Research

Multi-Criteria Spatial Optimisation of Christchurch’s Urban Development

By November 7, 2020November 9th, 2020No Comments

How would you imagine a beautiful city for the next generations? Currently, local authorities are unequipped to quantitatively assess areas for future development in a manner that can consider multiple planning objectives. So let’s change that.

By Sam Archie & Jamie Fleming, with supervision by Tom Logan (2020). This report was first published in the November 2020 University of Canterbury Civil and Natural Resources Engineering Research Conference.

Background

The sustainable development of cities has recently been identified as an important way for us to adapt to and help solve climate change. In response, the National Policy Statement on Urban Development requires urban-planners to design future urban areas in New Zealand strategically for the next generations, primarily through intensification of existing residential areas. This paper continues the development of a multi-criteria spatial optimisation framework that uses a genetic algorithm. The framework is applied to the case study of Ōtautahi Christchurch to identify areas of priority for urban intensification, to aid decision-makers where to guide future growth whilst taking into account multiple hazard adaption and sustainability objectives.

This framework aimed to find an optimal, or a series of optimal, scenarios that are better for a range of attributes (known as objective functions).

Although the algorithm is currently programmed for Ōtautahi Christchurch, it is possible for this code to be adapted for other cities in New Zealand. Moreover, this case-study presented uses sample weightings and objective functions throughout the analysis to showcase the effectiveness of the framework. However, with consultation with major stakeholders, the analysis can be fine-tuned to better represent and locate development sites that satisfy their needs.

Planning Objectives

Planning Objectives Implemented

Icons of the example objective functions for Christchurch
Objective Name
Description
Parameterisation
Tsunami
Minimise exposure to inundation from a tsunami by a 1 in 2500-year earthquake in South America
The maximum inundation depth within a statistical are is normalised by the maximum inundation depth within any statistical area
Coastal Flooding
Minimise exposure to a 1 in 100-year coastal flooding surge. Includes increased exposure due to sea-level rise in increments of 0.1 meters, up to 3 meters
Statistical areas were assigned a value based on what incrementation of sea level rise causes the first instance of flooding in the statistical area, and which RCP path and upper/lower bound it corresponds to for predicted sea level risein 2120
River Flooding
Minimise exposure to a 1 in 500-year river flooding event for all rivers in Christchurch
Data did not provide inundation depths, just where it was affected by the flooding. As a result, the objective function returns a binary result of 0 if not in the flood zone or 1 if in the flood zone
Liquefaction
Minimise exposure to liquefaction susceptibility
Assigned a value in the range 0-1 based on categories of liquefaction susceptibility given by the dataset
Distance
Minimise the distance of new development to town centres to minimise travel
Each statistical area was assigned a value based on normalised distance by car-based transport to the nearest key activity area
Development
Minimise sprawl into rural areas
Each statistical area is assigned a value based on percentage of area which is currently allocated as residential, commercial or mixed use determined by the District Plan. The higher this percentage, the better the land is for densification and hence a lower score

Data Sources

Data Sources of Objective Functions

Data Provider
Data Type
Description
Source
Environment Canterbury (2019)
Raster
Ensamble maximum flow depth from a simulated earthquake-induced tsunami, caused by a 1 in 2500 year earthquake in South America
Private data sharing agreement with Environment Canterbury, from November 2019 consultancy report by GNS Science titled “Multiple scenario tsunami modelling for Canterbury”
National Institute of Water & Atmospheric Research (NIWA), NZ (2020)
Polygon
Extreme sea level (ESL) for 1% AEP coastal flooing, given a range from 0m to 3m of sea level rise (SLR) in 10cm increments
Private data sharing agreement
Christchurch City Council (2018)
Polygon
Land which is potentially suscetable to inundation during an extreme hydrological event (1 in 500 years), which may pose a risk to life or property due to water velocity and/or inundation depth experienced
Canterbury City Council Geospatial Public Portal. Layer: “FloodHazardHigh” through WFS dataset sharing accessible by clicking here
Canterbury Maps (2018)
Polygon
Vunerability to Liquefaction
Tonkin + Taylor July 2020 report titled “Christchurch Liquefaction Vulnerability Study” and maps accessible by clicking here
Canterbury Maps (2017)
Polygon
Key Activity Centres as shown in the Land Use Recovery Plan 2013. A new polygon features was manually added to represent the new development of the Riverside Market in the recent year
Data originated from Enivronment Canterbury by clicking here
Christchurch City Council (2018)
Polygon
Land use activity boundaries as identified in the District Plan. The zone is defined by type and code
Canterbury City Council Geospatial Public Portal. Layer: “Zone” through WFS dataset sharing accessible by clicking here

Data Sources of Constraints

Data Provider
Data Type
Description
Source
Canterbury Maps (2019)
Polygon
MBIE Technical Classes supplied by Canterbury Earthquake Recovery Authority (CERA)
Published in the gazetted Land Use Recovery Plan (6/12/2013) and republished by Environment Canterbury. Accessible by clicking here
Canterbury Maps (2017)
Polygon
Public parks in the Canterbury Region
Data originated from Environment Canterbury and is accessible by clicking here
Christchurch City Council (2018)
Polygon
Area of Christchurch City which means the contiguous urbanised boundary of the city only (with no discontiguous towns or settlements included). Reference to the Proposed Replacement District Plan: Appendix 2.2 Area of Christchurch City
Canterbury City Council Geospatial Public Portal. Layer: “ChristchurchCityUrbanExtent” through WFS dataset sharing accessible by clicking here
Christchurch City Council (2018)
Polygon
Land use activity boundaries as identified in the District Plan. The zone is defined by type and code
Canterbury City Council Geospatial Public Portal. Layer: “Zone” through WFS dataset sharing accessible by clicking here

Data Sources of Dwelling Counts

Data Provider
Data Type
Description
Source
Statistics New Zealand (Stats NZ) (2018)
Polygon
Occupied dwellings, unoccupied dwellings, and dwellings under construction, for private and non-private dwellings, 2006, 2013, and 2018 Censuses (RC, TA, SA2, DHB)
Published in DataFinder, commissioned by Stats NZ and accessible by clicking here

Summary of Results

With the following example objectives, using a balanced weighting scheme for a high dwelling projection

Icons of the example objective functions for Christchurch
Densities of Christchurch (2018) Densities of Christchurch envisioned to be more sustainable by the genetic algorithm
The height and colour of the extruded statistical areas indicate the relative urban densities; not the height of structures

Figures and Illustrations

Figures from the Report

Note: Clicking on any image will enlarge it.

Figure 1. Proportions of existing urban densities of Christchurch in 2018, by statistical area, indicating where different transport methods can be supported as outlined by Chakrabarti (2013). Note: A 3D interactive spatial plot of existing densities can be found here

Figure 2. Computational flowchart of the genetic algorithm used to implement the multi-objectional spatial optimisation framework. (Modified from Caparros-Midwood et al., 2016).

Figure 3. Demonstration of the Pareto front for two objectives. (Reproduced from Wang et al., 2015).

Figure 4. Parametrized spatial dataset for each objective function of the Ōtautahi Christchurch case study. A darker shade of red indicates that the statistical area has a high objective function score.

Figure 5. Performance of Pareto-optimal spatial plans that dominate in one objective across all objectives. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Figure 6. Ranked Pareto-optimal development sites. Darker blue signifies where statistical areas appeared more often in the MOPO sets. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Figure 7. Spatial variability of envisioned urban densities of Ōtautahi Christchurch, by statistical area, where the height and colour of the extruded statistical areas indicate relative urban density. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection). Note: A 3D interactive spatial plot of envisioned densities can be found here

Figure A1. Scatter plot of every spatial development plan’s fitness in two competing objective functions analysed in the entirety of the genetic algorithm for the Ōtautahi Christchurch case study. Highlighted is the Pareto-optimal plans along the Pareto-front curve. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Supplementary Figures

Note: Clicking on any image will enlarge it.

Figure S1. Combined map of overall objective scores of each statistical area in Ōtautahi Christchurch, using a balanced weighting scheme between objective functions. A darker shade of red indicates a higher total objective function score.

Figure S2. Pareto fronts. Each set of axis compares one of the objective functions to the other five. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Figure S3. Spatial plots of development plans in the MOPO set for the Ōtautahi Christchurch case study. Each plot represents the development plan that achieved the lowest score in the respective objective. (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Figure S4. All development sites for parent sets at selected generations for the Ōtautahi Christchurch case study (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Figure S5. The statistical areas that most commonly appeared in the top 1% (of overall combined objective score) of  spatial development plans for the Ōtautahi Christchurch case study (Parents = 1000, Generations = 200, Balanced weightings, High dwelling projection)

Watch this space!

Further case-studies with more comprehensive weightings, objectives and accuracy are coming soon!