**Author details**

238 Genetic Programming – New Approaches and Successful Applications

identified.

management problems.

enumerated as follows;

**6. Conclusion** 

models is run to predict the outcome of implementing the alternative management options. For example, the authors developed simulation-optimization models to develop optimal management decisions for coastal aquifers. The optimal pumping from the coastal aquifer can be decided only by considering the impact of any alternative pumping strategy on saltwater intrusion. For this the numerical simulation model needs to be integrated with the optimization algorithm and the impact of each candidate pumping strategy is predicted by using the simulation model iteratively. This involve a lot of computational burden as thousands of numerical model runs are required before an optimal pumping strategy is

GP was used a surrogate model within the optimization algorithm as a substitute of the numerical simulation model in our study (Sreekanth and Datta, 2010). Previous studies have used artificial neural networks as surrogate models to replace groundwater numerical simulation models. Emily et a1 (2005) used genetic programming based surrogate models for groundwater pollution source identification. In our study (Sreekanth and Datta, 2010), it was found that genetic programming could be used as a superior surrogate model in such application with definite advantages. The study intended to develop optimal pumping strategies for coastal aquifers in which the total pumping could be maximized and at the same time limiting the saltwater intrusion at pre-specified limits. In doing so, the effect of pumping on the salinity levels was predicted using trained and tested GP models. The GP models were externally coupled to a genetic algorithm based optimization model to derive the optimal management strategies. The results of the GP based simulation-optimization was then compared to the results obtained using an ANNbased simulation-optimization model. The ability of GP in parsimoniously identifying the model inputs helped in reducing the dimension of the decision space in which modelling and optimization was carried out. The smaller dimension of the modelling space helped in reducing the training and testing required to develop the surrogate models. The study identified that GP has potential applicability in developing surrogate models with potential application in simulation-optimization methodology to solve environmental

The aim of this chapter is to introduce genetic programming as a potential modelling tool for hydrology and water resources applications. Genetic programming belongs to the broad class of evolutionary computational tools developed in recent years. Compared to the vast number of data mining and artificial intelligence applications in hydrology and water resources, the application of GP has been limited in spite of its potential applicability in a wide range of modelling applications. This chapter illustrates a few applications of GP as a modelling tool in the broad area of water resources modelling and management. The studies have found GP to be a useful tool for such applications with some advantages over other artificial intelligence techniques. The major findings reported in this chapter are J. Sreekanth1,2,3 and Bithin Datta1,2 *1CSIRO Land and Water, Ecosciences Precinct, Australia 2Discipline of Civil and Environmental Engineering, School of Engineering and Physical Sciences, James Cook University, Townsville, Australia 3CRC for Contaminant Assessment and Remediation of the Environment, Mawson Lakes, Australia* 
