Mapping Flood Disadvantage in Scotland 2015: Methodology Report

4. Indicators of social vulnerability to flooding

4.1 Selecting the indicators

The factors affecting social vulnerability to flooding are represented by either direct or proxy indicators. A total of 34 indicators were used in the assessment (Table 2).

Table 2. Indicators used in the assessment of social vulnerability to flooding.
For sources and more information see Appendix 1.

The initial list of indicators was the one used as a basis for the first flood disadvantage assessment for Scotland (Lindley and O'Neill, 2013). This initial set of indicators was critically reviewed, taking into consideration the following:

• The strength of existing evidence supporting them (see the main report and Appendix 1).
• Feedback from stakeholders, mainly within the ClimateJust project^[6], but also Scottish local authorities within the case studies carried out for this project and National Flood Forum.
• Consultation with the project's steering group, including representatives from the Scottish Government, SEPA, local authorities, and Joseph Rowntree Foundation.

The project team made changes to the original set of indicators, in collaboration with the Steering Group, due to the following reasons:

• Low level of confidence in some of the indicators, as suggested by the feedback offered by the local authorities and other intended users consulted within the ClimateJust project^[7]. For example, the indicator: 'percentage of people not providing unpaid care', was used previously as a proxy for the quality of social networks. However, it was seen by the stakeholders as difficult to understand and ambiguous: whilst the process of providing care helps to extend social networks for those being cared for, carers themselves may be socially isolated or in poor health (ClimateJust).
• Presence of additional data specific for Scotland. For example, the Scottish Charity Register was used as a source of data on the location of voluntary organisations with a specific interest in the local community, which is being used as a proxy for local networks.
• Emergence of more relevant data. For example, whilst in the original assessment of social vulnerability to climate change (Lindley et al., 2011), percentage of people born outside the UK was used as a proxy for the ability of people to speak English, census 2011 collected data on actual English language proficiency, which more accurately reflects the ability to use information related to flooding.
• Specificity of Scotland in comparison to England and the rest of the UK. For example, in the climate disadvantage assessment for England, the proportion of all pensioner households was used as one of the proxies for low income households. However, in Scotland, the proportion of pensioners in low income is lower than in any of the other regions of the UK and pensioners are much less likely to be living in low-income households than non-pensioners^[8]. Thus, this indicator was not considered to be relevant as a proxy for low-income households.
• Unintentional overlap between like indicators (in relation to income). The redundant indicators, i.e. those likely to represent the same aspect of vulnerability as other indicators, were removed based on the results of a correlation analysis (see Appendix 2).

At the same time, it was not possible to update some of the indicators. Census 2011 did not collect data on the lowest floor level of the household which is needed to estimate the enhanced exposure index. Therefore, the data for 2001 has been used, which may not be accurate for areas that have had significant turnover of housing stock, i.e. areas of urban regeneration (see also section 3 in relation to data zones with zero population). This emphasises the importance of local authorities treating this dataset as indicative of the flood disadvantage in their area and using more detailed, locally available information to assess the disadvantage with higher accuracy.

The final set of indicators used in the vulnerability assessment is provided in table 2, which shows how the different indicators have been grouped into thematic domains, and how they fit under different dimensions of vulnerability. Appendix 1 provides more detail about the datasets used, the processing involved in developing the indicators, and the modifications made since the first flood disadvantage assessment (Lindley and O'Neill, 2013).

The association of different domains with the dimensions of vulnerability is open to debate. For example, it could be argued that tenure does not only affect the ability to prepare (due to the limited power of tenants to make changes to the property they live in), but also affects the ability to recover, as evidence suggests that recovery is often hindered by the additional stress of dealing with frequently uncooperative landlords in the aftermath of flooding. As a result, tenants may suffer from more pronounced and lasting intangible impacts associated with physical and mental health than owner occupiers (Walker, 2006; Whittle et al., 2010). However, whilst the ability to prepare is similar for majority of the tenants, their recovery-phase situation can vary depending on the landlord's actions, thus it is more difficult to generalise.

The association of thematic domains with different dimensions of vulnerability was based on the existing evidence. As new evidence and data emerges, the balance of how different indicators contribute to different dimensions of vulnerability may shift. In addition, different locations, both within the UK and overseas, may have their specific circumstances affecting how the indicators and domains correspond with the ability to prepare, respond and recover.

The set of indicators presented in table 3 aimed to achieve a balance between the availability of data, consideration of stakeholders' expertise and opinions and evidence available from published research. Also, in some cases, trade-offs had to be made between the spatial accuracy of the data and its topical relevance. For example, the indicator used previously for the domain 'crime' was the SIMD 2004 Crime index, combining information on recorded instances of various types of crime (see also Appendix 1). Out of these types of crime, only domestic housebreakings bear relevance: fear of looting is a common reaction in the event of flooding (Bonkiewicz and Ruback, 2012), which may mean that people are hesitant to leave their homes. Further, with regard to property level protection (PLP) measures that would need to be in place when the resident was out of the home (e.g. flood barriers), some people may be concerned that these measures make it obvious that the occupiers are away and thus could make the houses more prone to burglary (Douglas, et al., 2010). Consequently, in this assessment, only 'domestic break-ins per 10,000 households' was used; however, this data was only available for the level of intermediate geography (IG; census data units that contain between 2,500 and 6,000 household residents, one level up in the spatial hierarchy from data zones), which meant that all data zones within a given intermediate geography unit were assigned the number of domestic breakings for this unit, whilst there may have been substantial spatial variation among the data zones.

Similarly, the number of charities specifically focused on working within local communities or neighbourhoods was recorded, based on their postcodes, for IG units. This was done using the logic that it is unlikely that charitable organisations' activities would be limited to the small unit of a data zone. The number of charities found for IG unit was applied to the data zones contained within the IG unit, assuming an equal coverage by the charities' activities within that area. However, the data underpinning the majority of the indicators are available at the data zone level, thus the generalisations made for the remaining indicators are unlikely to have a significant bearing for the overall assessment of social vulnerability to flooding.

The indicator that is subject to particular limitations relates to insurance availability, which estimates the proportion of properties built after 1^st January 2009 in flood risk areas as potentially uninsurable. This is based on the exemption of such properties from the commitment to provide insurance under the Statement of Principles between the Association of British Insurers (ABI) and the Scottish Government (2008) in order to discourage the development of properties on flood plains. The caveats associated with this indicator are as follows:

• The 1 in 200 years + climate change flood maps were considered, whilst the ABI Statement of Principles is largely based on the 1 in 75 years return period. This means that the flood extents that were used for the analysis here may overestimate the number of residential properties with potentially limited access to flood insurance.
• The information on the lowest level of dwelling was not available for the purposes of this analysis; therefore, it was not possible to separate houses and flats at/below ground level from those that are on higher floors and may not be affected by flooding. This means that an unknown proportion of these residential properties would never be affected by flooding (however, their residents may still be affected - e.g. all people living in an apartment block would experience some inconvenience if the ground floor was flooded). It is recommended that in future assessments, the Scottish Property Dataset is used to help identify the number of properties located at or below the ground level.
• Coastal and surface water is often readily managed to enable appropriate development.
• These figures do not take into account any resilience or resistance measures to mitigate flood damage that may be present (which under the Statement of Principles is required for properties built from 2009 onwards in areas exposed to flooding). Therefore, if there are PLP measures in place the exposure levels may be lower than indicated here.
• Many of the developments might have been subject to detailed flood risk assessments that demonstrate development as being appropriate to planning policy.
• There is also some uncertainty on the property dataset used, i.e. the column 'start date' may not be appropriate to identify new development.

Therefore, given the limitations of the method highlighted here, the outputs are best used to aid strategic planning to identify areas of concern; more detailed investigations into the nature of vulnerability are advised, to be carried out based on locally available, area-specific data, if possible, to guide specific decisions on targeting resources.

4.2 Standardising the indicators

In order to add all of the indicators together, they were standardised, which means presenting them on a uniform scale. This was done to avoid 'comparing apples and pears', as the indicators are expressed in different units (e.g. number/percentage of people; number/percentage of households; percentage of the area). Z-score standardisation was used, which means that all standardised indicators have a mean (average) value of zero and standard deviation value of one. For the standardised indicators, values above the mean for all data zones in Scotland are positive, and values below the mean for all the data zones in Scotland are negative. The further the original values are from the mean for all data zones, the more extreme the positive and negative values of z-scores.