What issues need to be considered in using the maps?
1. There is no plausible single measure of the welfare impacts of climate-related events such as heatwaves and floods. People suffer in different ways and some of these will be more important than others for influencing how seriously or quickly a person will be affected, how long these impacts last, and what knock-on effects there will be,1 issues which indicators cannot fully capture.2 However, the clustering of different dimensions of disadvantage means in practice it may not be difficult to identify the most vulnerable and climate-disadvantaged individuals and neighbourhoods based on the evidence we have already on the social impacts of extreme weather.
2. Some factors are difficult to represent as indicators and selecting and interpreting indicators is inevitably contestable. Indicators have been selected on the basis of an interpretation of the factors affecting social vulnerability identified in the existing literature.3,4 However, these interpretations are sometimes open to debate. Indeed, in some cases there is disagreement about whether factors act to enhance or reduce the potential for harm, for example, the influence of previous experience of flood events. Rather than acting to improve resilience, previous experience of flooding could be more indicative of people living in fear of the ‘next’ event (even though it may not occur in their lifetimes) or include people who have not fully recovered in other ways. There are other examples of factors which are difficult to represent, such as social networks or the availability of insurance. Proxy indicators have been used here in some cases; these are imperfect and provide only an indirect representation of the factor which they are intending to represent.
3. There is a general lack of suitable indicators at a fine geographical scale for some factors. This restricts the reliability of the index maps. For example, a limited number of indicators were available to use as a proxy for social networks at the neighbourhood scale. There is a low level of confidence associated with the social networks indicators. The new revised flood disadvantage data12 from work published in 2017 use finer scale geographical units and give additional metrics of community support. Nevertheless, some of the same data limitations apply.
4. There are some important caveats for some of the flood-related indicators.
-
The following limitations apply for the original flood disadvantage data:
-
For England, where both flood zone and historical flood data were made available, the former has been used to represent local knowledge and the latter insurance availability. It could equally be argued that both represent a lack of insurance availability and that this is more important than the local knowledge associated with previous flooding in a locality.
-
The relative percentage cover of a neighbourhood by a flood zone representing significant (1-in-75-year) likelihood of river or coastal flooding has been used to represent the relative potential for insurance availability problems. A similar measure has been used to represent the potential for flood exposure. However, a neighbourhood may have a large proportion of its land area regularly flooded, but this could be in an entirely unpopulated area. Results should therefore be locally verified and treated with caution. Work for the Scottish Government has explored different ways of representing the potential for people and communities to be affected,5 e.g. by using the proportion of house addresses potentially affected.
-
Even for neighbourhoods estimated to have extreme flood disadvantage, there is not necessarily a geographical match between the places where socially vulnerable people live and the places where flood likelihood is thought to be high. This is particularly true in the case of highly geographically constrained hazards, such as river and coastal flooding.
-
The new revised flood disadvantage data overcome some of the limitations of the original flood disadvantage data. For example providing finer scale data and equivalent data across the UK. However, some limitations still apply:
-
For flood vulnerability data, the units used are relatively small, but social vulnerability can be extremely localised, to a specific street, household or individual. Some variability should be expected even within the smaller units used in the revised social vulnerability data. Furthermore, there will be some inconsistencies in data for England, Scotland and Wales due to slight differences in some of the indicators used, i.e. where no exact equivalent is available. There are also some differences in the sizes of units between the different nations. Many of the input datasets are based on the 2011 Census so changes are expected over time.
-
For flood hazard data, there is no consideration of ground water flooding, but this is considered to be far less important in the UK compared to fluvial, coastal and surface water flooding13. There is also no consideration of the joint probability of different forms of flooding.
-
For flood exposure data, the National Receptor Database (England and Wales) and Scottish Property Datasets are used to assess the number of affected properties and their characteristics. Although considered fit-for-purpose here, there are known errors associated with these datasets14.
-
For future flood exposure data it should be noted that the floodplain extent is defined by the present day 1:1000-year return period flood. This is assumed to remain unchanged in the future (although expected changes in flood probability are represented within the floodplain does change).
-
For future scenarios, there is no consideration of economic change or changing social characteristics. By their nature, scenarios can only present a broad picture of possible future characteristics. They are always prone to uncertainty.
5. In selecting indicators, efforts have been made to try to avoid unintentional double-counting while also being sensitive to some of the subtleties of the factors identified from the literature. Inevitably this is an imperfect process. A data reduction analysis identified the principal indicators from a much larger set of 80 considered in the original analysis.6 The outputs of this exercise helped to determine the most important socially vulnerable groups requiring consideration in the Climate Just resource.
6. Results of the original socio-spatial index appear to be reasonably insensitive to the weighting schemes used, i.e. how the individual indicators are combined into a single measure.7 In the current work all indicators are allocated equal weights irrespective of how many indicators are within each domain, such as age, income, social networks and local knowledge. However, it is still recognised that the weights used for the main analysis may not capture the true relative importance of different factors and each of the dimensions of socio-spatial vulnerability.
7. The size of units used in this study provides a broad picture of socio-spatial vulnerability. However, the physical size of units (e.g. in rural areas) and/or their internal variability will mean not all socially vulnerable places will be identified. Furthermore, it is still not possible to infer the broad characteristics of neighbourhoods to all individuals within them. Local knowledge and very fine-scale assessments using quantitative and qualitative methods are therefore a vital complement to national-scale analyses.8 The heat and original flood socio-spatial vulnerability and disadvantage data are mapped at Middle Super Output Area (MSOA) level but some indicators are available at Lower Super Output Area level (LSOA). A comparison of MSOA and LSOA versions of the index for 2001 can be found here.9 The revised flood vulnerability and disadvantage data use a new neighbourhood size based on LSOA units for England and Wales and Data Zones for Scotland.
8. Socio-spatial vulnerability results have been manipulated to a 25km grid in order to combine the results with some climate metrics. This introduces uncertainties since it involves modifying the geographies associated with the index data sources and results. Different geographies can be found in spatial data just through the effects of different zoning schemes and the use of different scales.
9. The heat disadvantage data in this map tool are developed using measures of current (2011) vulnerability with the current likelihood of being exposed to flooding and future high temperatures. Future projections of some of the vulnerability indicators may be available locally to supplement the information provided in the map tool.
10. The heat disadvantage data in the map tool have not been developed using a measure of future heatwave probability. This data can be created using the tools provided alongside the UKCP09 scenarios but is not made available in an easily accessible format for use in nationwide studies of this type. As an alternative, simple metrics of future heat hazard are used.10 It is recognised that they may not be representative of the geography of future heatwave probabilities across the UK. They are merely illustrative of the different pictures of relative climate disadvantage which may emerge and which are only a very small part of the full picture of climate change impacts in the country. Furthermore, temperatures are not the only climate determinant of heat stress, see the Heatwave Plan for England. The use of temperatures here is guided by the use of temperature thresholds for UK heatwave warning systems.
11. Heat disadvantage maps are created using the average (mean) of the score for social vulnerability and the score for potential exposure.11 Extreme values in either score will affect the overall average. Users can take account of this by reviewing the maps and data for each of the components of the overall index, i.e. exposure, vulnerability and the vulnerability dimensions.
12. The maps in this tool make a distinction between the likelihood and severity of exposure to a hazard, on the one hand, and vulnerabilities to the hazard, on the other, but in practice the difference is more blurred. The very fact of an increased threat makes a person more vulnerable. For example, in a risk-differentiated insurance regime, those with the greatest likelihood of flooding are also those who have greater problems in access to institutional protection against the consequences of flooding. Hence, vulnerability measures will not in practice be independent of likelihoods of exposure.
