**7. Conclusion**

A linked simulation-optimization method for source identification was developed based on adaptive simulated annealing. It was applied to an illustrative study area. The results obtained were compared with those obtained using genetic algorithm, a more commonly used optimization approach. It is evident from the limited numerical experiments that adaptive simulated annealing algorithm based solutions converge to the actual source fluxes faster than genetic algorithm based solutions. This results in substantial saving in computational time. The source fluxes identified by using adaptive simulated annealing are closer to actual fluxes when compared to the results obtained using genetic algorithm, even when the observation data are erroneous and the hydro-geological parameters are uncertain. It can be concluded that adaptive simulated annealing is computationally more efficient for use in simulation-optimization based methods for identification of unknown groundwater pollutant sources, specially in a time constrained environment. Use of ASA has the potential to reduce CPU time required for solution by an order of magnitude. However, with very large number of iterations in the linked simulation-optimization approach, it is possible that the solutions obtained using GA could converge to a marginally better solution compared to that ASA based algorithm. However, it appears that ASA based solutions converge very close to the optimal solution using only a small fraction of iterations required while using GA. The relevance of contaminant monitoring locations is demonstrated. Further studies are required to develop dedicated monitoring networks which can increase the efficiency of source identification.
