Case study of the cascading effects on critical infrastructure in Torbay coastal/pluvial flooding with climate change and 3D visualisation

Critical infrastructures (CIs) are commonly designed, built and maintained based on rigorous standards in order to withstand the climate and weather-related pressures. However, shifts in climate characteristics may result in increases of the magnitude and frequency of potential risks, or expose specific CI to new or increased risks not previously considered. As vital components of the normal functioning of modern societies, their resilience encompasses the operational elements, their structural integrity and the capacity to maximise business output under climate stressors. In this work, we apply an integrated and participatory methodological approach to assess the risk and enhance the resilience of interconnected CIs to urban flooding under climate change. The proposed methodology has been applied to an extended case study in Torbay to extend previous works, which seeks to protect coastal communities from future events through using the proposed methodology to justify future investment in coastal defences, as a part of the validation of EU-CIRCLE projects developed methodologies. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/hydro.2019.032 ://iwaponline.com/jh/article-pdf/22/1/77/642308/jh0220077.pdf M. J. Gibson (corresponding author) A. S. Chen M. Khoury L. S. Vamvakeridou-Lyroudia D. A. Savić S. Djordjević Centre for Water Systems, University of Exeter, Harrison Building, North Park Road, Exeter EX4 4QF, UK E-mail: m.j.gibson@exeter.ac.uk D. Stewart M. Wood Torbay Council, Town Hall, Torquay TQ1 3DR, UK D. A. Savić KWR Water Cycle Research Institute, Nieuwegein, The Netherlands

exceeds the order of decades, infrastructures are heavily exposed to natural disasters, often with devasting consequences to the society, economy and the environment. The increasingly dependent, interdependent and interconnected nature of CIs may expose societies to previously unseen risks, new vulnerabilities and opportunities due to disruption across multiple CI networks. Meanwhile, the spatiotemporal evolution of hazard impacts is critical for crisis management and strategic planning to prevent cascading impacts (Mazzorana et  Acknowledging that infrastructure's vulnerabilities and impacts go far beyond physical damage (Hokstad et al. ), the work presented here is concerned with an assessment of the impacts on the services provided by CI. We considered not only the impacts associated with the repair and/or replacement of services but also included the externalities of the CIs operation, societal costs, environmental effects and economic consequences due to suspended activities. Interdependencies among infrastructures dramatically increase the overall complexity of the 'systems of systems'.
There is therefore a need to consider multiple interconnected infrastructures and their interdependencies in a holistic manner (Galbusera et al. ).
The EU H2020 funded project EU-CIRCLE (a pan-European framework for strengthening CI resilience) developed a holistic framework for identifying the risks of multi-climate hazards to heterogeneous interconnected and interdependent CI (Sfetsos et al. ). In the work presented in this paper, the resilience framework was applied for urban flooding.

METHODOLOGY
The EU-CIRCLE methodology is detailed in the literature (Sfetsos et al. ), but the primary features are also presented here. Shown in Figure 1 is the resilience framework developed in the EU-CIRCLE project and is applied in this work. The generic resilience framework has three main sections which are interconnected and allows for tailoring to the local specific CI, or the area or challenges of interest in the given case study it is applied to. The generic resilience framework consists of three main interacting and interconnected sections (shown in Figure 1): 1. A step-by-step precedure climate risk management framework; 2. An interactive risk and risk modelling framework; 3. Outputs to the end-users.
A key to the implementation of this framework is the interaction of stakeholder and end-users at both the development and tailoring of the framework to the given case and the output required. This framework is very flexible and allows for the inclusion of multiple or different modelling methodologies given the stakeholder and end-users interests in the given target area.
The climate risk management framework begins with establishing relevant climate change resilience policies and decisions which are of interest to stakeholders and end-users in the given area. What policies are in place and what questions are to be answered? These issues may have life spans of many years and may cross many different service providers. Climate data may come from any number of sources, including global climate models at higher and lower resolutions, statistically downscaled climate information and historical data. Given the target area, it is possible to establish which infrastructures are critical and which systems interconnect these, and what the risks to these systems are. Some areas are prone to forest fires, or typhoons/hurricanes, and others as in the case study presented in this paper are susceptible pluvial and coastal flooding. Each area will have different risks stemming from the combination of current risks and how climate change may exacerbate these, for example, increased sea levels or more extreme storms. At this stage, any protective programmes, including adaptation options to be investigated as part of the framework, are established.
The risk and resilience modelling section details the structure of risk modelling and its different components and interactions. A number of scenarios are constructed using the climate data and projections, which are combined with a list of relevant CI and risks established in the climate risk management section. The system over- The last section involves the display/visualisation of analysis and result and is connected to all stages, as stakeholder and end-user involvement is so critical. To this end, there is a large focus on using modern computing techniques to fully visualise the results and analysis in order to engage and effectively communicate to stakeholder and end-users. The work in this paper presents a single case study using the EU-CIRCLE methodology.

CASE STUDY SETUP
Torbay case study background  the major cities of Exeter and Plymouth, Torbay has developed with its major industry being that of tourism, due to its relatively warm climate for the UK and beautiful coast lines, and is known as the English Riviera. Due to the steep hill leading down to the coast, it suffers from frequent pluvial (surface) flooding, but also from extreme storms causing the sea to overtop or breach the coastal defences.
There are three large towns of Torquay, Paignton and Brixham which constitute the majority of the catchment area, meaning that the vast majority of the catchment is urban in nature.
There are three target areas in Torbay centred around the three main towns: Torquay, Paignton and Brixham (shown in Figure 2). In order to process comprehensive cli-  flooding are exacerbated due to the lack of the capacity of the surface water outfalls, which discharge to the sea.
These flood events threaten a large number of commercial and residential properties on a regular basis and also affect numerous roads to some extent. A number of roads are closed due to flooding from such storm events, and this can have critical knock-on effects on the road network when large capillary roads are closed, resulting in long traffic diversion and delays.

Terrain data/parameters
For each of the three geographical locations considered, a high-resolution 1-m grid was applied to simulate flooding scenarios that include coastal, pluvial and combined con- Road cells, including those which touch a road polygon, are lowered by 12.5 cm from their existing terrain levels, which leaves their slopes intact. However for buildings, it would be expected that there would be a uniform level across each building, i.e. assuming all buildings have flat roofs. Whereas apart from the terrain smoothing, buildings built on the steep slopes tend to take on these characteristics. Therefore, the cells within each building polygon are set to the same uniform level, based on that of the highest elevation plus a 15 cm threshold, which is designed to simulate the doorstep level of the building.
After the doorstep level, water is allowed to enter the building polygon, but in order to account for the structural walls and content effects on flow within buildings, further parametrisation is required. To account for the reduced flow into, out of and inside of buildings, the caFloodPro application allows for Manning's roughness factor to set for individual or groups of cells. In this case, Manning's roughness is increased from 0.015 to 0.1, which slows down the flow within building areas.
CaFloodPro also allows for an infiltration rate to be set on any particular cell or group thereof, which is utilised to simulate the effects of the sewer system to drain water from the urban surfaces, while also being used to account for the capacity of the green area to absorb water. The infiltration rates used for these areas are shown in Table 1.
Torbay's sewer systems are designed to cope with up to a 1 in 30-year return period pluvial event; however, it is thought due to blockages of vegetable matter, gullies and inlets along roads do not provide the equivalent capacity.
Therefore, road drainage is reduced to the equivalent of a 1 in 5-year return period event, as well as a rainfall reduction of 12 mm/h to green areas, which is in line with the EA procedure.
A smaller domain was utilised when modelling the coastal events, as these events only affect the low lying area next to the coastal line, by delineating the areas which are lower than 30 m above sea level. Whereas pluvial events are largely affected by the area upon which they fall, especially when designed storms are utilised where rain falls the same amount everywhere in the given area. Therefore, a terrain analysis is used to determine the area which is capable of contributing to towns flooding.

Overtopping conditions
The AMAZON model ( Given that there is a 12-h tide cycle, the highest tides affect the rate at which the waves overtop the sea defences, and this can be seen in the inflows to caddies, as shown in Figure 3.

Pluvial conditions
A designed rainfall was utilised for pluvial events, which uses a spatially uniform distribution across the terrain. for 50 years and 40% for 100 years of climate change.

Combined pluvial and overtopping conditions
There is a reasonably low chance of either the extreme pluvial or overtopping conditions to occur independently, and while there is obviously some casual connections in that a storm may create the conditions for both cases, it is still considered to be less likely that the conditions for both exist at exactly the same time. Therefore, a more likely pluvial and coastal set of storms are modelled together, using a 1 in 50-year design rainfall for 1 h which was aligned with the peak of a 1 in 50-year coastal event, at the 36th hour of the simulation.  Paignton's coastal flooding represents the largest risk discovered in this case study and can be clearly seen when contrasted to Brixham ( Figure 10) and Torquay ( Figure 11).
An adaptation plan has been proposed to build additional sea defences to mitigate this risk.
Shown in Figure 9           The results in Figure 14 are in stark contrast to Figure 13, showing that Torquay is at greatest risk of damage from pluvial flooding, and that Paignton is the least at risk from pluvial flood damage. The highest overall damage comes from the pluvial cases, due to the steep terrain of the area which limits coastal flooding and exacerbates pluvial cases. Furthermore, Torquay and Brixham have storage tanks to help alleviate this, which was not modelled. The next biggest risk is that of Paignton, which further justifies the work on the additional sea defences. However, this is assuming that a design rainfall occurs, that is rain falls equally for the temporal duration and spatial extent, which as discussed earlier is the maximum size of the catchment.
Real-world storms have more variable spatiotemporal distributions which would lead to different spatial distribution of risks. While this work highlights all the areas at risk from pluvial flooding in this fashion, it is likely to overestimate the overall risk.

VISUALISATION
Visualisation of the resulting water depth and damages with realistic, descriptive and accurate mapping is a key to stakeholder engagement with such projects as the EU-CIRCLE framework and requires the collaboration of computer science and hydrological disciplines that form the core of the hydroinformatics discipline. Stakeholders need to be able to see the scale/breadth and be able to easily locate/ relate these to real-world locations that can only be provided by a 3D realistic mapping solution. Stakeholders then need to be able to focus on the accurate details of locations of interest of heavy damage, as well as visualising the dynamics of the flood and damages both direct and indirect, in quicker than real time. This is not a trivial challenge, and the EUcircle framework includes methodologies to achieve this high visualisation and is detailed in the following section.
Two main software, Google Earth Pro and the Unity3D game engine, are utilised.

Google Earth Pro
To demonstrate the modelling results in a more user-friendly way that improves risk communication, we exported the  1. The degree of precision used to render the terrain heightmap is insufficient as it cannot render a mesh with subtle differences in height. Presently, the terrain engine only takes 16 bits raw images as input, which limit the precision to 2 16 or 65,536 unique values. Assuming the terrain height range was set to 2,000 m, the setting would provide an approximate resolution of 3.05 cm (2,000 m/65,536) which is too coarse as a 3 cm difference in the water level is not negligible.
2. The default terrain is unable to display consistently overlapping meshes at different altitude levels as shown in Figure 16. As soon as the user 'zooms out', high detail meshes are substituted with simplified one automatically via a forced optimisation process that is suitable for ordinary games, but not for visualising flood.
3. Finally, the default process by which detailed terrain and flood meshes are animated is too slow to allow real-time changes, especially on machines with a less powerful GPU. In the next section, we describe how our system addresses these shortcomings and then show added functionalities regarding the visualisation of a narrative framework as well as the impact of flood on CIs failure.

Visualisation of 3D animated floods
The 3D rendering engine has been modified to allow the terrain/flood height to be rendered with a higher degree of The Shaders are not only engineered to deform 3D meshes efficiently but also change the flood mesh colour gradient from clear blue to dark blue with increasing flood depth ( Figure 17). This works seamlessly even on fairly modest machines or small laptops, animating meshes with more than 1,440,000 triangles in real time with ease, as long as they have a DirectX 11 compatible graphic card.

Visualising consequences of flooding on CIs
All objects such as, for example, 3D buildings can be easily changed in an interactive manner (shown in Figure 18). The Figure 16 | By default, the further away the observer is from the terrain, the more the Unity3D terrain object does deform meshes by substituting highly detailed meshes with simplified ones.  Figure 18).

CONCLUSIONS
In the paper, we presented an extended case study utilising the EU-CIRCLE metholodogy, so as to understand better how future climate regimes might affect the normal operation of interconnected CI in urban areas and how to assess the effectiveness of adaptation measures. The methodology was applied to analyse the flood impacts to Torbay due to pluvial, coastal overtopping events and their combinations under the present and future climate scenarios. Both the direct flood damage and the cascading effects due to CI failures were evaluated using flood hazards information, building characteristics, depth-damage relationship and the interdependencies among CIs and properties. By combining computational modelling with advanced 3D visualisation techniques, we also developed an effective tool for communicating with stakeholders during the decision-making process. The application can help local stakeholders and operators in the co-design of the approach, the assessment and the evaluation of adaptation measures.