It's hard to think of a time in the recent past when people have been so interested in geography.
Coronavirus restrictions have changed that.
Here in Melbourne, travel and administrative restrictions have reduced the size of our worlds to a 5km radius.
Many people are querying the geographic definition of metropolitan Melbourne and regional Victoria.
It's great to see so much interest in geography, even if it has resulted from such a devastating pandemic.
This blog will look at why good geographic boundaries are important by looking at some of the quirkier aspects of spatial units.
The state boundary - NSW and Queensland
Border closures that have occurred in order to restrict movement in the wake of coronavirus have highlighted where state boundaries can be cumbersome.
A case in point is the NSW-Queensland border.It mostly runs along the 29th parallel south, but in the east it follows a series of rivers and other landmarks.
Things become interesting where it runs through Coolangatta (Queensland) and Tweed Heads (NSW) on the coast.
There are many streets where the neighbours across the road live in a different state (not to mention a different time zone in summer!).
Progressively tighter border controls imposed by the Queensland government have resulted in a great deal of inconvenience for residents of these localities.
Although some special considerations have been granted, there are nevertheless tight restrictions on how far Gold Coast and Tweed Heads residents can travel.
The figure below shows how the border runs through Coolangatta and Tweed Heads.
The NSW-Queensland border was established well before what we now know as the Gold Coast even existed.
Urban growth over many decades has resulted in contiguous development across the state border on both sides, such that there is now an urban conurbation stretching more than 200km along the coast.
Furthermore, the Gold Coast airport straddles the state boundary, with parts of the terminal in NSW.
Aside from road signage welcoming you to either Queensland or NSW, there's not much in the way of differentiating the two states.
The Gold Coast mayor, Tom Tate, has even gone so far as to suggest that the state border be moved south to the Tweed River.
In this COVID-19 era it may reduce some of the border control issues as it would align to a visible feature, but it does raise an interesting possibility.
Tweed Heads and the surrounding suburbs operate as part of the wider Gold Coast region, hence a major reason why some special considerations have been granted to residents living near the border.
Some fun facts to ponder.
Firstly, if the NSW-Queensland border followed the 29th parallel south all the way to the coast, the actual border would be located in the locality of Broadwater.
This is a small locality located just north of Evans Head, which is about 130km south of the actual border at Tweed Heads.
Secondly, how many people would be affected if the border was moved as per the suggestion made by Tom Tate?
If we assume the Tweed and Cobaki Rivers form a natural boundary, around 10,600 people (2016 Census) would be added to the Queensland population.
An interesting prospect, but let's face it - State borders are unlikely to change, even when the lines drawn up over 100 years ago have been shown to be impractical.
Postcodes
Postcodes are designed for the delivery of mail, not for the dissemination of statistics. So went the mantra when I worked at the ABS many years ago. While many people may identify with, or can conceptualise a postcode, they are actually not good units for spatial analysis. The reasons include:
- they can cover very large areas, particularly in rural Australia, incorporating towns and surrounding areas. In some cases, postcodes straddle state borders.
- similarly, postcodes can cover more than one suburb in metropolitan areas. Data from the Victorian Dept of Health and Human Services shows that postcode 3029 has the highest number of COVID-19 cases, but this postcode includes three suburbs - Hoppers Crossing, Tarneit and Truganina.
- many postcodes are non-contiguous, ie they are in separate parts. This is very common in rural areas but the example of postcode 3030 shown in the picture below also shows it can occur in metropolitan areas too (note - sourced from here).
- postcode data is prone to human error. Because it's typically collected through administrative data, accuracy is reliant on people knowing their correct postcode. Anyone who has had to clean administrative data can attest to the number of people who think they live in a PO box number such as 2001 or 3001. However people will know what suburb or town they live in.
- postcodes really are designed for the delivery of mail! In the initial stages of the stage 3 lockdown in Melbourne, only certain postcodes were affected. One of these was 3055, which is essentially the suburb of Brunswick West. To avoid confusion, suburbs were included in the media information, but ironically this lead to more confusion. Many Brunswick West residents were surprised to learn they shared their postcode with Brunswick South and Moonee Vale - which are post offices, not localities.
There has been a lot of noise made on social media that publishing COVID-19 cases by LGA wasn't sufficient - people were concerned about more localised data on case numbers.
This is probably driven by a need to see where exactly the COVID-19 cases are located, especially in larger LGAs.
However, suburb and even SA2s (an ABS construct) are better spatial units for most purposes.
School zones
The Dept of Education and Training is the custodian of school zone boundaries in Victoria.
These are based on mathematical computations that produce Voronoi polygons.
Confused? Try being a parent working out what school zone you live in.
Not only are the zones monitored on an annual basis and are subject to change, they generally do not follow visible features, nor do they consider natural or physical boundaries.
Instead, the focus is on measuring distance in a straight line.
In some cases it means children have to attend a school that is much further from their home, simply due to the way the polygons have been created.
This example from the suburb of Seabrook shows that some children are required to attend a school in Altona Meadows, even though they pass Seabrook Primary School on the way.
The zones themselves also have wider impacts than inconvenienced families.
House price data indicates that properties in the zones of sought after schools sell at a premium compared to those outside the zone.
This is particularly interesting given the fact that school zones are subject to change - a classic example of buyer beware?
No meaningful data can be created for school zones in their current, as boundaries do not relate to any other geographies (therefore data can't be validated as no benchmark), but importantly, they are not useful in a practical sense.
Which school do you attend if the boundary runs through your living room?
School zones would be far more practical if they used visible boundaries and considered impassible natural features like creeks, as well as potential walkability barriers such as major freeways.
The example below concerns Flemington Primary school and shows that children in the surrounding area, as well as those who live east of the Tullamarine Freeway, are zoned to attend this school.
However, access to the school for children east of the freeway is difficult as there are no crossovers, precluding walkability for younger children in particular.
On the other hand, Brunswick South West Primary School is much closer, but in a different zone.
Concluding comments
So what's the best geographic boundary to use?
Ultimately, it depends on your needs, data availability and statistical rigour/fit for purpose.
Some data, such as that produced by surveys, is not available for smaller geographic areas.
Census data has the advantage of many data items being available for small geographies such as SA1s, but even then the data is subject to confidentiality.
As with all data, geography has its strengths and weaknesses.
Understanding this, and how it impacts data outcomes, is all part of the spatial analysis toolbox.
