamis catastróficos ao longo dos últimos dois milénios. Dada esta herança histórica, e a crescente consciencialização devida a fenómenos recentes com mediatização mundial, o presente trabalho procura fornecer informação relevante quanto à exposição das zonas ribeirinhas do estuário do Tejo a um tsunami semelhante ao ocorrido em 1 de Novembro de 1755.
Novas abordagens, teóricas e metodológicas, foram usadas para modelar um cenário de Tsunami, com especial enfase para a propagação em terra. Um dos aspetos mais relevantes desta fase de um tsunami é a sua capacidade de incorporar detritos, sejam eles de origem natural, como sedimentos ou vegetação, ou artificial, como ruínas de infraestrutura construída ou outros equipamentos do quotidiano humano.
Um modelo conceptual específico para escoamentos deste tipo - debris flows - foi revisto e ampliado com um flexível esquema de condições de fronteira. O modelo numérico contempla as mais recentes técnicas de discretização matemática e uma implementação computacional eficiente. Os seus fundamentos teóricos são apresentados a um nível introdutório enquanto que as novas funcionalidades são exaustivamente explicadas, demonstradas e discutidas.
Um esforço muito considerável foi direcionado para a discretização da cidade de Lisboa. O ambiente con- struído deixa de ser um simples pano de fundo para visualizar resultados e passa a definir, explicitamente, a superfície do terreno. Todas as estruturas relevantes foram adequadamente modeladas com modelos digitais de terreno e malhas computacionais muitíssimo finas para a escala do problema em mão.
Os resultados são apresentados em vários níveis diferentes, desde perspetivas abrangentes a análises detalhadas ao nível de cada arruamento. Nestas últimas, é fornecida informação pormenorizada quanto aos tempos de chegada do tsunami, as suas propriedades hidrodinâmicas e os padrões de deposição de detritos subsequentes à inundação.
Este estudo demonstrou que vários locais na frente ribeirinha do estuário do Tejo são especialmente preocupantes no que a um cenário de Tsunami diz respeito. Este documento fornece informação relevante para apoiar o planeamento das respostas de socorro e eventual evacuação da cidade, constituindo mais um passo na promoção da segurança dos cidadãos de Lisboa.
Palavras-Chave: Tsunamis, modelação matemática, estuário do Tejo, Lisboa
iii iv Abstract
A recent revision of the catalogue of tsunamis in Portugal has shown that the Tagus estuary has been affected by catastrophic tsunamis numerous times over the past two millennia. Provided this historical heritage, and the growing awareness due to recent events worldwide, the present work aims at providing relevant data on the exposure of the Tagus estuary waterfront to a tsunami similar to the one occurred on the 1st November of 1755.
New theoretical and methodological approaches to tsunami modeling were employed, with special empha- sis given to propagation over dry land. One of the most relevant features of this stage of the tsunami is its ability to incorporate debris, either natural sediment incorporated from the bottom boundary or remains of human built environment. The increased mass and momentum of the run-up can inflict tremendous damage and, regrettably, severe human losses.
A conceptual model specifically suited for describing debris flows is reviewed and extended with flexible boundary conditions. The employed numerical tool features the most up to date mathematical dis- cretization techniques and an efficient computational performance. The theoretical fundamentals of the numerical model are presented at an introductory level while the new features implemented are given a thorough explanation and discussion.
A very considerable effort was directed to Lisbon’s topology discretization. The built environment is no longer seen as a simple background layer for result visualization, it actually explicitly defines the terrain surface. All relevant structures are adequately modeled with very fine elevation models and computational meshes, provided the scale of the problem at hand.
The results are provided at different levels, ranging from generic overviews to street-level analysis on designated areas. At these locations detailed data on wave arrival times, hydrodynamic features of the advancing inundation front and urban debris scattering patterns are provided.
This study has shown that several locations on the Tagus estuary waterfront are especially worrisome in what respects to a tsunami impact scenario. The present document provides relevant information to support the design of emergency response plans, comprising one further step in promoting the safety of Lisbon’s citizens.
Keywords: Tsunamis, mathematical modelling, Tagus estuary, Lisbon.
v vi Notation
A Cell area [L2] C Sediment concentration [−]
CL Homogenous, depth-averaged sediment concentration in layer L [−]
Cf Friction coefficient [−] c Shallow water wave velocity [ms−1] −1 c˜ik Approximate c in k edge [ms ]
ds Reference sediment diameter [m] (n) −1 e˜ik n eigenvector [ms ] E Flux vector
fsij Stress tensor [P a] F~ Generic force [N] F Flux vector in x G Flux vector in y g Gravitic acceleration [ms−2] H Source terms vector h Fluid height [m] H Average depth [m]
hL Thickness of layer L [m]
hL Fluid height on the left side of a shock [m]
hR Fluid height on the right side of a shock [m] ~n Unit normal to a plane [m] 1/3 −1 Ks Manning-Strickler coefficient [m s ] L Wavelength (KdV notation) [−]
Ll Lower interface of a given layer [−]
Lu Upper interface of a given layer [−] p Bed porosity [−] p Hydrostatic Pressure [P a]
PL Depth-averaged hydrostatic pressure in Layer L [P a] 3 −1 qs Solid discharge [m s ] ∗ 3 −1 qs Solid discharge capacity [m s ] R Friction source term vector s Specific sediment gravity [−] S Shock speed [ms−1]
tik Unit tangent to k edge, in natural rotation from nik [m] T Bottom slope source terms numerical flux matrix
Tij Depth integrated turbulent tensions tensor [P a] −1 uφ Velocity associated with the vertical mass flux [ms ] −1 uI Interface velocity [ms ]
uL Velocity on the left side of a shock [m]
uR Velocity on the right side of a shock [m] −1 u∗ Friction velocity [ms ] −1 UiL Depth averaged velocity in layer L, in the xi direction [ms ] −1 UL Depth averaged velocity in layer L, in the x direction [ms ] U Independent variables vector
vii V Primitive variables vector −1 VL Depth averaged velocity in layer L, in the y direction [ms ]
wi Weight factor for cell i [m] −1 ws Sediment settling velocity [ms ]
Zb Bed elevation [m] (n) αik Wave strengths in k edge [m] (n) βik Bottom source flux coefficient in k edge [m]
δij Kronecker delta [−] Depth-averaged turbulent kinetic energy rate of dissipation [m2s−3] −1 ΦLu Net vertical flux across the lower interface [ms ] −1 ΦLu Net vertical flux across the upper interface [ms ] −1 Φa,b Net vertical flux from layer a to layer b [ms ] φ Generic function [m3s−1] η Free-surface elevation (KdV notation) [−] θ Shields parameter [−] κ Depth-averaged turbulent kinetic energy [m2s−2] ˜(n) −1 λik n eigenvalue of [ms ]
λt Wavelength [m]
λL Characteristics speed on the left side of a shock [m]
λR Characteristics speed on the right side of a shock [m] Λ Adaptation length [m] µ Dynamic viscosity of the fluid [P a.s] ν Poisson coefficient [−]
νT Turbulent viscosity [P a.s] ρ Mixture density [kgm−3] −3 ρL Depth-averaged density of the mixture on layer L [kgm ] ρ(w) Clean water density [kgm−3]
sigmaij Stress tensor [P a]
τij Turbulent stress tensor [P a]
τb Bed shear stress [P a]
τy Yield stress [P a]
τv Viscous stress [P a]
τt Turbulent stress [P a]
viii Acronyms
BC Boundary Conditions BKBC Bottom Kinematic Boundary Condition CFD Computational Fluid Dynamics CFL Courant-Friedrichs-Lewy CPU Central Processing Unit DEM Digital Elevation Model FEM Finite Element Method FDM Finite Difference Method FSDBC Free Surface Dynamic Boundary Condition FSKBC Free Surface Kinematic Boundary Condition FVM Finite Volume Method GIS Geographical Information Systems IC Initial Conditions KdV Kortweg-de Vries (equations) PDE Partial Differential Equation RH Rankine-Hugoniot (conditons) RP Riemann Problem STAV-2D Strong Transients in Alluvial Valleys 2D
ix x Contents
Acknowledgments ...... i Resumo ...... iii Abstract ...... v Notation ...... vii Acronyms...... ix List of Figures ...... xv
1 Introduction 1 1.1 Motivation and framework ...... 1 1.2 Proposed objectives ...... 2 1.3 Structure ...... 2
2 Tsunamis 5 2.1 Introduction ...... 5 2.2 What is a tsunami? ...... 5 2.3 Generation mechanisms ...... 7 2.4 Tsunami dynamics ...... 10 2.5 Tsunami consequences ...... 11 2.5.1 The 1755 Lisbon tsunami ...... 11 2.5.2 Recent tsunamis in the Indian Ocean and Japan ...... 13
3 Conceptual Model 17 3.1 Basic equations of fluid and sediment dynamics ...... 17 3.1.1 Material derivative and transport theorem ...... 17 3.1.2 Conservation of mass ...... 18 3.1.3 Conservation of momentum ...... 19 3.2 Governing equations for stratified flows ...... 21 3.2.1 Conservation of mass ...... 24 3.2.2 Conservation of momentum ...... 27 3.3 Closure equations ...... 29
4 Numerical Model 33 4.1 Discretization scheme ...... 33
xi 4.1.1 Introduction ...... 33 4.1.2 Finite volume schemes: The case of the 2D shallow-water equations ...... 34 4.1.3 Stability region ...... 37 4.1.4 Wetting-drying algorithm ...... 38 4.1.5 Entropy correction ...... 38 4.2 Boundary conditions ...... 39 4.2.1 Introduction ...... 39 4.2.2 Rankine-Hugoniot conditions ...... 39 4.2.3 Shocks ...... 40 4.2.4 Expansion waves ...... 41 4.2.5 Analytical solutions ...... 42 4.2.6 Physical feasibility of the mathematical solutions ...... 43 4.2.7 Results ...... 44
5 Integration with Geographic Information Systems 49 5.1 Introduction ...... 49 5.2 Pre-Processing utilities ...... 49 5.2.1 Generating boundaries ...... 49 5.2.2 Refining meshes ...... 50 5.3 Post-processing utilities ...... 51 5.3.1 Resampler and raster converter ...... 51
6 A 1755 tsunami in today’s Tagus estuary topography 55 6.1 Pre-Processing ...... 55 6.1.1 Lisbon and Tagus estuary topography ...... 55 6.1.2 Initial Conditions ...... 58 6.1.3 Boundary Conditions ...... 58 6.2 Brief framework for result presentation and discussion ...... 61 6.3 Overview of a tsunami scenario in Lisbon ...... 62 6.3.1 General description of tsunami propagation ...... 62 6.3.2 Impact at Carcavelos ...... 63 6.3.3 Impact at Caxias and Cruz Quebrada ...... 64 6.3.4 Impact at Belém and Algés ...... 65 6.3.5 Impact at Caparica and Trafaria ...... 66 6.4 Street-level results for inland propagation ...... 67 6.4.1 Results for Alcântara ...... 67 6.4.2 Results for Downtown ...... 73 6.4.3 Results for Almada ...... 80 6.5 Morphological impacts ...... 87 6.5.1 The relevance of flow-morphology interaction ...... 87 6.5.2 Vale do Zebro Royal complex ...... 87 xii 6.5.3 Urban debris washed at Downtown and Alcântara ...... 89 6.5.4 Morphological impacts on other locations ...... 93
7 Conclusions and recommendations 95 7.1 Main conclusions ...... 95 7.2 Directions for future work ...... 96
Bibliography 97
Appendices 101
xiii xiv List of Figures
2.1 Oscillatory and translatory waves ...... 6 2.2 Three main stages of tsunami propagation ...... 7 2.3 Schematics of a seismic generation mechanism ...... 9 2.4 A soliton or solitary wave ...... 10 2.5 A regular wave evolving into a bore ...... 11 2.6 1755 copper engraving depicting the 1755 Lisbon earthquake and tsunami (City Museum) 12 2.7 Tide gauge measurements of the 2004 tsunami at several locations (Rabinovich & Thomson, 2007) ...... 13 2.8 Morphological impact of the 2004 Indian Ocean at Lhoknga, northwest coast of Sumatra, and Phuhket, Thailand ...... 14 2.9 Pressure gage readings for the 2011 Tohoku tsunami (Maeda et al., 2011) ...... 15 2.10 Morphological impact of the 2011 tsunami at Natori and Fukushima, south of Tohoku (ABC News, 2012) ...... 15
3.1 Layer set featured in the conceptual model ...... 22 3.2 A non-material surface ...... 23 3.3 Layer set featured in the conceptual model ...... 27
4.1 A shock (compressive) at the boundary, ∂Ω ...... 41 4.2 Physical admissibility with respect to the entropy condition ...... 44 4.3 Comparison results for the implemented boundary conditions ...... 45 4.4 Steady conditions at time t = 100 [s] and t = 2000 [s] ...... 47 4.5 Steady conditions at time t = 100 s (still the same at t = 2000 s) ...... 47
5.1 Pre-processing samples: example of ArcMap isoline defined boundary (left) and conversion to Gmsh (right) ...... 50 5.2 Example of mesh refinement procedure ...... 51 5.3 Point-in-polygon (PIP) problem examples ...... 52 5.4 Examples of VTK file resampling results ...... 53
6.1 DEM Raster with 5x5 [m] resolution ...... 56 6.2 Bathimetry with 10x10 [m] resolution ...... 56
xv 6.3 Example of DEM built environment enhancement ...... 57 6.5 Mean annual flow duration curve (Ómnias - Sacavém hydrometric station) ...... 58 6.4 Warm-up setup and results ...... 59 6.6 Supplied data for a 1755-like tsunami ...... 60 6.7 Supplied data for a 1755-like tsunami (2) ...... 61 6.8 Relevant locations for result presentation and discussion ...... 61 6.9 Tsunami propagation overview at the Tagus estuary ...... 62 6.10 Tsunami run-up at Carcavelos ...... 63 6.11 Tsunami run-up at Caxias and Cruz Quebrada ...... 64 6.12 Tsunami run-up at Belém and Algés ...... 65 6.13 Tsunami run-up at Caparica and Trafaria ...... 66 6.14 One-way traffic zones (top) and tsunami approach to Alcântara (bottom; t=10’20”) . . . . 67 6.15 Detailed results for Alcântara, t = 12 minutes 20 seconds ...... 69 6.16 Detailed results for Alcântara, t = 14 minutes 20 seconds ...... 70 6.17 Detailed results for Alcântara, t = 16 minutes 20 seconds ...... 71 6.18 Detailed results for Alcântara, t = 20 minutes 20 seconds ...... 72 6.19 One-way traffic zones (top) and tsunami approach to Downtown (bottom; t=13’20”) . . . 73 6.20 Detailed results for Downtown, t = 15 minutes ...... 75 6.21 Detailed results for Downtown, t = 15 minutes 40 seconds ...... 76 6.22 Detailed results for Downtown, t = 16 minutes 20 seconds ...... 77 6.23 Detailed results for Downtown, t = 17 minutes 20 seconds ...... 78 6.24 Detailed results for Downtown, t = 20 minutes 20 seconds ...... 79 6.25 Detailed results for Cacilhas, t = 14 minutes 30 seconds ...... 81 6.26 Detailed results for Cacilhas, t = 16 minutes 10 seconds ...... 82 6.27 Detailed results for Cacilhas, t = 17 minutes 20 seconds ...... 83 6.28 Detailed results for Cacilhas, t = 22 minutes 50 seconds ...... 84 6.29 Detailed results for Cacilhas, t = 24 minutes 20 seconds ...... 85 6.30 Detailed results for Cacilhas, t = 27 minutes 40 seconds ...... 86 6.31 Tsunami approaching the Royal Complex at Vale do Zebro, t=25’ ...... 87 6.32 Tsunami run-up at Coina river (Vale do Zebro) ...... 88 6.33 Tsunami impact at the Royal complex ...... 89 6.34 Idealized setup for urban debris, mainly composed of vehicle-like solid materials ...... 90 6.35 Mobile material distribution along the riverfront ...... 90 6.36 Debris scattering patterns in Alcântara at the aftermath of the tsunami ...... 91 6.37 Debris scattering patterns in Downtoen at the aftermath of the tsunami ...... 91 6.38 Comparison between debris and clearwater formulations ...... 92 6.39 Morphological impacts on remaining locations ...... 93
xvi Chapter 1
Introduction
1.1 Motivation and framework
Often considered as rare events when compared to other natural hazards tsunamis have an enormous devastating potential. The frequency concept associated to these phenomena is to be understood within consistent timescales, namely the geological timescale. At this level, tsunamis and earthquakes, as well as many other geophysical occurrences such as volcanic eruptions or major landslides, become relatively common events whose predictability, at finer timescales, becomes practically impossible. Other difficulties arise when studying tsunamis since one of the main uncertainties lies in the potential of an oceanic earthquake to generate tsunamis. Not all oceanic earthquakes occur in such a way that the seabed gets vertically displaced and vertical momentum is transferred to the water column.
All these variables highlight a much required awareness of the stochastic nature of tsunamis and other geophysical phenomena, in general. Furthermore, seismic risk mitigation is presently a standard in most structural design regulations and public safety procedures. In coastal countries like Portugal, where tectonic activity is likely to concentrate in oceanic faults, accounting for seismic risks while negletcing potential tsunami impacts seems, at least, inconsistent and incautious.
The large ratio between seafront extension and continental territory made Portugal hugely dependant on its maritime Economic Exclusive Zone. Whether it is due to turism, fishing activities or transoceanic transport, the economic contribution of the seafront to the global wealth of the country has been very significant and drove most of Portuguese towards the coast. Over centuries, this generalized demographic behaviour ultimately converged into a country with two noticeably distinct development levels: a rela- tively unpopulated countryside on interior regions contrasting with densely populated and commercially active coastal cities. Although the last decades had a significant contribution in fading such dissimilari- ties, drawn by major communication and technological progresses, the heritage left by recent centuries is unavoidable: most of the population, infrastructure and valuable historical assets are located in coastal areas. This emphasizes the relevance of seismic risk mitigation procedures that involve structural damage as well as inundation extents.
1 1. Introduction
Recent revisions of the Portuguese catalog of tsunamis (Baptista & Miranda, 2009) have also shown that the seafront has been struck by numerous events over the past two millenia.
Presently, numerous shallow water models are used within the scientific community for tsunami propa- gation (Imamura et al., 2006). While these include all the relevant components for offshore propagation, their ability to account for very specific features in waterfront and overland propagation is limited. One of the most relevant features of tsunami propagation over solid boundaries is its ability to incorporate debris, either natural sediment incorporated from the bottom or remains of human built environment. Despite acknowledged, this feature is commonly neglected in most tsunami inland propagation studies, probably by virtue of a relatively incomplete understanding of the involved phenomena when sediment transport imply significant changes in the rheological behaviour of the flow.
1.2 Proposed objectives
A computational fluid dynamics (CFD) model based on the finite volume method (FVM) (Canelas, 2010) - STAV2D (Strong Transients over Alluvial Valleys 2D) - featuring enhanced sediment transport capabilities (Canelas, 2010) and up to date numerical discretization schemes (Murillo, 2006) was employed to analyze a tsunami scenario in the Tagus estuary. The tsunami parameters resemble the widely known event that struck Lisbon on the 1st November of 1755.
Not only is the present work based on improved conceptual and numerical models, but also, the physical environment discretization was taken to a level not known in other studies. The human built environment was detailed at a very refined scale, featuring buildings, streets and even urban mobile debris.
The main objective of this work lies in providing detailed data on the exposure of designated areas along the Tagus estuary riverfront. Lisbon’s Downtown, Alcântara and Cacilhas were thoroughly discretized in order to perform a street level description, a significant improvement over the present knowledge of a tsunami impact on these locations. Morphological changes and debris scattering were also analyzed by taking full advantage of the solid transport capabilities of the numerical model.
A set of secondary objectives was also delineated, consisting of computational work on the CFD code: a new scheme of boundary conditions was implemented and pre/post-processing toolkits were developed to promote functional integration with commercial Geographical Infromation Systems (GIS).
1.3 Structure
The first chapter is devoted to a brief bibliographical review on the subject of Tsunamis. Particular emphasis is given to the hydrodynamic concepts and numerical modeling trends. Historical context is also provided on a very relevant event for the present work: The 1755 Great Lisbon Earthquake and Tsunami. Recent events that were given great interest in worldwide media and scientific communities
2 1.3. Structure are also revised.
Theoretical background on the conceptual model that supports the present work and a revision of the em- ployed closure models are provided in Chapter 2, with special care being given to particular mathematical formulations inherent to multiphase flows.
A brief description of the discretization scheme is presented in the following chapter. The implemented boundary condition scheme is thoroughly explained and its performance assessed with a set of functional tests.
The fourth chapter introduces the developed toolkit for pre and post-processing tasks. It is particularly useful for future reference and also highlights the expected ease of use of these tools.
The last chapter features a comprehensive description of the simulated scenario. All details concerning the setup and outsourced inputs for these simulations are provided, including boundary and initial con- ditions and the employed topographic datasets. The results are displayed at different levels: a tsunami scenario overview on the whole Tagus estuary, a detailed street-by-street analysis on designated areas and morphological impact assessment.
3 1. Introduction
4 Chapter 2
Tsunamis
2.1 Introduction
Human communities living near the ocean have always experienced the devastating effects of a tsunami. These waves have surely claimed hundreds of thousand of lives and have caused incalculable material losses. However, since they are not frequent in industrialized western countries, they have not been given relevant media attention. The morning of the December 24th changed this forever when an earthquake of magnitude 9.1 stroke the coast of Sumatra in Indonesia. The tremendous energy released in this oceanic earthquake generated a vertical displacement of the seabed, rising the sea surface to form a devastating tsunami that killed 280,000 people and affected the lives of several million others (Helal, 2008).
While people living along the coastline of Sumatra had little time and no warning on the approaching tsunami, those living along the further coasts of Thailand, Sri Lanka, India and East Africa had plenty of time to escape to higher grounds. Unaware, they stayed at lower ground, as no tsunami warning system was implemented in the Indian Ocean at that time. Unfortunately, it has taken a disaster of such magnitude to awake the scientific communities to this silent but fearful natural hazard. The present chapter aims to describing some of the most common physical features and modeling trends on tsunamis.
2.2 What is a tsunami?
The term tsunami originates from the Japanese tsu and nami meaning wave and harbor, respectively. It may have been originally used by fishermen, who would sail out, encountering no visually unusual waves while out at sea, and later returned to find an overwhelmed hometown.
Tsunamis are waves with large wavelengths that are originated by some kind of sudden disturbance in standing 1 water and are characterized as shallow-water waves. These are different waves from the ones most commonly observed on a beach, which are caused by the wind blowing across the ocean’s surface.
1 By standing water one means a water mass whose√ velocity is much lower than the velocity of a propagating tsunami, which is close to that of shallow-water waves celerity ( gH).
5 2. Tsunamis
Wind-generated waves usually have periods of five to thirty seconds and a wavelength of 100 to 200 [m] whereas tsunamis can have periods and wavelengths that are much higher (Helal, 2008).
A major difference between these waves and tsunamis is the mass transport carried by the wave. While wind-generated waves induce nearly-closed trajectories 2 to the particles a tsunami displaces them several meters from their initial positions. The volume of water that gets transported and the speeds reached by tsunami waves allows them to carry enough energy to wipe out entire towns and cities (Helal, 2008). Wind-generated waves are called oscillatory waves whereas tsunamis are translatory waves (Figure 2.1).
Oscillatory Wave Translatory Wave
Stokes Drift
Figure 2.1: Oscillatory and translatory waves
Tsunami propagation comprises three main stages: oceanic propagation, near shore shoaling and inland propagation (Figure 2.2 on page 7). At the first stage, within deep ocean waters, the period of a tsunami can range from 5 minutes to 2 hours and the wavelength, λt, can be as high as 500 [km].
Waves are governed by shallow-water equations if the ratio between the wavelength and the depth of the ocean is over 20. Given that the maximum mean depth of the sea-floor, H, is about 5 km, the ratio of the
flow depth to the wavelength is small: H/λt ≈ 0.001, a common value, for instance, in river flows. Hence, tsunamis will always behave as shallow-water waves no matter where on the globe they are being modeled √ or observed. The velocity of a shallow-water wave is also equal to gH, where g is the acceleration of gravity (9.8 [m/s2]) and H is the depth of the water column.
Thus, in very deep water, a tsunami will propagate at high speeds: in the Pacific Ocean, for instance, the average depth is approximately 4000 [km], which means that the velocity of a tsunami in this water mass is nearly 200 [m/s]. Considering that the distance across this ocean, for instance from Seattle to Hong Kong, is roughly 8000 [km] this would yield a total of 11 hours for travel time.
2Trajectories are not exactly closed because the final position of the particles is not coincident with the initial position. This is called the Stokes Drift (Stokes, 1847). Furhter details on surface wave dynamics can be found in the Appendix 1. Nonetheless, the mass transport in this drift is negligible when compared to that of a tsunami.
6 2.3. Generation mechanisms
Offshore tsunami amplitudes are likely below 1 [m] and wavelengths are certainly a few hundred kilometers long. Given that the order of magnitude of the hydraulic gradient can be given by the wave amplitude to wavelength ratio, this leads to practically null energy dissipation: the hydraulic gradient is 10−5 for a rather small wavelength of 100 [km]. This yields basically no head loss for a tsunami propagating over deep waters, reaching distant shores with high energy levels.
As a tsunami approaches shallower waters near shore the depth of the water decreases, meaning that the velocity of the tsunami will also decrease. However, its energy and mass are maintained, suggesting that the height of the tsunami must grow enormously. It is at this stage that the tsunami becomes visible, resembling the atrocious footage spread out on the media in 2004 (Indian Ocean) and 2011 (Japan).
The third stage of a tsunami deals with breaking as it approaches shore and depends greatly on the bathymetry. Overland tsunami propagation may form bores: abrupt wave forms mathematically repre- sented as discontinuities. These discontinuous and highly turbulent flows are likely to incorporate debris, whether natural sediment or human built environment remains. The dynamics of this final stage of a tsunami is somewhat similar to that of flood waves caused by dam breaking, dyke breaking or overtopping. Very high concentrations of debris are carried on the water column and may change the hydrodynamic properties and rheological behaviour of the fluid (Dias, 2006).
Translatory wave low amplitude mass is conserved bores may form high wavelength amplitude increases shore mean sea level
continental shelf velocity
sea bed
Figure 2.2: Three main stages of tsunami propagation
2.3 Generation mechanisms
Tsunamis can are mainly generated by seismic activity (seismicity), landslides or volcanic activity. The seismic generation mechanism is responsible for most tsunami occurrences and, as a major source, must be discussed with further detail. Tsunamis can be generated by seismicity when the sea floor abruptly rises and vertically displaces the overlying seawater. Tectonic earthquakes are a particular kind of earthquake that are associated with the continental drift. When these earthquakes occur in the ocean, the water above the deformed area is displaced from its initial position (Dias, 2006). Namely, a tsunami can be generated when thrust faults, associated with convergent or destructive plate boundaries, suddenly slip and displace large volumes of water due to the vertical component of momentum involved (see Figures 2.3a to 2.3d).
7 2. Tsunamis
Mathematical modeling of this mechanism requires a displacement field to be parametrized at the sea bottom. Volterra (1907) originally solved this problem in the case of an entire elastic space. Hence, and for simplicity’s sake, that will be the approach herein presented.
Considering the thrust mechanism as a fracture, the discontinuities in the displaced components across the fractured surface can be described using the theory of elasticity, namely surfaces across which the dis- placement field is discontinuous (Steketee, 1958). For simplicity’s sake some assumptions are introduced.
The curvature of the earth, its gravity, temperature, magnetism and non-homogeneity are neglected. A semi-infinite, homogeneous and isotropic medium is considered instead. Further details on how the earth’s curvature, topography, homogeneity and isotropy influence the displacement fields can be found on the works published by McGinley (1969), Ben-Mehanem et al. (1970a,b) and Masterlark (2003). Furthermore, one assumes that the the linear elasticity theories are valid within the fractured region.
Let O be the origin of a Cartesian coordinate system, xi the Cartesian coordinates with i = 1, 2, 3, and ~ei ~ k a unit vector in the positive xi direction. A force F = F~ek at O generates a displacement field ui (P,O) at point some point P , given by,
2 2 k F ∂ r ∂ r ui (P,O) = δik 2 − α (2.1) 8πµ ∂xn ∂xi∂xk
where δik is the Kronecker delta, r is the distance from P to O and α is a coefficient defined as,
λ + µ 1 α = = (1 − ν) (2.2) λ + 2µ 2 where λ and µ are the Lamé’s constans and ν is the Poisson’s ratio 3 . Hooke’s Law easily yields the stresses due to the displacement fields in equation (2.2)
k ∂uk ∂ui ∂uj σij = λδij + µ + (2.3) ∂xk ∂xj ∂xi
+ − The dislocation is defined as a surface Γ across which one finds a discontinuity ∆ui = ui − ui in the displacement fields. This yields the following formula (Volterra, 1907) describing the displacement field for a particular point (x1, x2, x3)
1 ZZ u (x , x , x ) = ∆u σk · n dS (2.4) k 1 2 3 F i ij j Γ
. In general, the displacement field represented by equation (2.4) is reproduced in the ocean free surface (Figure 2.3c on page 9), posing the initial conditions for tsunami propagation (Imamura et al., 2006).
Tsunamis often have a small amplitude offshore which is why they generally pass unnoticed at sea, forming only a slight swell usually less than half a meter above the normal sea surface. A tsunami can occur in
3Transverse to axial strain ratio when a material is compressed.
8 2.3. Generation mechanisms any tidal state and even at low tide can still severely inundate coastal areas. The 1964 Alaska earthquake (Magnitude, Mw, 9.2), the 2004 Indian Ocean earthquake (Mw 9.2) and the 2011 Tohoku earthquake (Mw 9.0) are recent examples of powerful thrust earthquakes that generated powerful tsunamis, capable of crossing entire oceans and still inflict severe devastation in coastal areas. Other potential generation mechanisms may occur, although with fewer recorded occurrences. These include landslides, explosive volcanic eruptions, large atmospheric depressions and even meteor impacts.
Tsunamis generated by landslides are called sciorrucks. These phenomena rapidly displace large water volumes, as energy from falling debris or expansion is transferred to the water at a rate faster than it can absorb. Their existence was confirmed in 1958, when a giant landslide in Lituya Bay, Alaska, caused the highest wave ever recorded, which had a height of 524 [m]. The wave didn’t travel far, as it struck land almost immediately. Two people fishing in the bay were killed, but another boat amazingly managed to ride the wave (Fine et al., 2006). Scientists named these waves megatsunami and discovered that extremely large landslides from volcanic island collapses can also generate these kind of waves.
Meteotsunamis are generated by deep depressions that cause tropical cyclones, generating a storm surge often called a meteotsunami that can raise tides several meters above normal levels. The displacement comes from low atmospheric pressure within the centre of the depression.
SWL - Still Water Level SWL - Still Water Level
Overriding Plate Overriding Plate Subducting Plate Subducting Plate
(a) Converging plates before earthquake (b) Overriding plate bends under strain
Overriding Plate Overriding Plate Subducting Plate Subducting Plate
(c) Accumulated energy is released (d) Tsunami propagates radially
Figure 2.3: Schematics of a seismic generation mechanism
9 2. Tsunamis
2.4 Tsunami dynamics
A simple mathematical formulation to demonstrate the physics behind a tsunami wave is herein pre- sented Helal (2008). In this approach a tsunami is conceptualized as a non-linear bi-dimensional wave propagating over a horizontal ground.
The governing law for this solution is the Korteweg-de Vries equation (KdV). It is a nonlinear, dispersive partial differential equation for a function of two real variables, space, x, and time, t. For some function φ the Korteweg-de Vries equation is stated as,
3 ∂tφ + ∂xφ + 6φ∂xφ = 0 (2.5) where the constant 6 in front of the last term is merely conventional: multiplying t, x, and φ by constants can be used to vary the constants preceding each of the featured terms.
For hydrodynamic applications this equation is often presented in the form,
∂η ∂η 3 ∂η ξ ∂3η + + η + = 0 (2.6) ∂t ∂x 2 ∂x 6 ∂x3 and featuring the following non-dimensional variables,
δ − h h h2 η = ; = ; ξ = (2.7) H H L2 where is the wave amplitude corresponding to the depth h; L is the wavelength; η(x, t) is the relative movement of the free surface relating to the maximum height and ξ is the squared depth to wavelength ratio. This form of the KdV equation is an numerical simulation standard when it comes to offshore propagation and its solutions are provided in the form of solitons (Figure 2.4) or other dispersive waves.
L
H mean sea level
h δ
sea bed
Figure 2.4: A soliton or solitary wave
The dispersive term (fourth in equation (2.6)) counterbalances the quasi-linear terms (second an third terms in equation (2.6)) allowing for the self-similar propagation of the wave-form shown in Figure 2.4. As the wave approaches shore non-linear terms will become more important and the wave will cease to be decribed by the KdV equation.
10 2.5. Tsunami consequences
When a tsunami reaches the seashore and propagates overland ( last stage in Figure 2.2) it may develop into an abrupt wave form, or bore, which is mathematically considered as a discontinuity. Bores are non-linear wave forms withour relevant dispersive terms, especially if debris are incorporated into the water column. t +2dt t +dt 0 t 0 0
Figure 2.5: A regular wave evolving into a bore
In this case, the dynamics of the tsunami is governed by shallow-water equations. In fact, dispersive effects are often disregarded and shallow-water equations are frequently used to describe the propagation of tsunamis in open waters (Imamura & Shuto, 1990).
The shallow-water equations will not be described in this section. They constitute the core of the mathematical simulation tools employed in this dissertation and will be carefully derived in Chapter 3, p. 17 to 21.
2.5 Tsunami consequences
2.5.1 The 1755 Lisbon tsunami
Around 9:40 AM on the 1st November of 1755, the All Saints’ Day Catholic holiday, Lisbon was struck 4 by a violent earthquake with a magnitude ranging from 8.5 Mw to 9 Mw , (Abe, 1989) with an epicenter in the Atlantic Ocean 5 that became known as the Great 1755 Lisbon Earthquake. This earthquake was followed by several fires and most importantly, for the purpose of this dissertation, by a major tsunami.
At that time Lisbon was inhabited not only by Portuguese people but also by a considerable foreign population, mainly French, Spanish, English and German. This contributed to the wide diversity of publications, in both Portuguese and English, regarding the Great 1755 Lisbon Earthquake and made it the most well documented historic tsunami event (Baptista et al., 1998b).
The following are amongst the most relevant historic reports (Baptista et al., 1998a, 2006),
"There was another great shock after this that pretty much affected the river, but, I think not so violent as the preceding, though several people afterwards assured that as they were riding on horseback, in the great road to Belem, one side of which lays upon the river, the waters rushed
4 Moment Magnitude Scale, Mw 5Located south of the west coast of Algarve and northeast of the Archipelago of Madeira(Baptista et al., 1998b)
11 2. Tsunamis
in so fast that they were forced to gallop as fast as possible to the upper grounds for fear of being carried away" (British Accounts, 1755)
"A large quay, piled up with goods near the Custom House sunk the first shock, with about 600 persons upon it, who all perished" (British Accounts, 1755)
"(...) a lot of people run to the river bank trying to escape from the ruins. Suddenly the sea came in through the bar and flooded the river banks..." (Moreira de Mendonça, 1758).
"(...) waters with their ebb and flow flooded the Customs, the Square and the Vedoria".
Figure 2.6: 1755 copper engraving depicting the 1755 Lisbon earthquake and tsunami (City Museum)
To avoid interpretation and subsequent distortions Baptista et al. (1998a) carried a direct analysis of historical data between 1755 and 1759 where all the documents were original from the 18th century and were analyzed in their original language. This study pointed out very useful aspects about the earthquake and tsunami that destroyed Lisbon on that morning of the November 1st.
The earthquake shook the city at about 9h30 a.m. and lasted about 9min. When most people were heading to the riverfront in a desperate effort to avoid the fires and ruining buildings in the inner city and also out of curiosity about the unusual low tide, which was in fact a drawback and the only warning of an incoming tsunami, a huge water wall hit the riverfront. The tsunami traveled for about 90 minutes in the open sea and should have hit the city at about 11h00 a.m. climbing the shores up to a mean run-up level of 5 [m] and penetrating the city as deep as 250 [m]. The death toll only due to the tsunami is estimated in around 900 people whereas the earthquake itself should be held responsible for a number lives ranging from 10.000 up to 100.000 (Álvaro S. Pereira, 2006).
About eighty-five percent of the city’s built environment collapsed, including famous palaces and libraries constituting some of the most brilliant examples of the unique Manueline architectural style. Even buildings that had suffered little damage from the earthquake were destroyed in the subsequent fires. The Royal Ribeira Palace, which was located near the riverfront in the modern square of Terreiro do Paço, was destroyed by the earthquake and tsunami. Inside, the 70.000 volume royal library, also filled
12 2.5. Tsunami consequences with hundreds of works of art by notorious painters and sculptors of that time were forever lost. The scientific and historian communities were also deprived of priceless intelligence on the early campaigns led by the famous navigators Vasco da Gama, Fernão de Magalhães and Pedro Alvares Cabral.
All the major churches in Lisbon were ruined by the earthquake, namely the Lisbon Cathedral, the Basilicas of São Paulo, Santa Catarina, São Vicente de Fora, and the Misericórdia Church. The Royal Hospital of All Saints in the Rossio Square was consumed by fire and many patients burned to death. Visitors to Lisbon may still walk the ruins of the Carmo Convent, which were preserved to remind Lisbonners of that day’s destruction.
2.5.2 Recent tsunamis in the Indian Ocean and Japan
Two recent tsunami events in the last decade have surely awaken people and governments to the reality and unpredictability of these natural hazards, namely the 2004 Indian Ocean and the 2011 Tohoku tsunamis. The 2004 Indian Ocean earthquake was an undersea megathrust earthquake 6 that occurred on Sunday, 26 December 2004, with an epicenter off the west coast of Sumatra, Indonesia. The resulting tsunami is given various names, including 2004 Indian Ocean tsunami or Boxing Day tsunami.
The earthquake was caused by subduction and triggered a series of devastating tsunamis along the coasts of most landmasses bordering the Indian Ocean, killing over 230.000 people in fourteen countries, and inundating coastal communities.
Asia Hillarys 20˚N Salalah Cocos Africa Hanimaadhoo Male Colombo 200 cm 0˚ Lamu Port Louis Gan M w =9.3 Zanzibar Colombo Pointe Diego Garcia La Rue Cocos 3 hr 20˚S Port Louis Pointe La Rue Australia 5 hr Hanimaadhoo Hillarys 7 hr St. Paul Sea level 40˚ 9 hr Salalah 11 hr Kerguelen Male Indian Ocean 13 hr 60˚ Lamu Gan
Mawson Zanzibar Diego Garcia Casey Dumont d'Urville SyowaSyowa Zhong Shan Davis Antarctica 0˚ 20˚ 40˚ 60˚ 80˚ 100˚ 120˚ 140˚E 26 27 28 26 27 28
Figure 2.7: Tide gauge measurements of the 2004 tsunami at several locations (Rabinovich & Thomson, 2007)
Figure 2.7 displays sea level measurements from several buoys along the Indian Ocean. As expected, offshore amplitudes were relatively small. For instance, the tide gauge located at the Cocos Islands, 1700 [km] from the epicentre, revealed a small 30 [cm] high wave followed by a long sequence of oscillations,
6A megathurst earthquake is an earthquake generated by seismicity at convergent plate boundaries, where the subducting plate shallowly dips under the overriding plate. All recorded major earthquakes, with Mw over 9.0, are of the megathurst type.
13 2. Tsunamis
Higher readings were provided by the gauges located at Gan and Diego Garcia Islands where amplitudes were just below 1 [m]. As the tsunami approached continental platforms the wave height grew consider- ably. Gauge data yields amplitudes of nearly 3 [m] (Titov et al., 2005) at Pointe La Rue, Salalah and Colombo. At Banda Aceh, Indonesia, water was reported to have risen by 3 [m] in under 1.5 minutes (Chanson, 2005).
Important morphological changes were observed in Sri Lanka and Indonesia with some bed variations reaching over 5 [m] (Chanson, 2005; Goto et al., 2011) at relevant harbors, for instance, the Lhoknga harbour depicted in Figure 2.8, and at most seafront mobile reaches. This highlights the solid transport capacity of the overland propagation stage, clearly visible on the aftermath pictures in Figure 2.8, and the importance of effectively accounting for this feature in numerical simulations. Waste and debris are mostly deposited inland but can also be carried offshore, when draw-down occurs, dispersing along the shore.
(a) Lhokonga before (b) Lhokonga after the (c) Phuket before the (d) Phuket after the the tsunami impact tsunami impact tsunami impact tsunami impact (Chanson, 2005) (Chanson, 2005) (NASA) (NASA)
Figure 2.8: Morphological impact of the 2004 Indian Ocean at Lhoknga, northwest coast of Sumatra, and Phuhket, Thailand
This tsunami was one of the deadliest natural disasters in recorded history. Indonesia was the hardest- hit country, followed by Sri Lanka, India, and Thailand. With a magnitude of Mw 9.1 – 9.3, it is the third largest earthquake ever recorded on a seismograph. The plight of the affected people and countries prompted a worldwide humanitarian response.
The 2011 earthquake off the Pacific coast of Tohoku and also known as the the Great East Japan Earthquake or the 3/11 Earthquake, was a magnitude 9.0 (Mw) undersea megathrust earthquake off the coast of Japan that occurred on Friday, 11 March 2011, with the epicenter approximately 70 kilometres east of the Oshika Peninsula of Tohoku and the hypocenter at an underwater depth of 32 [km]. It was the most powerful known earthquake ever to have hit Japan, and one of the five most powerful earthquakes in the world since modern record-keeping began in 1900. The earthquake triggered powerful tsunami waves that travelled as far as 10 [km] inland with up to 4 [m] high.
14 2.5. Tsunami consequences
Tohoku
Fukushima
Figure 2.9: Pressure gage readings for the 2011 Tohoku tsunami (Maeda et al., 2011)
Pressure gauge readings for this earthquake, such as the ones shown in Figure 2.9, show offshore heights up to 1.2[m]. As previously referenced these small waves are prone to shoaling and the water height that hits the shore is expected to be much higher. Several amateur footage was spread out on the media showing flow depths ranging from 2 to 4 [m] high (Biggs & Sheldrick, 2011).
Significant morphological impacts occurred throughout all Japanese coast. Mobile reaches suffered severe bed configuration changes as depicted in Figure 2.10. Built areas were completely wiped by the advancing front of a turbulent, debris laden, flow.
(a) Natori before the tsunami impact (b) Natori after the tsunami impact
(c) Fukushima before the tsunami impact (d) Fukushima after the tsunami impact
Figure 2.10: Morphological impact of the 2011 tsunami at Natori and Fukushima, south of Tohoku (ABC News, 2012)
15 2. Tsunamis
The Japanese National Police Agency confirmed 15,867 deaths. The tsunami also caused a number of nuclear accidents, primarily the level 7 meltdowns at three reactors in the Fukushima Daiichi Nuclear Power Plant. Many electrical generators were taken down, and at least three nuclear reactors suffered explosions due to hydrogen gas that had built up after the cooling system failed.
Early estimates placed insured losses from the earthquake alone at US$14.5 to $34.6 billion.[31] The Bank of Japan had to offer $15 trillion (US$183 billion) to the banking system on 14 March in an effort to normalize market conditions. The World Bank’s estimated economic cost was US$235 billion, making it the most expensive natural disaster in world history.
16 Chapter 3
Conceptual Model
3.1 Basic equations of fluid and sediment dynamics
The purpose of the present section is to introduce the fundamental equations of fluid dynamics within the continuum hypothesis (Atkin & Crane, 1976). Under this hypothesis individual sediment grains are ignored and the mass of sediment is uniformly distributed over the volume of the system. Furthermore, this assumption is only valid at scales larger than the grain diameter and assumes homogeneity within the sediment-fluid mixture. Along these lines, variables such as density, velocity or suspended solids concentration are to be described as continuous time-dependent scalar or vector fields on R3.
The physical phenomena that govern flow dynamics are to be mathematically represented by a set of three laws, namely conservation of mass, momentum and energy. The derivation of these equations can follow one of two distinct paths: a space-averaged estimation of flow effects over a finite region, usually referred to as the integral form, or a more refined point-to-point description of the flow, the differential form. The conceptual model, described in section 3, is significantly more consistent with the differential form and so this will be the undertaken approach to derive the referred conservation equations.
3.1.1 Material derivative and transport theorem
The time-derivative of a given property φ(xi, t) of a fluid element within a velocity field ui(xi, t) can be obtained by tracing this same element along its trajectory. In such case there will be two distinct contributions for the total variation of φ: the first is due to the time dependence of the φ field, named the local term, and the other is induced by the motion of the fluid element through spatial gradients of φ, the advective term. The space coordinates xi are x1 = x, x2 = y and x3 = z, where x, y are the horizontal coordinates and z is a vertical coordinate aligned with the gravity field. The time coordinate is t.
The material derivative, denoted by Dφ/Dt, is thus given by,
Dφ ∂ = φ(t, x (t)) (3.1) Dt ∂t i
17 3. Conceptual Model which in turn can be expanded, by means of the chain rule, to,
Dφ ∂φ ∂φ dxi ∂φ ∂φ = + = + ui (3.2) Dt ∂t ∂xi dt ∂t ∂xi
where t stands for time and ui are the velocity vector components. This form clearly distinguishes the two contributory terms for the total variation of φ: the local term, ∂tφ, and the advective term, ui ∂xi φ.
Another rather important concept in fluid mechanics is the Reynold’s Transport Theorem which states that for any material volume V (t) and differentiable scalar field ϕ it is valid that,
d Z Z ∂ϕ ∂ ϕ dV = + (ϕui) dV (3.3) dt ∂t ∂xi V (t) V (t) which provides the necessary framework to formulate the conservation laws for mass, momentum and energy (Wesseling, 2009).
3.1.2 Conservation of mass
This conservation law states that the rate of change of mass must equal the rate of mass production in an arbitrary material volume V (t). Mathematically this is expressed by,
d Z Z ρ dV = σ dV (3.4) dt V (t) V (t) where ρ is the apparent density of an homogeneous granular-fluid mixture and σ is the rate of mass production per unit volume. The definition of ρ is as follows,
δM ρ = lim (3.5) δ∀→0 δ∀ where δ an arbitrary difference operator, δM is the mass of the fluid-granular mixture and δ∀ is the volume occupied by that mass. The mass of the mixture is M = Mw + Ms where M is the total mass and Ms is the sediment mass and hence equation (3.5) becomes,
δM δ∀ δM δ∀ ρ = lim w w + s s (3.6) δ∀→0 δ∀ δ∀w δ∀ δ∀s
where ∀w and ∀w are the volumes of water and sediment, respectively. Note that ∀w = ∀ − ∀s and (3.6) converges to ρ = ρw (1 − C) + ρsC. Defining the specific gravity as s = ρs/ρw on obtains the final form of the mixture density,
ρ = ρw [1 + (s − 1)C] (3.7)
18 3.1. Basic equations of fluid and sediment dynamics
By using the transport theorem, equation (3.3), one can expand equation (3.4) to,
Z ∂ρ ∂ Z + (ρui) dV = σ dV (3.8) ∂t ∂xi V (t) V (t)
where ui is the velocity vector field and t stands for time. The mass production rate per unit volume, σ, accounts for singular sinks or sources inside V (t). In this case there are no sinks or sources since all inputs and outputs to the control volume can be expressed in terms of incoming or outgoing fluxes, respectively.
Provided that equation (3.8) holds for every arbitrary V (t), the differential form of the mass conservation law can be obtained by extracting the left-hand integrand,
∂ρ ∂ + (ρui) = 0 (3.9) ∂t ∂xi which is commonly known as the continuity equation as it requires no further assumptions other than that both density and velocity are continuum functions (White, 2002).
The conceptual model is aimed at river flows which are greatly influenced by the transported sediment in what concerns hydrodynamics. Since both water and sediment are incompressible so must be their mixed flow but this alone does not imply a constant density value. Following the material derivative definition, equation (3.2), one further obtains,
Dρ ∂u + ρ i = 0 (3.10) Dt ∂xi
3.1.3 Conservation of momentum
The momentum conservation equation is derived from Newton’s 2nd Law, the fundamental physical prin- ciple governing dynamics. Employing this law to balance the rate of change in linear momentum with all the applied external forces on a fluid element will yield,
d Z Z Z ρu dV = ρg dV + f n dS (3.11) dt i i Sij j V (t) V (t) ∂V (t)
where ∂V (t) is the boundary surface of V (t), gi is the gravitational force per unit mass, nj is the outward unit vector and fSij are the surface forces per unit area. Expanding the left term according to the Reynold’s Transport Theorem will further demonstrate the implications of this conservation equation,
Z Z Z ∂ρui ∂ρui + uj dV = ρgi dV + fSij nj dS (3.12) ∂t ∂xj V (t) V (t) ∂V (t)
where ui stands for the velocity component being conserved.
19 3. Conceptual Model
The source terms on the right hand side are gravity (body force), pressure and viscous stresses (surface forces). It can be shown that the surface forces are defined by the following 2nd order tensor, (Aris, 1989)
fSij = −p δij + τij (3.13)
where p is the hydrostatic pressure, δij is the Kronecker delta and τij is the deviatoric stress tensor (viscous stress tensor, in the case of pure water). By applying the Divergence Theorem to the surface integral in (3.12) one obtains,
Z Z ∂fSij fSij · njdS = dV (3.14) ∂xj ∂V (t) V (t) which, in further combination with equation (3.12), must give,
Z Z ∂ρui ∂ ∂ + ui (ρuj) dV = ρgi + fSij dV (3.15) ∂t ∂xj ∂xj V (t) V (t) which, being valid for any material volume V (t), can be condensed into a differential form by extracting the integrands. Expanding the surface forces according to equation (3.13) will further yield,
∂ρui ∂ρuiuj ∂p ∂τij + = − + ρgi + (3.16) ∂t ∂xj ∂xi ∂xj which completes the momentum conservation law. For the special case of incompressible flows, as the null divergence hypothesis still holds, (3.16) can be further simplified to,
∂ρui ∂ρui ∂p ∂τij + uj = − + ρgi + (3.17) ∂t ∂xj ∂xi ∂xj
The Navier-Stokes equations
In order to complete the system of equations it is still necessary to relate the deviatoric stress tensor
τij with the velocity field by means of a constitutive relation. For clear water, the simplest of these constitutive relations establishes that the viscous stresses are proportional to the strain rates and to the dynamic viscosity, µ, (newtonian fluid) and is then given by, (Batchelor, 1967)
∂ui ∂uj 2 τij = µ + − ∇ · ~uδij (3.18) ∂xj ∂xi 3 which, for incompressible flows, simplifies to,
∂ui ∂uj τij = µ + (3.19) ∂xj ∂xi
20 3.2. Governing equations for stratified flows
Combining the momentum conservation equation for incompressible flows, equation (3.17), with the previous consititutive relation (3.19) will yield the widely known Navier-Stokes equations.
∂ρui ∂ρui ∂p ∂ ∂ui ∂uj + uj = − + ρgi + µ + (3.20) ∂t ∂xj ∂xj ∂xj ∂xj ∂xi which describe the flow in case of absence of suspended sediments.
3.2 Governing equations for stratified flows
The conceptual model underlying the present dissertation idealizes river flow as a stratified planar flow of both water and sediment. Holding the hypothesis that the layers composing this stratified flow are continua, the conservation equations suitable for describing such media are also applicable to each layer and ultimately to the whole layer set. Each of these layers is a mixture of sediment and water characterized by its respective sediment concentration and ensuing density.
As the model is designed in 2D further work is required on the equations presented in Chapter 3.1. Specifically these equations must be brought from their generic 3D arrangement down to a 2D nearly equivalent form. A consistent path to performing such transition is depth-averaging those conservation equations. For a given function, for instance f(x, y, z, t), the depth-averaging is defined as
1 Z y0+hL FL(x, y, t) = f(x, y, z, t) dz (3.21) hL y0
where FL(x, y, t) is the depth-averaged f(x, y, z, t) function within a layer L with depth hL.
It can be shown that integrating any function f(x, y, z, t) over a finite domain of the dependant variable z will result in some function of the form Fz(x, y, t) which may not depend on z itself. Hence,
∂F L = −1 (3.22) ∂z
Neglecting the vertical velocity (2D) in the conservation equations can be acceptable if the studied phenomena have an horizontal scale that happens to be much greater than the vertical scale, which is consistent with the hydrodynamics of a tsunami wave. Although the propagation of this type of wave occurs within the deep waters of the oceans, its wave length is great enough so that the tsunami will always interact with the seabed, thus generating friction, leading to a shallow-water wave behaviour.
The applicability of this conceptual model is restricted to flows where there is a clear distinction between suspended sediment, transported within the water column, and bedload, composed by coarser sediment transported near the riverbed Ferreira (2005). The layers featured in this conceptual model are those depicted in Figure 3.1. The first layer, from herein onwards Layer [1] or clear water layer, is the upper water column in which fine suspended sediments are transported. Underneath lies the contact load layer,
21 3. Conceptual Model
Layer [2] or bedload layer, which has a high concentration of coarse and suspended sediment. One must note that the thickness of Layer [2] is expected to be much smaller than the one of Layer [1], and so the set-up presented in Figure 3.1 only serves illustrative purposes. The bed, or Layer [b], was modeled as a passive, non-moving repository of sediment that is allowed to undergo either deposition or scour.
Clearwater Layer Layer [1]
Bedload Layer Layer [2] Bed Layer [b]
Figure 3.1: Layer set featured in the conceptual model
As previously introduced in section 3.1 the governing equations are a set of partial differential equations (PDEs) that describe fluid dynamics. The featured conservation equations consist of two mass conser- vation equations, for either total and solid phases, two momentum conservation equations, one for each direction of the x−y plane and a set of closure equations. Since two distinct phases are featured, the bedload layer and the clear water layer, the 2D conservation equations must be manipulated with special care, so that all fluxes are accounted for and all quantities remain conserved.
Each layer is bounded by two interfaces. These are defined as continuum surface functions of the form,
SL(x, y, z, t) = IL(x, y, t) − z = 0 (3.23)
These surfaces can be defined as material or non-material. Material surfaces are not prone to fluxes which implies that the fluid velocity vector at every point of these surfaces must equal the velocity vector of the surface itself. On the contrary, non-material surfaces can be crossed and so no equality is imposed between the velocity vectors of the fluid and surface (see Figure 3.2).
At some instant, t, the interface is defined by the surface z = St and the fluid particles lying on it are 0 defined by the surface z = St. After an infinitesimal time step, dt, both of these surfaces will change their 0 configuration into St+dt and St+dt, respectively. The particles initially lying on the interface may now be completely detached from it, meaning that the surface and these particles will have different velocities, respectively ~uI and ~u. However these can be related to one another by an equivalent flux velocity, ~uφ, according to ~u = ~uI + ~uφ. The particular case where ~uφ = 0 defines a material surface, as it is the case of the free-surface.
Whether a surface is defined as material or non-material it should always be a continuous surface and by such definition one can use equation (3.2) and obtain,
DS S (x, y, z, t) = 0 =⇒ L = 0 (3.24) L Dt
22 3.2. Governing equations for stratified flows
uF
u n uS S uF u S‘ F,t = t+dt uF,n F
St+dt uS
St=S’t
Figure 3.2: A non-material surface
that leads to the corresponding interface continuity equation,
DS ∂I ∂I ∂I L = L + u L + v L − w = 0 (3.25) Dt ∂t I ∂x I ∂y I
If fluid velocity is the vector sum of the surface velocity and the equivalent flux velocity then ~uI = ~u − ~uφ and one can rewrite equation (3.25) as,
∂I ∂I ∂I L + (u − u ) L + (v − v ) L − (w − w ) = 0 (3.26) ∂t φ ∂x φ ∂y φ which will simply yield,
∂I ∂I ∂I ∂I ∂I L + u L + v L − w + u L + v L − w = 0 (3.27) ∂t ∂x ∂y φ ∂x φ ∂y φ
The flux across the interface, ΦIL , is defined as the projection of the equivalent flux velocity ~uφ on the interface normal since the tangential component of ~uφ represents a displacement along the surface and not through it, thus carrying no flux, as displayed in Figure 3.2. The surface normal is herein defined as a unit vector collinear with the surface gradient.
∂I ∂I ∇~ I L + L − 1 Φ = ~u · ~n = ~u · L = ~u · ∂x ∂y (3.28) IL φ IL φ φ r k∇~ I k 2 2 L ∂IL ∂IL ∂x + ∂y +1
For the sake of simplicity an additional hypothesis is proposed: the interfaces are surfaces who happen 2 ∂IL to be smooth enough so that ∂x ≈ 0 and so equation (3.28) can be written as,
∂I ∂I Φ = ~u · ∇~ I = u L + v L − w (3.29) IL φ L φ ∂x φ ∂y φ
Merging equations (3.29) and (3.27) and multiplying all the resulting terms by the fluid density at the
23 3. Conceptual Model interface will yield
∂IL ∂IL ρ|z=IL + [ρui]|z=IL − [ρw]|z=IL = ΦIL ρ|z=IL (3.30) ∂t ∂xi
3.2.1 Conservation of mass
Subscript notation can be used to write a more compact form of the mass conservation equation for incompressible flows derived in Chapter 3.1, (equation (3.10))
∂ρ ∂ρu + i = 0 (3.31) ∂t ∂xi
To integrate the terms of this equation over the depth of each layer it is necessary to make systematic use of the Leibniz’ Integral Rule, which states that,
Z b(β) ∂ ∂ Z b(β) ∂a(β) ∂b(β) f(x, β) dx = f(x, β) dx + f(a(β), β) − f(b(β), β) (3.32) a(β) ∂β ∂β a(β) ∂β ∂β and where all variables are generic. For presentation and readability purposes the depth-averaging will be carried out separately for each term. The integration sequence will also follow the order in which these terms appear in the mass conservation equation.
Recalling the depth-averaged function definition and the Leibniz Integral Rule, equations (3.21) and (3.32), respectively, the depth-averaging of the local term yields,
Z Lu Z Lu ∂ ∂ ∂Ll ∂Lu ρ dz = ρ dz + ρ|z=Ll − ρ|z=Lu Ll ∂t ∂t Ll ∂t ∂t ∂ ∂L ∂L = [ρ h ] + l ρ| − u ρ| (3.33) ∂t L L ∂t z=Ll ∂t z=Lu
where t is time, xi are the spatial coordinates (x, y, z) assuming that z is a vertical coordinate, ~ez = ~g/|~g|,
ρL is the depth-averaged density within a layer L enclosed by lower and upper interfaces, denoted by Ll and Lu respectively. This procedure can be also used to integrate the convective term, thus obtaining,
Z Lu Z Lu ∂ ∂ ∂Ll ∂Lu ρui dz = ρui dz + [ρui]|z=Ll − [ρui]|z=Lu Ll ∂xi ∂xi Ll ∂t ∂t ∂ ∂Ll ∂Lu = [ρLUiL hL] + [ρui]|z=Ll − [ρui]|z=Lu (3.34) ∂xi ∂xi ∂xi which is accurate for both xx and yy axis. The zz axis is a particular case of this integration,
Z Lu Z Lu ∂ ∂ ∂Ll ∂Lu ρw dz = ρw dz + [ρw]|z=Ll − [ρw]|z=Lu (3.35) Ll ∂z ∂z Ll ∂z ∂t because the first term on the right hand side of equation (3.35) is obviously null. Following equation
24 3.2. Governing equations for stratified flows
(3.22) and provided that the interfaces are defined as surfaces of the form FL(x, y, z, t) = IL(x, y, t) − z, equation (3.35) simplifies to,
Z Lu ∂ ρw dz = [ρw]|z=Lu − [ρw]|z=Ll (3.36) Ll ∂z concluding the depth-averaging of the continuity equation. Finally, summing equations (3.33), (3.32) and (3.36) will yield,
∂ ∂ ∂Ll ∂Ll [ρLhL] + [ρLUiL hL] + ρ|z=Ll + [ρui]|z=Ll − [ρw]|z=Ll ∂t ∂xi ∂t ∂xi ∂Lu ∂Lu − ρ|z=Lu [ρui]|z=Lu − [ρw]|z=Lu = 0 (3.37) ∂t ∂xi which is solely an extended form of the mass conservation equation when integrated over the depth of a given layer L,
Z Lu ∂ρ ∂ρu ∂ρv ∂ρw + + + dz = 0 Ll ∂t ∂x ∂y ∂z
Equation (3.37) must still undergo further manipulation to clearly reveal the vertical fluxes through the interfaces of overlying layers. Most of the interfaces featured in this model are non-material, being the free-surface, the topmost surface, the only exception. Finally, by recalling equation (3.37)
∂ ∂ ∂Ll ∂Ll [ρLhL] + [ρLUiL hL] + ρ|z=Ll + [ρui]|z=Ll − [ρw]|z=Ll ∂t ∂xi ∂t ∂xi ∂Lu ∂Lu − ρ|z=Lu + [ρui]|z=Lu − [ρw]|z=Lu = 0 ∂t ∂xi one can notice that the terms inside the brackets are in fact the vertical fluxes, as defined in (3.30). Performing such substitution will yield the mass conservation equation for the conceptual model in its final differential form,
∂ ∂ ∂ [ρ h ] + [ρ U h ] + [ρ V h ] = Φ ρ| − Φ ρ| (3.38) ∂t L L ∂x L L L ∂y L L L Lu z=Lu Ll z=Ll
where ΦLu ρ|z=Lu and ΦLu ρ|z=Ll are the net mass fluxes between Layer L and the layer above, through interface Lu, and also with the layer underneath, through interface Ll.
Another particular feature of this formulation is that every layer that happens to have an appreciable concentration of suspended sediment justifies two distint mass conservation equation: one for the solid phase - suspended sediment - and another for the liquid phase - clean water. Recalling the definition of mixture density, introduced in (3.5),
ρL = ρw (1 + CL(s − 1)) (3.39)
25 3. Conceptual Model and bearing in mind that it evidences two distinct phases,
ρL = ρw(1 − CL) + ρsCL (3.40) substituting (3.40) in (3.38) will result in a single equation that implicitly contains both solid and water mass conservation equations. For the water mass conservation equation one has,
∂ ∂ ∂ [(1 − C )ρ h ] + [(1 − C )ρ U h ] + [(1 − C )ρ V h ] = ∂t L w L ∂x L w L L ∂y L w L L
= (1 − CL)ΦLu ρ|z=Lu − (1 − CL)ΦLl ρ|z=Ll (3.41) while the solid mass conservation equation is simply written as the supplementary part,
∂ ∂ ∂ [C ρ h ] + [C ρ U h ] + [C ρ V h ] = C Φ ρ| − C Φ ρ| (3.42) ∂t L s L ∂x L s L L ∂y L s L L L Lu z=Lu L Ll z=Ll
Applying equations (3.41) and (3.42) to the layers featured in the conceptual model will complete the final set of mass conservation equations.
For the suspended sediment layer, Layer [1] in Figure 3.1, one can assume that the sediment concentration is negligible and therefore the depth-averaged layer density is roughly the same as clean water density and so only equation (3.41) is relevant. If one considers the water density to be constant in both time and space then it can be removed from the conservation equation,
∂ ∂ ∂ h + [U h ] + [V h ] = − Φ (3.43) ∂t 1 ∂x 1 1 ∂y 1 1 1,2
where U and V are the depth-averaged velocities in the ~ex and ~ey directions, respecively, and I1,2 is the interface between Layer [1] (clear water) and Layer [2] (bedload). High concentration of sediment is expected in Layer [2] which leads to using both (3.38) and (3.42),
∂ ∂ ∂ [h ] + [U h ] + [V h ] = Φ − Φ (3.44) ∂t 2 ∂x 2 2 ∂y 2 2 1,2 b
∂ ∂ ∂ [C h ] + [C U h ] + [C V h ] = C Φ − (1 − p)Φ (3.45) ∂t 2 2 ∂x 2 2 2 ∂y 2 2 2 2 1,2 b
where b is the bed, p is its respective porosity, Φ1,2 is the mass net flux between layers [1] and [2], Ib is the bed interface and Φb is the net flux between the bed and Layer [2]. Figure 3.3 can further clarify the physical meaning of the previous variables.
Since the bed is assumed to be saturated and to have no velocity at all there is only need for a sediment
26 3.2. Governing equations for stratified flows mass conservation equation in this layer. Hence the only relevant equation is (3.42)
∂ (1 − p) Z = (1 − p)Φ = D − E (3.46) ∂t b b
where p is the bed porosity, Zb is the bed elevation and D and E are equivalent deposition and erosion rates, respectively.
U1 Clearwater Layer C1 ≈ 0 h1 Layer [1] Φ 1,2 I I1,2 1,2 U Bedload Layer h C Φ 2 Layer [2] 2 2 b I I Z b b b Bed Layer [b]
Figure 3.3: Layer set featured in the conceptual model
3.2.2 Conservation of momentum
In this section the momentum conservation equation derived in Chapter 3.1, equation (3.16), will undergo the same depth-averaging technique as the mass conservation equation, in the previous section. Recalling the momentum conservation equation,
∂ρui ∂ρui ∂p ∂τij + uj = − + ρgi + ∂t ∂xj ∂xi ∂xj one can use it to justify the hydrostatic pressure field. If the vertical velocity is negligible so must be the vertical acceleration which leads to that both left-hand side of equation (3.16) and stresses τzj are null. This will simply resume (3.16) to,
∂p Z h ∂p Z h = ρg ⇔ dz = ρg dz ⇔ p(h) = ρgh (3.47) ∂z 0 ∂z 0 while showing why the pressure field is assumed to be hydrostatic.
Again, to improve the readability and understanding of the following steps, all integrations will be performed separately for each hand of the momentum conservation equation, and also a term at a time whenever deemed necessary. One should also note that the horizontal components of gi are null. Applying the Leiniz’ Rule to the integration of the left-hand member over the depth of a given layer L will yield,
27 3. Conceptual Model
Z Lu ∂ρui ∂ ∂Ll ∂Lu dz = [ρLUiL hL] + [ρui] |z=Ll + [ρui] |z=Lu (3.48) Ll ∂t ∂t ∂t ∂t Z Lu ∂ρui ∂ ∂Ll ∂Lu dz = [ρLUiL hL] + [ρui] |z=Ll + [ρui] |z=Lu (3.49) Ll ∂xj ∂xj ∂xj ∂xj where the i index now stands for the 2D horizontal space directions - i = [1, 2]. One should note the existing similarities between these two equations and equations (3.33) and (3.34). The steps used to derive the vertical mass fluxes can also be applied to full extent on equations (3.48) and (3.49). After those manipulations one obtains,
∂Lu ∂Lu [ρui] |z=Lu + [ρuiuj] |z=Lu − [ρuiw] |z=Lu = ΦLu [ρui] |z=Lu (3.50) ∂t ∂xj ∂Ll ∂Ll [ρui] |z=Lu + [ρuiuj] |z=Ll − [ρuiw] |z=Ll = ΦLl [ρui] |z=Ll (3.51) ∂t ∂xj
which represents the vertical net flux of momenta, in the ~ei direction, through the interfaces of some layer L. The depth-averaged left-hand member of (3.16) can be therefore written as
∂ ∂ [ρLUiL hL] + ρLUiL ULj hL + ΦLl [ρui] |z=Ll − ΦLu [ρui] |z=Lu (3.52) ∂t ∂xj
For the right-hand member of (3.16), the applied external forces, the same logic must be followed in order to achieve a consistent set of depth-averaged quantities.
The first term to be integrated is the pressure field, which was shown to be hydrostatic,
Z Lu Z Lu ∂p ∂ ∂Ll ∂Lu dz = p(z) dz + p|z=Ll − p|z=Lu (3.53) Ll ∂xi ∂xi Ll ∂xi ∂xi
The depth-averaged pressure is obtained by using the variable substitution ξ = z−Ll while also accounting for for the weight of all the nL layers overlying layer L,
n Z Lu Z hL 1 XL p(z) dz = p(ξ) dξ = γ h2 + γ h h = P (3.54) 2 L L Ll i L L Ll 0 i=L+1
and γL is the same as ρLg. The depth-averaging of the viscous stresses is performed by integrating the deviatoric stress tensor over the depth of a layer, τij, yielding,
Z Lu Z Lu ∂τij ∂ ∂Ll ∂Lu dz = τij dz + τij|z=Ll − τij|z=Lu Ll ∂xj ∂xj Ll ∂xj ∂xj ∂ ∂Ll ∂Lu = TLij hL + τij|z=Ll − τij|z=Lu (3.55) ∂xj ∂xj ∂xj
Adding equations (3.52) to (3.55) will yield a 2D depth-averaged equation of the momentum conservation
28 3.3. Closure equations equation that, even while apparently more complicated, is in fact simpler to solve as it involves fewer unknowns, due to the removal of the vertical velocity, and thus reduces the governing system to three conservation laws. So, the 2D version of equation (3.16) is given by,
∂ ∂ ∂ [ρLUiL hL] + ρLUiL ULj hL + ΦLl [ρui] |z=Ll − ΦLu [ρui] |z=Lu = − PL ∂t ∂xj ∂xi ∂Ll ∂Lu ∂ ∂Ll ∂Lu − p|z=Ll + p|z=Lu + TLij hL + τij|z=Ll − τij|z=Lu (3.56) ∂xi ∂xi ∂xj ∂xj ∂xj
Applying equation (3.56) to Layer [1] will yield the following result,
∂ ∂ [ρ1U1i h1] + ρ1U1i U1j h1 + ΦI1,2 [ρui] |z=I1,2 = ∂t ∂xj ∂ ∂I1,2 ∂ ∂I1,2 = − P1 − p|z=I1,2 + T1ij hL + τij|z=I1,2 (3.57) ∂xi ∂xi ∂xj ∂xj and similarly for Layer [2] one should obtain,
∂ ∂ ∂ ∂Zb [ρ2U2i h2] + ρ2U2i U2j h2 − Φ1,2 [ρui] |z=I1,2 = − P2 − p|z=Zb ∂t ∂xj ∂xi ∂xi ∂I1,2 ∂ ∂Zb ∂I1,2 + p|z=I1,2 + T2ij h2 + τb,j − τij|z=I1,2 (3.58) ∂xi ∂xj ∂xj ∂xj
A total momentum conservation equation, accounting for both layers, is obtained if one sums equations (3.57) and (3.58). Considering the total values of momentum, momentum flux and applied external forces for both layers as a whole one obtains,
∂ ∂ ∂ ∂Z ∂ ∂Z [ρU h] + [ρU U h] = − P − b gh + T h + b τ (3.59) i T i j T T T ijT T b,j ∂t ∂xj ∂xi ∂xi ∂xj ∂xj where all the variables with index T represent the sum of their equivalents over both layers 1 and P2 2: generically ϕT = i=1 ϕi, where ϕ can be ρUih, ρUiUjh, P , h or Tijhi. Due to the hydrostatic pressure field it is also equivalent to write ∂xi Zb ghT instead of ∂xi Zb p|z=Zb . Provided that there are no discontinuities in the velocity field across the interface that bounds both layers 1 and 2, Ferreira et al.
(2009) showed that the vertical fluxes of momentum Φ1,2 [ρui] |z=I1,2 and the terms ∂xi I1,2 p|z=I1,2 are cancelled when (3.57) and (3.58) are summed.
The governing system of equations has thus been formally derived and its complete definition depends only on adequate closures or closure equations, which will be derived in the next section.
3.3 Closure equations
The proposed system of governing equations, namely (3.43), (3.44), (3.45), (3.57) and (3.58), requires closures for the dynamics of the bedload layer, h2 and U2, for the evolution of bed morphology, Φb or
D − E, and finally for flow resistance, τb,j and TLij (Canelas, 2010; Ferreira et al., 2009).
29 3. Conceptual Model
The closures for the dynamics of the bedload layer were derived under the premise that greater kinetic energy fluxes will lead to an increase of the bedload layer thickness, thus allowing for a complete dissipation of the kinetic energy fluctuating components (Ferreira, 2005).
The proposed closure for the bedload layer thickness, h2, is given by, (Ferreira, 2005)
h2 = m1 + m2θ (3.60) ds
where ds is the significant particle diameter, m1 and m2 are parameters that are dependent upon the mechanical properties of the particles and fluid viscosity and lastly θ is the Shield’s parameter, calculated 2 as θ = Cf k~uk / (gds(s − 1)) where Cf is the friction coefficient. The bedload layer depth-averaged velocity,
U2, is given by a power law derived by Ferreira (2005),
h 1/6 U = U 2 (3.61) 2 h
Bed morphology closures relate the existing imbalance between solid discharge and transport capacity with bed elevation changes. The solid discharge, qs, is given by qs = CρU while several formulas can be ∗ used for the transport capacity, qs , namely Meyer-Peter & Muller (1947), Bagnold (1966) and a specific formula for stratified flows and debris flows by Ferreira (2005). The closure equation for ∂tZb is thus given by,
∂Z q − q∗ b = s s (1 − p)−1 (3.62) ∂t Λ where p is the bed porosity and Λ is an adaptation length defined as a two horizontal asymptotic curve
Λ(Λmin, Λmax, θref , θ) whose arguments Λmin, Λmax and θref are parameters used to ensure that the value of Λ will never grow too large nor will it decrease to nearly zero.
Closures for flow resistance feature bottom friction, viscous stresses and turbulent stresses. The bedload layer is expected to carry a high concentration of sediment leading to the same rheological behavior of an hyperconcentrated flow.
In such conditions the total shear stress is the sum of three main components: the yield stress, τy, the viscous stresses, τν , and the turbulent stresses, τt. The cohesive nature of very fine particles is taken into account through constant values of the yield stress, since most of the presently available formulae are only supported by very singular experiments.
A useful property of a Newtonian fluid is that the viscous stresses can be expressed in terms of both dynamic viscosity, µ, and vertical velocity gradients, ∂zui. As for the turbulent stresses induced by bottom interaction they can be associated to the second power of the vertical velocity gradient as these
30 3.3. Closure equations are related to the kinetic energy of the flow (Takahashi, 2007). Total shear stress is thus given by
∂u ∂u τ = τ + τ + τ = τ + µ i + ρ d2 c i (3.63) i y ν t y ∂z s f ∂z
The bed shear stress is given by,
τbi = Cf ρk~ukui (3.64)
where the friction coeficient Cf can be given by the Manning-Strickler formula,
1 gh 3 Cf = 2 (3.65) Ks or by a specific formula concerning stratified flows or debris flows (Ferreira et al., 2009),
k~ukds Cf = (3.66) hωs
where ωs is the settling velocity of the sediment particles (Jimenez & Madsen, 2003) and Ks is the Manning coefficient.
For the turbulent stresses a simple κ − model is applied (Pope, 2000). This model ignores small-scale eddies in the motion and calculates a large-scale motion with an eddy viscosity that carries and dissipates the energy of the smaller-scale flow.
Under these assumptions one can define the depth-averaged stress tensor Tij as,
∂ui ∂uj Tij = ρνT + (3.67) ∂xj ∂xi
where νT is the eddy viscosity defined in terms of the depth-averaged kinetic energy, κ, and turbulent kinetic energy dissipation rate, ,
2 22 κ 4.375u∗ νT = Cµ = 0.09 3 (3.68) 5u∗/h
√ where u∗ is the shear velocity or friction velocity, calculated as u∗ = τb/ρ.
The system of equations is now completely defined featuring governing equations (3.43), (3.44), (3.45), (3.57), (3.58) and closure equations (3.63) to (3.68).
31 3. Conceptual Model
32 Chapter 4
Numerical Model
4.1 Discretization scheme
4.1.1 Introduction
Eulerian mesh-based methods featuring finite difference, finite element or finite volume schemes are amongst the most widely used tools in computational fluid dynamics (CFD) (Hircsh, 1988). The employed numerical model (Canelas, 2010; Murillo & García-Navarro, 2010) stands on a fully conservative finite volume method (FVM), that has now became a standard within the CFD world. Its important property of providing a proper discrete representation of the conservation laws when discontinuities are present in the domain (Hircsh, 1988; LeVeque, 1990; Toro, 2009) is a valuable advantage over other methods.
Only one flow layer is currently implemented in the computational model and hence the governing equa- tions derived in Chapter 3 can be simplified. The absence of adjacent flow layers vanishes all the vertical flux terms obtained when depth-averaging the governing equations with Leibniz’s Integral Rule, as demon- strated for the total momentum conservation equation (3.59) in chapter 3. The only exceptions are the total and sediment mass fluxes between the flow layer and the bed. The governing system features total mass conservation, equation (4.1), total momentum conservation for the x and y directions, equations (4.2) and (4.3), respectively, and sediment mass conservation, (4.4).
∂th + ∂x (hu) + ∂y (hv) = −∂tZb (4.1)