1 GIS analysis - Part 3

1.1 Spatial joins

Spatial Joins are prevalently used when solving some real world problems in GIS. This functionality is available through the Join Attributes by Location tool from Vector > Data Management Tools. We are going to use this function to identify retail centres most at risk to consumers who prevalently shop online. More specifically, we are going to apply a spatial join to the retail catchments (both Buffers and Huff) and the IUC targets (the Online Shopping prevalence layer). Follow the steps below:

  • Make sure you have saved Huff_catchments, Buffer_Hierarchy and Online Shopping prevalence layers to your working directory

  • Open a new blank map in QGIS

  • Load the Huff_catchments, BufferHierarchy and Online Shopping prevalence layers

  • Open the Attribute table of the Online Shopping prevalence layer and examine all the attributes. Since we need to identify those retail centres that are at high risk to online shopping (exposure to consumers with high prevalence to online shopping), we can use the binary variable that was created earlier - Online_Shop. This can be joined with the catchment layers

  • Go to Processing Toolbox (by clicking Toolbox or right-click anywhere near the top of QGIS in the grey area and enabling Processing Toolbox). Then type Join Attributes by Location and double click on Join attributes by location (summary). This will allow you to generate various statistics for the joined layer.

  • In the dialog window, select BufferHierarchy as your input layer - Join to features in and Online Shopping prevalence as your By comparing to

  • Use Intersect as your Where the featuresfunction

  • Click the … button next to Fields to summarize and select Online_Shop. Click OK.

  • Save your Output shapefile (name it SpatialJoin_buffer) to your working directory and click Run and Close

  • A new layer should be added automatically to your Layers Panel

  • Open the Attribute table of the new layer and see what columns have been added

  • There are all possible statistics calculated, but the stats of an interest to us are the Sum and Count, so you can delete any other newly created columns.

Note: If your column names have been saved as, e.g. Online_shp1, then the columns you will need to keep are Online_Shop and Online_shp5. These are the count and sum columns, respectively.

  • To do so press the Delete column button DeleteCol and select the required field from the Delete Attributes dialog
  • Next you need to decide how the new columns can be used to identify those retail centres most at risk to consumers who prevalently shop online…

One way is to display the Online_Shop_sum column which shows the total number of LSOAs characterised by higher online shopping prevalence for each retail catchment. As such, by using the buffer rings as retail catchments we can see that Allerton Rd (with the sum of 31) is the most exposed retail centre in Liverpool, followed by Old Swan, Gatecare and Liverpool.

  • Create a map showing particular catchments exposure to consumers who prevalently shop online
  • Use for classes of exposure Low, Medium, High and Very High
  • Remember to include labels, north arrow, scale bar and legend
  • Compare your map against the one below:
    Exposure Buffers

Nevertheless, the results shown above only account for the number of LSOAs with high exposure. A more comprehensive way of identifying centres most at risk to consumers who prevalently shop online could be to look at a number of LSOAs with high exposure relative to total number of LSOAs within each catchment. This can also easily be calculated within QGIS.

  • All you need to do is:

  • Go to Properties>Symbology of the SpatialJoin_bufferlayer and change to Graduated

  • Click Expression next to the Value drop-down menu

  • In the expression window, type Online_Shop_sum / Online_Sho_count

  • Click OK and style

  • By using Online_Shop_sum / OnlineSho_count we will display the proportion of the LSOAs with higher exposure within each catchment, instead of a static column value

  • Create another map and compare it to the previous one
    Risk buffer

1.2 Spatial joins (optional)

  • Repeat all the above steps to perform a spatial join for the Huff_catchments
  • Create the two equivalent maps for Huff catchments
  • Compare the results against those obtained for Buffer catchments
  • Comment on the differences and similarities
    Risk_Huff

2 Siting a discount supermarket - location analysis

Since the global economic crisis of 2008-09, market share of the hard discounters in the UK has been growing, often at the expense of the major grocery retailers - the Big 4 (Tesco, Sainsbury’s, Asda and Morrisons). This trend has been exacerbated by the recent cost of living crisis. The aim of this exercise is to find the most suitable location(s) (could be more than one) for a discount supermarket in Liverpool using GIS analysis. Typically, an analyst would use some sophisticated location analysis tools such as traffic pattern information, demographics, lifestyle data and footfall, and would carry out an analyses of its competitors. A location analysis often requires looking at footfall traffic generators (in particular when siting convenience stores), as other retailers in the neighbourhood may draw customers from various employment sites that are nearby. These could include industrial or office parks, schools, colleges and hospital complexes.
In the case of a discount supermarket, it is important to choose the right neighbourhood as typically, the most affluent catchments tend to shop more in the upmarket stores (not heavy discounters). First, we will explore the potential competition and different catchment estimation techniques.

2.1 Store catchments

  • Add the Liverpool_Foodstores shapefile (located in the Retail folder) to the Layers Panel
  • Check the Attribute table > Store Type column.

As we want to find the most suitable location for a discount supermarket, we can make an assumption that there will be more competition from the supermarkets than convenience stores. As such, for our analysis we can remove the convenience stores from our database.

  • Display only the supermarkets by creating a spatial query for the Liverpool_Foodstoreslayer by using the Select features using an expression tool in the Attribute Table
  • Create a query: “Store Type” = ‘supermarket’). Click on the Select features button
  • Save the selected features as a shapefile to your working directory, name it Liv_Supermarkets.

However, there are 73 supermarkets in Liverpool and therefore creating that many buffers is not very useful as the map would be very cluttered. Therefore, in our analysis we will focus only on the direct competitors - the heavy discounters - Aldi and Lidl stores. - Create a new layer for Aldi and Lidl supermarkets by using the Select features using an expression tool in the Attribute Table (“retailer” = ‘Aldi’ OR “retailer” = ‘Lidl’), name it Discount supermarkets. How many discounters there are in Liverpool? - Then create 1500m buffers for the Discount supermarketslayer and name it 1500buffer_Discounters Buffer_Discounters

  • You could also check how similar the distance catchments are when compared to the buffers.
  • Open ArcGISPro and create drive distance catchments for the discounters - name the layer Disc_DriveTimes1500
  • Once created and saved in your working directory, add the Disc_DriveTimes1500layer to your map in QGIS.
  • Make the drive time catchments transparent, but keep the catchment boundaries. Make them red and set the stroke width to 0.3
  • Compare the extent of buffer catchments vs. drive distances. Are there any under-served areas?
    Buffer_DriveDist

It is clear from the map that some areas fall outside the delineated catchments, indicating that these areas are not effectively served by the analysed discount supermarkets. From a competition perspective, these areas could be potential locations for the new discount supermarkets. However, this would depend on demand-related factors like total population, affluence etc.

Locating a new discounter in a vicinity of a larger supermarket doesn’t necessary have to be viewed as a disadvantage, providing of course, that the level of market saturation is not very high and you know the catchment’s demographics. One way of obtaining new insights to decide where to locate a store is to examine the analogue stores. As such, we will examine the catchments of other discount supermarkets in Liverpool.

2.1.1 Catchment affluence

Understanding the potential consumers and their spatial distribution that discount supermarkets are likely to target is vital here. Research shows that the predominant locations are either city/town centres or other urban areas, especially those with less affluent catchments. So let’s explore that - we will examine the relationship between Income deprivation in Liverpool and the discount supermarket locations.This exercise will use 1500m buffer catchments to obtain the statistics.

Assignment tip - think what areas and population/catchment characteristics would be preferable for a corporate convenience store

  • Create a Chloropleth Map of Income deprivation variable from the IMD 2015 layer (or use it the one from Section 2.3 if you have it open)
  • Use the Select by Location tool (Vector > Research Tools), select all the LSOAs that intersect with the 1500buffer_Discounters (Select features from: IMD2015, Where the features: intersect, By comparing to the features from: 1500buffer_Discounters)
  • Go to the Basic statistics for Fields tool and check what is the mean/median Income deprivation value for the selected LSOAs (is it 0.259 mean/0.27 median with StdDev 0.135?)
  • Now check what is the mean/median value for the Income deprivation within the boundary of Liverpool City. It should be 0.256/0.255 respectively, and to put that into a broader context the mean value for Great Britain is 0.15. Also the StdDev of 0.135 suggest that there is a relatively large degree of variation within those catchments.

IMD_Discount So based only on this very simple analysis we can make an assumption that the new discount supermarket should be located in a neighbourhood that is slightly less affluent than the city/regional average and has mixed demographics (population characteristics).

Assignment tip - check what the mean value for IMD 2025 is using drive times Disc_DriveTimes1500

2.1.2 Catchment total population

  • Next, we will examine total population for our catchments (this could be extended to checking other population-related factors such as density or difference in total population between various catchments, as well as the demographic/employment composition of each catchment)

  • Similarly to the above section, select any catchment area and check what is the total population of it (this can be referred to as potential patronage of that store).

  • Again, use the Select by Location tool to select all the LSOAs in your residential/night time population layer that intersect with a discount store catchment area of your choice

  • Repeat this for other catchment areas

  • The total population should vary between 8,000 and 60,000 people per catchment

Nevertheless, it is important to bear in mind that many of these catchments are overlapping so in some cases we count the same residential population twice - a phenomenon also called cannibalisation. So for this reason it would be useful to calculate the average population count per catchments.

  • Select all LSOAs that intersect with the discounter catchments, generate basic statistics for all selected LSOAs and divide the total population (SUM 406581) by 14 (no of discount stores)
  • Is the average population per catchment around 29,000? (if your average varies dramatically please ask for help)

On the other hand, the day time population within the city centre is much higher that the residential population (ONS estimates) so examining the workplace population may be useful.

Assignment tip - This is, in particular important for the location of corporate convenience stores, as the larger employment sites play important role in creating demand for convenience goods retail.

2.1.3 Location of a new discount store

Having done some analyses, we can now decide on the most suitable location for our new discount store. It is pretty obvious from the map that there are no discount stores in the north-east and especially in south Liverpool. The map of income deprivation suggests that both areas could be considered. The population distribution and the relatively low affluence levels suggest that Garston or Speke area (South Liverpool) can be most suitable.

  • Taking into consideration the existing road network, create a new point indicating the potential location of a new discounter; To create a new point, use the steps from practical 1. If you need some help follow the steps below:

  • Go to Layer > Create Layer > New Shapefile Layer

  • In the window that pops up give it a name: New Discounter

  • Choose Point from the drop down menu as your Geometry Type

  • Set Your CRS to British National Grid and Click OK

  • The newly created (empty) layer will be added to your Layer Panel

  • Right click on it > Toggle Editing and then from your Toolbar menu use the Add Point feature tool Add Point to create a point in Speke Boulevard

  • Open the Attribute Table of the new layer

  • Toggle Editing > New Field

  • Type Name as your field name, choose Text (String), length 50

  • You can now enter a name for your newly created store in the Attribute Table

  • Call it New Discounter

  • Toggle Editing Mode off > Save Edits

  • Then create 1500m buffer and drive time catchments New Discounter

  • Now you can check the total population and income deprivation levels for the New Discounter catchment.

Assignment tip - Is it in line with other similar stores? Check the mean values of income deprivation, total population, and online shopping prevalence for the new store and compare them against similar existing stores. Such data would also make a very useful and insightful table.

Please note that the LSOA level is too coarse for the real world location analysis, OAs would produce much more accurate estimates; however we use free data only and most of our variables are available at LSOA level only.

  • You have completed the final practical, so well done!!! Now, apply your GIS skills and contextual understanding to successfully complete the second assignment.