An Artificial Immune System for multi-objective land use allocation (AIS-MOLA) Kangning Huang, Xiaoping Liu, Xia Li and Jiayong Liang School of Geography and Planning, Sun Yat-sen University, Guangzhou, RP China Keywords: multisite land use allocation (MLUA), multiobjective optimization, Pareto font, artificial immune system; Objective: An Artificial Immune System for multi-objective land use allocation (AIS-MOLA) is developed to find the Pareto-optimal alternatives of multisite land use allocation problem. Background: Multisite Land Use Allocation Problem (MLUA) [1] refers to the optimal allocation of different types of land use units to a specified region, which is proved to be an NP Complete problem. It often involves multiple conflicting objectives, such as suitability of land use, cost of development and compactness of the allocated areas. Recent researches emphasize the capability of Pareto-optimal alternatives [2] that it can well represent the possible trade-off among those conflicting objectives. However, instead of seeking the evenly distributed set of Pareto-optimal alternatives, most of the existing MLUA solutions [1, 3-5] simplify the problem by optimizing the linear weighted sum of those objectives. Because these simplifications disregard the concave parts of the Pareto font, they inevitably fail to produce well distributed solutions [6]. Data: The proposed approach is implemented in the case study of land use allocation in Panyu, a region of Guangzhou, RP China to test its validity. This specific allocation problem includes two major conflicting objectives – the spatial suitability and compactness of the allocated land use area. Methodology: Many Immune Algorithms in literature are proposed for multiobjective optimization[7], including MISA, IDCMA, VAIS and NNIA. To apply Immune Algorithm in MLUA, several significant modifications should been made. First, a novel crossover operator is employed to generate better solutions. Further, a heuristic hyper mutation according to the distribution of current Pareto font is introduced to improve the efficiency of the algorithm. Finally, the nondominated neighbor-based selection and proportional cloning method make the immune system focus more on the less-explored region of the objective space. Result and Discussion: Figure 1 shows the spatial suitability of the four convertible land uses: agriculture, industry, commerce and residence. The proposed algorithm was applied to simultaneously optimize the spatial suitability and compactness of land use in the study area. It took AIS- MOLA about 1 hour and 20 minutes to generate a satisfactory set of Pareto alternatives, which are shown in Figure 2. Moreover, the land use patterns of those solutions labeled as (a)-(d) are shown in Figure 3. As can be seen from Figure 2, under different conditions, the improvement of the spatial suitability will result in different degrees of impairment on the spatial compactness. The distribution of Pareto alternatives successfully revealed the complex trade-offs between these two objectives. Figure 1. Spatial suitability of the four land uses: (a) agriculture, (b) industry, (c) commerce and (d) residence. Figure 2. Pareto alternatives of MLUA problem of Panyu City generated by AIS-MOLA. The land use patterns of those solutions labeled as (a)-(d) are shown in Figure 3.