to analyse data on actual urban travel patterns. The combined analysis of flow and travel cost matrices can be used to explore relationships between
accessibility and travel flows, i.e. between potential and actual interactions. The analysis of travel data is the basis of measuring travel sustainability, which in this research is the CO₂ emissions indicator. This allows the sustainability impacts of travel patterns to be quantified and relationships analysed.
Aggregate travel pattern data is based on a matrix structure defining flows between origin and destination zones. In the context of the UK census, travel costs such as travel distance and time are not included in the source data, and need to be calculated using the network analysis methods described in the previous sub-section. As the number of zones increases (with generally thousands of zones needed for meso-scale city-region analysis) the number of flows increases exponentially. Therefore techniques to summarise the properties of matrices and understand general patterns are useful. This could involve macro-city scale analysis where a single value for an entire city is calculated (see Sub-Section 3.3.3). This approach is useful for analysing city dynamics as a reduced number of headline indicators are calculated (Frost and Spence, 2008), though it in unsuitable where intra-urban spatial patterns are the focus.
For sub-regional analysis Plane (1995) proposes collapsing the matrix into a three-fold inner-city, outer-city and hinterland structure, leaving a greatly reduced 3x3 matrix. A version of this sub-region technique is calculated for mode-choice in the London region in Section 6.2. The problem with this
monocentric structure, thus cannot adequately explore the sub-centres and polycentric forms that are the main focus of this research.
Instead this research is based on calculating the properties of flows for the entire city-region matrix. The advantage of this approach is that it makes no prior assumptions about urban structure, and is highly comprehensive with all the city-region flows analysed (flows with external origins/destinations should be included as well). The challenge for this approach is the computational
overhead, and thus the GIS software and database technologies described earlier in this chapter are an essential prerequisite for this method. The computational demands depend on the size of the matrix, which in turn depends on the study extent and zonation selected. The only two available 2001 census zonations sufficiently fine-grained to identify urban sub-centres in the UK are wards (approximate residential population 12,000) and output areas (approximate residential population 300). As the extent of this study is relatively large, the corresponding matrices are very large, with the ward matrix containing approximately 3 million values and the output area matrix containing over 4 billion values! Whilst the majority of the values in these matrices are zero (about 15% of potential flows are greater than zero in the study area), the size of the output area matrix was found to be infeasible for current desktop GIS software. Therefore the ward zonation is used as the basis of the travel analysis in this research. This scale proves to be sufficient in identifying detailed intra-urban patterns (see Chapter 6). Note also that some key variables such as employment class have not been released for the UK census at output area scale, which would also be problematic for the relationships analysed in this research.
Once the matrices for the study area have been completed- including relevant mode-choice and socio-economic data, and the calculation of mode-specific travel time and distance for each flow- the sustainability properties of each flow can be estimated. The advantage of composite indicators such as CO₂ emissions and energy use is that they integrate mode-choice and distance patterns into a single measure. Furthermore CO₂ and energy concerns are ultimately central to sustainability policy. The method used here is to multiply the distance travelled by each mode by a mode-specific emissions factor modelled by the UK
government. Similar approaches has been used in existing studies (Banister et al., 1997; Frost and Spence, 2008) and the novelty in this research is the intra-urban scale and level of detail in the network routing analysis, as well as a CO₂ rather than an energy focus. The emission factors used are shown in Figure 4.11. These are estimates of the direct emissions resulting from a trip, not including less specific indirect emissions from vehicle manufacture and infrastructure construction (note that the direct/indirect distinction is more problematic for public transport networks as discussed below).
a Figure 4.11: Estimates of Carbon Dioxide Emissions Per-Passenger-km by Private and Public
Transport Modes (London values shown where available). Source: DEFRA (2010).
DEFRA has developed a detailed methodology for producing the above emission coefficients including profiling of the UK vehicle fleet; empirical analysis of typical road conditions and driving behaviour; integrating the results of public transport models from National Rail and Transport for London;
including emissions resulting from the production of fuels (e.g. crude-oil to petrol); and additionally including CO₂ equivalent emissions from other greenhouse gases such as methane and carbon monoxide (DEFRA 2010). The challenges for this distance based mode-coefficient methodology relate to the degree to which spatial and temporal patterns diverge from these average values, particularly for key variables such as vehicle occupancy. For private
0 50 100 150 200 250 300
C02 equivalent (grams per passenger km)
speed/degree of congestion, and number of passengers. The method used here assumes that the car properties are evenly dispersed from the mean across the study area. For occupancy, the UK census records whether car users are car drivers or passengers. Thus we can calculate the average occupancy of vehicles in the study area, and calculate different average occupancies for car driver and car passenger trips (i.e. occupancy must be at least 2 for passenger trips). For congestion effects, a logical approach would be to calculate carbon emissions on a link-by-link basis in the network analysis stage. This technique would require the micro-level calibration of relationships between CO₂ emissions and link-based average speeds, which is unfortunately beyond the scope of this current work, but should be a priority for future research.
Public transport trip emissions are also affected by similar issues of spatial and temporal disaggregation. DEFRA provide data on London-specific tube, bus and taxi emissions (DEFRA 2010), which allows a basic level of spatial disaggregation to the London study area. National rail figures for London are not available, and UK figures have been used here. The issue is that occupancy is likely to be higher for London services and so public transport emissions may be over-predicted. The occupancy of public transport varies temporally between peak and off-peak services, with commuting journeys taking place during peak times, and only average business transport values are available through the DEFRA statistics used here. Using average occupancy means that per-capita carbon emission estimates for public transport are likely to be over-estimated.
Public transport is problematic in a wider sense for this methodology as
emissions are less „journey-specific‟ and could be considered as a product of the public transport system as a whole. The relationship between public transport supply and demand is less direct and immediate than private transport. Yet public transport is demand responsive in the longer term, and the alternative of ignoring carbon emissions from public transport would be highly misleading for policy.
In summary, the meso-scale city region analysis of travel patterns involves handling large matrices with potentially millions of flows, and subsequently GIS technologies are essential in calculating indicators based on these matrices.
Composite indicators such as CO₂ emissions and energy use integrate trends in mode-choice and travel distances, and provide data directly relevant to policy.
The emission coefficients are central to these indicator calculations. This research makes significant progress in spatial disaggregated analysis of intra-urban travel flows, but errors remain in the spatial and temporal disaggregation of emission coefficients regarding factors such as occupancy and congestion.
Addressing these issues should be a priority for future research in this field.
4.8 Chapter Conclusions
This chapter has addressed Research Aim 3, which is to develop the methodology for the intra-metropolitan analysis of employment geography, urban structure and travel patterns. Data and methods to analyse employment geography, the built-environment, accessibility and travel sustainability have been detailed, and these methods are applied to the study area of the London region in the next two chapters of this thesis. Geographical analysis typically involves a trade-off between study extent and level-of-detail, but improvements in data extent and availability, in tandem with the analysis and visualisation capabilities of GIS, are enabling this trade-off to be overcome with the linking of extensive and intensive analysis. A meso-scale intra-metropolitan analysis, which is at a regional extent and is relatively fine-scale, has been advocated here to provide an appropriate balance for the study of city-region structure.
Several methodological hurdles remain however, and a series of methodological innovations have been proposed to enable the intra-metropolitan spatial analysis of cities to be effective. These are the detailed analysis of business survey data for the fine-scale measurement employment geography; the inclusion of real-estate data in geographical analysis to allow property market processes and urban development to be analysed; and more accurate accessibility measures based on the network analysis of detailed transport infrastructure and timetable data. Finally a CO₂ travel emissions methodology has been developed in a format that can be calculated for millions of trips in a city-region. This approach of including the entire trip matrix at an intensive zonal scale allows
comprehensive intra-metropolitan analysis which is capable of identifying trends connected to decentralised, polycentric and monocentric forms. This is opposed to the more commonly applied aggregate approaches that mask intra-urban variation. GIS tools are an essential prerequisite to handle the
computational demands of this approach.
The urban structure analysis methodology is very demanding in terms of data, and there are some gaps in data availability. These can be minimised by drawing on multiple data sources to allow dynamic analysis and link socio-economic and built-environment geographies. For the study area of London there are issues regarding the availability of detailed real-estate data for the
wider region which restricts the real-estate analysis in Chapter 5. There are also data gaps in the travel pattern indicators. The UK census does not contain information on travel distances and times, or more complex sustainability indicators such as CO2 emissions. A methodology to derive these travel cost measures using GIS analysis has been detailed. The main shortcoming in the travel data is that it is restricted to journey-to-work travel in 2001. The methodology could be applied to any time period or trip type where data is available, but the analysis of the London region travel in this research is restricted to journey-to-work.
The second aim of this chapter (Research Aim 4) was to provide a methodology for the empirical analysis of monocentric and polycentric forms. We concluded that monocentricity and polycentricity are scale-dependent relative terms that describe patterns of dominance within an intra-urban hierarchy or network of centres (similar to central place theory described in Chapter 2). The technique developed to identify these forms uses a combination of centralisation and clustering spatial statistics to provide a transparent method of classifying urban forms. These methods are applied in the London context in the next Chapter.