When interpreting the results it is important to note that when generalising at the LSOA scale, some data will be masked in a small number of cases. For example, the Sutton Park area in the north of the city that contains the actual park has to be extended to include an area with approximately 1,500 people in order to match the LSOA geography. As this LSOA is physically one of the biggest by area within Birmingham, maps can look skewed.
Spatial Trend between the UHI and Exposed and Vulnerable
The UHI under heatwave conditions at LSOA level (Figure ) reflects the results (from [41
]) and gives confidence that the generalisation to LSOA has not compromised the dataset. A full discussion of the spatial trends is available [41
] but in summary, the highest temperatures are found in the city centre where as the Sutton Park area in the north of the city is the coolest area. As expected, there is a general trend towards lower temperatures in the suburban areas.
Birmingham UHI under heatwave conditions at LSOA level. 18th July 2006 from MODIS remotely sensed data. Shown with contour lines for validation.
The four main "exposed and vulnerable" layers were displayed in a GIS with natural breaks (Jenks) symbology (Figure ) in order to view groupings inherent in the data. Concentrations of old people are scattered throughout the city, with distinct clusters in the north. This is not surprising as the northern Sutton Coldfield area is generally regarded as having a slower pace of life, with close proximity to countryside being appealing to the older generation. This also helps explain the lack of elderly people in the city centre, where they are conspicuously absent. There are additional concentrations of older people in the east and towards the south.
Four "exposed and vulnerable" layers at LSOA level. Named areas are detailed in text.
Conversely when looking at flats, there is a significant concentration in the city centre, a result of high land costs forcing the development of high rise flats. This property type is unappealing for the majority of elderly people, given the difficulties of access (e.g. stairs/lifts) and greater noise levels. Away from the centre, there are other LSOA's with high levels of flats, including small numbers in the north, and even less in the south. For example, clusters can be found in student areas, such as the high rise student housing located on Birmingham City University campus (Area Z, Figure ).
There is less of a visible range when looking at density (detailed in HH per km2). Again, the highest density LSOA's are located in the city centre, extending north westwards into areas renowned for having a high immigrant population. Conversely, density reduces heading south from the city. For example, Edgbaston (Area Y, Figure ) is an affluent area that also includes the University of Birmingham, Edgbaston golf course and other land uses not associated with households. The north east quarter of the city centre (Area N, Figure ) is also low density, and is an area traditionally associated with industry. However, the overall density levels across the city are generally similar, with local variations between LSOA's dependent on the presence of greenspace (which increase the size of the LSOA area but not numbers of HH).
Finally, significant concentrations in the spatial pattern of people with ill health exist. This is particularly evident across the city centre and in a belt north east of the city centre and towards the cities eastern edge. Pockets are also visible in the south, after noticeable lows in the affluent area of Edgbaston and the transient student population of Selly Oak (Area S, Figure ), who are unlikely to stay in the same place long enough for reliable health statistics to be compiled.
A Spearman's rank order correlation was carried out to determine the statistical relationships between each "exposed and vulnerable" group and the UHI at the LSOA level (n = 641). Table shows that the results generally agree with the visual interpretation and all relationships are statistically significant (p < 0.01) except density vs flats. There is a weak positive correlation between density, flats and illness with the UHI, showing that as the UHI increases, the number of "exposed and vulnerable" groups also increases. There is a stronger negative correlation between old people and the UHI that agrees with the visual interpretation already discussed.
Spearman's rank correlation coefficient matrix
When the above four vulnerable groups are combined and equally weighted (Figure ) it is clear to see that the very high risk areas are concentrated around the city centre. This is to be expected due to the individual distributions already discussed, and agrees with previous work in the USA which has found that vulnerability increased in warmer neighbourhoods [45
] and that these neighbourhoods had a tendency to be located within the inner city [71
]. Although equal weightings for all layers have been used in this study, it is recognised that features of urban form (e.g. density) can also act as predictors for the UHI. As a result, this can impact the output risk, and is an area that could be explored more in the future when considering different weightings for layers.
Combined (equal weighting) "exposed and vulnerable" layer at LSOA level.
The Final Risk Layer
Figure shows that the majority of the "very high" risk LSOA's are grouped together in the city centre. It is here where the highest temperatures are experienced as well as the highest number of ill people, number of flats and density. However, additional pockets of "very high" risk also exist and these require additional explanation. As already discussed, a high concentration of flats increases the density of a LSOA. Outside of the city centre, these flats are frequently high rise social housing that is often associated with increased illness in the poorer sections of communities. A typical "high risk" pocket has significant high rise social housing which increases the density, scores highly for flat and often for illness as well.
The lowest risk areas are found in the north west (Sutton Park area) and north east of the city. This is explained by the low and very low UHI risk coupled with very low "exposed and vulnerable
" populations. An anomaly of this area is that it actually has the highest concentration of elderly people, but they are less vulnerable to heat due to their distance from the city centre. Other very low risk areas are evident west of the city centre and scattered south of the city centre. In general these are heavily linked to greenspace; which has the dual effect of ameliorating the UHI and reducing the number of people living in an area. Indeed, a more explicit look at the distribution of greenspace within the conurbation could be useful (e.g. using surface cover analysis [92
] or energy exchange models [93
]), given the benefits of reducing the UHI [94
] and improving health inequalities [95
A strength of the methodology detailed in this paper is that once the risk areas have been identified, a subsequent detailed analysis down to HH level can be conducted. Such high resolution work within urban areas is a logical development of previous broader scale work, such as the province wide analysis carried out in Quebec, Canada [70
]. A GIS was used to identify 37,477 HH's (or ~8.76% of 427,914) that fall within the "very high" risk LSOA's (33 out of 641). These HH's can then be profiled using Mosaic type (Figure ), which illustrates the vast majority are either 47 (Deprived view) or 64 (Bright young things), accounting for ~7,000 HH each. This illustrates a clear divide within the "very high" risk area which is only able to be explored by having access to high resolution underlying datasets such as Mosaic. Type 47 are "poor people who live in high rise blocks of socially owned housing...many have disabilities...characterised by extreme poverty". Type 64 are "well educated young high flyers...live in smart inner city areas...mostly modern, purpose built or converted apartments". Despite living in broadly the same area, the populations are generally separated (Figure ) and are at polarised levels of heat risk. Type 64 typically live in new apartments located within the inner city. These dwellings may have good insulation, air conditioning or even passive cooling. This is a contrast to type 47, who live in older, social apartments located in less desirable areas surrounding the urban core. Unlike type 64, this group is unlikely to have the finances available to make themselves comfortable or safe.
Mosaic type within "very high" risk LSOA. Filterered to only include HH counts > 1000.
Mosaic type 64 and 47 spatial distribution within "very high" risk city centre area.