Mapping flood disadvantage in Scotland 2015: report

3. Methods and data

This section begins by explaining the assessment framework for social vulnerability to flooding and flood disadvantage used in the project ( section 3.1) before discussing the personal, social and environmental factors that affect vulnerability to flooding, introducing the corresponding indicators, and describing how the indicators were combined into the vulnerability index ( section 3.2). The datasets pertaining to flood hazard in Scotland are described in section 3.3. The details of the datasets used, processing methods and map development are provided in the methodology document.

3.1. Assessment framework for social vulnerability to flooding and flood disadvantage

This project applies the approach to social vulnerability to flooding and flood disadvantage developed in the assessment for the UK (Lindley et al, 2011) and the first disadvantage assessment for Scotland (Lindley and O'Neill, 2013). The text in this section therefore refers to these reports and the subsequent Climate Just website ( www.climatejust.org.uk).

The assessment framework developed by Lindley et al (2011) is based on the 'risk triangle' (Crichton, 1999), originally applied as a method for estimating risk (the probability of loss) in the insurance industry. In the risk triangle, the magnitude of risk depends on:

• The extent of vulnerability, or the extent to which the object or system in question would suffer damage or loss;
• The frequency and severity of hazard; and,
• The level of exposure, for example the location that results in contact with the hazard.

Lindley et al. (2011) used this framework to carry out the assessment of climate disadvantage of communities, focusing on neighbourhoods as the elements at risk, or at disadvantage from climate change impacts. As with the 2011 study, the assessment is carried out at the level of communities, or neighbourhoods (represented by data zones), therefore the geographical characteristics are also important in the assessment. Thus, social vulnerability to flooding considers how the characteristics of individuals, communities and places affect the chance of a neighbourhood being negatively affected by flooding, should it happen in that location.

Social vulnerability to flooding in this report is understood as the varying degree to which people's health and well-being would be negatively affected if they came into contact with flooding. The higher the vulnerability, the greater the negative effect of flooding.

Social vulnerability to flooding is the combination of sensitivity, enhanced exposure and the adaptive capacity, which comprises the ability to prepare for, respond to, and recover from, flooding (see Figure 1).

Sensitivity reflects the personal characteristics, namely age and health status, that increase the likelihood that a flood event will have negative health and well-being impacts on people. In the neighbourhood-level assessment, a higher proportion of older people, young children and those in poor health would increase the sensitivity.

Enhanced exposure refers to the aspects of the physical environment, which accentuate or offset the severity of flood events. For example, neighbourhoods with little green space (and thus low flood water infiltration rates) and a high proportion of houses with basements would have higher enhanced exposure.

Adaptive capacity is the ability of people to prepare for, respond to and recover after flooding, related mainly to their social and material situation. For example, areas of high material deprivation, poor access or where social networks are weak, are likely to have lower adaptive capacity.

Figure 1. The framework of socio-spatial vulnerability and flood disadvantage (after Lindley et al., 2011; adapted to flood hazard).

Thus, this assessment framework recognises that vulnerability is influenced by a mix of personal ( e.g. disability or age), environmental ( e.g. elevation of housing, presence of green space) and social factors ( e.g. levels of income, tenure or extent of social networks) which, when combined, affect the degree to which flood events may affect the well-being of individuals. These factors are discussed in section 3.2.

In the risk triangle , hazard is the type (or source) of flooding - coastal, river or surface water flooding and the likelihood of such flooding happening. The likelihood is expressed in return periods (for example 1 in 200 years), which estimate the average length of time between flood events of similar magnitude. The likelihood may also be expressed as the Annual Exceedance Probability of a flood event of a given magnitude taking place - for example, a 1 in 200 years event has a 0.5% chance of happening in any given year; whilst a 1 in 10 years flooding has a 10% chance of occurring in any given year.

Exposure in the risk triangle refers to the geographical location of flooding. It is represented by flood extents, i.e. areas on flood maps where flooding is predicted to occur for a given likelihood. SEPA produces flood extents for different types and likelihoods of flooding (see section 3.3), which therefore represent the combination of hazard and exposure of the risk triangle and are referred to as hazard-exposure.

Flood hazard-exposure is the spatial extent of flooding of a given type and likelihood.

According to the risk triangle framework, if any one component or 'side' of the triangle is zero, then there is no risk (Crichton, 1999). Therefore, in locations where social vulnerability is high but the likelihood of flooding is close to zero ^[3] , the negative impacts of flood events on health and well-being will not be realised. Flood disadvantage, therefore, only occurs where social vulnerability coincides with hazard-exposure, i.e. where vulnerable communities live in areas that may be exposed to flooding. Therefore, the level of flood disadvantage reflects the magnitude of social vulnerability to flooding and the magnitude of hazard-exposure.

Flood disadvantage relates to the situation where neighbourhoods assessed as vulnerable coincide spatially with areas which may be affected by flooding.

3.2. Factors influencing social vulnerability to flooding

This section describes the personal, social and environmental factors that make individuals or households vulnerable to flooding and provides evidence that supports the use of direct or proxy indicators. These factors can be grouped into domains corresponding to different dimensions of vulnerability (Table 1).

Table 1. Thematic domains against the dimensions of social vulnerability to flooding.

Factors	Domains	Dimensions of vulnerability
		Sensitivity	Enhanced exposure	Adaptive capacity
				Ability to prepare	Ability to respond	Ability to recover
Personal	Age	√
	Health	√
Environmental	Housing		√
	Green space		√
Social	Income			√	√	√
	Information use			√	√	√
	Insurance			√	√	√
	Local knowledge			√	√
	Social networks			√	√
	Tenure			√
	Mobility				√	√
	Physical access				√
	Crime				√
	Access to services					√

The association of different domains with the dimensions of vulnerability is based on existing evidence and represents the strongest links found between the thematic domains and the dimensions of vulnerability. For example, tenure is considered to affect the ability to prepare due to the limited power of tenants to make changes to the property they live in. Yet, the literature also suggests that tenants' ability to recover may be hindered by the additional stress of dealing with landlords in the aftermath of flooding (Whittle et al., 2010). However, whilst the ability to prepare is similar for the majority of tenants, their recovery-phase situation can vary depending on the landlord's actions, and thus is more difficult to generalise.

3.2.1. Personal factors

Personal factors affecting the sensitivity of individuals to flooding include age and health. The impacts of floods on health are more likely to be felt by the old, the young, and those with pre-existing health problems. Older people tend to experience greater impacts from flood events. This includes higher rates of mortality due to drowning, hypothermia and heart problems (Green et al., 1994; Vardoulakis and Heaviside, 2012), and a potentially greater incidence of flood-related disease (for example, the gastro-intestinal infections associated with coming into contact with contaminated water ( HPS, 2011), posing a risk in areas where sewage is mixed with flood water). Conditions such as dementia and Alzheimer's disease can affect how a person views the dangers associated with a flood and their behavioural responses ( DEFRA, 2012).

Flooding has been associated with increased mental health and behavioural problems in children, as well as increases in the incidence of a range of diseases (Ahern et al., 2005; Norris et al., 2002). Cold or damp housing is known to increase the incidence of some minor illnesses and exacerbate the severity of others in children (Marmot Review, 2011). Both older people and children have been found to suffer considerable psychological trauma following flood events (Fernandez et al., 2002; Rygel et al., 2006; Tapsell et al., 2002).

For people in poor health, flooding may restrict an individual's access to medicine, e.g. due to loss or damage, and make it difficult to obtain appropriate medical attention in an emergency. Flood events can directly impact local medical services and also affect the wider community given that it may be necessary for hospitals to postpone routine or other non-urgent medical treatments. Vulnerability can be particularly high during flood events. For example, power-cuts can impact on life support equipment, such as oxygen generators or ventilators, or affect people's mobility given that they may be reliant on electric wheelchairs requiring recharging and/or access to lifts (Fernandez et al., 2002).

3.2.2. Social factors

One of the main social factors affecting vulnerability to flooding is related to people's financial situation. People on low incomes living in areas exposed to flooding may not be able to afford property level protection ( PLP) measures (Bichard and Kazmierczak, 2012). They are also less likely to have home contents insurance (Tapsell et al., 2002). Also, lower skilled workers and those not in work were found to have lower awareness of being exposed to flooding than those in higher socio-economic groups (Fielding, 2012).

The rate of poverty tends to be higher among renters than homeowners, with social tenants having the lowest incomes (McInness, 2013). Thus, tenants tend to have fewer resources to invest in PLP measures, and either require the permission of property owners and managers to implement them or may be reluctant to fully or partly contribute to costs or to suffer the associated disruption of implementing flood resistance or resilience measures when they are living in a property that does not belong to them (ClimateJust, 2015).

Also, tenants are less likely to have home contents insurance compared to owner occupiers. The 2007 Scottish Household Survey found that 56% of local authority tenants and 50% of housing association, cooperative or private tenants had contents insurance, compared to 98% of owner-occupiers with a mortgage (Hayton et al., 2007). Buildings insurance is usually the responsibility of the owner, thus tenants are reliant on their landlord to ensure they live in a building which is appropriately insured (ClimateJust, 2015). As it is the occupiers who may bear most of the cost of flood damage, landlords are less motivated to invest in property-level resilience measures ( ASC, 2011). Finally, private tenants may have less local knowledge as they tend to have shorter length of residence in an area compared to owner occupiers ( DCLG, 2013).

People living in areas with a high turnover of population may be less aware of the likelihood of being affected by floods, how to respond and where to seek support. They may also lack social connections to friends and neighbours in the local community (Zsamboky et al., 2011) who can improve knowledge bases, and provide social support and a response network (Lindley et al., 2011). As a result, those without family and friends within their local area, especially the lower income groups, are the most likely to need to use public shelters in the event of evacuations (Scawthorn et al., 2006). Conversely, where social networks are relatively well-established there is evidence of a better response to emergency situations and quicker recovery (Preston et al., 2014). The World Health Organisation ( WHO, 2013) identifies poor social networks as a vulnerability factor which is particularly associated with: older people, people in poor health or with disabilities, people reliant on social services for home care, people living alone, ethnic minorities, people who are homeless, people who are substance abusers and people living in rural areas. Isolated and housebound people (especially older people) may wait longer for help when service providers cannot reach them due to impassable roads affected by floods (Fernandez et al., 2002). On the other hand, people with children at school age have, in general, better local social networks (Corcoran et al., 2010) and in many cases locally-focused charities reduce the social isolation of individuals (Leisure Futures, 2011).

Other issues affecting the ability to prepare for, respond to, and recover after flooding include the ability to understand information, i.e. being literate (Cutter et al., 2003) and having knowledge of the official language (McGeehin and Mirabelli, 2001). Ability to respond is also influenced by mobility; for example, difficulties with balance or strength may mean that taking recommended flood measures is more challenging (Vardoulakis and Heaviside, 2012). Having access to a car, rather than relying on public transport and the general good connectivity of the area by roads, influences people's ability to respond to flood events quickly. Physical isolation presents a particular challenge for responding to floods, especially if critical transport infrastructure is also affected by the event. People working far away from home may also have limited capacity to assist others, move their belongings or deploy any property-level protection ( PLP) measures at home, such as door guards, in the case of rapid onset events like surface water flooding (ClimateJust, 2015). Also, the deployment of door guards as a precautionary measure when a person leaves home, may be affected by the fear of crime and the anxiety that they could indicate that the residents are away (Douglas et al., 2010).

3.2.3. Environmental factors

The physical characteristics of the neighbourhood can affect the extent to which people are impacted by a flood event. Increased " surface sealing" by roofs, roads, car parks, walkways and paved-over gardens reduces the ability of drainage systems to remove runoff created during intense rainfall events or as a result of flooding. Where water cannot be absorbed into the ground because of built surfaces, it forms surface runoff, which is then channelled into any drainage system. If the rates of rainfall and subsequent runoff are higher than the capacity of the drainage system it can cause surface water flooding (ClimateJust, 2015). Conversely, the presence of vegetation reduces surface runoff and thus can support the reduction of the risk of flooding (Armson et al., 2013).

The type of housing also plays a role in mitigating the effect of flooding on people. Houses with the lowest floor at or below ground level are more exposed than dwellings located on higher floors, and occupants and their belongings may be more significantly affected by a flood event (Thieken et al., 2005). Single level properties favoured by older people and constructed as retirement developments (Pannell and Blood, 2012), may mean that older people living in such houses are disproportionately affected by flooding (nowhere to move their belongings to protect them from flood water; remaining on the floor affected by the flood water in the aftermath of flooding). Also, solid masonry buildings can withstand flooding without suffering major structural damage, whilst lightweight constructions - in particular mobile or temporary structures - may be more easily damaged (Sanders and Phillipson, 2003). In addition, these lightweight and temporary structures also tend to be occupied by people on lower incomes (Benzie et al., 2011).

3.2.4. Indicators of social vulnerability to flooding

As the evidence above indicates, the personal, social and environmental factors affecting vulnerability to flooding are strongly interconnected. It is extremely challenging to identify with full certainty the indicators or proxies that would reflect the complexity of the problems of social vulnerability to flooding emerging from the literature. Therefore, in order to ensure that vulnerability to flooding is represented adequately, stakeholder input was sought in defining the list of indicators in an iterative process. The organisations involved included the Scottish Government, SEPA, JRF, local authorities in Scotland and the National Flood Forum. Table 2 presents the list of indicators used in the assessment of social vulnerability to flooding.

The assessment of social vulnerability to flooding and flood disadvantage presented in this report is based on a quantitative geospatial assessment. Therefore, the values of the indicators were obtained for census units - data zones. Scottish census 2001 data zones were used, since the majority of the data underpinning the assessment has been reported for these census units. Data zones are compact areas with around 500-1,000 residents that contain households with similar social characteristics ( SG ATOM Feed, 2014).

Social vulnerability to flooding and flood disadvantage are assessed at the data zone level. In this report data zones are also referred to as neighbourhoods.

3.2.5. Developing the index of social vulnerability to flooding

In order to add all of the indicators together, the indicators were standardised, which means presenting all the indicators on a uniform scale. Z-score standardisation was used. Then, the indicators were equally weighted within each of the thematic domains (Table 2) they were assigned to. This was done to avoid over-representing the domains with a larger number of indicators. The weighted indicators were added together to develop the dimensions of sensitivity, exposure, and ability to prepare, respond and recover. The dimensions of sensitivity, exposure, and ability to prepare, respond and recover were standardised and summed to form the vulnerability index. The vulnerability index was then standardised. The methodology document provides more details on the procedure.

Table 2. Indicators used in the assessment of social vulnerability to flooding.

Domain	Indicator
Age	% people under 5 years old 1
	% people over 75 years old 1
Health	% people whose day-to-day activities are limited 1
	% households with at least one person with long term limiting illness 1
Income	% people in routine or semi-routine occupations 1
	% of people who are long term unemployed or who have never worked 1
	% households with dependent children and no adults in employment 1
	Number of Income Support claimants 2
	Number of Job seeker allowance claimants 2
	Total pension credit claimants 2
	Total number of families receiving tax credits ( WTC and CTC) 2
Information use	% people with <1 year residency in the UK 1
	% people who do not speak English/no not speak English well 1
Insurance	% new addresses (01.01.2009) in flood risk areas (insurance availability) 3
	Number of historic flood events (insurance cost) 4
Local knowledge	% addresses in Flood Warning Target Areas 5
	% new residents (< 1 year) arriving from outside the local area 1
Tenure	% social rented households 1
	% private rented households 1
Mobility	% of Incapacity Benefit/Severe Disablement allowance claimants 2
	% people living in medical and care establishments 1
	% households with no car or van 1
Social networks	% children of primary school age 1
	Number of voluntary organisations focused on local community 6
	% single pensioner households 1
Physical access	% people working further than 30km from home 1
	Low road density 7
Crime	Number of domestic break-ins 2
Access to services	Travel time to GP surgery (private transport) 2
	Travel time to GP surgery (public transport) 2
Housing Characteristics	% households with the lowest floor level: ground floor 8
	% households with the lowest floor level: basement or semi-basement 8
	% caravan or other mobile or temporary structures in all households 1
Physical environment	% urban land cover 9

Sources of data: 1 - Scottish census 2011; 2 - Scottish Neighbourhood Statistics; 3 - OS AddressBase and SEPA flood extents; 4 - SEPA Historic Flood Data; 5 - SEPA Flood Warning Target Areas and OS AddressBase; 6 - Scottish Charity Register; 7 - OS MasterMap Integrated Transport Network Layer; 8 - Scottish census 2001; 9 - Land Cover Map 2007

The index of social vulnerability to flooding has been categorised into six classes (Table 3). The negative values of the index indicate lower than the average social vulnerability to flooding. The higher the positive values, the higher the social vulnerability to flooding; the highest values of the index are in the 'acute' category. The values oscillating around zero are near the national average. The same categories have been used for flood disadvantage.

Table 3. Classes of social vulnerability to flooding and flood disadvantage

Value of the standardised index	Level of vulnerability / disadvantage
≥ 2.5	Acute
1.5 - 2.5	Extremely high
0.5 - 1.5	Relatively high
-0.5 - 0.5	Average
-1.5 - -0.5	Relatively low
-2.5 - -1.5	Extremely low

3.3. Flood hazard-exposure

Three types of flooding were considered in this assessment: coastal, river, and surface water flooding. SEPA provided flood maps for each source of flooding for different return periods (Table 4) in consultation with the project's steering group. These flood hazard maps were developed with nationally applied methodologies. Further information about the datasets is provided in the methodology document.

The flood return periods were chosen to reflect a range of flood event probabilities. They included:

• 1 in 25 years (coastal) and 1 in 30 years (river and surface water). In this report they are referred to as 'high probability';
• 1 in 200 years (all types of flooding), referred to as 'medium probability';
• 1 in 200 years (all types of flooding) which incorporates climate change projections, referred to as 'low probability' events. Considering this return period allows incorporating a future perspective.

The highest likelihood scenario considered by SEPA, 1 in 10 years return period, was not used due to considerable levels of uncertainty associated with its spatial extent in some locations. The 1 in 1000 years return periods were not considered as these events are extremely rare.

The defended extents were used where available in order to take the presence of flood defences into consideration ^[4] . For surface water flooding, a relatively low depth of 0.1 metres was considered, since even shallow water can cause significant damages and repair costs, thus making it difficult for people to recover after flooding (Kazmierczak and Cavan, 2011).

Table 4. Flood maps used ( SEPA, version 1.1, March 2015).

Type of flooding	Return period	Description	Code used in tables and figures
Coastal	1 in 25 year defended	A flood event is likely to occur in the defined area on average once in every 25 years (1:25). Or a 4% chance of happening in any one year.	C25
	1 in 200 year defended	A flood event is likely to occur in the defined area on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year.	C200
	1 in 200 year: climate change 2080H	A flood event is likely to occur in the defined area in the 2080s (2070-2099) on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year.	C200+cc
River	1 in 30 year defended	A flood event is likely to occur in the defined area on average once in every 30 years (1:30). Or a 3.3% chance of happening in any one year.	R30
	1 in 200 year defended	A flood event is likely to occur in the defined area on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year	R200
	1 in 200 year defended: climate change 2080H	A flood event is likely to occur in the defined area in the 2080s (2070-2099) on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year.	R200+cc
Surface water	1 in 30 year	A flood event is likely to occur in the defined area on average once in every 30 years (1:30). Or a 3.3% chance of happening in any one year.	S30
	1 in 200 year	A flood event is likely to occur in the defined area on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year.	S200
	1 in 200 year: climate change 2080H	A flood event is likely to occur in the defined area in the 2080s (2070-2099) on average once in every two hundred years (1:200). Or a 0.5% chance of happening in any one year.	S200+cc

The index of flood hazard-exposure represents the percentage of residential addresses exposed to flooding in each data zone. Residential addresses were obtained from Ordnance Survey AddressBase, supplied by the Scottish Government (version: 11 April 2015); therefore, as previously noted, this assessment is different to the NFRA, which also considered flood risk to commercial properties. The residential addresses were spatially overlaid with the flood extents listed in Table 4 and the proportion located in flood risk area was calculated for each data zone.

In principle, SEPA's flood extents should not be used to determine the risk associated with individual properties as they have been developed for strategic national mapping and broad scale analysis. However, this study aggregated the number of individual properties within the flood extents first to the data zone level and then to the local authority level. The metric representing the proportion of address points that may be at risk of flooding in a data zone was considered as more representative of the actual levels of exposure than the proportion of the land surface of the neighbourhood potentially affected by flooding. This is because a data zone may have a large land area potentially affected by flooding but this land area may not be associated with housing. The same approach is taken by SEPA for the strategic appraisal process and it has also recently been used by JBA Consulting (2014) in a study of the PLPs.

The percentage of residential address points within each flood extent, and in all the extents combined (any flooding) was calculated for each data zone. For each of the flood extents separately, and for all data zones, the percentage of address points exposed to flooding was standardised to develop the hazard-exposure indicator.

3.4. Developing the flood disadvantage index

The flood hazard-exposure indicator for each of the flood types and return periods (as well as for the flood map combining any type of flooding of low probability) was added to the standardised index of social vulnerability to flooding in order to calculate the flood disadvantage indices. These were then standardised and categorised using the classes presented in Table 3.

Consequently, for each of the 10 flood extents considered, the disadvantage index compares the values in data zones to the average Scottish disadvantage for each particular flood extent. Therefore, each of the disadvantage indices are presented on a slightly different scale. As a result, whilst some areas may for example have relatively high disadvantage with regards to 'any type of flooding', for some of the flood types and return periods the disadvantage may be extremely high or average. The users are advised to investigate the values of flood-hazard indicators and vulnerability in the data spreadsheet in more detail (also, see the case study section).

This methodology addresses some of the shortcomings highlighted by the previous report (Lindley and O'Neill, 2013). Firstly, the defended flood outlines are used in order to represent the risk of flooding taking into consideration the presence of flood defences. Secondly, surface water flooding as a substantial flood risk is included in the analysis alongside flooding from the rivers and the sea. Further, the list of vulnerability indicators has been modified to take into account new data that more accurately reflects the diverse aspects of social vulnerability to flooding and the most up-to date sources of information have been used including census 2011 data (see Appendix 1). The details of the modifications to the methodology can be found in the methodology document. The assessment methodology still has some limitations, which relate to the aggregation of households at data zone level, data paucity e.g. for certain vulnerable groups and enhanced exposure, and no consideration of potential PLP. These are discussed in more detail in section 5.3.