Sustainable land use optimization using Boundary-based Fast Genetic Algorithm Kai Cao a,b,c,⇑ , Bo Huang a , Shaowen Wang c,d , Hui Lin e a Department of Geography and Resource Management, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong b Center for Geographic Analysis, Harvard University, 1737 Cambridge Street, Cambridge, MA 02138, USA c CyberInfrastructure and Geospatial Information Laboratory, Department of Geography, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA d National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA e Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong article info Article history: Received 6 March 2010 Received in revised form 10 August 2011 Accepted 11 August 2011 Available online 22 September 2011 Keywords: Land use optimization Genetic algorithm Sustainability Spatial compactness Reference point Tongzhou Newtown abstract Under the notion of sustainable development, a heuristic method named as the Boundary-based Fast Genetic Algorithm (BFGA) is developed to search for optimal solutions to a land use allocation problem with multiple objectives and constraints. Plans are obtained based on the trade-off among economic ben- efit, environmental and ecological benefit, social equity including Gross Domestic Product (GDP), conver- sion cost, geological suitability, ecological suitability, accessibility, Not In My Back Yard (NIMBY) influence, compactness, and compatibility. These objectives and constraints are formulated into a Multi-objective Optimization of Land Use (MOLU) model based on a reference point method (i.e. goal pro- gramming). This paper demonstrates that the BFGA is effective by offering the possibility of searching over tens of thousands of plans for trade-off sets of non-dominated plans. This paper presents an appli- cation of the model to the Tongzhou Newtown in Beijing, China. The results clearly evince the potential of the model in a planning support process by generating suggested near-optimal planning scenarios con- sidering multi-objectives with different preferences. Published by Elsevier Ltd. 1. Introduction The well-known World Commission on Environment and Development (WCED, 1987) defined sustainability as ‘‘development that meets the needs of the present without compromising the ability of future generations to meet their own needs’’. This notion of sus- tainability refers to a specific type of development for the society. As the WCED Commission states ‘‘in essence, sustainable develop- ment is a process of change in which the exploitation of resources, the direction of investment, the orientation of technological develop- ment and institutional change all are in harmony’’. Land use alloca- tion, as a type of resource allocation, can be understood as the process of allocating different activities or uses (e.g., residential land, industries, recreational facility, and green land) to specific units of area within a geospatial context. Sustainability often rep- resents a primary goal for land use planning. Comprehensive sustainability in land use allocation can be de- fined as a long-term balance between economic development, envi- ronmental protection, efficient resource use, and social equity. To pursue this balance can be treated as a multi-objective optimization problem. Generally, there are two types of methods: Pareto-front-based (Balling, Brown, & Day, 1999) and weighted sum (Aerts, Herwijnen, & Stewart, 2003) for such multi-objective optimization. All multi-objective optimization models are based on one of these two types of methods. Both types have their advan- tages and disadvantages. The Pareto-front-based methods focus on exploiting the diversity of the solutions, but often have issues of inadequate efficiency and effectiveness; while the weighted sum methods are straightforward to implement with superior efficiency and effectiveness, but requires prior knowledge. The problem of land use allocation optimization is a complicated process as it involves determining not only what to do (selection of activities) and how much to do, but also where to do the selection. It also adds a whole extra class of variables to the problem when combined with the consideration of indispensable spatial optimiza- tion. The utility of optimization as a normative tool for spatial prob- lems is widely recognized (Church, 1999, 2002; Cromley & Hanink, 2003; Malczewski, 1999). The complexity of the problem is attrib- uted to the inclusion of multiple objectives that may not be linear or simple. The objectives within a spatial context must incorporate location information to all attributes which further increases the complexity of the problem. Moreover, geographic units and associ- ated neighboring features are not independent. Such complicated non-linear multi-objective optimization problems as a type of Non-deterministic Polynomial (NP ) hard problem require heuristic methods for executing optimization processes. Various heuristic algorithms have been developed, such as simu- lated annealing (SA) algorithm, ants algorithm, and genetic 0198-9715/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.compenvurbsys.2011.08.001 ⇑ Corresponding author at: CyberInfrastructure and Geospatial Information Laboratory (CIGI), Department of Geography, University of Illinois at Urbana- Champaign, Room 324 Davenport Hall, Campus Mail Code: MC-150, 607 South Mathews Avenue, Urbana, IL 61801, USA. E-mail address: kcao@illinois.edu (K. Cao). Computers, Environment and Urban Systems 36 (2012) 257–269 Contents lists available at SciVerse ScienceDirect Computers, Environment and Urban Systems journal homepage: www.elsevier.com/locate/compenvurbsys