**2.3. Optimization algorithms**

Of the various simulated annealing implementations, it is evident in literature that the adaptive simulated annealing algorithm converges faster [16] while maintaining the reliability of results and hence it was preferred over traditional Boltzmann annealing implementation [19]. Its application to the unknown pollutant source identification has been limited but it is potentially a good alternative because its convergence curve is steep, thereby producing better results when execution time is limited.

Currently, the most widely used optimization algorithm for solving groundwater source identification problem using linked simulation-optimization model is Genetic Algorithm and its variants. The effectiveness of ASA in solving this problem is compared against the effectiveness of GA. Genetic algorithms (GAs) are population based search strategies which are popular for many difficult to solve optimization problems including inverse problems. GAs emulate the natural evolutionary process in a population where the fittest survive and reproduce [12]. GA-based search performs well because of its ability to combine aspects of solutions from different parts of the search space. Real coded genetic algorithm was used with a population size of 100, crossover probability of 0.85 and a mutation probability of 0.05. The values were chosen based on a series of numerical experiments.
