1.0 Executive summary1.1 IntroductionOverthe years, several numerical models have been applied to simulate flow andwater problems in different coastal environments. However, these models are inadequatelyuser-friendly and as a result lack knowledge transfers in model interpretation.The primary aim of this paper will be to review the development as well as thecurrent progress of AI being integrated into water quality modelling.1.2 Process and planning stageInitially,the need of AI integration was assessed in terms of current issues, reasons andtendency.
Currently, models require the use of model manipulation, which aimsto provide an adequate simulation, this poses a problem as it was found outthat many modellers do not posses this skill at an acceptable standard. AI techniquescan carry out the manipulations which gave sufficient reasoning to integratethe AI. After looking at the tendencies, it could be seen that current modellingtechniques do not incorporate few features to facilitate other users andmethods. Identifying these parameters in the planning stage provided a set ofcriteria which would need to be satisfied by the AI in the experimental stage. 1.
3 Assessing the AI techniquesFollowingthe identification of the criteria, several different AI techniques were assessed.All AI techniques employed different algorithms and the different applications inwater quality testing were assessed to see which method would fit best for aspecific task. The four different AI methods employed were Knowledge-basedsystems (KBS), Genetic algorithms (GA), Artificial neural networks (ANN) andFuzzy inference systems. 1.4 ResultsThetable below shows the results that were obtained after the assessments werecompleted. The table indicates which AI technique is best fitted for a specificapplication. Technique Water quality applications Knowledge-based systems (KBS) Selection and manipulation of various numerical models on water quality.
Genetic Algorithms (GA) Optimization of calibration of the parameters of numerical models on water quality. Artificial Neural Networks (ANN) Determination of underlying physical/ biological relationships that are not fully understood. Fuzzy inference systems Quantification of the semantemes of the expertise and determine the confidence factors of the semantemes. Table 1- Different applications foreach AI technique1.5 Future DirectionsWhileall assessed AI methods are capable of carrying out specific tasks in relationto water quality testing, a more versatile approach would be to combine thesemethods together. Demand for more improved and more efficient AI techniqueswill likely rise and as a result research is being carried out to address thekey issues with current AI systems. These new tools will be developed to havemore user-friendly systems as well as clearer knowledge representation.
1.6 ConclusionThecurrent techniques used in water quality testing result in many constraints,and as a result those with little experience in the field will have difficultyselecting a suitable model. As such, improvements in AI technologies willminimisie the gap between a developer and practitioner.
The assessment of the variousAI techniques into modelling has demonstrated that they can contribute invarious aspects. Furthermore, it is possible that two or more of thesetechniques can be combined to produce a more versatile modelling system allowingfor a simpler experience for those new to the practice. Finally, research inthis field will certainly result in better AI systems which would enhance the applicationsof this AI in water quality modelling.