Engineering Applications of Artificial Intelligence 16 (2003) 543–554 An integrated multi-criteria decision analysis and inexact mixed integer linear programming approach for solid waste management S. Cheng a , C.W. Chan b, *, G.H. Huang c a Faculty of Engineering, University of Regina, Regina, Saskatchewan, Canada, S4S 0A2 b Department of Computer Science/Energy Informatics Laboratory, University of Regina, Regina, Saskatchewan, Canada, S4S 0A2 c Faculty of Engineering/Energy Informatics Laboratory, University of Regina, Regina, Saskatchewan, Canada, S4S 0A2 Abstract This paper reports on an integration of multi-criteria decision analysis (MCDA) and inexact mixed integer linear programming (IMILP) methods to support selection of an optimal landfill site and a waste-flow-allocation pattern such that the total system cost can be minimized. Selection of a landfill site involves both qualitative and quantitative criteria and heuristics. In order to select the best landfill location, it is often necessary to compromise among possibly conflicting tangible and intangible factors. Different multi-objective programming models have been proposed to solve the problem. A weakness with the different multi-objective programming models used to solve the problem is that they are basically mathematical and ignore qualitative and often subjective considerations such as the risk of groundwater pollution as well as other environmental and socio-economic factors which are important in landfill selection. The selection problem also involves a change in allocation pattern of waste-flows required by construction of a new landfill. A waste flow refers to the routine of transferring waste from one location in a city to another. In selection of landfill locations, decision makers need to consider both the potential sites that should be used as well as the allocation pattern of the waste-flow at different periods of time. This paper reports on our findings in applying an integrated IMILP/MCDA approach for solving the solid waste management problem in a prairie city. The five MCDA methods of simple weighted addition, weighted product, co-operative game theory, TOPSIS, and complementary ELECTRE are adopted to evaluate the landfill site alternatives considered in the solid waste management problem, and results from the evaluation process are presented. r 2003 Elsevier Ltd. All rights reserved. Keywords: Decision-support; Environmental impact; Multi-criteria decision analysis; Landfill selection 1. Introduction A solid waste management program often involves conflicting economical, environmental, and socio-ecolo- gical impacts. For example, locating a new site for landfill development at minimal cost is feasible, but the tradeoff could be the likelihood of groundwater pollu- tion. The question then arises as to how the decision maker can reach a compromise among the conflicting impacts and select the optimal landfill location. The landfill selection problems have often been tackled using multi-criteria decision analysis (MCDA). For example, Hipel (1982) proposed an early version of multi-criteria modelling that incorporated fuzzy set theory and applied the method to a solid-waste disposal problem in Canada. Chen et al. (1997) developed the fuzzy DRASTIC for landfill siting which used a geological information system (GIS) to evaluate potential sites in Taiwan. A different approach is adopted by Hokkanen et al. (1997) who applied a MCDA method called PROMETHEE for facility allocation. Hokkanen et al. (1994) and Vuk et al. (1991) also demonstrated effectiveness of using MCDA methods to solve the selection problem in solid-waste management. In this paper, we present our work in locating a new and optimal landfill site in the city of Regina that would cause the least negative economical and socio-environ- mental impacts. Multi-objective linear programming (MOLP) models are applicable for handling this type of problems, but the complexity of constructing a MOLP model and the lengthy computation involved render such models inconvenient. A weaknesses in MOLP modelling is that after the result from MOLP has been generated, further analysis may still be ARTICLE IN PRESS *Corresponding author. Tel.: +1-306-585-5225; fax: +1-306-585- 4745. E-mail address: chan@cs.uregina.ca (C.W. Chan). 0952-1976/03/$-see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0952-1976(03)00069-1