SEARCH

SEARCH BY CITATION

Urban landscapes typically experience higher air temperatures than the surrounding countryside due to absorption of solar heat by the fabric of buildings and trapping of outgoing longwave radiation within enclosed spaces. There is also less evapo-transpiration from paved areas compared with vegetated surfaces, leaving a greater fraction of solar energy for surface heating. Artificial heat sources from transport, industry, air conditioning and space heating can further contribute to local warming. Although such phenomena are well-understood, urban heat islands (UHIs) can be contentious because of their potential to corrupt estimates of global near-surface temperature trends (Parker, 2010). There are also concerns that UHIs may intensify under future climate conditions, thereby amplifying heat stress on urban residents and habitats (Wilby, 2008; McCarthy et al., 2010).

Thanks to the pioneering work of Luke Howard (1833), the UHI of central London has probably attracted more attention than that of any other city. The earliest digitized temperature series for the London area originate from Kew Observatory (1881) and St James's Park [SJP] (1907). The longest digitized records for nearby sites regarded as rural are for Rothamsted [ROTH] (1872) and Wisley [WIS] (1931). These data show that the UHI varies over a range of timescales: on average peaking at night, towards the end of the week, and in summer (Wilby, 2003a; 2008). The locus of the maximum temperature anomaly is known to be highly mobile as it drifts across central London in response to prevailing wind speed and direction (Graves et al., 2001; McGregor et al., 2006).

London's UHI has also varied over recent decades. Using temperature differences between SJP (urban) and WIS (rural) during day and night, Lee (1992) found that the nocturnal UHI intensified during the period 1962–1989. The same index was used by Wilby (2003b) to show that the intensity of London's nocturnal UHI in summer increased by 0.12 degC per decade over the period 1959–1998. However, the most recent assessment suggests that temperatures at SJP have not increased any more than those at ROTH or WIS since 1907 (Jones and Lister, 2009). This is a finding of some consequence for urban planners and building designers; it further opens the debate about which meteorological stations can be included in global temperature datasets.

The purpose of this article is to reconcile the apparent discrepancy between these studies by uncovering the causes of multidecadal variations in London's UHI. Accordingly, we investigate two groups of meteorological factors: first, potential influences of record length, homogeneity and quality on long-term trends, and secondly, possible changes in the frequency and properties of weather patterns favouring intense heat-island episodes. We pay particular attention to the summer nocturnal heat island because the climatic signature is strongest and has the greatest health implications at these times.

Data and methods

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References

We employ both daily and monthly mean maximum (Tx) and minimum (Tn) temperatures measured at SJP and WIS. As the name suggests, SJP is not located in a highly developed area of central London and is known to be cooler than other nearby sites such as the London Weather Centre (LWC) (now closed, and, as a rooftop site, less than ideal). Furthermore, the WIS site has a slightly higher elevation than SJP (38 metres and 5 metres, respectively), whilst Burt and Eden (2004) suggested that it may have become more sheltered in recent years. These factors suggest that the urban-rural temperature gradient between the two sites could be less than might otherwise be expected. Jones and Lister (2009) addressed this by also using temperature data from ROTH (elevation 128 metres). This series shows the same long-term trends as WIS, and hence the same difference with SJP, clearly showing that WIS has not become any more sheltered over time. For this study, we chose the SJP and WIS stations for the longevity of their records and continuity with our earlier work.

Following Lee (1992) and Wilby (2003b), UHI intensity was estimated for every day in the 51-year period 1959–2009 by subtracting daily Tx and Tn temperatures at WIS from the corresponding daily temperatures at SJP. These urban-rural differences were then aggregated by season and year to calculate simple indices of nocturnal (nUHI) and day-time (dUHI) heat island intensity. The same procedure was followed using monthly mean Tx and Tn for the period 1931–2006 as in Jones and Lister (2009). Comparison of the two sets of results (i.e. one based on daily series, the other on monthly records) helps identify potential discrepancies between data due to differing treatment of missing days. Linear regression analysis was applied to the nUHI and dUHI time series using different start and end dates to detect any significant trends in the UHI, and their dependency on analysis sub-period. In addition, annual and summer frequencies of intense heat-island episodes were extracted from the daily nUHI. For the present study, an intense heat-island event was defined as any night during which the urban-rural Tn difference exceeded 5 degC. Such events occur on average on fewer than five nights per summer.

Variations in the UHI were next analysed with respect to the prevailing synoptic weather conditions. The subjective Lamb (1972) weather type catalogue, for the British Isles, and the objective classification scheme of Jenkinson and Collison (1977) provide daily synoptic classes since 1861 and 1881 respectively. Each consists of seven pure weather types (anticyclonic [A], cyclonic [C], westerly [W], northerly [N], southerly [S], easterly [E], northwesterly [NW]), nineteen hybrid types (such as the anticyclonic southeasterly [ASE]), and one unclassified group [U]. The objective catalogue assigns each day to a Lamb weather type based on wind flow strength, direction and vorticity estimated from an array of sea-level pressure data points over the British Isles. Jones et al. (1993) compared the two approaches and showed that the major long-term differences relate to the number of W and S days. The objective scheme has considerably more S (and SW) days than does the subjective one. Lamb's (1972) classification of surface winds was more dependent on the steering of depressions and paid less attention to observed surface wind direction. The objective method is favoured because of consistency, as the Lamb (1972) scheme may have been affected by availability of upper air data from the 1940s onwards.

Daily values of nUHI were grouped by weather type and Analysis of Variance (ANOVA) statistics were calculated. These determine the extent to which a classification scheme is able to discriminate heat-island intensities under different weather patterns. Time series of summer and annual weather-type frequencies can be compared with time series of average and peak UHI episodes. The persistence of pure and hybrid weather types may also be estimated by examining long-term changes in the frequency of transitions from one state to another. This is of interest because changes in persistence of some weather types (e.g. successive days with anticyclonic conditions) could affect the likelihood of very intense heat-island episodes building over the course of several days (Meehl and Tebaldi, 2004).

Finally, a technique is applied that discriminates within weather-type and between weather-type variations in heat-island intensity (Comrie, 1992). This enables removal of the synoptic climate signal from UHI time series. Using historic frequencies of weather types, it is also possible to reconstruct the UHI for earlier periods. Analysis of these hindcasts shows the extent to which trends in nocturnal heat-island intensity since the late nineteenth century could have arisen from changes in atmospheric circulation alone.

Long-term trends

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References

The nocturnal heat island (since 1959) has intensified in all seasons except winter (December to February) but none of the trends are statistically significant at the 95% level (Figure 1). Although the nUHI in spring (March to May) and summer (June to August) is on average ∼0.3 degC greater than in the late 1950s, the change is much less than inter-annual variability and could be due to sampling variation only. Nonetheless, the modest increase in seasonal means translates into about seven more nights per year with intense (>5 degC) heat-island episodes – a trend that is barely significant (p∼0.10) (Figure 2). The summer heatwave of 1976 also stands out because of the large number of days with intense heat islands. In contrast, the dUHI has remained unchanged in spring and has weakened in the other seasons.

thumbnail image

Figure 1. Seasonal trends in London's daytime (solid circles) and nocturnal (open circles) heat-island intensity, 1959–2009.

Download figure to PowerPoint

thumbnail image

Figure 2. The frequency of nights with intense heat islands (urban-rural temperature difference > 5 degC) in summer (red line) and through the year as a whole (blue line).

Download figure to PowerPoint

These results concur with Jones and Lister (2009) that there has been no urban-related warming since 1907 in central London relative to surrounding rural stations. In fact, the declining dUHI in winter and autumn is due to greater warming at WIS than at SJP from the 1930s. Since the 1980s, warming at WIS has been greater than at SJP in both Tx and Tn for all seasons and times of day except winter nights. The WIS results are consistent with those from ROTH. These findings, therefore, appear to be at odds with Wilby (2003b) who reported the opposite: more warming at SJP than WIS in all seasons except winter. However, as Table 1 shows, the inconsistencies are mainly due to the choice of study period. When analyses of monthly and daily temperature differences are based on the same years of record, the trends are well within instrumental error. The remaining discrepancies are explained in terms of the handling of missing data: Wilby (2003b) did not adjust seasonal statistics for missing days, whereas Jones and Lister (2009) infill with the long-term monthly mean whenever there are fewer than 20 days with data in a month.

Table 1. Dependence of the summer nocturnal UHI trends (derived from SJP and WIS) on period of record and method of analysis (degC per decade). Figures in bold are statistically significant (p<0.05).
PeriodJones & ListerWilbyHindcast
  • a

    Daily data are available from 1959.

  • b

    The Jenkinson catalogue ends in 2006.

1881–2006+0.01
1931–2006+0.04+0.02
1949–2006+0.10+0.02
1951–1980+0.32+0.32a+0.05
1959–2007+0.06+0.05+0.02b
1961–2000+0.11+0.10+0.03
1981–2006–0.02–0.05+0.01

Both studies report statistically significant (p<0.05) intensification of summer nUHI during the period 1951–1980 (Table 1). As subsequent years have been added to the record, the strength of the trend has declined steadily since the late 1970s (Figure 3(a)). Had there been no data prior to the late 1960s, all trend analyses would have reported a weakening of the summer nUHI (Figure 3(b)). With hindsight, the data available at the time of the studies of Lee (1992) and Wilby (2003b) fell within the period of summer heat-island intensification. The following section considers the extent to which multidecadal changes in the synoptic climatology of the British Isles have also influenced variations in the intensity of London's nUHI.

thumbnail image

Figure 3. Influence of meteorological record end (a) and start date (b) on the trend in summer nocturnal heat-island intensity.

Download figure to PowerPoint

Influence of changing weather patterns

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References

The strength of London's summer heat island is positively correlated with atmospheric pressure, but negatively correlated with near-surface wind speed (from the west) and relative humidity (a proxy for cloud cover) (Wilby, 2003b; McGregor et al., 2006). In other words, more intense nocturnal heat islands tend to develop in summer under high-pressure, low wind speed, and clear sky conditions associated with anticyclonic weather patterns. These general observations were confirmed by stratifying daily nUHI values by weather type (Figure 4). Although there is variance of heat-island intensities within each Lamb weather type, it is apparent that the most marked episodes do indeed occur on anticyclonic days, whilst there are fewer of these under cy-clonic and/or westerly conditions. Air-flows from the E and SE (i.e. from continental Europe) yield significantly stronger nUHI in summer compared with the year as a whole. The ANOVA analysis confirms that the various pure and hybrid Lamb weather types of the objective scheme produce statistically distinct (p<0.001) distributions of nUHI. Overall, the most intense nUHI develop under A, ASE, AS, S and CSE weather types, and the weakest under CNE and CW types. Scatterplots of weather pattern frequency versus average nUHI further support this finding (Figure 5). Hence, years with a greater number of anticyclonic days (or fewer cyclonic or westerly days) are expected to have stronger nUHI.

thumbnail image

Figure 4. Summer (red bars) and annual (blue bars) nocturnal heat-island intensity under different objective Lamb weather types for the period 1959–2006. T-bars denote the 95% confidence range of the sample means.

Download figure to PowerPoint

thumbnail image

Figure 5. Relationship between summer nocturnal heat-island intensity and the frequencies of pure (upper panel) and basic (lower panel) weather types. FollowingLamb (1972)frequencies of basic weather types were weighted as: 1 for pure types (such as A); 0.5 for hybrids comprising two elements (such as AN) and 0.33 for hybrids comprising three elements (such as ASW).

Download figure to PowerPoint

The Lamb and Jenkinson schemes exhibit marked inter-annual and multidecadal variations in all three weather types (Figure 6). Both catalogues indicate an abrupt increase in the frequency of summer anticyclonic days from the mid-1960s and a gradual rise in the annual frequency of cyclonic days since the 1940s. There is less agreement about the long-term behaviour of the westerly weather pattern, but both schemes concur that the frequency of westerly days was at historically low levels between the late 1960s and early 1980s (Jones et al., 1993). All other factors remaining constant, the rapid increase then stabilization in anticyclonic conditions, accompanied by gradual strengthening of cyclonic and zonal airflows, would have favoured more intense nUHI up to the 1980s and a decline thereafter. This resembles the trends shown in Figures 1 to 3 and Table 1.

thumbnail image

Figure 6. Frequencies of the most common pure Lamb (red symbols) and Jenkinson (blue symbols) weather types during summer (left column) and annually (right column) for the period 1861–2006.

Download figure to PowerPoint

Comrie (1992) describes a simple technique for isolating the synoptic climate signal within time series of environmental data. First, the long-term mean of the variable of interest (in this case summer nUHI) is estimated for each of the different weather types (as shown in Figure 4). Seasonal frequencies of the weather types are then multiplied by the respective weather pattern's mean to give a weighted average. The resulting series shows the extent to which year-to-year variations in summer weather type frequencies force variations in the summer nUHI. In a second step, the procedure is reversed. This time the long-term average frequency of the different weather types is multiplied by the annual series of nUHI for each weather type. The weighted average now reveals the extent to which variability within each weather type accounts for trends in the nUHI.

Figure 7 shows that inter-annual variations in the summer nUHI are better explained by the time-varying properties of weather types (blue dotted line) than their frequency (blue solid line). The within-type signal accounts for 76% of the variance in summer nUHI compared with 48% explained by the frequency of each weather type. Figure 8 shows that there has been a long-term (but statistically insignificant) rise in the average nUHI within the anticyclonic weather pattern. Conversely, there has been a statistically significant (p<0.05) decline in dUHI such that, for anticyclonic days, average summer Tx has been greater at WIS than at SJP since 1980. There have also been some subtle changes in the day-to-day persistence of the most common weather types. During the period 1951–1980 there was a statistically significant (p<0.05) rise in the persistence of the pure A-type. This would have favoured the build-up of more intense heat-island episodes.

thumbnail image

Figure 7. Inter-annual summer nocturnal heat-island intensity showing observations (black line), variation due to air mass properties (blue dotted line) and weather pattern frequencies (solid blue line).

Download figure to PowerPoint

thumbnail image

Figure 8. Annual variations in summer daytime (solid circles) and nocturnal (open circles) heat-island intensity within the anticyclonic weather type, 1959–2006.

Download figure to PowerPoint

Unfortunately, it is not possible to factor all of these changes into hindcasts of the nUHI. This is because within weather-type statistics can only be computed for the period of digitized daily temperatures. However, it is feasible to hindcast the synoptic signal of the nUHI using the objective Lamb weather-type frequencies. The reconstructed nUHI is remarkably stable given the large multi-decadal variability in the A, C, and W weather types (Figure 9). Over the full 126-year period there has been a slight tendency towards more intense nUHI but the change is statistically insignificant except when the trend is measured from the 1930s. The period of maximum intensification of the nUHI (1951–1980) is found in the hindcast but the rate of change is much less than observed (Table 1). From this we infer that the rise of nUHI between 1951 and 1980 was due, in large part, to changes in weather patterns. Future studies of UHIs in other cities might similarly benefit from analysing potential influences from changing synoptic patterns. As we have demonstrated, the Jenkinson and Collison (1977) scheme provides a relatively straightforward method for classifying then normalizing such data.

thumbnail image

Figure 9. Summer nocturnal heat-island intensity reconstructed from seasonal variations in objective Lamb weather type frequencies 1881–2006. The correlation between overlapping observations and reconstructions is r = 0.69.

Download figure to PowerPoint

Concluding remarks

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References

Our review of the long-term behaviour of London's UHI provides a salutary reminder that the appearance and disappearance of trends in environmental data can depend very much on the segment of data analysed. Nonetheless, we can confirm – using both daily and monthly temperature records – that the summer nUHI did intensify between the late 1950s/early 1960s and the 1980s. This period coincided with an abrupt increase in the frequency of summer anticyclonic weather. There is also evidence of a slight rise in the annual number of intense heat-island events that can be linked to more persistent anticylonic weather systems at that time. A weak decline in summer nUHI since the 1980s coincides with a rise in the frequency of cyclonic weather. Since 1931, the summer nUHI has risen slightly, but not significantly. The overall annual mean nUHI does, however, show a weak but significant (p<0.05) rise when the monthly SJP record is compared to that of WIS.

Over the 50-year daily record, less than half of the variance in the summer-mean nUHI signal is explained by synoptic weather patterns. This could be due to a number of factors. The weather types describe conditions across the British Isles generally, rather than for southeast England specifically. The conditions experienced within a given weather class are known to vary from day to day. There have also been marked changes in regional air quality in the wake of the notorious winter ‘smogs’ of the 1950s and the summer stubble burning of the 1970s and 1980s. Other time-dependent factors (such as artificial heat sources, building albedo, thermal mass, sky-view factors, surface roughness, and vegetated area) may be locally important (McGregor et al., 2006). Furthermore, censuses show that the population of Greater London peaked in 1939 then fell until 1991 and has since risen again1.

Paradoxically, London's nocturnal heat island was actually strengthening during the decades of declining population. During that time, though, there was a substantial increase in the number of cars. However, the counter-trend is further evidence of a temporary synoptic signal superimposed on a heat island in central London that was largely established by earlier phases of urban development. Jones and Lister (2009) report that urban-related warming has shifted from central locations to the periphery of London. This is consistent with observed behaviour of the UHI in other global cities. For instance, long-term temperature data for Toronto show that the most recent increases in temperature are at urbanizing suburban stations (Mohsin and Gough, 2009). Recent warming in Japanese cities is largely attri-buted to regional temperature rises, but the urban signature amplifies where there is greatest population density (Fujibe, 2009). Clearly, any long-term changes in UHI behaviour should be interpreted on a case-by-case basis.

We conclude that decadal variability in the intensity of London's nUHI reflects background variations in the occurrence and persistence of weather types favouring day-time build-up and night-time release of heat by the urban landscape. However, these patterns are superimposed upon a long-term rise in regional air temperatures affecting both urban and rural sites. Changes in population density, associated waste heat from buildings, commerce and traffic further add to the surface energy budget. Urban redevelopment could also have affected the local thermal and radiative properties of the city. Predicting future changes in air temperatures requires a modelling framework that can embrace the complex interplay between all of these geographical and physical factors. However, any future era of more persistent anticyclonic weather would predispose London to more intense nocturnal heat islands.

Acknowledgements

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References

Daily temperature series for SJP and WIS were obtained from the Met Office MIDAS Land Surface Observation Stations Data held by the British Atmospheric Data Centre.

References

  1. Top of page
  2. Data and methods
  3. Long-term trends
  4. Influence of changing weather patterns
  5. Concluding remarks
  6. Acknowledgements
  7. References
  • Burt S, Eden P. 2004. The August 2003 heatwave in the United Kingdom: Part 2 – The hottest sites. Weather 59: 239246.
  • Comrie AC. 1992. A procedure for removing the synoptic climate signal from environmental data. Int. J. Climatol. 12: 177183.
  • Fujibe F. 2009. Detection of urban warming in recent temperature trends in Japan. Int. J. Climatol. 29: 18111822.
  • Graves HM, Watkins R, Westbury P, Littlefair PJ. 2001. Cooling Buildings in London. BR 431, RRCR Ltd., London.
  • Howard L. 1833. The Climate of London: Deduced from Meteorological Observations made in the Metropolis and at Various Places around it. Harvey and Darton: London.
  • Jenkinson AF, Collison FP. 1977. An initial climatology of gales over the North Sea. Synoptic Climatology Branch, Memorandum No. 62, Meteorological Office, Bracknell.
  • Jones PD, Lister DH. 2009. The urban heat island in central London and urban–related warming trends in Central London since 1900. Weather 64: 323327.
  • Jones PD, Hulme M, Briffa KR. 1993. A comparison of Lamb circulation types with an objective classification scheme. Int. J. Climatol. 13: 655663.
  • Lamb HH. 1972. British Isles weather types and a register of the daily sequence of circulation patterns 1861–1971. Meteorological Office Memoir 116, HMSO, London.
  • Lee DO. 1992. Urban warming&quest; An analysis of recent trends in London's urban heat island. Weather 47: 5056.
  • McCarthy MP, Best MJ, Betts RA. 2010. Climate change in cities due to global warming and urban effects. Geophys. Res. Lett. 37: L09705.
  • McGregor GR, Belcher S, Hacker J, Kovats S, Salmond J, Watkins RW, Grimmond S, Golden J, Wooster M. 2006. London's Urban Heat Island. A report to the Greater London Authority. Centre for Environmental Assessment, Management and Policy, King's College London, 111pp.
  • Meehl GA, Tebaldi C. 2004. More intense, more frequent, and longer lasting heatwaves in the 21st century. Science 305: 994997.
  • Mohsin T, Gough WA. 2009. Trend analysis of long-term temperature time series in the Greater Toronto Area (GTA). Theoretical and Applied Climatology 10.1007/s00704-009-0214-x
  • Parker DE. 2010. Urban heat island effects on estimates of observed climate change. WIRES Climate Change 1: 123133.
  • Wilby RL. 2003a. Weekly warming. Weather 58: 446447.
  • Wilby RL. 2003b. Past and projected trends in London's urban heat island. Weather 58: 251260.
  • Wilby RL. 2007. A review of climate change impacts on the built environment. Built Environment Journal 33: 3145.
  • Wilby RL. 2008. Constructing climate change scenarios of urban heat island intensity and air quality. Environ. Plan. B 35: 902919.