Overtopping on Grass Covered Dikes: Resistance and Failure of the Inner Slopes

(1)

Overtopping on grass covered dikes

(2)

(3)

Overtopping on grass covered dikes

Resistance and failure of the inner slopes

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft,

op gezag van de Rector Magnificus prof. ir. K. C. A. M. Luyben, voorzitter van het College voor Promoties,

in het openbaar te verdedigen op maandag 19 mei 2014 om 15:00 uur

door

LÊ Hai Trung

Master of Science in Civil Engineering, Delft University of Technology

(4)

Prof. dr. -ing. H. Schüttrumpf RWTH Aachen University

Assoc. prof. dr. T.Q. Tuan Water Resources University, Vietnam Prof. dr. ir. S.N. Jonkman Technische Universiteit Delft, reservelid

This research program has been financially supported by the Royal Netherlands Embassy in Hanoi through the project ‘Upgrading Training Capacity in Coastal En-gineering’ at Water Resources University in Vietnam.

This thesis should be referred to as: Trung, L.H. (2014). Overtopping on grass cov-ered dikes - Resistance and failure of the inner slopes. Ph.D. thesis, Delft University of Technology.

Keywords: damage, dike, erosion, flow, grass, overtopping, resistance, root, soil

Printed by: Ipskamp Drukkers B.V., the Netherlands

This document was typeset using the TU Delft template (2013/07/08 v1.0 disser-tation class) and house style fonts in XƎLA_TEX.

ISBN 978-94-6259-170-7

An electronic version of this dissertation is available at

(5)

Preface

More than five years ago, I took part in a project supporting the Vietnamese govern-ment to build up new guidelines on design of sea dikes. The project aimed to cover a large number of issues tide regimes, wave conditions, dike materials, foundation, dike stability, … For centuries, our sea dikes had been built manually and mainly with accumulated experience. Under attack of some 6 to 7 typhoons every year, dike failures often start with erosion on the landward slopes due to overtopping. Little was known about the strength and stability of these slopes, especially which were covered with grass. It was this concern that brought me to Delft again to start my research.

It was a blessing that the research topic was clear at the beginning. However, it took me a couple of years to clearly define the scope. The research aims to look more comprehensively at a grass covered slope working under wave overtopping flows. To this end, several issues need to be explored. First, observations on the real dikes certainly help to better understand how damage takes place. By chance, the simulator tests started earliest giving an overall impression of the phenomenon in question. Second, the overtopping flows which are the destructive factor need to be quantified. Thanks to the available data, a set of new formulas could be estab-lished to estimate these flows. Third, inspired by previous studies measurements were conducted to determine how grass roots enhance the soil strength. All these works stimulate the research to go further. As a result, the obtained findings were integrated to classify different manners of damage to a slope. And to predict its re-sistance against overtopping, a model was developed using some ideas originated in the field of materials engineering.

In line with many studies, the research is contributing to improve the under-standing of how a slope behaves when subject to overtopping. In general, it underlines the relationship between the loadinduced by the water flow and the

strength produced by the grass cover when evaluating a sea dike during a storm surge. Furthermore, it shows that grass is a durable material. Hopefully, the thesis will encourage the application of grass in protecting the dike slopes in developing countries like Vietnam.

To conduct this study, I have obtained guidance and support from my promoters Prof. drs. ir. J.K. Vrijling and Prof. dr. ir. M.J.F. Marcel Stive and my co-promoter Assoc. prof. ir. H.J. Verhagen. I would like to express my gratitude to them. I highly appreciate all the other committee members for reviewing my work and giving critical comments. Besides, Dr. M.L. Beliën, M.A. S.F. Johnson and Ir. J. van Veen are acknowledged for their help to correct and improve the English text of this thesis.

Many studies are based upon data, specially those in engineering. In this the-sis, a large part of the data has been derived from the project ’Technical Assistance

(6)

and the Dike Department (Vietnam Ministry of Agriculture and Rural Development) for their coordination and help. Finally, I would like to thank my colleagues who always show me a great hospitality at the Department of Hydraulic Engineering.

Lê Hải Trung Delft, February 2014

(7)

1

Introduction

Sea dikes are essential in protecting people and land across the world. How-ever, this kind of coastal defence is considered and understood differently depending on the development level of each country. This thesis is an at-tempt to increase the knowledge about the sea dikes in Vietnam. To start, the current chapter underlines the rationales behind the research and formu-lates the main objectives.

1.1. Motivation

I

n Vietnam, the first sea dikes were built to create extra land for the members ofthe royal family in the Thai Binh and Nam Dinh provinces during the Ly (1010 -1225) and Tran dynasty (1225 - 1400) (e.g.,Kim,1971). Later on, sea dikes were continuously enlarged and lengthened by local communities. The present system including estuary and sea dikes is clearly defined along the coastline of Quang Ninh, Hai Phong, Thai Binh, Nam Dinh and Ninh Binh provinces with a total length of more than 700 km as shown in Figure1.1.

In general, the master plan of the sea dikes is based on the coastline retreat strategy. In the areas of some estuaries such as Van Uc, Thai Binh, Ba Lat and Ninh Co, where the coastline is subject to accretion, polders have been created. Dike stretches were built one after the other to extend towards the sea protecting thousands of hectares, like for instance the Binh Minh dike in Kim Son district, Ninh Binh province. In some other areas, where the coast is subject to erosion, dikes have been built in parallel pairs (a primary line and a secondary line) and transversal dikes have been added in order to reduce damage in the case of flooding (Cong,

2010). Furthermore, dikes have also been constructed around some islands, such as the Ha Nam dike in Quang Ninh and the Cat Hai dike in Hai Phong. The fundamental function of these systems is to protect agricultural land from (sea) flooding and to prevent salt water intrusion.

A comprehensive report on the sea dikes in Vietnam was submitted for the first 1

(12)

Figure 1.1: The northern coast of Vietnam from Quang Ninh to Ninh Binh provinces (source: National Hydro - Meteorological Servicehttp://www.nchmf.gov.vn/).

time in 1991 (MWRI,1991). Between 1996 and 2000, about 308 km of sea dikes were improved and 76 km of revetments were constructed from Quang Ninh to Ninh Binh within the framework of the project PAM 5325, which was funded by the World Food Program and the Vietnamese government (MARD,2004). During the period from 2006 to 2010, the government invested more than 3000 billion Vietnam Dong (approximately 140 million US Dollars) in strengthening 272 km of sea dikes and planting 132 hectares of mangrove from Quang Ninh to Quang Nam (MARD,

2010).

Vietnam is located in the north west of the Pacific Ocean, where about 30% of the typhoons in the world are formed (afterThao et al.,2000;Thuy,2003). In the period from 1951 to 2004, on average, 28 tropical typhoons originated from the Northwest Pacific Ocean every year and 10 of them passed over the South China Sea, which is known as ’Bien Dong’ in Vietnamese (e.g., Tuan,2007). As a result, there are some 6 to 7 typhoons attacking the coastline of some 3000 km long each year, distributed evenly over three areas: the north, the northern and middle central part, and the south central and the south (e.g.,Tuan,2007;Villegas,2004;Bache and MacAskill,2012). The maximum wind field radius is relatively small in the range of 40 to 100 km while the maximum wind speed may reach 50 m/s (Thuy,2003).

Along the northern coast, tidal ranges can be over 3 m and storm surges higher than 2.5 m might occur in most areas (e.g.,Thuy,2003;Sao,2008). Besides, the sea dikes had been designed with exceedance probabilities of design water level of about 1/20 according to the design guidelines applied before 2012 (e.g.,14TCN 130, 2002; Cong, 2010; Kanning, 2012). In fact, the crest levels are commonly found to reach 5.0 m above Mean Sea Level (MSL). Apparently, dikes are mostly overtopped during storms with a mean discharge varying between roughly 30 and 300 l/s per m of dike length (e.g., Trung et al.,2008). In general, flood defences usually fail due to insufficient height and insufficient strength. Similarly, damage

(13)

1.2.Research objectives ..

1

3 induced by overtopping on the crests and the landward slopes has led to a large number of dike failures for the last decades. For example, Figure1.2shows a cross-section of the Hai Hau dike after a severe storm surge in 2005. Overtopping was estimated to contribute considerably up to 46% of the total failure probability of the sea dikes in Nam Dinh province (Cong,2010).

Figure 1.2: Cross-section of Hai Hau dike after breaching due to a storm surge in 2005, Nam Dinh province (photo: Trung, L.H.).

Under the severe attack of overtopping, grass is found as the most popular ma-terial when protecting (fully or partly) more than 80% of the Vietnamese landward slopes. However, little is known about the strength or the resistance against ero-sion of these slopes. For example, a discharge of 10 l/s per m is acceptable on a slope covered with good grass according to the new sea dike design guidelines (14TCN 1613,2012). In fact, this suggested value is basically based upon the de-sign codes currently applied in western countries such as the Netherlands, but not on any measurement or experiment conducted in Vietnam. Furthermore, how over-topping causes damage to a grass covered slope is not sufficiently understood. It is these issues that merely stimulate the prime concern of this thesis. The research objectives will be formulated in the next section.

1.2. Research objectives

The main research question addressed in this thesis is:

How does a grass covered slope fail when exposed to overtopping flows during a storm surge?

To provide the answer to the main question, the research is going to explore four key questions:

1. What is thedamagepattern of a grass covered slope subject to overtopping flows? (chapter 3)

(14)

standing potentially inspires how to design more durable sea dikes.

1.3. Thesis outline

T

he thesis is organised as follows. Chapterhydraulic loads and the resistance of a grass cover. Subsequently, results of2 reviews existing theories on the the simulator tests from 2009 to 2012 in Vietnam are described in Chapter 3(key question 1). Chapter 4 analyses existing data on overtopping to establish new formulas predicting the flow velocity and the water-layer thickness (key question 2). The new formulas will be compared to the measurements performed with the simulators. Chapter 5 investigates some properties predominantly governing the

strength of a grass cover (key question 3). Chapter 6 explores the damage to a grass covered slope due to overtopping, i.e. it reveals how thestrengthand theload

interact (key question 4). To appraise the resistance of a grass cover, the chapter will develop a model using theories originated in materials engineering. Together with conclusions, appropriate suggestions to design and evaluate a dike slope are presented in Chapter7.

(15)

2

Literature review

Under moderate overtopping conditions, grass has been used to protect sea dikes for ages. This chapter reviews the most relevant theory on overtopping flow and resistance of grass covered slopes. After that, it discusses the as-pects which should be covered when assessing the damage to a landward slope. By doing so, the scope of the current thesis will be revised. Conse-quently, several methods will be specified to achieve the research objectives. First, the chapter classifies failures induced by overtopping to clarify the fo-cused phenomenon.

2.1. Failure mechanisms due to overtopping

W

ave overtopping on a sea dike may cause damage to its crest and landward slope especially when grass is the protecting material. Various mechanisms of damage have been reported and explored in previous research works. For exam-ple, by performing extensive field studies,Cooling and Marsland(1954) found that seepage through the embankment might cause sliding of the landward slope in Es-sex and Kent flooding areas. Meanwhile, fine fissuring had developed to a depth of up to 1.2 m for embankments constructed of highly plastic clays. The mechanism ’turf set-off’ described in Hewlett et al.(1987) was observed as the first stage of the failure due to heavy overtopping (Schüttrumpf and Oumeraci,2004). Notably, severe erosion on the landward slopes has led to many dike failures in Nam Dinh province (e.g.,Trung et al.,2012).

Figure2.1distinguishes two failure mechanisms induced by overtopping. The first one mainly occurs due to water infiltration and sliding on too steep slopes, especially with an inclination between 1/1.5 and 1/2.0. This sliding may directly cause dike collapse. However, it is not likely to be induced by wave overtopping as such, but rather due to large quantities of water infiltrating the dike body. This process may be aggravated by heavy rainfall. It is also possible for such failures to take place on slopes with an inclination smaller than 1/2.

(16)

The present thesis focuses on the second mechanism induced by fast overtop-ping flows that may cause damage to a grass cover. If initial erosion occurs, it can extend to the material layers underneath and may possibly lead to a dike breaching. The next section describes existing theory on overtopping flow.

2.2. Hydraulic loads

C

onsequences of overtopping-induced dike erosion can be considerable. To un-derstand that, a large number of studies has been conducted encompassing field, laboratory and numerical methods. As a result, guidelines on overtopping have been issued across the world (e.g., Pullen et al.,2007;USACE,2008). Using these guidelines, various parameters can be quantitatively estimated for different purposes of design and assessment. Among those, an overtopping discharge is often considered as one of the most representative quantities. However, some au-thors have discussed the shortcomings of using an average flow rate to design and to estimate the resistance of a landward slope against overtopping (e.g., Young,

2005; Dean et al., 2010). For example, the same value of a time-average dis-charge𝑞 can be produced by a small number of high waves with large overtopping volumes or by a great amount of low waves with small volumes as depicted in Fig-ure2.2. Recent insitu tests have also revealed that the effect of large and small water volumes are different on a dike (e.g.,Van der Meer et al.,2010).

Figure 2.2: A mean overtopping discharge can be produced by either a great number of small waves or a small amount of large waves.

Therefore, considering overtopping flows with associated parameters such as a velocity and a water-layer thickness seems to be more relevant rather than a time-averaged discharge. And the flow characteristics are known for mainly depending

(17)

2.2.Hydraulic loads ..

2

7 on the flow regime. Earlier studies assumed that the regime of an overtopping flow is critical or supercritical on the landward side of a dike crest (e.g., Tega and Kobayashi, 1996). For most protected-side slopes, critical conditions will exist at the landward edge of the crest causing the flow to be supercritical on the inner slope (Hughes,2008). However, he also argued that this condition is unsteady and the peak velocities are sustained for only a brief time. The coming paragraphs will show how overtopping flows are estimated.

2.2.1. Flow velocity and water-layer thickness

Experiments were performed in wave flumes with small and large scale models to estimate the intermittent flows generated by overtopping (e.g., Schüttrumpf,

2001;Van Gent,2002;Schüttrumpf and Oumeraci,2005). The difference between the fictitious run-up level and the crest level(𝑅 , %− 𝑅 ) is widely considered as a measure of the (potential) energy of the overtopping water. This energy measure is therefore used to quantify the values of the flow velocity and water-layer thickness. The formulas reviewed here are only for the maximum values at the leading front of a flow. In fact, the velocity and thickness often decrease after the passage of the wave front at a certain position.

On the seaward slope

At the seaward crest edge, the flow parameters are determined from the peaks of the overtopping wave time series. Sometimes, these parameters represent the levels exceeded by only 2% of the total waves during the tests. The thickness is proportional to (𝑅 , %− 𝑅 ) while the velocity is related to √𝑅 , %− 𝑅 by empir-ical coefficients (Van Gent, 2002; Schüttrumpf,2001). However, the coefficients given by the two authors are different because measurements were performed on slopes with different inclinations 1/4 and 1/6. Bosman(2007) reanalysed their ex-perimental data and suggested new coefficients which are related to the seaward slope angle𝛼. In another work, the thickness directly depends on the wave height and is in inverse proportion to the run-up level (Schüttrumpf and Oumeraci,2005).

On the crest

Apparently, the roughness of the crest surface affects the flow. However, the friction effect is taken into account when estimating the velocity but not the thick-ness (Van Gent, 2002; Schüttrumpf, 2001). Besides, these parameters are ex-pressed as exponential functions of the horizontal distance to the seaward crest edge.

Van der Meer et al.(2010) reanalysed the data of Schüttrumpf and Van Gent and also included the results from the FlowDike project leading to different equations. On a smooth crest, the thickness remains constant and equals to about 1/10 of (𝑅 , %− 𝑅 ). Meanwhile, the velocity declines as an exponential function of the dimensionless distance to the seaward crest edge.

On the landward slope

Schüttrumpf and Van Gent based their works upon the assumption of steady flows. The authors derived theoretical expression for the wave front depth-averaged, slope-parallel flow velocity down the inner slope by simplifying the momentum equation. The former presented an iterative solution (alsoSchüttrumpf and Oumeraci,

(18)

sudden change of geometry (direction), thus resulting in higher turbulence. The water flowing on the slope now directly hits the horizontal part as a jet impact. It would appear that the jet impact and the turbulence enhance the shear stress around the transition area. The toe is exposed to more severe attacking forces, damage consequently gets higher opportunity to happen. Theory and observation of turbulent jets are treated in great detail in a book of Rajaratnam(1976). Valk

(2009) adapted the characteristics of a turbulent wall jet for overtopping flows to investigate the head-cut mechanism at the inner toe. Main features of her work will be discussed later in this chapter.

2.2.3. Turbulence

With specifications of leaves and stems, grass normally retards flow causing energy loss and turbulence. The roughness of a grass covered bed can be characterised by Chézy coefficient 𝐶. In the meanwhile, the depth-averaged relative turbulence intensity𝑟 presents how turbulent the flow is (e.g.,Hoffmans et al.,2008).

The maximum pressure fluctuation𝑝 [N/m ] near the bed under uniform tur-bulent flow conditions is found to be six times the instantaneous pressure𝜎 which is three times the mean bed shear stress𝜏 (afterHoffmans et al.,2008). Hence, we derive

𝑝 = 18𝜏 (2.1)

where the mean bed shear stress is defined as𝜏 = 𝜌(𝑢∗) [N/m ] with 𝑢∗ = 𝑈 √𝑔/𝐶 [m/s] is the bed shear velocity. The maximum pressure fluctuation 𝑝 is considered as the predominant load acting on the turf-element model (Hoffmans et al.,2008) which is described in the next section.

2.2.4. Discontinuous flow

Different from overflow, overtopping flows are unsteady in practice. Besides, every flow with its associated thickness and velocity are likely to vary over the crest and the landward slope. Using the concept of progressive collapse,Young(2005) indi-cated that a grass turf can be gradually weakened by repetitive loading. In case of consecutive overtopping events, load is applied and removed, and then reapplied again and again. Accordingly, the overtopping flows would exert more critical im-pact than the steady overflow. This process reflects the response of a grass turf to

(19)

2.3.Grass covers ..

2

9 a rapid load fluctuation and how it re-distributes loads.

2.3. Grass covers

G

rass covers would seem to be one of the most prevalent types of surface pro-tection on earthen structures around the world such as dikes, dams, embank-ments, walls, … In coastal engineering, primary functions of a grass cover is to work as a revetment protecting the dike body against erosion induced by loading mainly from waves (seaward and landward slopes) and currents (seaward slope). There-fore, a grass cover should be resistant against erosion. This property is mainly gained from the interaction between the soil and the grass (root system) (e.g.,

Burger,1984). Besides, other values can be considered and encouraged when fea-sible such as nature, recreation, landscape and historical/ cultural values. In this section, we shall see below how grass and soil composition contributes towards the performance of a grass cover.

2.3.1. Composition

A grass mat showed a high resistance against erosion during the tests at Knardijk, the Netherlands in 1970. This resistance mainly comes from the structure of the root system and not from the leaves and stems above the ground (Burger,1984). Additionally, some factors influence a grass in preventing erosion are how it cover the soil surface, the seeppage flow in the direction of slope, and surface irregu-larities (Hewlett et al., 1987). In general, a cover consists of grass and the soil layers on which grass is based. Figure2.3gives a definition of a grass cover with fundamental composition as usually found in literature (TAW,1997; Muijs,1999). The associated information is also rewritten in the present subsection as follows.

Figure 2.3: Composition of a grass cover (TAW,1997).

Herbage is composed of two parts: one on the ground including sward and stubble; and the other under the ground, roots. Sward and stubble are the green part that one can visually see and stand on. It is the grass sod or turf where roots mostly concentrate providing additional strength to the soil. The turf lives on and is supported by a substrate, in which the number of roots declines over depth.

(20)

Figure 2.4: A grass turf on a sea dike in Friesland, the Netherlands (drawing: Trung, L.H.).

A network of roots is often dense as shown in Figure 2.4, which graphically sketches a piece of grass turf taken from a sea dike in Friesland, the Netherlands. The roots act as anchor to keep the soil particles together. Fine aggregates are kept by very fine root hair and symbiotic fungal threads in the soil. Large and small particles are held together by coarser roots in a sort of network. It is the network of these fine and coarse roots that makes the sod a strong, springy and flexible and permeable layer. Two important characteristics contributing to the protective property of the turf are its flexibility and springiness, and the retention of soil par-ticles by the root network. Without the anchoring effect, individual parpar-ticles are easily swept away by water flows. Moreover, exposed parts like the stubble and the sward are effective in limiting the direct impact of flows on soil.

In this thesis, the term ’grass turf’ is basically understood as depicted in Fig-ures2.3and2.4. The turf includes the top soil layer of about 20 to 30 cm thick on a slope, where most roots concentrate. Sometimes, a ’grass cover’ may be found as well. In the context of erosion resistance, these two terms are exchangeable. Note that the thesis does not consider the sward and the subsoil normally associated with a grass cover.

(21)

2.3.Grass covers ..

2

11

2.3.2. Root reinforcement to soil

The root reinforcement to soil is governed by the size, density, Young modulus, and tensile strength of roots (e.g., Greenway, 1987). Root diameters might vary widely depending on species, e.g. 0.3 mm for Late Juncellus grass and 1.7 mm for Vetiver (Cheng et al.,2003). The decay of root volume is often expressed by an exponential function related to depth under the soil surface (Sprangers,1999;

Stanczak et al.,2007). Meanwhile, many claim the tensile strength depends on the diameter.

To estimate the strength of soil permeated with roots, experiments were con-ducted encompassing direct shear tests (Waldron,1977;Waldron and Dakessian,

1981;Stanczak et al.,2007;Mickovski et al.,2010) and insituones (Fan and Chen,

2010). The increase in the shear strength is often claimed to be due to the root system via several mechanisms such as stretching, sliding, pulling out and breaking (Waldron and Dakessian,1981). Some authors have modified the Mohr Coulomb theory to model how roots enhance soil (e.g.,Wu et al.,1979;Schmidt et al.,2001;

Pollen and Simon,2005; Hoffmans et al.,2008). The theory describes soil failure in terms of shear stress and effective normal stress along an frictional sliding plane (Lambe and Whitman,1969) and can be expressed as

𝜏 = 𝑐 + (𝜎 − 𝑝 ) tan 𝜙 (2.2)

where𝑐 is the effective cohesion [N/m ], 𝑝 is the pore water pressure [N/m ], 𝜎 is the soil normal stress [N/m ], 𝜏 is the shear strength [N/m ] and 𝜙 is the internal friction angle. Cohesion is the result of cementation, electrical bonding of clays and organic colloids and capillary tension, while 𝜙 represents the frictional interaction of individual particles and their interlocking. The effective normal stress is caused by the soil weight and by the pore water pressure.

It is commonly accepted that roots contribute to the soil strength by providing an artificial (apparent) cohesion and exert minor effect on the frictional component of strength (e.g.,Gray and Megahan,1981;Waldron and Dakessian,1981;O’Loughlin and Ziemer,1982;Pollen and Simon,2005). Similarly for root permeated soil like a grass turf, the shear strength is enhanced by roots and therefore can be modified as

𝜏 = 𝑐 + 𝑐 + (𝜎 − 𝑝 ) tan 𝜙 (2.3) where the additional term𝑐 is the artificial cohesion produced by roots [N/m ].

Wu et al.(1979) presents a model describing the reinforcement effect due to roots of grassland vegetation. Roots are presumed to have sufficient bond so that they only break at a failure cross-section but will not be pulled out, and they are originally perpendicular to the cutting plane. The increase in the shear strength, i.e. the apparent cohesion𝑐 , due to the bonding action of roots is

𝑐 = 𝜏 𝐴

𝐴(cos 𝜃 tan 𝜙 + sin 𝜃) (2.4) where 𝜏 is the root tensile strength [N/m ], 𝐴 is the area of roots on the failure plane [m ],𝐴 is the cross-section area of soil under investigation [m ] and

(22)

1982). At 41 sites in the Oregon Coast Range,Schmidt et al.(2001) estimated the cohesive reinforcement to soil due to roots of ≥ 1 mm in diameter. They found that median lateral root cohesion varies from 6.8×10 - 23.2×10 N/m in indus-trial forests to 25.6×10 - 94.3×10 N/m in natural forests. Additionally, Simon and Collison(2002) investigated the hydrologic and mechanical effects of different riparian tree and grass species on bank stability. They found that grass roots con-tributed 6×10 to 18×10 N/m to the soil strength and consequently increased the slope stability.

The importance of a grass cover to the slope stability has also received con-siderable attention. Stanczak et al.(2007) measured the root volume ratio of ten soil samples taken from sea dikes giving a good agreement with a distribution sug-gested earlier bySprangers(1999). Besides, the tensile strength of ten single roots was measured showing a narrow range between 3×10 and 8×10 N/m . Further-more, direct shear tests were performed at different depth below the soil surface in order to determine the increase in the shear strength induced by roots (apparent root cohesion). The measurements confirmed the Wu’s model.

In the model ofWu et al.(1979), all roots are assumed to break simultaneously at failure, i.e. their tensile strength are mobilized instantaneously. In practice, each root is likely to have a different strength and therefore will break gradually one by one. As a result, load is redistributed to intact ones progressively till the failure of the whole system. Applying a fiber bundle model for the progressive failure of roots,

Pollen and Simon (2005) estimated the mechanical effects of riparian vegetation on stream bank stability. The study revealed that Wu’s model overestimates the magnitude of root reinforcement to soil.

2.4. Experiments on grass covered slopes

E

xperiments have been performed concerning the resistance against erosion and failure mechanisms of a grass covered slope over the past 50 years. Various layouts and boundary conditions were tested from small spillways under overflow to dike slope subjected to wave overtopping. Reviewing these tests will suggest a approach to the first research question.

(23)

2.4.Experiments on grass covered slopes ..

2

13

2.4.1. Overflow and overtopping tests

Protection afforded to spillways by natural grass covers were studied in a flume and on fields in Australia within a research program aimed at improving the design and construction of farm dams (Cornish et al.,1967). In the first phase, Couch, Kikuyu and Rhodes grasses with ages varying from 5 to 30 weeks were grown from seeds in boxes filled with unconsolidated loam (very silty sand) and then transferred to the test flume. The test results indicated that a good cover of grass could protect the soil surfaces against relatively high flow velocities. The critical velocity was likely to increase with the age of grass. For example, 28-week-old Kikuyu grass which had a thick even surface cover gave the highest protection with a velocity of up to 5.5 m/s.

In the second phase, Couch, Kikuyu and Rhodes grasses were placed and main-tained over a winter on nine spillway channels with slope inclinations increasing gradually from 1/10 to 1/3 (Yong and Stone,1967). The grass quality varied from channel to channel but was considered ’moderate’ at the time of testing. The tested flow was entirely supercritical (thinner than 0.1 m) and aerated for half the length of these spillways. A grass cover was deemed to have failed when the flow over a duration of 2 or 3 hours (peak of a flood from a small catchment), hence leading to damage that the grass would not able to regenerate and recover. The resistance of these one-year-old grasses did not vary noticeably in the range tested with slope, nor with the flow velocity or depth. Permissible velocities for the three selected grasses with a dense and even cover was about 2.7 m/s. Two failure mechanisms were observed, Kikuyu grass mat was rolled up and then fell off the slope while the development of scour holes undermined the root formation of the other grasses.

As an attempt to reduce overtopping on the sea dikes,Anh(2007) observed and measured the performance of water flows through Vetiver grass. The test program included different densities of grass. And the results showed that Vetiver is able to withstand backwater flows with depths of up to 0.4 m.

Large scale model tests were performed to investigate the failure of a sea dike covered with grass in the Large Wave Flume (LWF) of the Coastal Research Centre (Piontkowitz,2009). Excluding the seaward slope, the model represented a typical sea dike commonly built in the Netherlands, Germany and Denmark. The dike body was constructed of a sand core and a clayey layer protected with grass sods which were excavated at the Ribe sea defence in Denmark. No major damage to the landwards slope was recognised with a time-averaged overtopping rate of up to 30 l/s per m.

Increase in pore pressure in and underneath the outer layer may decrease the shear strength, thus resulting in slope sliding. Van Hoven et al.(2011) investigated the sliding stability of a clay slope using field measurements during prototype-scale tests on real dikes. To stimulate sliding, water flowed 56 hours over two slope sections protected by grass. Each section was 30 m long. However, the expected mechanism did not take place. Nevertheless, observations and field measurements help to better understand the phenomenon as well as estimate the pore pressure building up.

(24)

spill-in the Netherlands, spill-includspill-ing various slope specifications. These test cites are pre-sented here by reproducing information given in previous proceedings (Van der Meer et al.,2009;Steendam et al.,2008,2010). In Delfzijl, Groningen, where the dike is merely constructed of clay, three sections were tested a normal grass cover, a Geo textile reinforced slope and a bare clay slope (the grass sod was removed). At the Boonweg dike in Friesland, four slopes had been managed and maintained in different ways for more than 15 years. The resistance of these sections was tested with the same discharge. At a single point in St Philipsland, Zeeland, a 1/2-inclination slope was covered with bad grass. In Kattendijke, Zeeland, the dike consists of a sandy core and an outer clayey layer of 60 cm thick. Two grass covered slopes were naturally or manually damaged; and two other sections were reinforced with a resin fixed gravel layer Elastocoast ® and open stone asphalt, respectively. The Afsluitdijk is constructed of two layers, 40 cm of clay on 1 m of boulder clay. Three separate sections were evaluated to investigate the influence of obstacles such as a staircase and a transition between the grass cover and a horizontal park-ing lot pitched with stone bricks. In Zwolle, the Vechtdijk is a 100% sand dike and covered with good grass. At the first section, there was a maintenance road crossing the slope while at the second one, a tree with a trunk diameter of 0.8 m was located in the berm. At the third and the forth sections, the gentle slopes were covered with grass; and a significant mole activity was evident. At Tholen, an erosion sensitive clay of about 0.4 m thick was present on a sandy core. The grass cover was very bad with open areas due to the mowing method; and a location was heavily used by sheep. The experimental program includes a staircase and a fence running from the crest to the toe.

The overtopping rate was gradually increased from small to large at every site. For example, 0.1, 1, 10, 30, 50 and 75 l/s per m were used for the grass covers; and 125 l/s per m for the specially reinforced slopes. The same hydraulic conditions with a significant wave height of 2 m and a peak period of 6 s (wave steepness of 4%) were applied with the exception of two sections at Vechtdijk which used a wave height of 1 m and 3 m, respectively. These conditions were considered to be representative for the Dutch sea and river dikes. In general, each certain discharge lasted for 6 hours.

In Belgium, tests were conducted on a 1/4-inclination slope of the Sigma dike and at three sections of a ring dike with a slope inclination of 1/2.5. These sections were different in terms of vegetation types, grass cover appearance, management

(25)

2.5.Erosion models ..

2

15 regime and the material components of the outer layer. The Sigma slope was regular and covered with good grass; some minor damage and bald spots were found. Meanwhile, the ring dike was irregular and covered with rugged grass including a lot of Stinging nettles and Thistles. Besides, some sheep trails were found to exist as well.

A significant wave height of 0.75 m (discharges of 1 and 10l/s per m) and 1.0 m (30 and 50 l/s per m) were used for the Sigma dike and two sections of the ring dike. Each discharge lasted for 2 hours as long as the expected peak of the local high water tide on the river. The third ring dike section was tested with a wave height of 3 m and flow rates increasing from 1 to 10 l/s per m. More details of the Belgian tests can be found inSteendam et al.(2011).

2.5. Erosion models

O

vertopping flows are likely to cause damage to a grass covered slope at dif-ferent positions under certain circumstances. In other words, a slope might probably be damaged at different positions in different manners. Several models are discussed in this section giving a brief overview of how to evaluate the resis-tance of a grass dike against overtopping flows expressed in velocity, turbulence and jet impact. Names and associated application areas of these models are shown in Figure2.5.

Figure 2.5: Models predicting damage and resistance of grass covered slopes due to overtopping flows.

For the sake of clarity, every model is described in a separate paragraph and only key parameters are presented. For further details and explanations the readers are referred to the original works. The views expressed by individual authors are not necessarily those of the other authors. A simple model is introduced first giving maximum velocity that a grass cover is able to handle with the course of time.

(26)

however little information of the application duration could be found. In principle, the resistance is solely depended on the cover quality. A good cover is a dense, tightly-knit turf developed for no less than two growing seasons. A poor one con-sists of uneven tussocky grass growth with bare ground exposed or a considerable part contributed by non-grass weed species. And new grass seems to have poor quality after the first season.

The diagram was developed for grass covers subject to overflow in the course of time. The overflow on a spillway or in an irrigation channel is obviously different from the overtopping flow on a dike slope. Nevertheless, these limit curves can provide a reference when studying the performance of a grass cover under impact of the intermittent flows generated by overtopping.

2.5.2. Turf set-off model

Based on a series of prototype tests on a typical clay bank,Marsland(1957) found that fissuring in the top layer made the clay ’as permeable as gravel’ due to the dominant effect of soil structure (also seeThorn,1966, chap. 24). He suggested when checking the stability of the top structured clay layer (without root), seepage is assumed parallel to an infinite long slope. By analysing failures due to overtopping in a number of observations and experiments,Young(2005) emphasised that direct surface erosion might contribute to the slope failure but not the sole mechanism. Young considered a shallow sliding failure ’turf set-off’ with the assumption of a full saturated top clayey layer. Two elements loadand strengthof the limit state function are determined as follows.

At about 30 cm under the slope surface, where there are only few roots, the shear stress generated by overtopping flows is combined together with the satu-rated weight of the soil yielding theloadto the top layer. Thestrength(resistance) is a modification of the Mohr-Coulomb model including cohesive contribution in-duced by roots whilst the top layer is assumed to have no cohesion. However, soil structure seems likely to decrease over depth, thus resulting in a gradually increasing cohesion which will improve the stability of the grass cover.

’Turf set-off’ failure occurs when theloadexceeds thestrength. A dike stretch of 30 m long was subject to overflow in 56 hours in the Netherlands (Van Hoven et al.,

2011). The dike would be completely saturated at the end of the test, however the sliding mechanism did not take place.

(27)

2.5.Erosion models ..

2

17

2.5.3. EPM model

Van den Bos (2006) proposed the model EPM (in Dutch: Erosiegevoelige Plekken Model) to assess the surface erosion on a landward slope. Variance over the slope often causes bare spots where grass does not protect soil. These spots are consid-ered weak points that facilitate erosion due to overtopping. The erosion process is assessed using the theory developed for scouring of a bed protection. The assump-tions are valid as long as the erosion depth does not exceed the grass sod thickness. The total depth is derived from the summation of every individual overtopping event in a given duration, e.g. during a storm.

In the EPM model, the strength mainly depends on the covering rate of a grass turf. Influences of the grass quality were not determined. And the model did not investigate the distribution of the bare spots or the spatial variance of the cover. In comparison with the model’s assumptions, Sprangers(1999) represents the strength of a grass turf with a distribution of roots over depth (alsoVTV,2006).

2.5.4. Turf-element model

Hoffmans et al.(2008) introduced the turf-element model to estimate the erodibility of a grass cover (see alsoHoffmans,2012). In the saturated zone, a turf aggregate with the dimensions of ℓ × ℓ × ℓ is subject toloadand strength. Theloadis the lift force𝐹 induced by the pressure fluctuations 𝑝 perpendicular to the grass cover. Thestrengthis the total of the submerged soil weight𝐹 , the frictional forces 𝐹 and the tensile force 𝐹 . On the sidewall, the critical frictional force 𝐹 depends on the critical rupture strength of clay𝐶clay, c and the critical mean grass shear stress 𝜏grass, c of the intersected roots. The critical mean tensile force is a combination of soil and grass roots on the bottom-element. The turf aggregate is unstable when the loadovercomes the strength, 𝐹 ≥ 𝐹 + 4𝐹 + 𝐹 , with 𝑝 at the top of the turf aggregate considerably declines over depth. The pressure fluctuation 𝑝 is compared with the soil normal stress𝜎_soilto determine whether a) the grass turf is gradually eroded if𝜏_{grass, c}gets minimum within the turf thickness; orb) a bulging mechanism takes place if theloadis large enough.

Furthermore,Hoffmans et al.(2008) proved that the critical Shields parameter applied in the horizontal movement of loosely packed materials can be used in evaluating the critical stage of turf aggregates. This means turf aggregates can move if𝜏 reaches or exceeds the critical mean bed shear stress 𝜏 . The strength of a grass cover is assumed to depend upon the root area ratio, the root diameter and associated tensile strength.

The critical depth-averaged velocity of a turf 𝑢 is derived from the relation-ship between the mean bed shear stress and the depth-averaged flow velocity 𝑢 (see subsection2.2.3). In line with work ofVan den Bos(2006), the Dutch scour approach is deployed in estimating the erosion development in time due to highly turbulent flow by comparing the flow velocity𝑢 and the critical one 𝑢 .

2.5.5. Head-cut model

Valk(2009) studied erosion induced by overtopping on a landward slope. The outer layer is about 0.5 to 1.0 m thick with a top soil of 0.2 m where most roots

(28)

con-which happens faster will take over the surface erosion to play a predominant role. The estimate of the surface erosion (stage 2 and 3) is partly based upon works of

Hoffmans et al.(2008) with an introduction of a depth dependency. The Site Spill-way Erosion Analysis (SSEA) model is adopted to evaluate the head-cut mechanism. When soil becomes unstable due to the shear stress exerted by the flowing water and the soil dead weight, the head-cut mechanism will take place. Valk (2009) expressed the head-cut velocity as a function of a hydraulic load and a material head-cut coefficient. The hydraulic load depends upon a specific discharge and a head-cut height varying between 50 and 100 cm.

2.5.6. Cumulative hydraulic load

A prediction method was developed to appraise the stages of damage to a slope sub-ject to overtopping (Van der Meer et al.,2010). The method compares the velocity 𝑢 (load) produced by every overtopping wave with a critical velocity𝑢 (strength) using a parameter called ’cumulative hydraulic load’ as ∑(𝑢 − 𝑢 ) [m /s ], in which only large waves with (𝑢 > 𝑢 ) are taken into account because smaller ones are supposed to negligibly stimulate the damage. Four criteria are defined as ’first damage’, ’various damaged locations’, ’failure’ and ’non-failure’ regarding the final situations of some slopes after being tested with the simulator. Depending on the strength, each slope is represented by a corresponding critical velocity𝑢 . Test re-sults in Vechtdijk were analysed to quantify the four criteria, for example ∑(𝑢 − 𝑢 ) is approximately 3500 [m /s ] at the ’failure’ stage.

The method contributes to assess the overall situation of a grass covered slope after being attacked by a given number of overtopping waves, i.e. after a certain storm. The method did not intend to shed new light on the failure mechanism of a grass cover. However, it partly proved that a small amount of big waves exerts a cumulative hydraulic load that comparable to what does a great number of small waves. Additionally, a stronger cover is likely to withstand higher flow velocities, i.e. a higher critical velocity.

2.5.7. Cumulative erosional work units

Using knowledge of erosion due to steady flow,Dean et al.(2010) determined the physical basis for the damage to a levee slope. In addition to a mean discharge, overtopping duration was taken into account to estimate three thresholds including

(29)

2.6.Discussion ..

2

19 velocity, shear stress and work for ’good’, ’average’ and ’bad’ grass. These erosion quantities were considered with respect to criteria relating velocities and duration for tolerable erosion as suggested byHewlett et al.(1987). A method was developed to evaluate the cumulative erosion work index, which should exceed a threshold to cause unacceptable damage to a levee slope, e.g. 0.492×10 m /s for a good cover. In general, dike crest elevations are lower when calculated applying the method than using the concept of an average discharge given in the Overtopping Manual (Pullen et al.,2007).

Similar to the preceding model (Van der Meer et al.,2010), the authors estab-lished thresholds of damage by solving mathematical equations rather than relating to mechanical properties of a grass cover. Unique values of critical velocity, shear stress and erosional work were attributed to a certain quality of grass. If the thresh-olds exist, these should be determined with regards to measurable parameters such as shear strength and thickness of a grass turf. Besides, the model did not distin-guish clearly the two situations acceptable and unacceptable erosion. Furthermore, together with the cross-section material composition may considerably influence the performance of a slope as explored later in the thesis.

2.6. Discussion

G

rass mat has been used to naturally protect the surface of soil structures acrossthe world for ages. A well established grass cover should mitigate the de-structive effects on the structures and enhance their strength. The chapter has presented a number of research works concerning overtopping flows, grass covers, resistance and performance of grass slopes exposed to overflows and overtopping flows.

In case of a landward sea dike slope, the hydraulic loads are mainly induced by the overtopping flows. The pressure fluctuations due to the flow turbulence are often considered as the acting load that causes damage to the matrix of soil and roots. Several authors made use of the continuity equation to estimate the velocity and water-layer thickness. However, overtopping flows are very turbulent and intermittent rather than steady and continuous. For this reason, more reliable and realistic estimates of these flows are necessary.

Research on grass performed earlier will serve as useful reference but cannot be directly applied in Vietnam. Besides, there is limited research into the resistance of grass covers on sea dikes. Therefore, grass covers on Vietnamese sea dikes absolutely deserve some serious investigation.

From early experiments on grass spillways in Australia to late field observations on sliding tests in the Netherlands, a grass cover always shows a certain resistance against erosion. This protective property is mainly derived from combining the strength of soil and roots. In Vietnam, there are limited studies on the reinforcement effects of roots to soil, especially on sea dikes. Therefore, root permeated soil obviously warrants more serious investigation.

Several models have been developed to quantitatively assess the resistance of a grass covered slope against overflows and overtopping flows. As one of the early models, velocity-duration curves were proposed by Whitehead et al. (1976) for

(30)

flow impacts and grass turf into a model to predict the head-cut mechanism. In short, these works concern individual mechanisms that might lead to the failure of a grass turf on the steep part (surface erosion) or at the toe (head-cut) of a landward slope. The strength of a turf is principally represented by a critical velocity depend-ing upon the specifications of soil and roots. Further, the damage magnitude is usually considered as a process of time and can be given in an erosion depth.

Additionally, some models have been established in line with the concept of a gradual process of damage. For example, the cumulative destructive effects are determined by adding every overtopping flow that exceeds a certain threshold such as a critical velocity or a critical erosional work (Van der Meer et al.,2010;Dean et al.,2010). The final results are compared to several criteria corresponding to the distinguishable stages of the slope like failure or non-failure.

In the context of failure investigation, it is first necessary to formulate the mech-anisms correctly. Under attack of overtopping, a slope can be initially damaged at different positions in different manners. For example, erosion might take place on the slope while head-cut often starts around the toe. Existing literature hardly provides any information of which circumstances facilitating which mechanism. To address the remaining issue, the present research attempts to distinguish types of damage with regard to the flow regimes and the slope specifications. To this end, the classification should be based on observations during experiments or real storms (more difficult). The hypothesis of bare spots (Van den Bos,2006) might properly help determine the probability of damage. Suspect areas therefore should be investigated to position the potential damage.

Last but not least, a tool (model) needs to be built up to appraise the perfor-mance of a grass covered slope subject to overtopping. Inspired by the cumulative effects (Van der Meer et al.,2010;Dean et al.,2010), the model should consider the damage as a progressive process in regard to successive attacking flows (over-topping events). To evaluate the destructive effects induced by these flows, critical parameters should be determined using findings on the properties of soil and roots, and root permeated soil (grass turf).

The above arguments help to revise the scope of the present thesis. Besides, the discussion suggests the approaches to be applied. To address the fours research objectives, several methods are specified as follows.

(31)

2.7.Method ..

2

21

2.7. Method

I

n coastal engineering, basic processes are usually studied with numerical and/ or physical models depending on which one is investigated. The phenomenon of wave overtopping on coastal defences has been studied intensively for decades using physical models, e.g. in wave flumes. On the basis of a large number of phys-ical experiments, formulas have been established to characterise the phenomenon, such as mean discharge, probability of overtopping, volume of a single overtopping event and distribution of overtopping waves during a storm surge (e.g.,Pullen et al.,

2007;USACE,2008).

However, the effects that overtopping has on grass covered slopes are not well understood. On the one hand, that is mainly due to the fact that these effects cannot be studied in small wave flumes, as it is impossible to scale down the prop-erties of soil and grass. On the other hand, if it is possible to build up a prototype dike cross-section in a large wave flume, the grass cover needs one or two years to become mature and suitable for testing. Besides, only one type of the grass cover can be tested for each cross-section.

To address this issue, dike slopes were tested in situ under normal weather conditions by means of a wave overtopping simulator (e.g., Van der Meer et al.,

2006) for the first time in Vietnam (key question 1). By doing so, it became possible to perform observations and measurements to provide insight into the formation and development of damage to different types of slope.

Root permeated soil (grass covers) and overtopping flows were separately stud-ied to better understand how they interact. An available set of data was re-analysed to establish new formulas which should be applicable for estimating flows gener-ated by both natural waves and the simulator (key question 2). Distinctive methods in plant protection and soil mechanics were deployed to determine the properties that govern the resistance of grass covers (key question 3).

The evidence derived from the simulator tests served to classify and predict damage to inner slopes subject to overtopping. The resistance of grass covered slopes was evaluated with regard to new findings on root permeated soil and over-topping flows (key question 4). Note that the present study is limited to the grass cover (or a grass turf); erosion of the underneath layers (e.g. clay and sand) or the residual strength of a dike body are not investigated.

(32)

(33)

3

Overtopping simulator tests

Little is known about the strength of a landward dike slope under overtopping attack. Some grass covered slopes were tested by a wave overtopping simu-lator in Vietnam in 2009 and 2010. The chapter describes the simusimu-lator and presents the test results. The tested grass covers could withstand a mean overtopping discharge varying from 20 to 100 l/s per metre of dike length during some hours. Damage usually started at bare spots, at the transition between the slope and the toe, at the transition between different materials, and around objects (e.g., big trees). These features reduce the strength of a grass cover and therefore should be avoided as much as possible.

(34)

number of sea dike failures. For example, overtopping was estimated to contribute up to 46% to the total failure probability of a sea dike in Nam Dinh province (Cong,

2010).

To date, little is known about the strength and stability of the landward slopes of the present Vietnamese sea dikes under the impact of overtopping. Trung et al.

(2008) estimated that overtopping discharges could vary roughly between 30 and 300 l/s per meter of dike length. According to the sea dike design guidelines14TCN 1613recently approved, a mean discharge of 10 l/s per m is accepted on a landward slope with good grass cover. However, overtopping discharge was not mentioned in previous guidelines because the dike crest was designed with regard to wave run-up criteria. In fact, the strength of a grass covered slope can only be quantified by real storm surges and by evaluating the post-storm condition of the structure. Fur-ther, due to difficulties in performing observations and measurements, it is almost impossible to investigate the damage process during storm surges. The critical rate of overtopping at sea dikes also remains an issue that needs to be further studied. Basic processes in coastal engineering are usually studied by using numerical and/ or physical models depending on which is investigated. There are two main concerns: first, the available theory describing the process and the corresponding solutions; and secondly, whether it is possible to physically model the process.

The phenomenon of overtopping on coastal defences has been studied inten-sively for decades using physical models. On the basis of a large number of physi-cal experiments, formulas have been established to characterise the phenomenon such as mean discharge, probability of overtopping, volume of a single overtopping event and distribution of overtopping waves during a storm surge (e.g.,Pullen et al.,

2007).

However, the effects that overtopping waves have on grass covered landward slopes are not comprehensively understood. On the one hand, that is mainly due to the fact that these erosion effects cannot be studied in a small wave flume as it is impossible to scale down the properties of soil and grass. On the other hand, if it is possible to build up a prototype dike cross-section in a large wave flume, a grass cover needs one or two years to get mature and be suitable for testing. Besides, only one type of grass cover can be tested for each cross-section. To address this issue, a device that can generate overtopping tongues on real dikes was developed in the Netherlands (Van der Meer et al.,2006;Van der Meer,2007;Van der Meer et al.,2008).

(35)

3.2.Operation and design of the simulator ..

3

25 The wave overtopping simulator was used to test the strength of various grass covered slopes of sea dikes and river dikes in the Netherlands and Belgium between 2007 and 2011. These tests revealed that a grass cover could withstand a time average overtopping rate in the order of 10 l/s per metre of dike length (Van der Meer et al., 2009). Damage was observed to start at weak areas such as the transition between the slope and the horizontal toe, and around obstacles such as a tree or a staircase (Steendam et al.,2010). The erosional resistance against overtopping of a undisturbed grass cover is determined mainly by the root system rather than the soil characteristics (Steendam et al.,2008).

In the north of Vietnam, the sea dikes differ in cross-sections, material compo-nents and grass species from those either in the Netherlands or in Belgium. Despite the various results obtained after the Dutch and Belgian tests, it is dangerous to apply these findings to Vietnam. Therefore, testing sea dikes with a simulator is strongly encouraged. In late 2008, the original design was adjusted and improved to make a second one. Since 2009, this new simulator was utilised to test the strength of several dike slopes in Hai Phong, Thai Binh and Nam Dinh provinces. Insituobservations and measurements help to understand better the performance of a grass cover subject to overtopping flows and to improve the present sea dike structures.

This chapter gives an overrall description of the tests conducted in Vietnam so far. To do so, the chapter incorporates materials presented in the previous techni-cal reports (Trung,2011,2012b) and proceedings (e.g.,Trung et al.,2010,2011,

2012). First, it outlines the operational principle of the simulator and the design. Next, the test sections and the experimental set-up are described. Following is the main results. After that, the chapter underlines some aspects regarding a mean discharge. Finally, it discusses the formation and development of the damage to a grass covered slope due to overtopping.

3.2. Operation and design of the simulator

U

nderstanding the damage to a slope due to overtopping is important in both designing and assessing a sea dike, especially during a storm surge. To explore this problem, attempts have been made at performing observations insituas well as testing with the wave flumes. However, researchers might have encountered many difficulties with scale factor, time consuming, expenses, … One solution is to produce overtopping flows at a real dike. By doing so, one can easily investigate the phenomenon in question. A device namely the Wave Overtopping Simulator was developed to realise this idea (Van der Meer et al.,2006).

The simulator, which is basically a water reservoir, is easy to be transported from one place to another and assembled or dismantled in a reasonable time (one to three days). When working, it is continuously filled with a certain and constant discharge of water and is emptied at predefined moments through two butterfly valves at the bottom. After getting out of the two valves, water is guided by a transition chute to change from vertical (in the tank) to horizontal direction to sim-ulate overtopping flows on the dike crest. The underlying principle is that it is not necessary to simulate the entire overtopping wave but only the water tongue on

(36)

(a) Operation principle of the simulator (b) Actual performance at Yen Binh dike

Figure 3.1: Wave overtopping simulator, principle of operation (after Van der Meer,2007) and the Vietnamese copy with a capacity of 100 l/s per m and a width of of 4 m from left to right.

The Vietnamese simulator was constructed in 2008 and 2009. Its maximum dimensions are 4 m wide (along the dike stretch), 2 m thick and 5.5 m high giving a maximum unit volume of 5.5 m per m of width and a total volume of 5.5×4 = 22 m . With this size a mean discharge of 100 l/s per m can be generated. The simulator is divided into two parts for convenience of transporting. The open width of the butterfly valve is 0.8 m giving good simulation of the overtopping flows on a dike crest. The hydraulic pump that opens and closes the two valves is attached to the simulator body. The vertical position can be adjusted by six hydraulic cylinders. Two water pumps with respective capacities of 22 and 55 kW are used to fill the tank. Last but not least, a 250 kVA generator supplies power to all devices and equipments deployed during the tests.

When considering a slope subject to overtopping, the grass cover is considered to bear the strengthwhile flows play as the hydraulicload. Damage is regarded if the grass cover is eroded, e.g. at one or several positions. The simulator is used to generate (simulate) the fast wave overtopping on a real dike so that the grass cover can be tested. A simulator test is usually limited to the extent that the erosion is not so serious that might probably threaten the slope stability.

3.3. Test sections and test scenarios

W

ith experience acquired after a couple of years, an adjusted and improved de-sign was transfered to Vietnam to manufacture another simulator via a

(37)

tech-3.3.Test sections and test scenarios ..

3

27 nical assistant project. For the first time, our dikes were tested in situ giving an impression of how strong the sea defences are. The section describes the dike slopes to be tested, the set-up and the scenarios.

3.3.1. Thinh Long and Thai Tho sea dikes

In 2009 and 2010, tests were performed in Nam Dinh and Thai Binh provinces. Positions were selected to satisfy operational and test requirements. Operation requires that heavy trucks (20 tons) and a 25-ton crane can approach the site easily and safely. A crest of 4 m wide is required for positioning the simulator, the generator and the two pumps. The selected slope has to be partly or entirely covered with grass. A water source, either a canal or a pond is necessary to supply water for testing. A volume of roughly 9000 m is estimated sufficient for testing in 4 to 6 hours with a discharge of up to 100 l/s per m1. The test sites in Thinh Long and Thai Tho are described in the following paragraphs.

In Nam Dinh, the Thinh Long (TL) dike was constructed of sand (core) and clay (outer layer). The crest level was at +5.5 m above Mean Sea Level (MSL) , 4.5 m wide and paved with 20 cm of concrete. Next to the toe, the horizontal part was about 10 m long and covered with Ray grass (Panicum Repens). The 1:3-inclination slope was about 9.5 to 10 m long, and mainly covered by Bermuda grass (Cynodon Dactylon), sometimes in combination with Casuarina treesCasuarinaceae. Three sections were selected to reflect different slope conditions and grass quality. Section TL1 was regular and protected by good grass, TL2 had a big Casuarina tree with a trunk diameter of 7 to 10 cm, and TL3 had been eroded at several points before testing. Water was pumped from a brackish water canal running along the dike toe. Figure3.2illustrates the Thinh Long cross-section.

Figure 3.2: Cross-section of the Thinh Long dike.

The second test site was situated on an estuary dike in Thai Tho (TT) commune, Thai Thuy district, Thai Binh province. A combination of Bermuda and Vetiver grass (Vetiver zizanioides and recentlyChrysopogon zizanioides) on the river-side slope was tested at three positions: TT1, TT2 and TT3. The dike body was constructed of good clay; and there were concrete beams running along and across the 1:3-inclination slope. The sections were chosen to have different configuration of beams

1_{Discharge 100 l/s per m}_{× 4 m (width of the simulator) × 3600 s × 6 hours = 8640 m the required}

(38)

and grass quality. Connected to the slope, the horizontal berm was often covered by a dense sward of Ray grass. Water was pumped from Tra Ly river. Figure 3.3

sketches the Thai Tho dike.

3.3.2. Yen Binh dike model

A dike stretch of 30 m long with a slope inclination of 1/15 was built in Yen Binh (YB) commune, Thach That district, Hanoi. After the completion of construction in June 2011, the dike height was about 3 m and the crest was 5 m wide. Next to the dike model, there was a pond supplying and discharging water for testing. Figure

3.4illustrates the dike model with its main design parameters.

Figure 3.4: Yen Binh dike model has a slope inclination of 1/15 and a height of 3 m.

On two halves of the gentle slope, respectively two local grasses were planted, Bermuda (Cynodon dactylon) and Carpet (Axonopus compressus). These species are popularly found on river and sea dikes and meadow in the north of Vietnam. The slope was divided into four sections YB1, YB2, YB3 and YB4. Two structures were erected at YB1 and YB4 to investigate the influence of obstacle on a grass cover subject to overtopping. Figure 3.5shows the cross-section of the Yen Binh dike.

(39)

3.3.Test sections and test scenarios ..

3

29

Figure 3.5: Cross-section of the Yen Binh dike.

Table 3.1: Description of the slope sections to be tested. Notes horizontal = hor., concrete = con., vertical = ver., and obstacle = obs..

Section Length Regularity Soil Cover Structures Grass age TL1 10.0 m regular moderate clay good Bermuda no Bermuda: 4 years TL2 10.0 m irregular with small holes poor mixture of Bermuda + Car-pet, a small Ca-suarina no TL3 9.5 m relatively_regular very poor Bermuda + Ray at toe no TT1

9.0 m regular good clay

good Bermuda + Vetiver hor. con. beam Bermuda: 5 years, Vetiver: 6 months TT2 poor Bermuda + Vetiver hor. and ver. con. beams TT3 YB1

30.0 m regular good clay

good Bermuda round-head obs. Bermuda: 1 year YB2 no YB3

good Carpet no Carpet: 1 year

YB4

square-head obs.

Where soil and grass were evaluated qualitatively. For quantitative assessments readers are referred to other reports (Trung,2011,2012b). The next part presents wave conditions and test scenarios.

3.3.3. Test scenarios

A wave condition was selected to represent the storm characteristics of the northern coast of Vietnam. At Thinh Long and Thai Tho, wave height𝐻 is 1.5 m and peak period𝑇 is 6 s. The seaward slope of the two dikes was assumed to have a tan 𝛼 of 1/4. Depending on the slope strength, different mean discharges were used, varying from 10 to 120 l/s per m. The same condition was also used at YB1, YB2 and YB3. Specifically, a wave height of 2 m was deployed at YB3 and YB4 because

Overtopping on Grass Covered Dikes: Resistance and Failure of the Inner Slopes

Overtopping on grass covered dikes

Overtopping on grass covered dikes

Resistance and failure of the inner slopes

Proefschrift

LÊ Hai Trung

Preface

Contents

1

Introduction

1.1.

Motivation

I

1

1.2.

Research objectives

1.3.

Thesis outline

T

2

Literature review

2.1.

Failure mechanisms due to overtopping

W

2.2.

Hydraulic loads

C

2

2.2.1.

Flow velocity and water-layer thickness

2.2.3.

Turbulence

2.2.4.

Discontinuous flow

2

2.3.

Grass covers

G

2.3.1.

Composition

2

2.3.2.

Root reinforcement to soil

2.4.

Experiments on grass covered slopes

E

2

2.4.1.

Overflow and overtopping tests

2

2.5.

Erosion models

O

2.5.2.

Turf set-off model

2

2.5.3.

EPM model

2.5.4.

Turf-element model

2.5.5.

Head-cut model

2.5.6.

Cumulative hydraulic load

2.5.7.

Cumulative erosional work units

2

2.6.

Discussion

G

2

2.7.

Method

I

3

Overtopping simulator tests

3

3.2.

Operation and design of the simulator

U