ECSMGE-special, september 2019

Page 1





S O E T A E R T. B E


Soetaert-Soiltech is an expert in trenching and foundation techniques with years of experience in soil improvement techniques. Soetaert-Soiltech has the necessary equipment and expertise to complete your project successfully and on a solid basis. MV-piles | Kelly-piles | Slurry walls | Anchors | Vibroflotation | Stone Coloms | Bored double rotary piles | Soilmix | Berliner walls | Combi walls | Sheet pile walls


KIVI Engineering Society P.O. Box 30424 2500 GK Den Haag The Netherlands Tel. 0031 (0)70 391 99 00

Waterpas 98 2495 AT Den Haag Tel. 0031 (0)88 255 62 64

IJzerweg 4 8445 PK Heerenveen The Netherlands Tel. 0031 (0)513 63 13 55

Haven 1025 Scheldedijk 30 2070 Zwijndrecht Belgium Tel. 0032 (3)250 52 11

Boussinesqweg 1 2629 HV Delft Tel. 0031 (0)88 335 82 73

Industrieweg 3 8531 PA LEMMER The Netherlands Tel. 0031 (0)514 588 000

Eijkelkamp GeoPoint SoilSolutions B.V. Rijksstraatweg 22-F 2171 AL Sassenheim Tel. 0031 (0)71 301 92 51

Veurse Achterweg 10 2264 SG Leidschendam The Netherlands Tel. 0031 (0)70 311 1333

Limaweg 17 2743 CB Waddinxveen The Netherlands Tel. 0031 (0)182 640 964

Esperantolaan 10A B-8400 Oostende Belgium Tel. 0032 (59) 55 00 00

P.O. Box 812 3300 AV Dordrecht The Netherlands Tel. 0031 (78) 654 62 00













Ir. B.J. Admiraal Volker Staal en Funderingen

PNEUMATIC SINKING 2.0, OR ‘HOW DEEP CAN YOU SINK (FROM A GEOTECHNICAL PERSPECTIVE) ’ Old and new Kattwyk Bridge The port city of Hamburg is built in the delta of the Elbe River. There are several bridges to reach the city regarding which the existing Kattwyk Bridge is located on the central railway route between the eastern and western port area. The bridge also has a local but important traffic function including for the connection with the A7. The bridge was built in 1973 and bridges the southern section of the Elbe. It is one of the largest vertical lift bridges in the world. The high lifting towers create a characteristic architectural focal point. The combination of railway and road, however, represents a problem these days due to the relatively narrow bridge. If a train needs to pass the bridge, road users cannot use the bridge since the railway runs over the road lanes. This leads to frequent traffic jams.

A new bridge was therefore urgently needed. The decision was taken to build an extra bridge with the same shape that is especially intended for trains. The tender was won by the construction combination Max Bögl-H.C. Hagemann-Heijmans. The new bridge that has a total length of 287 m is being built approximately 60 m to the north of the existing bridge. The bridge will be lifted more than 50 m above the water when it is open, which is the same height as with the existing bridge. There is a distance of 131 m between the piers and they are approximately 75 m from the river bank. The decision has been taken to build a microtunnel between the two piers at a depth of approximately 24.5 m for the cabling to assure a symmetrical movement during the lifting movements of the bridge.

Figure 1 – New Kattwyk Bridge sinking location, Hamburg. GEOTECHNIEK ECSMGE 2019 SPECIAL



The piers themselves have an area of 29 x 14 m2 and a foundation level at 30.0 m below standard zero ('Normalnull').

Geotechnical conditions The soil where the piers are located has been surveyed by means of borings and CPT’s. Performing CPT’s, however, was difficult. The reason for this may be a very hard soil or the presence of stones in the top stratum. The soil survey has shown that the soil can be characterised as follows: The soil survey has shown that the soil structure under both the piers shows large differences in this basin area. The location at the south westerly pier is, for example, characterised by a melt water sand layer that is approximately 8 m thick, gritty in nature with possibly larger stones and underneath strongly layered clay and silt formation. At the

SUMMARY river piers of the New Kattwyk Bridge, has therefore mechanised the sinking process innovatively. The soil under the caissons is excavated in a remotely controlled manner and disposed of while deploying occasionally. This article discusses the implementation concept for the river piers, the development of sinking in a remotely controlled manner and the expected and unexpected challenges that must be resolved during implementation.

The New Kattwyk Bridge in Hamburg is currently being realised by the Hamburg Port Authority (HPA). A special construction method was chosen for the realisation of the river piers. The pneumatic sinking technique, among other specialties, was used in relation to this. Pneumatic sinking means working under an overpressure with all the safety measures that this entails. It also means that working hours are seriously limited when sinking up to a water depth of 32 m. Volker Staal en Funderingen (VSF), the subcontractor for sinking the

north easterly pier, the sand stratum formed by glaciers is not present, but the clay/silt formation is found at a higher level with a relatively small thickness and underneath a thick older sand layer.

Table 1 Description of the soil structure at the pier locations Stratum

SW pier [m below standard zero] River bed -9.8 +/- 0.7 Backfill material N/A

River sand

-11.2 +/- 0.6

NE pier [m below standard zero] -12.1 +/- 0.5 -13.9 +/- 0.3

-14.7 +/- 0.6

Melt water sand -19.3 +/- 0.9


Basin silt

-30.9 +/- 2.2

-19.1 +/- 0.6

Basin sand

-37.3 +/- 0.1

-36.5 +/- 0.5

Soil description

Sand with mussel shells and with varying peat, silt and rubble content. H2S was clearly released from the samples. Varying silt, peat, wood, shells and sludge content. Possibly strong gravel and stone content with silt layers. Predominantly compacted. Very different in thickness. Built up from very thin clay and silt layers with a regular sand content. Contains much fine organic material (approx. 10%). Predominantly rigid behaviour and very cohesive. Fine sand with a varying silt and clay content. Also strata with lignite. Predominantly qc up to 25 MPa.

Pier construction method Constructing at a great depth in the middle of a busy industrial area sets quite some requirements and preconditions in relation to the design and implementation. The average water depth amounts to approx. 11 m above the river. This increases by more than 2 m during a normal high tide, but this can even increase to nearly 6.5 m when there is a heavy storm or spring tide. Therefore, something to be taken very seriously. The client therefore decided to use a very special construction method. Figure 3 provides insight into the construction method. A cofferdam using combined sheet pile walls was realised for each pier in the river but with the one side still open. On top of this, a steel structure frame was installed with a height higher than 10 m above standard zero. The concrete floor area of the bottom plate including the edges of the cuts of the pier was made on a pontoon near the quay at that time and subsequently sailed into the cofferdam during high tide (1). The bottom plate was lifted in by means of 24 heavy Gewi bars connected on to the steel structure and the pontoon could, in this way, again be sailed out of the cofferdam during low tide. Once the cofferdam had been fully closed, the construction of the pier could really get started (2). The concrete pier was constructed suspended from the Gewi rods in sections of approximately 5 m alternating and it was jacked downwards (3+4). After the structure was placed on the cleaned up soil, the equipment for sinking could be built and the suspension could be removed in a controlled manner (5). The sketches 5-7 only show the installation of water cannons and a pump as conventionally used in caisson sinking. The characteristic access airlocks and air pressure system have been left out in the sketches. A spectacular construction method on the water and the pneumatic sinking had not even started yet!

Figure 2 – New Kattwyk Bridge impression.

Figure 3 – Building the concrete pier suspended from bars.

Sinking method Sinking caissons and certainly pneumatically is not a very well-known technique in the world of foun-




dations. This technique, however, already exists 200 years and, in the Netherlands, since the end of the 19th century. The word 'pneumatic' refers to the application of increased air pressure in the working chamber under a caisson to thus create a dry working space making the controlled removal of soil possible. Until halfway through the last century, serious accidents occurred regularly in relation to workers that were tasked with the job. This is why the term 'caisson disease' (decompression sickness) was used without people knowing the cause, how to prevent it or how to treat it. Legal precaution measures, however, were already taken in the Netherlands at the start of the 20th century. This was known as the Caisson Decision that was part of the Mining Act. Knowledge and insight only really changed with the development of diving after World War II, in particular in the navy. And, in the Netherlands, also due to the professional deployment of large hydraulic construction projects such as the well-known Delta Works (in Dutch: Deltawerken) in the 1970s and 1980s for dyke safety along the North Sea. Nowadays, specialised and certified diving doctors are trained. Decompression and treatment tables have been developed with specific procedures for caisson workers. VSF has proven with thousands of 'dry diving' instances at many projects in the Netherlands that caisson work can be performed safely. However, working under overpressure is far from ideal, even though the risks to health are minimal.

The work is labour intensive, in particular in cohesive soil and at a greater depth. A greater depth, after all, means a higher water pressure and therefore a higher air pressure to keep the working chamber dry. This results in seriously restricted working hours.

Sinking 2.0 Up to now, sinking in Europe took place by using caisson workers who worked in the working chamber under the caisson. After atmospheric lockingin up to the set overpressure, the soil is usually jetted in a targeted manner to remove it using water cannons. The mixture of water and soil is, subsequently, taken outside the caisson in containers or a sludge depot. Sometimes, small excavators are used, in particular in corosic soil of hard strala. The soil is taken outside, in this case, using buckets lifted each time through a material airlock. Workers climb up to the people airlock after every shift. Decompression takes place, in steps, in a controlled manner until obtaining atmospheric pressure and by using respiratory equipment with pure oxygen. Outsiders often believe that this is dangerous and dirty work, but those immediately involved believe that they are the greatest project experiences that you can have. Playing with sand and water is fun at any age and, as a geotechnical engineer, it is unique to really see and feel the soil to a great depth!

Figure 4 – Phasing of the bridge pier realisation.




As already said, however, working under an overpressure with teams is most definitely not ideal. It is labour intensive and therefore less competitive than building in a building pit. The communication with the caisson workers is also difficult because of the sealed chamber and this is at the expense of efficiency. A development to obtain a new sinking version was therefore required. With projects in its portfolio such as the New Kattwyk Bridge and also the lock heads of the New IJmuiden Sea Lock, central to the port of Amsterdam Volker Staal en Funderingen was given the opportunity to give its market leader role in this technique an extra boost. The objective is to remotely control the sinking so that caisson workers only have to work under an overpressure with regard to the installation and maintenance. The entire process of releasing the soil, transporting it in the working chamber and removing the soil from the working chamber is considered in many variants during the development process but also the drive: hydraulically, electrically or pneumatically for the movements, and cameras and sensors for monitoring. Close collaboration took place with regard to this, an engineering firm and machine builder. A significant pitfall that is known from much practical experience is to want to build a machine that can do nearly everything and is suitable everywhere: What is commonly referred to as expecting the impossible. The trick was therefore to design and build a machine that is suitable for approximately 80% of the contemplated market area that is

new with regard to its type, but that is robust and reliable. In East Asia, they already started to mechanise sinking over 20 years ago. They developed excavators that are suspended from rail structures there. Remotely controlled, soil is deposited in buckets that are lifted up from the working chamber through a material airlock, emptied and again locked in. This system works in basically any soil type and is applied on a large scale by Japanese companies in particular. It does, however, have a considerable disadvantage. Productivity is very low. The aspect that needed to be improved was productivity. VSF decided to check what already exists in the field of technology related to earthmoving and dredging soil conditions related to delta regions. This finally led to a concept with cutting head pumps installed on an arm that can be moved in three directions. Turning arms have been developed that can be extended up to 14 m! The reason for this is that this allows the direct dredging of soil material with a reach that is the largest possible. Outside of this reach, there is a combination with powerful remotely controllable jets that are installed on to the ceiling of the

Figure 5 – Pier sinking situation. Figure 6 – Excavation arm in the working chamber.




working chamber. Dredging pumps could be used that were already available on the market. The most important criteria when making this choice was limiting the weight, having sufficient pumping capacity for sufficient flow speed in the long pipes and an outlet opening that was the largest possible for pumping gravel. In addition to sensors in the hydraulic drive of the different motors and cables, investments were also made in cameras that are suitable for the special conditions that can occur in the working chamber. It is not just the increased air pressure that is a factor that must be considered, but mist will already be created when small pressure reductions occur. Droplets rain down with soil when jetting and there is a risk that the lighting may disrupt instead of support visibility due to shimmer or glare. In short: compressive strength, cleaning systems and high light sensitivity formed the key conditions when developing the cameras in combination with a good light plan.

Figure 7 – Remotely controlled excavation.

Extensive testing is, of course, an integral part of such a development. The design process was developed based on extensive reviews by different experts. A testing set-up was made to develop the combination of jets, camera and lighting. When the first machine was delivered, a test location was built to install the equipment airlock and build the arm structure through this. This was done to train employees in operating but also in installing and disassembling in a sealed working chamber.

Figure 8 – Stones and rocks - a mountain of obstacles.

Project experiences And then it really had to happen! The piers had already been installed on the river bed of the Elbe; therefore at approx. 11 m below standard zero. The building of the air pressure equipment in the North-East pier started in January 2018. Next, the equipment airlock could be installed. The equipment was introduced into the working chamber in sections with a maximum length of 2 metres and installed. A temporary equipment airlock was installed on the floor for this. This ensured that sinking could be started in phases of at most 5 m. After every sinking phase, the pier could again be built up further after which another sinking phase followed. See figure 3, phases 6 and 7. In total, every pier was sunk in this way in five steps up to the final depth of more than 30 m below standard zero. All equipment was removed in the working chamber at this depth. Next, the chamber was filled with concrete as ballast to prevent uplift during the following construction phase. There were, naturally, a few start-up issues and teething problems, but these remained within what can be reasonably expected. The presence of obstacles and rubble in the top soil layer also led to damage to pumps occasionally. The real delay, however, was caused by the previ-

ously mentioned melt water sand layer under the South-West pier. The addition of sand to the name of this layer turned out to be fairly misleading. Significant resistance had already been noticed when the combined sheet pile walls were installed and many stones and rocks came up during predrilling. It was reported that an old creeck had been found filled with stones that runs obliquely through the location with a width of approx. 6 m. Mainly sand and gravel but with some stones and rocks. Some delay did then have to be taken into account although pumps with a clearance of 100 mm were specifically selected. 'Some stones and rocks' turned out to be an understatement because, in practice, it was a lot tougher. There was a stratum over virtually the entire area that was nearly 7 m thick consisting of mainly stones and rocks that had consolidated to a significant degree. Jetting the soil loose was not really possible, nor was pumping. In total, nearly 750 tonnes of rock were removed from the caisson by the caisson workers! A very heavy job but they did it! The NE and SW piers were installed on 22-8-2018 and 1-10-2018, respectively, at the final depth. Well within the set tolerances.




To conclude The sinking of caissons is a fantastic foundation technique intended for structures that must be built at larger depths. Usually, construction takes place at ground level and therefore a deep cofferdam with the risks that this entails is not required. Techniques to realise deep cofferdams have evolved significantly during the past decades. Volker Staal en Funderingen has taken a large step forwards with the development of a remotely controlled sinking method to offer a fully-fledged alternative. A development that was only possible in collaboration with various experts, suppliers and customers. Currently, the lock gate heads of the New IJmuiden sea lock are being realised. These concern caissons with mega dimensions. The outside head measuring 26 x 81 m2 is currently installed at a depth of 24.3 m. The inside head measuring 55 x 81 m2 will be the biggest caisson ever built and will be installed at 25.55 m depth in spring 2019. This will also be within unprecedented tolerances such as a torsion of 50 mm and tilting across and in the longitudinal direction of 0.3% but with a target value of <0.1%. That is our goal with sinking 2.0! 쎲

Traditional and basic if possible but innovative and advanced when necessary. 85( QHHGTU EWUVQO OCFG CPF EQUV GH¯EKGPV HQWPFCVKQP UQNWVKQPU YJKEJ ¯V KP KVU environment. That is our workmanship!



Erik Claassen

ANOTHER STEP CLOSER TO A 3D DIGITAL PLANT ABOUT FUGRO Fugro is the world’s leading Geo-data specialist, collecting and analyzing comprehensive information about the Earth and the structures built upon it. Adopting an integrated approach that incorporates acquisition and analysis of Geo-data and related advice, Fugro provides solutions. With expertise in site characterisation and asset integrity, clients are supported in the safe, sustainable and efficient design, construction and operation of their assets throughout the full lifecycle. Employing approximately 10,000 talented people in 65 countries, Fugro serves clients around the globe, predominantly in the energy and infrastructure industries, both offshore and onshore. The company is listed on Euronext Amsterdam.

Imagine having access to the status of your operational assets from the comfort of your office, whenever you want. This is essentially what you get from Fugro SITE-SPOT: up-todate digital 3D asset information hosted on a web portal to enable smarter, safer and more productive operations. Fugro is a global survey and asset data specialist founded in 1962 with headquarters in the Netherlands. The company supports customers in a wide range of sectors by providing integrated solutions in the fields of Geo-data and asset integrity. For network and plant managers, Fugro offers specialised digital solutions that give fast and efficient insights both above and below ground. These are based on innovative software and hardware, much of which is developed by a team of in-house experts. In terms of asset management, Fugro supports EPCs (engineering procurement and construction companies) and various owners and operators in the oil & gas, energy and chemical industries, including a number of companies based at the Port of Rotterdam. The company offers integrated tools for visualisation, modelling and analytics to optimise both new-build projects and the upgrade or expansion of existing plants.

Crucial asset knowledge Fugro’s Manager Digital Plants Europe Erik Claassen explains: “We see more customers opting for integrated digital solutions, because up-to-date and accurate asset knowledge is crucial to safety, cost reduction and maximising asset life.”

access arrangements. This shortens downtime and reduces management and maintenance costs. Once the situation has been captured digitally, any changes to the 3D model can be registered easily and reliably to avoid duplication of data.

Digital foundation Since the digital twin provided by SITE-SPOT is built on a high-resolution, 3D laser scanned dataset, a wide range of asset related data and other administrative information can be added to the model. These include date of data capture, asset attributes and historic or planned modifications. Various spatial survey tools enable users to make geography-based selections from the central database and visualise the results. For example, it is possible to see the location and status of all the inspection points in a particular pipe rack. A benefit of this type of web-based solution is that no specialised engineering software is required to access and utilise the data. In the case of a turnaround or regular site maintenance, work planners can easily determine precise asset location without leaving the office. Up-todate insights can pinpoint locations where work is needed and help to ensure it is completed safely and efficiently, for example, by planning suitable




With information hosted in a central cloud-based database, the user benefits from a single point of truth (SPOT), ensuring that all stakeholders access the same, up-to-date information at any time, from any location. The same user-friendly interface is available to groups such as training and operations departments and approved external partners. Clients list a range of benefits including accelerated response times as well as reductions in site visits, equipment clashes and the need for costly re-engineering. The digital foundation provided by SITE-SPOT puts crucial information and insight at the fingertips of those responsible for industrial plants and supports a smarter, safer and more productive way of working.

More information Erik Claassen +31 6 54391675

To T o fi find nd out out more more fu

Professional Pile TTesting esting e Equipment and a Geotechnical Monitoring Systems


VVibration ibration Monitoring

Pile Integ Integrity rity ty TTesting eesting

Structural Structural Monitoring

Groundwater Monitoring

Geotechnical Monitoring


M More information: information: Profound BV | TTel. eel. +31 (0)182 640 964 | | www


Creating land for the future



DEME is a world leader in the highly specialised fields of dredging, marine engineering and environmental solutions. By fostering a pioneering approach, DEME operates as a front runner in innovation and new technologies. With a strong presence in all of the world’s seas and continents, DEME provides solutions for global, worldwide challenges: a growing population, the scarcity of natural resources, polluted rivers and soils, the reduction of emissions and rising sea levels.

Dredging and land reclamation Marine and offshore energy solutions Infra marine solutions Environmental solutions Fluvial and marine resources

DEME Haven 1025, Scheldedijk 30 • B-2070 Zwijndrecht, Belgium T +32 3 250 52 11 • •

F. S. Tehrani Deltares, Delft

G. Santinelli Deltares, Delft

M. Herrera Delft University of Technology, Delft

MACHINE LEARNING FOR FORECASTING RAINFALL-INDUCED LANDSLIDES landslides (Segoni et al., 2018), hence the outcome suffers from geographical biases. To overcome the limitations of the conventional landslide forecasting methods, next to the rainfall intensity, duration and frequency, one needs to consider controlling factors, which include, among others, topography, lithology and geomorphology of slopes, soil type, ambient temperature, surface radiation, vegetation, soil moisture, land use and land cover. This was the subject of our study, for which we have set up a Machine Learning (ML) framework to better estimate the onset of rainfallinduced landslides. Figure 2 shows the forecasting framework that was adopted in this research project. We used the NASA Global Landslide Catalogue (Kirschbaum et al., 2010) to build the detailed database of landslides.

Datasets GLOBAL LANDSLIDE INVENTORY The global landslide inventory is derived from the global landslide catalogue (GLC), which was developed by NASA Goddard Space Flight Center. The GLC is based on various online news media, scholarly articles, and existing hazard databases. As of April 2018, the GLC consisted of 11,055 landslides with 10,988 landslides occurred after 2007. The GLC contains a limited number of landslides triggered by factors other than rainfall, such as earthquake and human action. In this study, we filtered these types of landslides out and focused only on rainfall-induced landslides.

Figure 1 – Sierra Leone landslide in 2018.

Figure 2 – Landslide forecasting framework of this study.

Introduction Landslides can pose serious threat to urban environment and to line infrastructures such as roads and pipelines. Among multiple triggering factors of landslides, precipitation is one of the most common ones, causing thousands of landslides in the past decade, some of which are amongst the deadliest landslides. For instance, the debris flow occurred in August 2017 in and around Freetown in Sierra Leone caused 1141 fatalities (Figure 1). Therefore, forecasting rainfall-induced

landslides can be extremely helpful to minimize mortalities due to landslides and plan mitigation and rescue measures. Forecasting rainfall-induced landslides is typically done based on rainfall thresholds (e.g. Guzzetti et al., 2007; Rossi et al., 2017). Although rainfall thresholds are widely used for predicting the occurrence of landslides, they suffer from certain limitations; one of them is that they have been mostly developed for region-specific prediction of




With regard to location accuracy, Kirschbaum et al. (2010) reported large uncertainties when assigning geographic coordinates to a landslide event. To deal with this uncertainty, they assigned a radius of confidence (which spans from tens of meters to tens of kilometres) to the location, indicating the estimated radius of a circle over which the landslide may have occurred. To reduce the uncertainty in finding the triggering and controlling factors associated with landslides events only with nearly exact locations and with short-term rainfall (to be explained later) greater than 20 mm are considered in this study [this is because the focus of this study is on rainfall-induced landslides]. For training a ML algorithm, non-landslide cases are also needed. We sampled non-landslides from landslide events with radius of confidence greater

SAM E N VAT T I N G SUMMARY landslides under the effect of precipitation. In this data-driven approach, ML methods are used to predict landslides by exploiting multiple Earth observation datasets, including rainfall data (e.g. TRMM 3B42) and Digital Elevation Models (e.g. SRTM), and the NASA Global Landslide Catalogue. A detailed inventory of landslides at a global level is built out of which a supervised ML algorithm is trained with landslide/non-landslide events. The trained ML model is then fed by rainfall data, topography features such as slope and elevation relief, soil and bedrock data, and vegetation index of target regions to assess the stability of the studied area.

Landslides are catastrophic geo-hazards that threaten urbanization. Growth in population besides construction of critical infrastructures such as roads and pipelines in landslide-prone areas elevates the risk associated with landslides. Therefore, a system that is able to predict landslides and issues warning in a timely manner is very appealing. It is widely accepted that precipitation is one of the most influential factors for triggering landslides. In this article, we present the preliminary results of a practical research study that has been carried out in Deltares. To that end, we have set up a framework that combines geo-engineering, remote sensing, hydrology with Machine Learning (ML) to predict the onset of

than 25 km and short-term rainfall less than 60 mm. Since every landslide in the GLC has a coordinate, it can be suggested that a landslide event with radius of confidence greater than 25 km did not happen on the reported coordinate. This was further verified by visual inspection. Applying these filters, the final dataset consist of 235 landslides and 1696 non-landslide events.

RAINFALL DATA As reported by Sun et al. (2018), currently there are approximately 30 available global precipitation datasets, including gauge-based, satellite-derived, and reanalysis datasets. These authors suggest that the reliability of precipitation datasets is mainly limited by the number and spatial coverage of surface stations, the accuracy of satellite algorithms, and the data assimilation models. For the scope of the current study, the maximum daily rainfall data from Tropical Rainfall Measurement Mission of NASA (TRMM 3B42) has been used for estimating the accumulated intensity of rainfall on the day of landslide event, the day before (short term rainfall) and nine days before these two days (long term rainfall) prior to the event. Figure 3 shows the frequency of the accumulated short term and long term rainfalls.

Figure 3 – Accumulated rainfall for the filtered landslide events based on TRMM3B42: (a) Short term and (b) long term.

Figure 4 – DEM properties for the filtered landslide events based on SRTM: (a) Slope and (b) Elevation relief.

Figure 5 – Soil fraction for the filtered landslides.

DIGITAL ELEVATION MODEL Digital elevation models (DEMs) are considered as one of the main datasets for analysing the controlling factors involved in the landslide hazard assessments (van Westen et al. 2008). These three-dimensional representations of the terrain are useful for extracting key topographical and geomorphological parameters including elevation, slope, and aspect of the ground surface. In this study, the NASA Shuttle Radar Topography Mission (SRTM, 2000) was used to obtain topo-graphical features of the terrains where landslide occurred. SRTM is selected due to the high spatial resolution (30 m) and its temporal coverage with an acquisition date before the occurrence of all the landslides recorded in the database. 4 shows the mean slope and elevation relief (difference between the maximum and minimum elevation within the landslide confidence area) for the filtered landslide data.

SOIL AND BEDROCK The comprising material of slopes and the depth of the bedrock can highly affect the hydro-geomechanical response of slopes to rainfall. Therefore, estimating the soil composition of hillslopes can potentially enhance the predictability of rainfall-induced landslides.




Soil composition was retrieved as raster data from the SoilGrids datasets (Hengl et al. 2014) at 250 m resolution with a global coverage. Among the information available of SoilGrid, the estimated fraction of sand, clay and silt and depth to the bedrock are used in this study. The average sand, silt and clay fraction of the seven standard depths

are calculated as features to be used later in the prediction stage. Figure 5 shows the fraction of these soil types for the filtered landslide events.

VEGETATION Vegetation is another controlling factor that can highly influence the stability of natural slopes and therefore play a vital role in predicting landslides. Leaves control soil moisture through evapotranspiration and roots can mechanically reinforce soil particles and increase shear strength of soil compound by increasing the matric suction. Therefore, it is accepted that in general lack or shortage of vegetation can increase the susceptibility of slopes to landslides. One way of quantifying vegetation density is through calculating the Normalized Difference Vegetation Index (NDVI).

NDVI quantifies vegetation by measuring the difference between near-infrared (NIR), which is strongly reflected by vegetation, and red (visible) light (R), which is strongly absorbed by vegetation. NDVI is calculated per pixel as a function of the red and near infrared bands:

MACHINE LEARNING For the current work, we used the Logistic Regression (LR) algorithm as a supervised ML method for classification of landslide and non-landslide events (binary classification). The LR algorithm is trained with training sub-sets, which include controlling and triggering factors as predictors (X) and labeled (landslide or non-landslide) output (Y). The perfor-

Table 1 - Example sets used in training the LR algorithm Example set/Features










x1 Short-term rain










x2 Long-term rain










x3 Mean slope










x4 Elevation relief




















x6 Soil and bedrock










Figure 6 – NDVI before landslide occurrence for the filtered landslides.

mance of LR models is measured on test sub-sets to evaluate the accuracy of predicting outputs. LR algorithm calculates the pro-bability that the predicted output belongs to a particular category or class (landslide and non-landslide in this study). Mathematically, the relationship between the probability p of landslide and the triggering and controlling factors (predictors or features) can be expressed using the sigmoid function: (1)

where z = w0 + w1x1+ w2x2+ w3x3+ ... + wnxn is a linear combination of predictors x1 to xn, w0 is the intercept or bias of the model, and wi (i =1, 2, ..., n) are the weights (fitting coeffects) of the features. These weights are derived by optimizing the cost function which measures the difference of predicted output and actual output. If the probability of occurrence is greater than 50%, the model classifies the output as 1 (landslide), otherwise 0 (non-landslide). As mentioned earlier 235 landslides and 1696 non-landslide events are used to build the ML dataset. However, the dataset suffers from imbalanced landslide and non-landslide events which can influence the performance of any ML algorithm. To overcome this issue, the non-landslide events are undersampled by random removal of 1000 non-landslides. The outcome is a dataset that consists of 235 landslides and 696 non-landslides.

Research outcomes LR algorithm was used to distinguish landslides and non-landslide cases. To train the LR model, nine example sets (E0 to E8) with different combination of triggering and controlling factors (model features) were built. Table 1 shows the combination of controlling and triggering features (x1 to x6) used for training the LR model, where 1 shows if the feature is used and 0 means otherwise.

Figure 7 - Accuracy of logistic regression model in classifying landslides and non-landslides.




The sample sets are split into training (70%) and test (30%) sets which then are used for training and assessing the LR model. The accuracy of the LR model in form of Receiver Operating Characteristic (ROC) curves and the associated Area Under Curve (AUC) is illustrated in 7. The ROC curve is a measure for evaluating a diagnostic test, where true positive rate (Sensitivity) is plotted against false positive rate (100 - Specificity) for various decision thresholds (between 0 to 100%). Every point on the ROC curve represents a sensitivity/ specificity pair that corresponds to a certain decision threshold. The area under the ROC curve (AUC) quantifies how well a group of features can be used to distinguish between two diagnostic groups (landslide / non-landslide). In general, higher AUC values (maximum = 1) indicate a more accurate classification. However, other metrics such as number of true positives and negatives

and false positives and negatives should be also checked for further verifications. 7 shows that, in general, the LR model can perform well in distinguishing landslide and non-landslide cases. By comparing the results of E1 (AUC = 0.66) and E2 (AUC = 0.92), it can be suggested that having the short-term rainfall fixed, elevation relief can be more effective than slope angle for landslide/non-landslide classification. This indicates that elevation relief on regional scales can be more representative of the topography of the region than slope angle. This has been suggested by other authors such as Lin et al. (2017). Looking into other cases with high accuracy (AUC = 0.91), namely E6, E8 and E10 it can be suggested that adding more features to a training set of an ML model might not necessarily result in better prediction. In this case, E2 prediction is as good as E6, E7 and E8. However, looking into number of true negatives (correctly predicted non-landslides), it seems E8 slightly performs better than the rest of example sets. This observation emphasizes the role of feature engineering in ML problems. Feature engineering can reduce the cost of prediction as less number of features may result in highly accurate ML models.

Summary conclusions In this paper, we presented the preliminary results of a practical research study on developing a datadriven framework for predicting rainfall-induced landslides. LR as an MR algorithm was used to predict landslides by exploiting multiple Earth Observation datasets. Although the database and forecasting framework that were reported in this study are at their initial stage, the results of the study (AUC greater than 90%) showed that such a framework, with enhanced datasets and perhaps more advanced ML algorithms, can be used for forecasting rainfall-induced landslides and landslide early warning systems at a global and regional scales.

References - Guzzetti, F., Peruccacci, S., Rossi, M., & Stark, C. P. (2007). Rainfall thresholds for the initiation of landslides in central and southern Europe. Meteorology and atmospheric physics, 98(3-4), 239-267. - Hengl, T., de Jesus, J. M., Heuvelink, G. B., Gonzalez, M. R., Kilibarda, M., Blagotić, A., ... & Guevara, M. A. (2017). SoilGrids250m: Global gridded soil information based on machine learning. PLoS one, 12(2), e0169748.




- Kirschbaum, D. B., Adler, R., Hong, Y., Hill, S., & Lerner-Lam, A. (2010). A global landslide catalog for hazard applications: method, results, and limitations. Natural Hazards, 52(3), 561-575. - Lin, L., Lin, Q., & Wang, Y. (2017). Landslide susceptibility mapping on a global scale using the method of logistic regression. Natural Hazards and Earth System Sciences, 17(8), 1411-1424. - Rossi, M., Luciani, S., Valigi, D., Kirschbaum, D., Brunetti, M. T., Peruccacci, S., & Guzzetti, F. (2017). Statistical approaches for the definition of landslide rainfall thresholds and their uncertainty using rain gauge and satellite data. Geomorphology, 285, 16-27-267. - Segoni, S., Piciullo, L., & Gariano, S. L. (2018). A review of the recent literature on rainfall thresholds for landslide occurrence. Landslides, 1-19. - Sun, Q., Miao, C., Duan, Q., Ashouri, H., Sorooshian, S., & Hsu, K. L. (2018). A review of global precipitation data sets: Data sources, estimation, and intercomparisons. Reviews of Geophysics, 56(1), 79-107. - Van Westen, C. J., Castellanos, E., & Kuriakose, S. L. (2008). Spatial data for landslide susceptibility, hazard, and vulnerability assessment: an overview. Engineering geology, 102(3-4), 112-131. 쎲

Sandro Katerberg A.P. van den Berg GeoTechnology

Harrie Dieteren CRUX


Guido Meinhardt CRUX

Tahl Tekofsky

THE BEST OF BOTH WORLDS 2016, a 320-tonne hydraulic pile pusher with a maximum pushing force of 3,200 kN has been used successfully in over 20 projects across the Netherlands with 3,000 pushed piles on a total lenght of 50,000 m. This article will go into experiences with the pile pushing method in the Netherlands so far, as well as into specific design aspects of using this new silent piling method with conventional precast concrete piles.

Description of the pile system and the hydraulic pile pusher The hydraulic pile pusher is made up of a body frame with long side beams and short cross beams and a rail structure over which the body can move. This combination of side and cross beams allows the rig to move forwards, backwards and sideways, and even to rotate 360 degrees around the pile position. The two side beams measure 11.8 m by 1.25 m, which, given the total rig weight of 320 tonnes, results in ground pressures of approx. 110 kPa per beam.

Introduction Traditionally, driving of precast piles has been among the more popular foundation methods in the European market, partly supported by the high quality and reliability of both the pile as well as the input method. However, because of regulations and environmental considerations the various methods that form the pile in the ground have grown in popularity. The downside of these methods is that they are more susceptible to quality variations in the piles (which are cast in situ), while these bored piling systems also tend to be more cost intensive. Pushing instead of driving highquality concrete piles into the soil combines the benefits of precast piles with a vibration-free and virtually silent pile pushing method, reducing nuisance for the surrounding area to a minimum.

This is particularly welcome when installing foundations in inner city areas. is a company within the IJB-Groep. The pile pushing rig that uses is supplied by Heerenveen-based A.P van den Berg GeoTechnology, the sole European distributor of hydraulic pile pushers built by China's T-Works. In Asia, this method has been used on a large scale for over a decade now. T-Works specialises in hydraulic pile pushers that deliver maximum pushing forces ranging from 600 kN to 12,000 kN. A.P. van den Berg adapts these rigs to ensure conformity to European legislation and regulations, particularly in terms of meeting European safety and environmental requirements, while also fitting T-Works’ rigs with a data acquisition system. Since late




While manoeuvring, the rig always rests on either the cross or side beams as the other beams are repositioned to move the rig. Besides the weight of the basic rig itself, counterweights are used to be able to get the maximum pushing force out of the rig. The rig is able to deliver this maximum force when a pile is pushed down through the centre of the rig. The pile pushing mechanism consists of four hydraulic cylinders and a hydraulic clamp that grips the pile. Owing to the relative simplicity of the design, the hydraulic pile pusher requires low-maintenance. An additional pushing system (the so-called side piler) can be fitted on the front end of the machine body. However, the maximum pushing force that the side piler can deliver is only about 50% of the centre piling system. The pushing mechanism (jack) can push a pile roughly 2 meters, followed by an interruption to move the clamp box up and grip the pile at a higher position. During the pushing process, depth and total pushing force are recorded automatically and continuously.

SUMMARY In late 2016, and A.P. van den Berg were the first companies in the Netherlands to start using the pile pushing method in combination with conventional precast concrete piles. Instead of traditional pile driving using a hammer, this system pushes precast concrete piles into the ground. This results in proven high-quality foundations that are installed without the usual noise and vibrations.Measurements performed so far (as every pushed pile is basically a compression test in its own right) seem to corroborate that the pile base coefficient (αp) is 1.0 (instead of the factor of 0.7 used in the Dutch standard). Thanks to continuous recording of data, the pile pushing method offers excellent

opportunities for quality improvement or responsible application of safety factors. For example, the fact that this piling method measures resistance from the moment of entry into the soil and throughout the pile pushing process, much like a cone penetration test, makes it possible to correlate the load-bearing capacity, cone penetration curve, and readings from the pushing force recording system to a demonstrable safety level through expert assessment. In other words, continuous pushing force recording will in future have to be considered a kind of (verification) cone penetration test that would make it possible to set the correlation factors (ξ ) at 1.0.

Figure 1 – Hydraulic pile pusher (Source: A.P. van den Berg).

Figure 2 – Calculated pushing force in sand (Rc;k;sand ) versus design value bearing capacity (Rc;d ) for a set of cone penetration tests and various pile tip levels.

So far, this new piling method has been used to push square 250 mm to 350 mm precast piles with lengths ranging from 6 to 22 m. Based on current insights, the pushing method will also be able to handle piles of up to approx. 25 m long, as well as segmental piles. The use of a steel extension piece will furthermore make it possible to push piles to ground level or even below.

Load-bearing capacity calculation At present, official pile load factors for this pile system (i.e. the type of pile combined with the installation method) are not yet available. In the absence of load test results, it is common practice to classify the ‘new’ pile system based on comparisons to pile systems that are already covered by

the Dutch standard. Analyses of several pushed piles from the early days of this technique and from projects that have been concluded show that existing pile load factors used for driven precast piles lend themselves well for use with this ‘new’ pile system. During the design phase, the calculation of the bearing capacity of a pushed precast concrete pile is, therefore, approached as if it concerned a driven pile using values from Tables 7c and 7d of the NEN 9997-1 (2016). It should be noted, however, that given current knowledge on pushed piles, a pile base coefficient (αp) of 0.7 seems to be on the low side (Van Baars 2018). Nonetheless, this value is still used until official load factors have been established based on pile load testing in accordance with NPR7201 (2017).




Pushing force prediction Given the piling method, it is important to accurately predict beforehand how deep into the soil piles are expected to be pushed. To do this, a CRUX-developed program is used to predict the pushing force needed to push a pile down to the required depth. In principle, the force needed to statically push a pile into the ground equals the expected value for pile tip resistance plus shaft resistance. Based on past analyses of recorded pushing forces, the maximum pile tip resistance is calculated using a pile base coefficient (αp) of 1.0 instead of 0.7, while pile tip strain (qb;max) is not capped at 15 MPa.

The influence zone below and above the pile tip is chosen between 2D/4D and 4D/8D. Although 4D/8D is a safe option for the design compressive resistance, it is potentially an overly favourable approach for the maximum expected pushing force, because the influence of unfavourable layers with diminishing cone resistance values is less significant. From the moment of insertion into the soil (at ground level), the pile is subjected to resistance on the tip and along the shaft. In light of this, skin friction (shaft friction) is also factored into the clay and unfavourable layers, from which, in line with the load-bearing capacity calculation, no shaft friction is derived. The pile load factor to use for the shaft in these unfavourable layers is determined based on Table 7d of NEN 9997-1:2016. In determining the maximum pile shaft resistance, cone resistance is not capped at 12 or 15 MPa, but actual values measured for cone resistance are used. Given that the principle of pushing precast concrete piles into the soil is based on local displacement of soil under the pile tip, the partial resistance factors (γb and γs ) and correlation factors (ξ) are set at 1.0. The total pushing force (Rc;k;total) needed is determined per level based on: Rc;k;total = Rpile tip + Rshaft It needs to be clear in advance which piles cannot be pushed through the centre of the rig, as this has great implications for the maximum pushing depth and is consequently a crucial factor to consider when designing the piling plan.

Assessment of pushing force As specified above, the pushing force prediction takes shaft friction into account from the moment of entry into the soil. When determining the design value bearing capacity, this is normally done from the top end of the zone where positive shaft friction is factored in, i.e. the foundation sand layer. To establish a relation between the actual pushing force achieved (force data recording) and the loadbearing capacity calculated, shaft friction down to the foundation sand layer must be deducted from the total pushing force calculated. Plotting the load-bearing capacity (Rc;d i.e. excluding negative skin friction) out against the calculated required pushing force in the sand (designated as Rc;k;sand) for each sub-area/CPT (Cone Penetration Test) and pile size will produce an equation as shown in figure 2. This should preferably be done for each sub-area/CPT over a length from at least pile tip level to approx. 2.5 m above that to obtain a more realistic comparison than when looking only at the pile tip level. It is important here that the variation coefficient over this length does not exceed 12% and that the load-bearing capacity increases as the pile is pushed deeper into the soil. The equation thus derived can only be used to make an assessment of the force recorded if the predicted force matches the actual value measured. Therefore, figure 3 plots the pushing forces measured (force data recording) for two piles out against the predicted pushing force.

Figure 3 – Calculated versus measured pushing force (top: total; bottom: in the sand).




According to the prediction the pushing force in the sand at pile tip level would be approximately 1,590 kN with an average value of approximately 1,280 kN over the final 3 metres. For two piles that were installed nearby, piles 105 and 106, values of 1,330 kN and 1,550 kN respectively were measured at pile tip level and values of 1,210 kN and 1,310 kN respectively along the pile length above the pile tip. Especially for the bottom section of the graph, the prediction is well aligned with actual values measured. Load-bearing capacities calculated along this length have a variation coefficient of 5% to 10%, as also reflected in the pushing forces measured (recorded data) in the sand. Based on the extrapolated comparison in figure 2, a reliable indication of the load-bearing capacity can be given for piles 105 and 106, namely approximately 780 kN and 900 kN respectively. The negative friction, as determined based on the design CPT’s, can (depending on the soil composition) subsequently be deducted, given that the pushing force graphs do not give grounds to adjust this value.

Experience gained in previous projects Every pushed pile basically constitutes a kind of compression test, as also shown by the example of assessment of readings from the pushing force recording system in this article and the example in figure 4. Based on projects completed so far, the pile base coefficient (αp) with this piling method is more likely to be 1.0 rather than the 0.7 from the

Figure 4 – Calculated versus measured pushing force (top: total; bottom: in the sand).

Dutch standard. This is despite the fact that a large number of pushed piles are over 8 metres in the sand. It is noted, however, that the tip-to-shaft ratio has not yet been determined exactly using sister bars/strain sensors, but is instead based on assumptions. This can be a subject for follow-up studies.

Design and execution aspects The pile pushing method (i.e. installation of precast concrete piles using the pushing method) combines the quality benefits of precast concrete piles with the silent and vibration-free characteristics of auger piling systems. These latter systems do, however, require extra effort in terms of quality monitoring, because they involve foundation elements that are cast in situ. Continuous total force recording during pushing gives this piling method further quality benefits. Focus points to consider with this piling method are the weight required and the dimensions of the hydraulic pile pusher, accurate prediction of the achievable pile tip level (also in relation to the

required load-bearing capacity) and incorporation of this system in foundation designs at an early stage. These aspects will be addressed successively. Rig weight affects the load-bearing capacity of the subsoil, as well as the possible influencing of objects in the immediate surroundings (such as adjacent buildings, embankments, sheet piling, and cables and pipes) and requires expert assessment through a risk analysis. In principle, the rig can push a pile positioned as close as approximately 0.8 m to an adjacent building (pushing force of approximately 50%). If the subsoil requires the use of dragline mats, it is advisable to align the piling pattern with the mats. During the design phase, there should also be extra focus on compacted to highly compacted sand layers, because these are difficult to pass. If the maximum allowable pushing force is reached above the required pile tip level, the pile will have to be cut off before work on the next pile can start. A robust assurance concerning feasibility of the




piling depth (pile tip level) is therefore recommended. Consequently, it may be advisable to, based on the pushing force prediction, use more shorter piles with possibly larger diameter rather than fewer but longer piles. Performing extra CPT’s when facing varying soil conditions will also reduce the risk profile with regards to the feasibility of the required pile tip level. The hydraulic pile pusher is suited particularly for projects of some scale (>50 piles). For projects with a regular piling pattern (for example, line infrastructure such as tunnels, approach slabs, noise barriers) or house and utility building projects, where early alignment of piling plans with this new piling method is possible, the pushing method offers clear benefits in combination with the absence of inconvenience caused by nuisance and vibration, a high speed of construction and proven high quality. 쎲

Prefab pile foundations without vibrations

+31(0)514 56 8000 is part of the IJB Group

P.O. Box 210


8530 AE Lemmer

I The Netherlands

Maria Luisa Taccari Deltares

Vahid Galavi Deltares

Faraz Sadeghi Tehrani Deltares

Ahmed Elkadi Deltares

IT IS WARMER, BUT ARE OUR ROAD EMBANKMENTS STILL SAFE? Extreme events and climate change Climate changes predictions for the Netherlands indicate increased frequency or intensity of extreme events and changes on precipitation and temperature [1]. Rain and extreme rain conditions will increase in winter while extreme rain will become more intense in the summer, with heavier hail and thunderstorm. The temperature keeps increasing, i.e. milder winters and hotter summers will happen more frequently. There will be more often dry periods, also in combination with changes in rain and evaporation. This will lead to an increase of potential evaporation, which depends on temperature and solar radiation. A report from the Royal Netherlands Meteorological Institute, KNMI, published in 2014, [1] gives the likely changes in the climate of the Netherlands around 2050 and 2085 compared to the climate in the period 1981-2010. The report translates the research results on the global climate in the Intergovernmental Panel on Climate Change IPCC report of 2013 to the Netherlands. Four scenarios describe the likely changes in the climate of the Netherlands. The scenarios differ in the extent to which the global temperature increases (“Mo-

derate” and “Warm”) and the possible change in the air circulation pattern (“Low value” and “High Value”), as shown in Figure 1. The increasing temperature, mean amount of precipitation, drought events and increasing intense precipitation are the major climatic changes that are likely to affect the stability of geotechnical infrastructures. The main climate-change-driven processes will be the generation of soil drying, the reduction in soil suctions, soil desiccation, soil erosion, flooding and hydro-mechanical failure [2].

Effect on road embankments Extreme climatic events (e.g. rain and drought) can lead to instability of road embankments. Rainfall induced slope failure is governed by the hydro-mechanical behaviour of the slope, external loads and environmental (weather) conditions. Infiltration of water influences predominantly this process, causing a change in the effective stresses in the slope [3]. The interdependent influence of weather conditions, soil permeability and surface vegetation dominates the pore water pressure regime within embankments. The dominant type and timing of embankment failure is influenced by the material composition and the difference in

construction of highway and railway embankments. Soil matric suction, which affects both the permeability of the soil and its shear strength, plays an important role in the slope failure process. Modelling the unsaturated zone, through the correct determination of saturated hydraulic conductivity and soil-water retention curve, is of main importance [4]. Above certain rainfall intensity, no more infiltration into the soil happens and only surface runoff increases. Not only rainfall, but also drought can weaken embankments and lead to their instability. Extreme drought conditions impose thermohydro-mechanical weakening mechanisms to slopes. These mechanisms include the following: influence on soil strength, desiccation cracking and soil softening, land erosion and subsidence, and soil organic carbon decomposition [5]. An overall increase in temperature and prolonged periods of drought can cause long-term soil drying. High temperatures during dry seasons can lower the water table to considerable depth in the soil profile. The unsaturated area dries out due to evaporation and plant transpiration, volume

Figure 1 – Collapse of a road embankment after heavy rain along the highway A74 near Venlo, the Netherlands. Photo by Fer Traugott [10].




SUMMARY Extreme weather events such as long and/or intensive rainfall can lead to instability of natural and man-made slopes. In the Netherlands, the changing climate will possibly impose increased frequency or intensity of such extreme weather events. These will likely influence the state of existing and mostly aging transport embankments that were not designed with climate change aspects in mind. Therefore, there is a need to better understand how these embankments will perform under climate change scenarios and if necessary, devise plans for adapting them to new climatic conditions. We provide a method to estimate the effect of climate change on geotechnical stability

change/shrinkage occurs as its water content decreases. Due to volume shrinkage and increase of suction, the tensile stresses in the dehydrated soil increases and when it exceeds the tensile strength of the material, shrinkage cracks initiate. As rainwater infiltrates less easily into the dehydrated soil, shrinkage continues, and cracks grow deeper. The process stops when the soil reaches its shrinkage limit and the void ratio remains constant with reduction in moisture content [6]. The initiation and propagation of cracks due to shrinkage depend on several factors, such as initial water content, mineral composition, clay content and plasticity index, layer thickness and size, surface vegetation cover, cyclic change of the climate [7]. Soil mechanical and hydraulic characteristics are significantly modified by the presence of desicca-

of road embankments. A series of fully coupled hydro-mechanical analyses under unsaturated condition are carried out on a typical road embankment. A selection of probable climate scenarios for the next 70 years is applied to the calculations in terms of rainfall intensity and duration. The results compare the current climate based on historical data with the climate scenario around 2085, as reported from the Royal Netherlands Meteorological Institute, KNMI. The authors also looked at how the development of soil desiccation cracks, after prolonged periods of drought, effect the stability of the road embankment.

tion cracks. The hydraulic conductivity increases, sometimes as large as three orders of magnitude. Water infiltration into open cracks can occur more rapidly, pore pressures increase and the effective stress and corresponding soil strength decreases [8]. Moreover, the sliding surface follows the path of least resistance. If an open crack, which has no shearing resistance, is located near the most likely sliding surface (without cracking), the sliding surface can follow the crack and lead to failure.

Numerical simulations This work focuses on short and long duration rainfall events with high and low intensity, respectively. To that end, a series of fully coupled hydromechanical analyses under unsaturated condition are carried out on a typical road embankment using the Finite Element package, PLAXIS 2D, version

2017 [8]. The model uses a 4 m high embankment with a slope of 1:3, as shown in figure 2. The core material consists of sand, covered by a 30 cm top layer. The road, consisting of impermeable crushed asphalt, lies on top of a permeable granular road base, both having thickness of 30 cm. An FE mesh consisting of six-noded triangular elements is employed to discretize the soil material. The mesh is refined close to the boundaries of the slope. The analysis uses the Mohr-Coulomb material model. To obtain a representative case, data obtained from practice have been used. The hydraulic properties of the soil are described with SoilWater Characteristic Curve (SWCC) and Hydraulic Conductivity Function (HCF). The left and right boundaries of the domain enable movement in the vertical direction and restrict movement in the horizontal direction, while the bottom boundary is fixed. For hydraulic boundary conditions, the lateral and the bottom boundaries are set as closed, while rainfall boundary is assigned to the top boundaries of the model. A selection of probable climate scenarios for the next 70 years is applied to the calculations in terms of rainfall intensity and duration. The current study considers rainfall duration and intensities for the winter and summer months, comparing the current climate based on historical data with the WH scenario around 2085, as given by [7]. Here only the results for the summer scenarios are presented. The simulation for the summer considers the effect of extreme precipitation events. Scenario WH around 2085 exhibits a decrease of mean amount of precipitation but an increase of extreme rain

Figure 2a – Geometry of the road embankment (top) and FE mesh (bottom) used in the analysis. Figure 2b – Results in terms of safety factor for the summer scenarios. Results are given with and without desiccation cracks, for three scenarios: no rain, current climate scenario and climate scenario around 2085.




Table 1 - Rainfall duration and intensity for summer considered in the analyses for current climate and climate scenario around 2085. The scenarios for summer consist of two phases: a longer duration rainfall event of 1 week, followed by an extreme precipitation event (2 hours).


Rainfall parameter

Summer [m/day]

Duration Intensity – current Intensity – WH 2085

Phase 1 longer duration rain 1 week 0.0078 0.0071

Phase 2 extreme precipitation 2 hours 1.54 2.24

Figure 3 – Results in terms of safety factor for the summer scenarios. Results are given with and without desiccation cracks, for three scenarios: no rain, current climate scenario and climate scenario around 2085.

Current climate without cracks

Climate scenario around 2085 without cracks

Current climate with cracks

Climate scenario around 2085 with cracks

Figure 4 – Failure surfaces for the summer scenarios: current climate on the left and climate scenario around 2085 on the right. Results are given for the case without (above) and with cracks (below).

events [1]. The simulations apply a week of mean intensity rain followed by two hours of intense rainfall to the model. Average quantity of rainfall per summer corresponds to 224 mm in the reference period 1981-2010 with a year to year variation of 113 mm. The amount of infiltration, applied for a week for the reference scenario is equal 0.0078 m/day, considering the upper bound of the variation and 43 days of rains. The climate change values for the climate around 2085 present a decrease of 23% according to scenario WH for 2085, a natural variation averaged over 30 years of 15%, a year-to-year variation increase of 2.3% and a decrease of rainy days of 5%. By combining these values as described by KNMI, rainfall intensity of 0.0071 m/day is applied to the future scenario for a week of rain. After that, two hours of intense precipitation are applied to the model, with values corresponding to 1.54 m/day and 2.24 m/day for the current and future scenarios respectively,

according to a return time of 1000 years. As according to design requirements, the calculation considers a traffic load equal to 20 kPa over the whole roadway [9]. The factor of safety is then calculated by means of a phi-c reduction procedure. Figure 3 shows the safety factor, which is calculated through shear strength reduction method, for the summer scenario. The safety factor decreases in the future scenario. It is equal to 2.70 in case of no rain and for the current climate, while it decreases to 2.1 for the climate around 2085. It implies that climate change has effect for stability, reducing the safety factor by 22%. According to KNMI scenarios for future climate, temperature will also increase. Increase in temperature and prolonged periods of drought can cause long-term soil drying. The effect of soil desiccation with largest impact on the stability of geotechnical




infrastructure is the development of dessication cracks. The presence of desiccation cracks in soil significantly modifies its mechanical and hydraulic characteristics. In order to account for this effect, a second calculation considers a top layer with a hydraulic conductivity one order of magnitude larger and soil water characteristic curve and hydraulic conductivity function of a medium type soil, as given by the Plaxis database. The resulting safety factor for the climate scenario around 2085 is equal to 1.80 (Figure 3). Similarly, soil desiccation cracking is simulated for the current scenario, leading to a safety factor equal to 2.40. The failure surface, for future climate scenario both in the case of presence or absence of cracks, starts from the top of the slope and it ends at its toe (Figure 4). However the presence of cracks reduces the safety factor by 33% compared to the case without precipitation. Desiccation cracking facili-

tates the infiltration of water, leading to significant decreases in the effective stress and corresponding soil strength. It should be noted that also calculations for the winter scenarios have been carried out. The safety factor decreases slightly when rainfall is applied to the model, while climate change doesn’t contribute to its reduction.

Conclusions A method to estimate the effect of climate change on geotechnical stability of road embankments has been presented. It is suggested to include the effect of climate change when designing and assessing the safety of road embankments. After definiting the critical cross section and determining the hydraulic and mechanical properties, climate scenarios should be chosen as suggested by national or local climate centers. The results of the calculations indicate that the criticality of the extreme events for geotechnical stability strongly depends on the hydraulic properties of the top layer. Due to drought, desiccation

cracking can decrease the hydraulic conductivity of the top soil; if then a longer period of rain happens, less run off occurs and more water infiltrates. When periods of drought are followed by low-intensity precipitation of long duration and sequentially by an extreme rainfall event, the stability may critically decrease. Higher temperature and prolonged periods of drought do also increase soil suction. Depending on the hydraulic conductivity of the material, the reduction of hydraulic conductivity with higher suction reduces the infiltration of water. In this case, extreme events could instead lead to erosion of the slope of the embankments. Contributions of climate changes on crack development and erosion are therefore critical and, for this reason, they are currently under study.

References 1. Koninklijk Nederlands Meteorologisch Instituut. KNMI’14 klimaatscenario’s voor Nederland. s.l. : KNMI, 2015. 2. Vardon, Philip J. Climatic influence on geotechnical infrastructure: a review. (2015). Environmental Geotechnics, pp. 166-174. 3. Briggs, K.Ml., Loveridge, F.A and Glendinning,

S. (2017). Failures in transport infrastructure embankments. Engineering Geology, pp. 107-117. 4. Birhan, Kenaglu Melih, et al. (2018). Effect of Unsaturated Soil Properties on the Intensity-. Teknik Dergi, 30 (2). 5. Robinson, Joe Dylan and Vahedifard, Farshid (2016). Weakening mechanisms imposed on California’s levees under multiyear extreme drought. Climatic Change, pp. 137(1-2). 6. Wijesooriya, Rathnayaka Mudiyanselage Sasika Dilrukshi (2017): Modelling of desiccation crack depths in clay soils. figshare. Thesis. 7. Albrecht B. A., Benson, C. H. s.l. (2011). Effect of desiccation on compacted natural clay.: Journal of Geotechnical and Geoenvironmental Engineering. Vols. 127(1), 67–75. 8. Brinkgreve, R.B., Kumarswamy, S. and Swolfs, W.M. PLAXIS (2016). Delft, The Netherlands : s.n. 9. Rijkswaterstaat. Eisen onderbouw (2016), version 4. 10. pompen-noodplan-voor-a74-bij-venlo. [Online] [Cited: January 19, 2019.] 쎲


GEOTECHNIEK ECSMGE 2019 SPECIAL P.O. Box 25296 3001 HG Rotterdam The Netherlands

In The Netherlands and Belgium Geotechniek is powered by many companies in the sector such as: KIVI Engineering Society, Deltares, Royal HaskoningDHV, Pagani Geotechnical Equipment, HUESKER, BESIX, Ballast Nedam, DEME/DIMCO, Fugro, Jan de Nul/Soetaert, Allnamics, BAM, A.P. van den Berg, Low & Bonar, KELLER, NAUE GmbH, Eijkelkamp GeoPoint SoilSolutions, and many more.

Publisher: Mr. Robert P.H. Diederiks Tel. 0031 (0)10 425 6544 © Copyrights Uitgeverij Educom September 2019

Please find previous editions of Geotechniek at:











High quality, innovative CPT equipment Eijkelkamp GeoPoint SoilSolutions is the leading Dutch manufacturer of CPT cones and equipment. Known for quality, accuracy and user-friendliness. The products that get the results you need! We offer an extensive range of innovative and practical survey equipment, developed to obtain very accurate and detailed geotechnical data. Passionate about CPT, our team is able to develop innovative solutions for every CPT challenge. Se Service lead times őɡƜ ɱǗȫɡʀࡊ őȚƉ ʸƜ žőȚ ɱʗɖɖȫɡʀ ˁȫʗ ǂɡȫȔ ȫǂ˛žƜɱ ʸȫɡȂƉʸǠƉƜࡑ al equipment equipm ment Geotechnical Inclinometers Piezometers Extensiometers

CPT Cones Digital CPT(u) SonicCPT Vision CPT (Videocone) Seismic CPT Magnetometer Environmental CPT

M Meet eet the e diff difference fer e ence

CPT units Compact CPT R Rig Drill’n CPT

T +31 71 301 925

E inffo