Spatially optimised sustainable urban development

Abstract

Tackling urbanisation and climate change requires more sustainable and resilient cities, which in turn will require planners to develop a portfolio of measures to manage climate risks such as flooding, meet energy and greenhouse gas reduction targets, and prioritise development on brownfield sites to preserve greenspace. However, the policies, strategies and measures put in place to meet such objectives can frequently conflict with each other or deliver unintended consequences, hampering long-term sustainability. For example, the densification of cities in order to reduce transport energy use can increase urban heat island effects and surface water flooding from extreme rainfall events. In order to make coherent decisions in the presence of such complex multi-dimensional spatial conflicts, urban planners require sophisticated planning tools to identify and manage potential trade-offs between the spatial strategies necessary to deliver sustainability. To achieve this aim, this research has developed a multi-objective spatial optimisation framework for the spatial planning of new residential development within cities. The implemented framework develops spatial strategies of required new residential development that minimize conflicts between multiple sustainability objectives as a result of planning policy and climate change related hazards. Five key sustainability objectives have been investigated, namely; (i) minimizing risk from heat waves, (ii) minimizing the risk from flood events, (iii) minimizing travel costs in order to reduce transport emissions, (iv) minimizing urban sprawl and (v) preventing development on existing greenspace. A review identified two optimisation algorithms as suitable for this task. Simulated Annealing (SA) is a traditional optimisation algorithm that uses a probabilistic approach to seek out a global optima by iteratively assessing a wide range of spatial configurations against the objectives under consideration. Gradual ‘cooling’, or reducing the probability of jumping to a different region of the objective space, helps the SA to converge on globally optimal spatial patterns. Genetic Algorithms (GA) evolve successive generations of solutions, by both recombining attributes and randomly mutating previous generations of solutions, to search for and converge towards superior spatial strategies. The framework works towards, and outputs, a series of Pareto-optimal spatial plans that outperform all other plans in at least one objective. This approach allows for a range of best trade-off plans for planners to choose from. ii Both SA and GA were evaluated for an initial case study in Middlesbrough, in the North East of England, and were able to identify strategies which significantly improve upon the local authority’s development plan. For example, the GA approach is able to identify a spatial strategy that reduces the travel to work distance between new development and the central business district by 77.5% whilst nullifying the flood risk to the new development. A comparison of the two optimisation approaches for the Middlesbrough case study revealed that the GA is the more effective approach. The GA is more able to escape local optima and on average outperforms the SA by 56% in in the Pareto fronts discovered whilst discovering double the number of multi-objective Pareto-optimal spatial plans. On the basis of the initial Middlesbrough case study the GA approach was applied to the significantly larger, and more computationally complex, problem of optimising spatial development plans for London in the UK – a total area of 1,572km2. The framework identified optimal strategies in less than 400 generations. The analysis showed, for example, strategies that provide the lowest heat risk (compared to the feasible spatial plans found) can be achieved whilst also using 85% brownfield land to locate new development. The framework was further extended to investigate the impact of different development and density regulations. This enabled the identification of optimised strategies, albeit at lower building density, that completely prevent any increase in urban sprawl whilst also improving the heat risk objective by 60% against a business as usual development strategy. Conversely by restricting development to brownfield the ability of the spatial plan to optimise future heat risk is reduced by 55.6% against the business as usual development strategy. The results of both case studies demonstrate the potential of spatial optimisation to provide planners with optimal spatial plans in the presence of conflicting sustainability objectives. The resulting diagnostic information provides an analytical appreciation of the sensitivity between conflicts and therefore the overall robustness of a plan to uncertainty. With the inclusion of further objectives, and qualitative information unsuitable for this type of analysis, spatial optimization can constitute a powerful decision support tool to help planners to identify spatial development strategies that satisfy multiple sustainability objectives and provide an evidence base for better decision making.