Chapter 4 – Spatially disaggregate population forecasts for SEQ
4.4 Results of spatial disaggregation
Figure 4.10: Land classification for 4-class dasymetric method
(The inset highlights the high density urban areas given by the ‘natural break’ classification. They are significant for many small SLAs around the Brisbane city).
4.4 Results of spatial disaggregation
The results of the population forecasts disaggregation for each forecast year are presented in Figure 4.11.
SLA boundary
Figure 4.11: Results of population forecasts disaggregation 2006–2026
A regional plan for the SEQ region introduced a balanced growth strategy to promote compact growth and decentralised polycentric development (OUM, 2006). The results of the spatially disaggregated population forecasts provide a valuable dataset for analysing the pattern of the population growth. The patterns of the population distribution for each forecast year are visualised by Figure 4.11, but they do not show a significant change in the overall spatial structure.
Exploratory spatial data analysis is used to measure the spatial clustering between the populations at the SLAs. Firstly, the univariate global Moran’s I (Moran, 1950) is calculated for the population at the SLAs. The univariate Moran’s I statistics is defined as:
2021 2026
2006 2011 2016
∑
indicates a strong positive spatial autocorrelation (a clustering of similar values), 0 indicates a random spatial ordering, and -1 indicates a strong negative autocorrelation pattern.The global Moran’s I statistics for the population forecasts are given in Table 4.3. The results show that global Moran’s I of spatially disaggregated population presents a growing positive spatial autocorrelation between 2001 and 2026 (all values are significant at 1%
level using permutation approach for inference with 999 permutations). This suggests that, from a global view, a more compact population growth will occur for the SEQ region.
Such compact patterns can be driven by the increasing restrictions on land availability and increase in intensification through urban development of residential areas. From a planning point of view this indicates a growing pressure on the housing marketplace, high demand for services such as schools or healthcare services in the future.
Table 4.3: Global Moran’s I statistics of population forecasts 2001–2026
Year 2001 2006 2011 2016 2021 2026
The population concentration always presents spatial variations. The specific location for the higher population concentrations in the future is a major subject of concern for the planners. There have been different spatial statistics for measuring the patterns of the urban structure (for example, density, proximity or clustering functions) (Galster et al., 2001). To predict a significant local concentration for the future population, a local exploratory spatial analysis method is applied. Local Indicators of Spatial Association (LISA) (Anselin, 1995) provide a measure for each unit of the region of the unit’s tendency
to have a value that is correlated with the values in the nearby areas. LISA indicators identify ‘hot spots’ that take into account not only the high or low values in a single place (a SLA) but also the values in the nearby places. Previously, LISA has been used as an effective method for detecting patterns of local dependence for a single dataset (Luo and Wei, 2006; Riguelle et al., 2007). In this research, LISA is applied to the multiple datasets created (forecasts) to analyse the patterns of the population growth over time.
The LISA statistics can be defined as:
∑
=
j j ij
i W z
z
I (4.8)
where ziand zjare standardised scores of the population values for the SLA i and j; it is among the identified neighbours of i according to the weight matrix Wij.
Each local area and its neighbours can be L = low (below the mean) or H= high above the mean). The LISA pair LH, for example, designates an SLA that has a low-level population (below the mean) locally and high level (above the mean) in the neighbouring SLAs. I tested whether a potential spatial diffusion of the population would cause changes in the levels of these local-neighbour LISA pairs over time.
A visual interpretation by the time series in Figure 4.12 shows the minor differences in the population settlement pattern (local autocorrelation) from 2006 to 2026. Basically, the future SEQ population will keep the existing concentrations and dispersions, with slight transitions in the population clustering moving from the eastern part of the region to the west. The only notable change is in the SLA (Ipswich West) presenting an LH to HH transition between 2016 and 2021, which indicates a potential urban expansion will occur. This is possibly caused by the higher pressure on the housing market, and many people will find home in less expensive suburban areas. As proposed in the South East Queensland Regional Plan (OUM, 2006), there is good uptake of vacant land in the western suburbs as preferred for housing development to reduce the growth pressure on the existing urbanized areas. The implication of the results for planning is that with current
healthcares, and public transport) should be well managed and planned to satisfy the new demands over the next 20 years.
Figure 4.12: local autocorrelation of spatially disaggregated of population forecasts 2006–2026
4.5 Conclusions
The spatial distribution of the population and the future urban structure for SEQ has become a focus of recent academic inquiry and planning concerns. This chapter investigates the patterns of the future population for SEQ using large regional population forecasts and the recent development in a spatial disaggregation method. The focus of this chapter has been on testing and the validation of the methods used for the spatially disaggregate population data from large forecast regions to smaller areas.
2016
2026 2021
2011 2006
The high level of spatial heterogeneity in the SEQ population requires the spatial disaggregation technique to be very sensitive to the density variations. Existing spatial disaggregation techniques that were previously used were based on the restrictive density assumptions and tested against simple geographical areas. To better resolve the spatial heterogeneity issue, the spatial disaggregation research was extended by increasing the total number of the density classes for the dasymetric method, which provided a wider range of density configurations for the study area.
The SEQ region is used as a larger and more complex study area to rigorously test the techniques. The result demonstrates that the traditional 3-class dasymetric method can be further refined by incorporating a more detailed land class. This indicates that the technique allowing a greater heterogeneity in the assumptions tends to be more accurate than the techniques based on restrictive assumptions. The results further illustrated that an excessive density subdivision is not suggested because of its limited capacity for improvement. The appropriate land class (4 class) can be determined by the complexity of the study area and the expected cost of the implementation.
The 4-class dasymetric technique was then applied to disaggregate the regional population forecast for every forecast year. The outputs of the spatially disaggregated forecasts for the smaller planning areas provided a more precise spatial scenario of the future population distribution in SEQ. To better understand the patterns of distribution, the exploratory spatial data analysis was applied to investigate the spatial clustering of the population in the local areas over time. The results tentatively indicate that the growing population in SEQ will maintain the existing structure but with a slightly more compact pattern across the region. The population concentration in the southwest of the region will increase, but such effects remain limited. I consider this is partly limited by the available data for the dasymetric disaggregation over time. A reclassification of the land areas including the planned new urban areas is needed, especially for the longer-forecast years. This work needs to be carried out in the future should the data be available to investigate those issues.
As stressed in the previous chapter, a full understanding of urban growth pattern not only
growth will influence the future employment locations is an interest of concern for both the academics and the planners. In the next chapter, the disaggregated population data are used as a major input to drive the regional employment disaggregation. The output of the continuing work is the spatially disaggregated employment forecasts in the local areas.
These forecasts provide the additional information on the underlying spatial structure of the economic activities as a result of the growing populations.