Abstract :
Genetic algorithms are globally oriented in searching and thus potentially useful in solving optimisation problems in which the objective function responses contain multiple optima and other irregularities. The usefulness of genetic algorithms in calibrating environmental models was investigated in the context of calibrating rainfall-runoff models. A genetic algorithm was introduced and used to calibrate a conceptual rainfall-runoff model with nine parameters. A hypothetical example, in which the true optimum set of parameter values was known by assumption, was used to examine whether the genetic algorithm was capable of finding that optimum. The performance of the genetic algorithm in model parameter calibration was then studied using real data from four catchments. The genetic algorithm was always able to find an objective function value close to the global minimum. In some runs, the search landed at a local optimum, but this happened only when the objective function value of the local optimum was similar to that of the global optimum. A combination of an initial search using the genetic algorithm and fine tuning using a standard search technique was shown to perform very effectively.
