ABSTRACT: The spatial distribution of hydrogeologic parameter is an important issue in ground water simulation. One of the methods is to divide an area into several zones such that parameters are assumed to be constant within zone. The purpose of this study is to apply Tabu Search (TS) to find the best zonation of parameters that can result in the best ground water simulation. The initial zonation can be determined as the Thiessen method, and then zonation is optimized by TS. The mean square error between simulated and observed hydraulic heads was used as the objective function. A designed confined aquifer with known zonation was used as an example to test the proposed method. Results indicated that Tabu Search can locate the optimal zonation successfully and avoid being trapped by local optimal zonations. Besides this, four other arbi- trary initial zonations can be directed to the optimal zonation by TS, which proves the robustness of the proposed method. The method proposed in this study is feasible and expected to work well in the field problems with sufficient sampling of concerned parame- ters. (KEY TERMS: water resources planning; modeling; groundwater hydrology; optimization; simulation.) INTRODUCTION Ground water simulation models play important roles in ground water management. The models need the spatial distribution of hydrogeologic parameters as input. Ground water simulation models are often developed to solve differential equations by numerical methods that normally divide an area into a grid net- work and require hydrogeologic parameters for each grid point as input. However, the hydrogeologic char- acteristics are seldom homogeneous in an area. Thus, it is not realistic to use the homogeneous assumption for those parameters for the whole study area. Hydro- geologic parameters such as hydraulic conductivity can be measured from well testing. It is impossible to have the observed values of parameters for all grids in a region. Thus, modelers have to perform the spa- tial analysis for the hydrogeologic parameters before running simulations. There are different methods for parameter identifi- cation. If there is no reference parameter available, inverse analysis (Yeh, 1975) could be applied to deter- mine parameters. On the other hand, if parameters of some locations in a study site are known, a spatial interpolation method such as the Kriging method or a zonation method such as the Thiessen method can be applied. The Kriging algorithm requires specialized training. Although some softwares provide the Krig- ing algorithm to interpolate parameters, inadequate application may be made without carefully examine whether assumptions of the algorithm are met. On the other hand, the Thiessen method gives geometric zonation with few physical meanings, and it does not guarantee the best simulation results. Yang and Yeh (1998) assigned ground water parameters according to geologic characteristics. However, spatial geology entails some uncertainty in this approach. Heuristic algorithms are popular for optimizing nonlinear problems in recent years. The most common algorithms include Tabu Search, Simulated Anneal- ing, Genetic Algorithm, etc. They have been applied to different research areas (e.g., Battiti and Tecchiolli, 1994; Dell’Amico and Trubian, 1993; Dougherty and Marryott, 1991; Gendreau et al., 1994; Zheng and Wang, 1996, 1999). Tabu Search was selected for this study, and other heuristic algorithms are expected to be comparable to TS. However, determining which heuristic algorithm is most efficient in the problem of 1 Paper No. 01187 of the Journal of the American Water Resources Association. Discussions are open until February 1, 2003. 2 Respectively, Associate Professor and Graduate Student, Department of Bioenvironmental Systems Engineering, National Taiwan Uni- versity, No. 1 Roosevelt Road, Sec. 4, Taipei, Taiwan 106 (E-Mail/Tung: cptung@ccms.ntu.edu.tw). JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 1115 JAWRA JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION VOL. 38, NO. 4 AMERICAN WATER RESOURCES ASSOCIATION AUGUST 2002 APPLICATION OF TABU SEARCH TO GROUND WATER PARAMETER ZONATION 1 Ching-Pin Tung and Chun-An Chou 2