Earth & E-nvironment 2: 54-83
University of Leeds Press
The impact of urban areas on climate in the UK: a spatial and temporal analysis, with an emphasis on temperature and precipitation
effects
Karen Hughes
School of Earth & Environment, University of Leeds, Leeds, W. Yorkshire LS2 9JT; Tel: 0113 3436461
Abstract
Regional and local consequences of climate change may be compounded by urbanisation. 48% of the world’s population live in urban areas (UN, 2006) and significant increases in this proportion are projected. Research on inadvertent urban climate modification in the UK is somewhat limited. The aim of this study was to clarify if urbanisation effects were evident at four cities in the UK: Bristol, Cardiff, Newcastle-upon-Tyne and Norwich.
The focus was on temperature effects, in particular whether urban heat islands were evident, and precipitation effects over a 10-year period from 1994 to 2003. Analysis of urban-rural differences was conducted using climate data from the UK observational network and statistical analysis performed on results. Further analysis was conducted to determine if a relationship existed between city size and temperature and precipitation effects.
Urban heat islands were evident at Norwich, Newcastle and Cardiff, averaging 0.5°C, 0.9°C and 1.0°C respectively. All were statistically significant at the 5% confidence level. No discernable heat island was evident at Bristol, although lack of data availability may have distorted results. Higher rural precipitation was evident at Norwich and Cardiff, which was unexpected as previous research suggests increased rainfall over and downwind of an urban centre. It was unclear whether urbanisation contributed to precipitation patterns at these cities and urban-rural differences were not statistically significant. A relationship between city size and magnitude of temperature gradients was evident, although not with precipitation. Further research is needed to clarify precipitation effects, although the study demonstrates urban modification of climate is taking place in the UK.
ISSN 1744-2893 (Online)
© University of Leeds
1 Introduction
It is generally accepted that the global climate is changing as a result of anthropogenic activities (Intergovernmental Panel on Climate Change (IPCC), 2001). Global mean surface temperatures have increased by approximately 0.6°C over the 20th century and climate models predict further increases, as well as changes in other climate variables in the future (IPCC, 2001). Climate change is not just a global phenomena, it has important consequences both regionally and locally. Inadvertent climate modification is thought to be taking place on a local scale as a result of urbanisation. The nature of surface characteristics in an urban environment differ greatly from those in non-urban areas, and can lead to changes in the energy budget within the urban boundary layer (Oke, 1987). It is important to gain a better understanding of the nature and magnitude of urban climate modification, in order to assess the contribution of urban areas to climate change and also what impact climate change may have on urban areas.
Approximately 48% of the world’s population live in urban areas (2003 figures) (United Nations (UN), 2004). This figure is projected to increase at an annual rate of 1.8% until 2030 (UN, 2004). Based on these estimates, the number and size of urban areas are likely to increase substantially in the future. In a recent report from the Global Atmosphere Research Programme (Department for Environment, Food and Rural Affairs (DEFRA), 2004), it is suggested that such increases in urbanisation may in fact compound the effects of climate change locally. In the United Kingdom (UK), climate model scenarios predict increases in average temperatures, drier summers, greater winter precipitation and an increase in the number of heavy rainfall days (United Kingdom Climate Impacts Programme (UKCIP), 2002).
Currently, evidence for urban climate modification is highly variable, particularly with regard to precipitation effects. Research also appears to be somewhat limited in the UK, although an early study demonstrated higher urban-rural temperatures in London (Chandler, 1965) and further research on climate change impacts at this city have been conducted (UKCIP, 2002a).
A higher urban-rural temperature forms the basis of the urban heat island (UHI) effect, which is a common urban phenomenon (Arnfield, 2003). Although a large number of observational studies have been conducted in this area, the majority appear to be focused in the United States of America (USA), with little in the way of similar studies in the UK. This is particularly the case with evidence for precipitation modification in urban areas, and other international studies have yielded variable results.
Evidence is unclear as to whether size of the urban area plays an important role with regard to the intensity of inadvertent climate modification. This could have major implications for climate change in light of projected increases in urbanisation.
1.1 Aims and Objectives
The overall aim of this study is to assess whether there is evidence for inadvertent urban climate modification in the UK. Specifically, it focuses on urban-rural comparisons over a 10-year period of mean temperatures and precipitation at four UK cities of varying size. The study also investigates trends in wind speeds and frequency of precipitation, as well as consideration of minimum and maximum temperatures. The hypotheses behind the study are higher urban-rural mean temperatures (the UHI effect) and higher urban-rural mean precipitation, based on existing research. Data from the weather station observational network in the UK will form the basis of analysis, subject to availability.
As well as complimenting existing research, the results of the study may benefit local government with regard to planning and design of urban areas to minimise effects on climate. Results could also be used in further research to assess impacts of urbanisation at these four cities in relation to climate change.
The following outlines the specific objectives of the study:
• To determine if urban heat islands are present and the magnitude of such at each city
• To determine the nature of apparent UHIs over time
• To determine if higher urban-rural precipitation is evident at each city and the magnitude of differences
• To determine if a relationship exists between wind speed and UHI magnitude
• To determine whether a relationship exists between city size and the magnitude of urban-rural differences in temperature and precipitation, using population as an indicator
• To confirm whether any apparent urban-rural differences are statistically significant
• To determine any spatial trends in urban-rural climate in the UK.
2 Urban Climatology: An Overview of the Research 2.1 Temperature Effects
2.1.1 The Urban Heat Island
The UHI effect has been discussed in a number of studies and appears to be generally accepted as evidence for anthropogenic climate modification (Arnfield, 2003). The basis of the effect is the existence of a steep atmospheric temperature gradient between the urban centre and rural surroundings, characteristic of a ‘dome’ of warmer air above the urban area (Oke, 1987).
The urban-rural vertical temperature gradient can be as much as 4˚C km
-1(Oke, 1987), although research suggests it can vary throughout the day, reaching a maximum 2-3 hours after sunset (Landsberg, 1981). The key driver for this temperature variance is thought to be the difference in surface characteristics, with many buildings and few areas of vegetation and water bodies in the urban centre compared to the rural counterpart. Attempts have been made to quantify the urban energy budget to assess the effects of surface properties and relevant fluxes on air temperature. A proposed equation for the surface energy balance is below (1), where Q
*is the net radiation flux, Q
Fthe anthropogenic heat flux from combustion processes, Q
Hsensible heat, Q
Elatent heat, ∆Q
Snet heat storage and ∆Q
Athe net heat advection (Oke, 1987, Oke and Cleugh, 1987):
Q
*+ Q
F= Q
H+ Q
E+ ∆Q
S+ ∆Q
A(Oke, 1987, Oke and Cleugh, 1987) (1)
Sensible and latent heat fluxes vary at an urban region, in contrast to those at a nearby rural area. For example, greater rural coverage of vegetation will lead to higher evapotranspiration rates, thereby losing more energy via latent heat with the atmosphere and reducing air temperature (Cotton and Pielke, 1995). Conversely, a greater sensible heat flux is likely in the urban district due to heat storage capacity of buildings, releasing heat in the late afternoon/evening (Oke, 1987).
Urban radiation fluxes can also differ, for example air pollution in an industrialised city can interact with incoming and outgoing radiation, with a reduction of up to 20% of incoming shortwave radiation if heavy industry is present (Oke, 1987). This would suggest lower urban temperatures if less radiation reaches the surface, however albedo values for urban surfaces tend to be much lower than typical rural ones (Oke, 1987), and heat can become trapped between structures and buildings. This implies some offsetting of any cooling brought about by the presence of air pollution particles, with a greater amount of incoming radiation being absorbed by surfaces. Urban air pollution can also have consequences on cloud condensation nuclei (CCN) and precipitation formation, which is discussed later.
2.1.2 Atmospheric Mechanisms of the UHI
The vertical atmosphere above the ground can be separated into two components, the urban canopy layer (UCL) beneath the average roof level, and urban boundary layer (UBL) above the UCL and extending up to the Earth’s boundary layer (Oke, 1976). Effects on atmospheric state within the UCL are entrained upwards into the UBL where larger scale processes dominate (Arnfield, 2003). The greater flux of sensible heat and lower evaporation rates over the urban area results in drier air and a deepening of the UBL (Cotton and Pielke, 1995), forming the dome of air associated with the UHI.
Observations and measurements indicate that the UHI is generally stronger nocturnally. This may be
due to heat releases from buildings and conduction from surfaces at night, reducing the amount of cooling and resulting in the air above becoming unstable (Cotton and Pielke, 1995). As a consequence, most urban areas have weak nocturnal inversion layers above the UBL in contrast to rural areas with relatively strong inversions, due to increased stability here (Cotton and Pielke, 1995). This has implications on convective processes and cloud development over the urban region (section 2.2).
Processes within the UCL depend on the urban structure and geometry, for example some cities may have many high-rise buildings in contrast to those of a similar size with low-level structures. Densely built urban areas tend to reveal a stronger urban-rural temperature difference, with gradients of up to 12˚C from observational measurements (Roth et al., 1989). The presence of street canyons, particularly within a heavily built-up area, can modify wind speeds and advection in and around the city (section 2.3), which will also have an impact on the magnitude of apparent temperature gradients. The UHI intensity is generally higher in calmer conditions when wind speeds are low and when anticyclonic weather dominates (Landsberg, 1981).
2.1.3 Synopsis of the Research
A number of studies have been conducted into the UHI effect around the world, most notably in the USA, but also in France, Switzerland, New Zealand, China and to a certain degree in the UK. Many of the early studies used climate data from observational networks and in-situ instrumentation to analyse urban-rural temperature differences. This was the approach taken by Oke (1973) in a study of city size effects on UHI intensities in the St Lawrence region of Quebec, Canada. A strong correlation was found between the size of an urban area, using population as an indicator, and the maximum urban- rural temperature difference. A regression formula (2,3) was derived for this relationship, where ∆T
u- r(max)is the maximum urban-rural temperature difference and P is the population.
∆T
u-r(max)= 2.96 log P – 6.41 (USA cities) (Oke, 1973) (2)
∆T
u-r(max)= 2.01 log P – 4.06 (European cities) (Oke, 1973) (3)
Some further observations of temperature gradients appear to correlate to these relationships (Landsberg, 1981), however care must be taken when using population as an indicator for city size.
There may be substantial variation in topography and background climate at different urban regions with similar population structures. Oke (1973) recognises such limitations in his work and proposes results as indicative rather than definitive. An alternative way of examining the relationship between UHI and city size is to conduct a temporal analysis of temperature changes at an urban area experiencing population growth. A similar study for urban areas in China revealed an increase in mean surface temperature of 0.05°C per decade from 1979 to 1998 for regions of rapid population growth over the period (Zhou et al., 2004). Although this approach demonstrates local scale effects of urbanisation and relationship with population growth, it does not reveal if similar trends are apparent at nearby rural locations or consider background climate variations.
Although limited, studies in the UK also support the existence of UHIs. In particular, analysis of minima and maxima temperatures between inner London and Wisley, a rural area on the outskirts, revealed a substantial difference in minima values between the sites (Chandler, 1965). Over the spring period, minimum temperature within the city averaged 11°C, compared to only 5°C on the outskirts.
Obviously London is a large city and densely populated, smaller urban areas in the UK may not exhibit such a significant temperature gradient.
The presence of a stronger UHI during the nocturnal period was confirmed by Tapper (1990) in a study at Christchurch, New Zealand. The average nocturnal urban-rural temperature difference was 2.5°C and it was also found that nocturnal temperature inversions over the urban area were weaker than over the rural location, which agrees with UBL theory. Research also suggests the UHI is more prominent during the summer season, with average urban-rural July differences of 1.3°C compared to 0.2°C in January for four cities in the USA (Geerts, 2002).
A somewhat different approach is to examine changes in temperature over time to identify trends
between urban and rural locations. Increasing the timescale of analysis may smooth out any sharp
peaks or troughs in data and also highlight seasonal trends. Comparative analysis of decadal data was conducted for 775 urban and 167 rural stations across the USA (Kalnay and Cai, 2003). Results revealed an average surface warming of 0.31°C for urban stations compared to 0.13°C for rural stations for the period. The larger increase in mean surface temperature for urban locations appears to support the UHI effect, with penetration of higher surface temperatures into the UBL, however the significant ratio difference between urban and rural locations used may introduce bias into the results.
The study does, however, highlight temporal trends and demonstrates the effect of urban areas on the magnitude of temperature changes.
One of the large-scale projects undertaken in urban climatology was METROMEX (Metropolitan Meteorological Experiment), which covered the city of St Louis in the USA (Changnon, 1981). Data was gathered over a 6-year period and climate variables analysed to investigate inadvertent climate modification, with a focus on precipitation. Results confirmed the existence of an UHI, with an average of 1°C urban-rural difference, and also a ‘doming’ effect over the urban centre caused by deepening of the UBL and mixing heights (Auer, 1981). A more recent large-scale project involving four cities in the USA used remote sensing data as a method to analyse and model UHI structures (Quattrochi et al., 2000). This technique was also employed in a survey of heat island temperatures over Vancouver, Canada and Los Angeles, USA (Roth et al., 1989), however it was found that, whilst it can be a useful tool to aid analysis, it provides information only on surface and ground characteristics, rather than the vertical temperature profile. Such a tool could be used in tandem with observational and instrumental data from ground-based stations in projects like METROMEX.
Principal research confirms the presence of urban-rural temperature gradients and, although some cities show fairly minor differences of 1°C or so, the existence of an UHI is still noticeable. A relationship with population size has been shown (Oke, 1973), however other factors need to be taken into account when comparing UHI properties between cities, for example urban geometry and local topography. Arnfield (2003) suggests UHIs may vary between urban regions displaying variation in local conditions. Local background climate will also play a role in the magnitude of any temperature gradients.
2.2 Precipitation Effects
2.2.1 Overview and Principal Mechanisms
The effects of urbanisation on precipitation processes have received little attention in comparison to temperature studies (Lowry, 1998). The processes that govern precipitation formation are much more complex and research has proved difficult due to its intermittent nature in space and time (Lowry, 1998). The state of current understanding with regard to temperature effects seems to be relatively well-defined, with generally higher temperatures observed over urban locations, however precipitation effects are not so clear. Much of the research to date suggests urban precipitation enhancement in comparison to rural locations (Braham, 1981, Landsberg, 1981, Oke, 1987). Little consideration has been given to the possibility of precipitation suppression. A study using satellite data confirmed substantial precipitation increases downwind of urban areas, with only slight increases over the centre itself (Shepherd et al, 2002). Lowry (1998) notes, with regard to urban-rural comparisons mentioned in the literature:
“…in these considerations of two-point differences, cities ‘enhance’ the precipitation-producing process, with scarcely a mention that, under some circumstances, even suppression may be the outcome.”
(Lowry, 1998, p 497)
Rain formation requires input of water vapour and cloud condensation nuclei (CCN) on which water
droplets can form and grow (Oke, 1987). Air pollution from industry and traffic in urban areas are
associated with an increase in CCN, leading to a greater number of smaller cloud droplets (Braham,
1974, Landsberg, 1981). The effects of anthropogenic aerosols on cloud physics, precipitation and
ability to act as CCN depends on their size range and type, for example their hygroscopic properties
(IPCC, 2001a). As a consequence, higher pollution levels in an urban region could act to either
enhance or suppress rainfall, depending on specific local conditions and pollution particle characteristics. Landsberg (1981), however, suggests that aerosol effects are insignificant in comparison to meteorological processes impacting on precipitation development at an urban region.
The study conducted by Braham (1974) at St Louis showed a relationship between CCN and cloud droplet characteristics. Results revealed precipitation formation was three times more frequent over the urban area and downwind compared to rural areas nearby (Braham, 1974). This demonstrates that aerosol emissions are noteworthy factors in the production of rain, and could play a contributory role in urban climate modification.
One of the major factors thought to lead to precipitation changes is the UHI effect (Braham et al., 1981, Landsberg, 1981, Cotton & Pielke, 1995). Higher air temperatures over the urban centre result in an increase in thermal convection here, with adiabatic motion of air parcels upward (Oke, 1987).
This increased convection and the fact that the air is drier over the urban centre can also lead to low level convergence, as cooler, moist air from the surrounding rural area is advected toward the city, resulting in conditions favourable for the formation of deep cumulus clouds (Cotton and Pielke, 1995).
Local topography and elevation will also impact upon cloud formation and rainfall amounts, for example if any mountainous regions or large water bodies, such as lakes, are located nearby (Landsberg, 1981). The geometry of urban centres is another major factor in modification of local climate (Landsberg, 1981, Cotton and Pielke, 1995). The shape and size of building structures can impact substantially on surface roughness, as they act as obstacles to the flow of air within the UCL and entrainment into the UBL above. Greater surface roughness can lead to an increase in mechanical turbulence (Huff and Changnon, 1973), with further impact on convection cells set up by the presence of an UHI.
Cotton and Pielke (1995) propose that this increase in surface roughness will slow airflow within the urban centre, causing uplift and convergence downwind. This implies that the formation of precipitation clouds may actually occur downwind, possibly in conjunction with convective cloud formation over the urban centre itself. Oke (1987) proposes that, even if increased convection leads to deep convective cumulus formation over the urban centre, by the time droplets form and grow to a sufficient size to fall as rain, advection will have carried these clouds downwind, resulting in a precipitation maximum of 5-30% occurring here. This would suggest the possibility of higher precipitation rates over a downwind region, perhaps resulting in lower precipitation amounts at the urban centre itself.
2.2.2 Synopsis of the Research
Earlier research was hindered by insufficient precipitation data from observational networks.
Changnon (1969) cited this, along with problems obtaining long-term data for rural stations, as reasons for limited study within the USA at the time. Lowry (1998) states that the complexity and high costs associated with such projects may have contributed to the lack of research in this field up to the late 1990’s. Despite these limitations, a number of studies have revealed interesting results. Project METROMEX was a significant large-scale study focussing on precipitation effects at St Louis, USA (Changnon, 1981). The study involved the use of a wide variety of instrumentation, including radiosondes, rain gauges and radar systems across a number of sites. The aim was to assess the frequency, intensity, amount and duration of precipitation for an urban climate, along with the mechanisms responsible for precipitation modification (Changnon, 1981).
Results revealed reduced precipitation during the morning and early afternoon, but a substantial
increase of up to 25% in the late afternoon and early evening, both over St Louis and downwind of the
city (Braham et al., 1981). Interestingly, the UHI gradient during this period was 0.5°C-1°C and it was
the vertical instability of air over the urban centre, inducing thermal convection, proposed as a causal
factor for the increased rainfall (Braham et al., 1981). Convective mechanisms appear to play a major
role in urban rainfall modification, producing significant urban-rural variances. One problem with this
idea is the presence of drier air beneath the UBL and lack of moisture availability for condensation
processes to take place. Moisture could be transported in from the rural surroundings by advection of
cool, moist air from low levels within the boundary layer. Although project METROMEX confirms
precipitation enhancement over and downwind of a city, only summer months were included in the study, therefore any seasonal trends cannot be observed.
METROMEX results seem to confirm some of the earlier studies carried out at cities across the USA.
Enhanced precipitation of 5-8% was found at four cities, with maximum rainfall occurring both over the urban centre and downwind (Changnon, 1969). Rainfall increases of 7% and 16% were also observed downwind of the urban centre for Washington DC and New York, respectively, although an element of these increases may have been attributable to topographic effects (Changnon, 1969). It is also important to consider background climate; rainfall rates could vary naturally with the passage of large-scale weather systems across a city, therefore urban-rural differences may not be entirely due to urbanisation. Major precipitation changes have been observed at rural locations both upwind and downwind of an urban centre (Diem and Mote, 2005). Examination of synoptic charts would aid in the identification of influential weather systems, although completely isolating an urban ‘signal’ may be problematic (Lowry, 1998).
Considerable interest has surrounded results obtained from precipitation studies at the city of Chicago, USA, termed the ‘La Porte anomaly’. Average precipitation increases of almost 50% were observed at La Porte, a region 40-55km downwind of Chicago, during the winter months over the period 1949- 1968 (Huff and Changnon, 1973). Much debate ensued as to whether urban effects could be responsible for such a substantial variance. Changnon (1969) proposed CCN particles emitted from an industrial complex upwind of La Porte and effects from nearby Lake Michigan as causal factors for the increase, in conjunction with a degree of urban effects. Analysis of cloud physics confirms changes in cloud patterns as a result of increased urban CCN at St Louis (Braham, 1974). A more recent study at this city suggests that urban effects are prominent when convective processes dominate (Changnon et al., 1991). It is a possibility that precipitation formation under convective conditions is enhanced by the presence of CCN from air pollution, although the link is not clear.
The predominant mechanism behind increased precipitation indicated in most studies is thermal convection, initiated by the existence of an UHI. This has recently been confirmed by numerical modelling (Thielen et al., 2000), taking into account the various flux parameters within an urban region.
Model outputs confirm enhancement of precipitation over urban areas and at a distance downwind, supporting results from the earlier METROMEX project. The time taken for precipitation formation to occur is reduced as the intensity of the UHI increases, and stronger updrafts are needed to initiate rain formation when convective processes are weaker (Baik et al., 2001).
Research indicates that urban areas do inadvertently modify precipitation characteristics, with evidence of significant urban-rural differences. The precise mechanisms behind this seem unclear and local conditions need to be taken into account, for example extent of air pollution present, prevailing weather conditions, such as wind direction, and local topography. Observations at La Porte illustrate this, with the presence of a large water body that may have influenced local rainfall patterns, with an increased availability of moisture for rain formation. Huff and Changnon (1973) suggest urban modification of rainfall may dominate when rain clouds are already present, as opposed to actually influencing precipitation formation.
2.3 Wind and Turbulence in the UBL
Airflow processes have received an increasing amount of attention in more recent urban climatology studies. Surface roughness levels within an urban region are higher due to the presence of buildings and tall structures, which modify the flow of air within the UBL (Arnfield, 2005). Modification of airflow can affect the intensity of the UHI and also impact upon rain formation.
Much of the research suggests urban-rural temperature differences are reduced as wind speed increases
(Oke, 1973, 1987, Landsberg, 1981). Wind speeds of 5ms
-1can reduce the maximum urban-rural
temperature difference by up to 8°C (Oke, 1973), depending on the maximum intensity at low or zero
wind speed. Turbulent eddies tend to be smaller in an urban region due to the roughness scales,
although it depends also on local topography (Landsberg, 1981). The overall increased turbulence
generated by high roughness characteristics may enhance sensible heat transport into the atmosphere above via mixing, thus influencing UHI development. Analysis of wind speeds and turbulence in the city of Basel, Switzerland (BUBBLE – Basel UrBan Boundary Layer Experiment) revealed that convective velocities were higher than the average wind speed under convective conditions for wind speeds up to about 2 ms
-1(Rotach et al., 2005), suggesting heat transport via convection dominates in calmer conditions and UHI intensity is greater.
Air stagnation can occur in certain parts of the urban region (Oke, 1987), where structures act as obstacles to airflow and frictional drag occurs. This can enhance the doming effect of the UBL over the urban centre as air converges and creates uplift (Oke, 1987). It is proposed that this doming of air can have effect further downwind, resulting in a ‘plume’ of rising air above the cooler rural surface air, which can be maintained for a significant distance away from the centre (Oke, 1987). This was confirmed in the study at Christchurch, New Zealand, where a plume of warmer air was observed at an altitude above 200m over the rural location (Tapper, 1990). This suggests that the UHI can actually be advected downwind of the urban centre at certain heights (Roth et al., 1989), creating a low level rural inversion.
Modification of airflow within the urban area is thought to contribute to changes in precipitation rates.
As well as creating a deeper UBL, reduced airflow within the urban canyon may lead to enhancement of precipitation downwind as a result of converging air and uplift (Thielen et al., 2000), possibly linked to advection of the UHI plume and enhanced convection. Thielen et al. (2000) propose that higher roughness heights within the urban canyon correlate to higher rainfall intensities downwind, with a maximum of 28.8mm of rain occurring here in model simulations.
3 Methodology 3.1 Overview
Secondary data formed the basis of the research, from the UK Met Office’s observational network.
This was accessed via the National Climatic Data Center (NCDC) website, www.ncdc.noaa.gov (NCDC, 2006) under agreement with the World Meteorological Organisation (WMO). Downloaded data was converted to metric units using formulas from an accompanying conversion file (table 3.1).
Table 3.1 Metric conversion formulas (adapted from NCDC, 2006) mperature: from °F to
°C d Speed: from knots t
ms
-1ipitation: from inches mm
ract 32 then multiply b
5/9 ultiply by 0.514800514 Multiply by 25.4
The locations included in the study were Norwich, Newcastle-upon-Tyne, Bristol and Cardiff. The latter three are in coastal regions, therefore not ideal choices due to the possibility of sea effects on local climate, however topographical and orography effects were minimal and sufficient data was available for urban-rural comparisons.
3.2 Study Locations
Wherever possible, rural stations were chosen with similar elevations to the urban station to maintain comparability, as variables such as temperature and wind speed vary with altitude. There was only one rural station at Bristol with suitable data for the study period, hence the elevation difference here.
Table 3.2 summarises location details for each station and maps of each city are shown further below
(figures 3.1 to 3.4).
Table 3.2 Station and location information (Source: NCDC, 2006) (urban-rural distances and station type based on maps, figures 3.1 to 3.4) Station
Number City/Station Name Station
Type Location (Lat &
Long in Degrees) Elevation
(m) Approx. Distance Between Urban and Rural
Station (km) 34920 Norwich
(Weather Centre) Urban +52.63
+1.31 18m 13km
34950 Norwich
(Coltishall) Rural +52.76
+1.35 17m 13km
37260 Bristol
(Weather Centre) Urban +51.46
-2.60 11m 6km
36283 Bristol
(Filton) Rural +51.51
-2.58 71m 6km
37170 Cardiff
(Weather Centre) Urban +51.48
-3.18 52m 14km
37150 Cardiff
(Airport) Rural +51.40
-3.35 67m 14km
32460 Newcastle
(Weather Centre) Urban +54.98
-1.60 52m 6km
32433 Newcastle Rural +55.03
-1.70 83m 6km
Key to Figures 3.1 to 3.4:
Red star: urban station Red circle: rural station
Figure 3.1 Cardiff station locations (Source: Microsoft Encarta Interactive World Atlas, 2000)
Figure 3.2 Bristol station locations (Source: Microsoft Encarta Interactive World Atlas, 2000)
Figure 3.3 Norwich station locations (Source: Microsoft Encarta Interactive World Atlas, 2000)
Figure 3.4 Newcastle-upon-Tyne station locations (Source: Microsoft Encarta Interactive World Atlas, 2000)
3.2.1 Background Climate
Cardiff is situated on the south east coast of Wales near the Bristol Channel (figure 3.5). Cardiff’s general climate is similar to that of the Midlands and classed as continental lowland (Wheeler and Mayes, 1997). As a coastal city, its weather may be influenced by proximity to the Bristol Channel, although it is somewhat sheltered from prevailing westerly winds over the Irish Sea.
Bristol is located in the southwest of England, slightly inland from the Bristol Channel (figure 3.5).
Prevailing air motion from the west/south west can bring strong winds to the city. Convectional precipitation can be dominant due to warmer sea surface temperatures, although rain can peter out further inland (Wheeler and Mayes, 1997).
Norwich is situated in eastern England, inland from the east coast (figure 3.5). The region is quite flat therefore topography has relatively little influence on climate. Due to passage over land, prevailing air motion from the south/south west changes characteristics by the time it reaches the region and becomes more ‘continental’ (Wheeler and Mayes, 1997).
Newcastle-upon-Tyne is located in the north east of England (figure 3.5). As it is a coastal city, the
North Sea may play a role in influencing local climate. The city may experience lower summer
temperatures and milder winters due to its proximity to a major river (Met Office, 2006), the River
Tyne. Frontal depressions from the west may become modified with passage over land, particularly
over the Pennine mountain range, therefore precipitation amounts may be reduced over the city.
Figure 3.5 Map of UK with city locations highlighted in yellow (Source: Microsoft Encarta Interactive World Atlas, 2000)
3.3 Research Design
Daily climate data for average temperature, precipitation, wind speed, and to some degree minimum and maximum temperatures, were analysed over a 10-year period from 1994 to 2003, inclusive, for urban and rural stations at each city. Oke (1973) used a similar two-point station method to analyse the relationship between city size and UHI intensity for ten urban areas, with populations ranging from 1,100 to 2 million. This method was also employed in a study of temperature trends at a region of substantial urbanisation in China (Zhou et al., 2004). A 10-year timescale was considered to be an appropriate minimum to provide adequate analysis of trends and to smooth out large fluctuations.
Some studies have covered much longer timescales (Zhou et al., 2004, Diem and Mote, 2005), although shorter projects (Huff and Changnon, 1973, Kim and Baik, 2005) have revealed notable results.
Average urban-rural temperature trends would indicate the presence of an UHI at the city, and also its magnitude in relation to the size of the city (as in Oke, 1973). Changes in minimum and maximum temperature may provide a further insight into the nature of any apparent UHI. Chandler (1965) found that changes in London’s urban-rural minimum temperatures were more significant than changes in maximum temperatures between 1921 and 1960, although proposed this was partly due to unsettled weather in some years. Trends in minimum temperature may reflect on the strength of the UHI during nocturnal hours, when it is believed to be more noticeable and at its strongest.
A significant proportion of precipitation data was missing, whereby either insufficient amounts were
recorded at the station to provide a valid value or precipitation may have occurred but was not
reported (NCDC, 2006). In order to correct for this, only days where rainfall was reported at both the
urban and rural station were included in the analysis. A substantial amount of data was also missing
for Bristol stations between 2002 and 2003, therefore these years were excluded from analysis. The
wind speed parameter was included within the study as a basic assessment of the airflow and turbulent
nature within each city (as in Rotach et al., 2005 and Grimmond et al., 2004), and also to further analyse
the relationship between wind speed and UHI intensity (Lee, 1984, Landsberg, 1981, Oke, 1987).
The following details the analysis procedures undertaken:
1. Comparison of Urban-Rural Trends at each City: Time series graphs were produced on a monthly and annual basis over the 10-year period for mean temperature, precipitation and wind speed. The graphs were used to analyse temporal and spatial trends of these variables at each city.
2. Comparison of Urban-Rural Differences between Cities: Time series graphs were plotted for urban-rural differences in average annual temperatures and wind speeds for the four cities, as well as average annual and monthly precipitation for Cardiff and Norwich. This enabled a comparison to be made of any apparent urban-rural differences between cities.
3. Precipitation Frequency: Frequency charts were plotted for light and moderate to heavy rain days at Cardiff and Norwich for the period.
4. City Size: Population figures from the 2001 national census (National Statistics, 2006) were used as indicators for city size. Scatter plots were constructed for overall urban-rural mean temperature and mean precipitation for the 10-year period. This enabled a comparative analysis of city size with UHI magnitude and precipitation differences. Regression and correlation analysis were also performed for a relationship between city size and overall urban-rural mean temperature. Population figures for each city are shown in table 3.3.
5. Statistical Analysis: Paired sample t-tests assuming unequal variances were used to test for significant differences at the 5% confidence interval between average annual urban and rural temperatures for individual years. The same method was used to test for significant differences between overall (all years) average temperatures. Results would accept or reject the hypothesis that there were significant differences between the urban and rural mean. This process was repeated for average precipitation.
Table 3.3 Population statistics (Source: National Statistics, 2006) (approximate area inferred from population number and density)
City Population
(Number) Approximate
Area (km
2) Population Density (per km
2)
Norwich 121,550 39 3,117
Bristol 380,615 109 3,482
Cardiff 305,353 137 2,222
wcastle-upon-Tyn 259,536 113 2,294
3.4 Errors and Limitations
Systematic errors should have been accounted for by proper calibration of station instrumentation as per Met Office and WMO standards. Random errors and variability around the mean were accounted for by calculating standard deviations for means and also for urban-rural differences. The actual precision error of instrumentation will be very small, 0.1°C for temperature, 0.1ms
-1for wind speed and 0.1mm for precipitation (NCDC, 2006). The following formula was used to calculate the error difference between urban and rural standard deviations, where σ
u-ris the urban-rural standard deviation, σ
uis the urban standard deviation and σ
ris the rural standard deviation.
σ
u-r= σ
u2+ σ
r2(Adapted from Lindberg, 2000) (4)
There is a possibility of human error involved in reorganising and consolidation of data, however
regular crosschecks back to the original data sets were conducted to minimise this.
Missing data limited the extent of analysis on precipitation effects, particularly as only two cities had sufficient urban and rural data for the time period. Also, stations were classified as urban and rural based on their location in relation to the urban border (figures 3.1 to 3.4), which may lead to inhomogeneous results (Zhou et al., 2004) with regard to urban-rural differences, as these classifications may not be correct.
4 Results and Discussion 4.1 Temperature
4.1.1 Cardiff
Figure 4.1 shows the average temperature at the urban and rural station over the period. The error bars represent the standard deviation or variation of daily values around the annual mean and are colour coded for each station. There is a consistent pattern of higher urban average temperatures, in the range of 0.7°C to 1.2°C. This indicates the presence of an UHI at Cardiff, particularly as the magnitude remains at a similar level over the period. The dip in temperatures in 1996 is consistent across all cities, signifying a generally cold year for the UK. Mean temperature observations at Maidenhead, Berkshire during 1996 were the lowest recorded here since 1987 (Brugge, 1997), confirming temperatures were below average for that year.
4 6 8 10 12 14 16 18
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Year Average Annual Temperature (Degrees Celsius)
Cardiff Rural Cardiff Urban
Figure 4.1 Average annual urban and rural temperature for Cardiff over 10-year period
The average monthly trend (figure 4.2) reveals a similar pattern of consistent higher urban temperatures. The error associated with these values is much less than with the annual averages, due to smaller seasonal variations in daily temperatures. The largest difference occurs in the summer, with an average urban-rural variation of 1.2°C, and the lowest occurs in the winter, averaging 0.7°C. This could be related to generally lower wind speeds during the summer season, effectively reducing turbulent mixing of air within the UCL. This correlates to research, which suggests the UHI is more prominent when wind speeds are low (Oke, 1973, 1987, Landsberg, 1981). Mean winter wind speeds at Cardiff were approximately 33% higher than during the summer, averaged over the 10-year period.
Mean rural wind speeds were generally higher than at the urban station for the entire study period, for
example a 10-year winter average of 5.6ms
-1compared to an urban equivalent of 4.9ms
-1. Modification
of airflow within the urban canyon can result in air stagnation and reduced flow (Oke, 1987) as a result
of urban structures. This is supported by wind speed variations at Cardiff.
0 5 10 15 20 25
Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec
Month
Average Monthly Temperature (Degrees Celsius)
Cardiff Rural Cardiff Urban
Figure 4.2 Average monthly urban and rural temperature for Cardiff over 10-Year period
4.1.2 Norwich
The trend at Norwich (figure 4.3) reveals a similar pattern of higher urban temperatures over the period. Again, there is consistency in the annual pattern, indicating the presence of an UHI. The magnitude of this ranges from 0.3°C to 0.8°C, with a larger variation during the latter two years. This later trend of larger urban-rural differences could be significant, however data for 2004 onwards would be required to analyse this further. A higher UHI strength could be associated with an increase in urbanisation, although population has only risen by less than 2% between 2001 and mid 2003 (National Statistics, 2006) at Norwich. The larger variation in 2002 and 2003 was due to an increase in the annual average temperature at the urban station, suggesting possible changes occurring within the city itself. A decrease in the ratio of vegetation to urban materials could lead to an increase in UHI strength (Best and Clark, 2002 cited in Collier, 2006).
3 5 7 9 11 13 15 17
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Year
Average Annual Temperature (Degrees Celsius)
Norwich Urban Norwich Rural
Figure 4.3 Average annual urban and rural temperature for Norwich over 10-year period.
The average monthly urban-rural difference (figure 4.4) for Norwich shows a similar seasonal pattern to that at Cardiff, with a reduction in the apparent UHI strength during the winter months. July and August exhibit an average difference of 0.7°C, compared to 0.3°C averaged for December, January and February. Mean wind speeds were approximately 41% higher during the winter period than in July and August. Again, rural wind speeds are generally higher than urban, although the difference is more marked in the winter season.
0 5 10 15 20 25
Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec
Month
Average Monthly Temperature (Degrees Celsius)
Norwich Urban Norwich Rural
Figure 4.4 Average monthly urban and rural temperature for Norwich over 10-Year period.
4.1.3 Newcastle-upon-Tyne
Average annual temperatures (figure 4.5) confirm previous results of higher urban values, consistent over the period. There is a greater urban-rural variation between 1994 and 1996; an average magnitude of 1.4°C during this period. From 1997 to 2003, this level decreases to 0.7°C. Mean temperatures at both stations generally increase from 1997 onwards, suggesting this anomaly may be related to changes in background climate. There were a number of extremely cold days during 1994 and 1995 where urban temperatures were much higher than at the rural station. An example was 27
thDecember 1995 when the urban average was -3.7°C compared to a rural value of -7.5°C. Large daily urban-rural variances such as these may have distorted the annual average for 1994 to 1996. It is not possible to determine whether these anomalies are associated with urbanisation effects or other factors, such as natural climatic variations.
The average monthly pattern (figure 4.6) is somewhat dissimilar to that for Cardiff and Norwich, with higher temperature differences occurring in the winter. November to January exhibits a mean urban- rural difference of 1.1°C, compared to 0.8°C between April and August. Large daily variations may have contributed to a distortion of monthly averages, which could explain this anomaly. The scale of this seasonal deviation is fairly small, only 0.3°C, and could be indicative of natural climate influences.
The River Tyne and proximity to the coast could be important factors in modifying seasonal
temperatures (Met Office, 2006), particularly at the urban station. In order to investigate this further, it
would be necessary to analyse temperature trends prior to 1994, possibly over a much longer timescale.
1 3 5 7 9 11 13 15 17
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Year
Average Annual Temperature (Degrees Celsius)
Newcastle Rural Newcastle Urban
Figure 4.5 Average annual urban and rural temperature for Newcastle-upon-Tyne over 10-year period.
0 2 4 6 8 10 12 14 16 18 20
Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec
Month
Average Monthly Temperature (Degrees Celsius)
Newcastle Rural Newcastle Urban
Figure 4.6 Average monthly urban and rural temperature for Newcastle-upon-Tyne over 10-Year period.
4.1.4 Bristol
Average annual trends at Bristol (figure 4.7) do not conform to patterns observed at the other cities, which is unexpected. Higher mean rural temperatures are observed for most of the period to 2000, with a changeover in this trend in 2001 (2002 and 2003 omitted as data not available for urban station).
A substantial amount of data was missing for the whole period, particularly with respect to the rural station. Despite this, average annual rural temperatures are still higher if only days where data was available for both stations is taken into account, with a difference of 0.4°C in 1994 and 0.3°C in 1995.
Where the trend reverses in 2001, higher mean urban temperatures of 0.1°C are evident under this
scenario. The results show little urban-rural differences, although it is difficult to obtain a complete
assessment without all the relevant data: a different pattern may emerge if missing data was available.
4 6 8 10 12 14 16 18 20
1994 1995 1996 1997 1998 1999 2000 2001
Year
Average Annual Temperature (Degrees Celsius)
Bristol Rural Bristol Urban
Figure 4.7 Average annual urban and rural temperature for Bristol over 8-year period.
4.2 Precipitation
Again, a substantial amount of data was missing therefore analysis was only possible for two of the cities. Also, as a significant proportion of daily data was missing for some stations, it was necessary to include only those days where precipitation was recorded at both the urban and rural station. As a result of this, the analysis period for Norwich was reduced to 7 years and 5 years for Cardiff. It is possible that rain occurred at one station and not at the other, however such days were excluded from analysis with the possibility that a zero value represented missing or insufficient data. The error bars represent the standard deviation difference around the mean, calculated using the Lindberg (2000) formula (4).
Figure 4.8 shows the mean annual urban-rural difference in precipitation at Cardiff and Norwich. A negative value indicates higher rural precipitation and a positive value higher urban precipitation.
Norwich exhibits higher average rural rainfall for most of the period, although there was no difference between mean urban-rural values in 1996, 1998 and 1999. Differences for the remaining years were all below 1mm. The overall rural average was approximately 7% higher than the urban mean at Norwich, as a proportion of the urban mean.
Again, the general pattern at Cardiff is higher mean rural precipitation, with the exception of 1996.
There were only two days where rainfall data was available for the urban and rural station during this year, therefore not representative of a true annual mean and should be considered as anomalous.
Differences for the remaining years were all below 1mm. Overall, the rural mean value was approximately 10% higher than the urban, as a proportion of the urban mean.
The average monthly trend (figure 4.9) reveals a greater amount of urban-rural variability at Cardiff, indicated by the larger error bars. This could be a result of limited data and a distortion of the true seasonal pattern. There is little variation in monthly mean differences at Norwich, but again this pattern could have been distorted by limited data. Despite this, the general trend over the period is one of higher rural rainfall at both cities, with no seasonal trends apparent.
Further analysis on the frequency of light and moderate to heavy rain days reveals interesting results
for both cities. Figure 4.10 shows that, over the 7-year period, there was a higher frequency of days
with light rainfall (below 4mm) at the urban station in contrast to the rural at Norwich. On days
where rainfall was greater than 4mm, this trend reversed, with a higher frequency of moderate to heavy
precipitation days occurring at the rural station. Interestingly, a similar pattern emerges at Cardiff (figure 4.11), although data only covered a period of 5 years.
The fact that this pattern is evident at both cities could indicate urbanisation effects, however other factors need to be considered also. The location of the rural station with respect to the urban station is different at each city, for example a north/northeast orientation at Norwich (figure 3.3) in contrast to a southwest orientation at Cardiff (figure 3.1). This would imply upwind and downwind effects are not factors contributing to this pattern, as the general prevailing wind direction is from the west/southwest (Wheeler and Mayes, 1997). Also, there is a greater amount of moisture availability for precipitation development at Cardiff due to its coastal proximity, which is not the case at Norwich.
-11 -6 -1 4 9
1994 1995 1996 1997 1998 1999 2000
Year
Average Annual Precipitation (mm)
U-R Norwich U-R Cardiff
Figure 4.8 Average annual urban-rural (U-R) difference in precipitation from 1994 to 2000.
-12 -7 -2 3 8
Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Month
Average Monthly Precipitation (mm)
U-R Norwich U-R Cardiff
Figure 4.9 Average monthly urban-rural (U-R) difference in precipitation from 1994 to 2000.
0 50 100 150 200 250 300 350 400 450
0-2 2-4 4-6 6-8 8-10 10-20 20-30 >30
Daily Precipitation Amount (mm )
Frequency (Number of days)
Norwich Urban Norwich Rural
Figure 4.10 Frequency (number of days) of urban and rural daily precipitation amounts for Norwich over 7-year period
0 10 20 30 40 50 60 70 80 90 100
0-2 2-4 4-6 6-8 8-10 10-20 20-30 >30
Daily Precipitation Amount (mm ) Frequency (Number of days)
Cardiff Urban Cardiff Rural
Figure 4.11 Frequency (number of days) of urban and rural daily precipitation amounts for Cardiff over 5-year period
The greater frequency of moderate to heavy rainfall days at the rural stations is in conflict with
research. Results from a study at Chicago, USA revealed a greater frequency of moderate to heavy
rainfall days occurring at the urban station, in the ratio of 22:20 annually (Changnon, 1969). The
intensity of precipitation varied seasonally and was more pronounced during the summer months
(Huff and Changnon, 1973). Previous research on the frequency of rainfall amounts seems to be
limited therefore results for Cardiff and Norwich need to be considered in this context, with further
investigation needed.
4.3 Wind Speed
Figure 4.12 shows the mean annual urban-rural difference in wind speed across the four cities.
Negative values denote higher rural averages and positive values correspond to higher urban averages.
The general trend is one of higher mean wind speeds at the rural stations. This correlates to research and theory that higher roughness levels created by urban structures reduce airflow within the urban canyon (Oke, 1987, Arnfield, 2005), as a result of increased frictional drag. An assessment of vertical updraft velocities could not be made from this data.
An interesting pattern of a reduction in urban-rural differences occurs over the latter part of the period for all cities (with the exception of Bristol), indicated by the decrease in negative values from 2000 onwards. Missing data for Bristol, particularly for the rural station, may have distorted annual averages here, although there appears to be little urban-rural fluctuations in mean wind speeds at this city. Mean urban-rural differences at the other cities have reduced by about 1ms
-1, comparing 2003 to 1994.
Although the magnitude of this change does not appear to be very significant, the pattern is consistent across the three cities, suggesting a possible link to urbanisation effects. Reductions in rural wind speeds are responsible for changes at Cardiff and Newcastle, although at Norwich increases in urban means are the main contributory factor. This inconsistency weakens the argument for urbanisation effects.
-6 -5 -4 -3 -2 -1 0 1 2 3 4
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Year Average Annual Wind Speed (ms-1)
U-R Newcastle U-R Norwich U-R Cardiff U-R Bristol
Figure 4.12 Average annual urban-rural (U-R) difference in wind speed for four cities over 10-year period.
4.4 Synthesis and Discussion of Results
Analyses of average temperature trends across all cities indicate the presence of an UHI (described by Oke, 1987), with the exception of Bristol. The limited data availability for Bristol stations may have distorted the true averages, however rural means are still higher if only days where data was available for both stations is included. One reason for this anomaly may be the classification of the rural station. It is located on the outskirts of the city, although is within two miles of two major motorways, the M5 and M48 (figure 3.2), therefore could actually be classed as suburban. The rural station is also at a much higher elevation than the urban station, in the order of 60m, which could also be a contributory factor in the temperature discrepancy, although this is likely to be minor. The suitability of the station location for urban-rural analyses could therefore be problematic. It is also a possibility that a discernable UHI does not exist at Bristol, which should not be discounted.
The spatial and temporal magnitude of apparent UHIs at the other cities is similar (figure 4.13). The
average urban-rural temperature gradient over the full period was 0.5°C, 0.9°C and 1.0°C for Norwich,
Newcastle and Cardiff respectively. These values are comparable to UHIs observed at other cities, for example the METROMEX project, a similar study, revealed an average UHI of 1°C over a 6-year period at St Louis, USA (Auer, 1981). Satellite images taken of Greater Manchester on one day during 2003 revealed an urban-rural temperature difference of 8°C (Collier, 2006), although a much larger gradient, the observation was made during the summer and at a larger metropolitan city. Winter gradients at this city were smaller, in the region of 2°C to 4°C (Collier, 2006).
-5 -4 -3 -2 -1 0 1 2 3 4 5
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Year
Average Annual Temperature (Degrees Celsius)
U-R Newcastle U-R Norwich U-R Cardiff U-R Bristol