A comparison of combinatory methods and GIS based MOLA (IDRISI®) for solving multi-objective land use assessment and allocation problems

Abstract

The aim of this study was to provide an informed choice among two combinatory methods and GIS based MOLA module in IDRISI® by comparing their performance in solving a hypothetical Multi-Objective Land use Assessment and Allocation (MOLAA)problem. Among the combinatory methods, Simulated Annealing and Tabu Search algorithms were chosen for study. The application of Simulated Annealing has already been demonstrated in solving a MOLAA problem but Tabu Search has not been used to a MOLAA problem before. The Kioloa Region of New South Wales, Australia was chosen for designing a hypothetical MOLAA problem due to availability and access to the digital datasets at the Australian National University. The MOLAA problem was formulated for accomplishing six land use objectives by allocating the area to four land use types, that is, conservation, agriculture, forestry and development, using altogether 1 7 criteria, including 16 factors and one constraint. The criteria maps were classified in ordinal, continuous and fuzzy scale and combined by using Weighted Linear Combination to produce land use suitability models for each land use type. The ordinal and continuous land use suitability models were used in solving the problem by applying the MOLA module. In order to apply the combinatory methods, all three land use suitability models, that is, ordinal, continuous and fuzzy, were transferred to cost suitability models where the lowest cost value represented the best suitability and the highest cost value represented the lowest suitability in the interval data set. Three initial input solutions generated by the random, cheapest and greatest difference methods were used for optimising by applying both algorithms. Both combinatory methods maximized overall land use suitability with better spatial compactness by allocating each land unit with the most suitable land use with the lowest cost. At the land use level, MOLA exhibited a bias towards land uses with lower area requirement and allocates more suitable land units to them. Though the MOLA module is highly efficient in solving large grid MOLAA problem, the combinatory methods deliver a solution close to the near-optimal solution with better compactness in an acceptable time frame. Hence, the combinatory methods have been shown to be appropriate choice to solve MOLAA problems. The solutions were not significantly different at their mean cost functions between Simulated Annealing and Tahu Search at the appropriate parameters. Among the cost suitability models, both algorithms performed better in the fuzzy models in the large MOLAA problem. The initial input solution influenced the performance of the algorithms. The algorithms produced better results in the cheapest and greatest difference initial input solution in the medium grid MOLAA problem whereas the cost function was more improved using the random initial input solution in the large grid. Although there is no significant difference in the mean cost functions between Simulated Annealing and Tahu Search, the previous one is found more efficient in solving large grid MOLAA problem. For the same values of compactness factors, Simulated Annealing produced more spatially compact land use allocation than Tahu Search. Thus decision makers/land use planners or consultants could obtain a better decision alternative to a land use allocation problem by applying Simulated Annealing with the recommended appropriate annealing schedule and initial input cost suitability model. This study recommends further research in Tahu Search to find an effective attribute for a Tahu list, to be applied to a MO LAA problem.