Today the use of computer models to investigate and solve hydraulic problems, simple or complex, almost goes without second thought. Although a relatively new discipline, the computer models encapsulate more than 2000 years of hydraulic engineering and fluid mechanics knowledge, which began with the Romans and ended with the development of numerical solution schemes in the 1960s. Since the 1960s considerable developments have continued in the design and refinement of numerical schemes and in software engineering. It is interesting and worthwhile to look at the evolution of this branch of engineering, chart its development and provide insight in to the future direction of where this challenging and exciting discipline is headed.



Mike B. Abbott (MIKE)


In 1991 Mike Abbott, one of the original founders of the Computational Hydraulics Centre (CHC) at the Danish Hydraulics Institute (DHI) back in the 1960s, presented a categorisation of computational modelling according to its technological level. Abbott et al (1991) described five generations of computational modelling:


1^st Generation – Computerised Formulae. The first generation has its origins in the 1950s and essentially consisted of the development of computer code aimed automating those methods of numerical calculation that were amenable to human computation. This generation thereby made undertaking those computations easier, quicker and more reliable to perform.


2^nd Generation – One-off Numerical Models. The second generation came to light in the 1960s. Second generation developments involved the construction, application and development of numerical models that were designed for one-off use within a specific study area. These models would have the description of the study area hard coded in to the numerical model and whilst being computer friendly (finite differences, finite elements, etc.) could be considered as user unfriendly! The modelling activities were primarily carried out by universities and research institutions with the modeller being the developer of the numerical model and having specialist skills not only in hydraulics but also in numerical methods, computer coding and computer systems (hardware and software).


3^rd Generation – Generalised Numerical Modelling Systems. The third generation focused on making the numerical model generic so that it could be applied repeatedly to different study applications. Study specific data would be provided to the generic numerical code (the engine) through simple input / output files. This was the era of the modelling system. The third generation modelling brought down the cost of modelling studies at the same time as making them more reliable. It was a facet of second and third generation modelling that these studies were undertaken mainly by developer-modellers employed by a select number of specialist engineering consultancies and academic and technological institutions, mainly in Europe and the USA. The term developer-modeller refers to a person or persons who not only undertakes the technical development of the numerical engine of the models but also applies it (with expert knowledge of the underlying computation hydraulics engines) to practical applications.


4^th Generation - The Industrial User-friendly Software Product. During the 1980’s, demand rose from the end users of modelling studies for access to the modelling systems themselves. At the same time the proliferation of personal computers and high-end workstations meant that the cost of computing power was brought down to such an extent that it became economic for end users to access these sophisticated modelling tools. As a result fourth generation modelling systems were developed. They are characterized as follows (Abbott et al, 1991):

• The modelling systems are designed for use by persons who are not computational hydraulics experts;
• A differentiation correspondingly occurs between the tool makers, who are computational hydraulics specialists, and the tool users, who are not;
• The fourth generation tool is a product, so that computational hydraulics specialists become makers of products, rather than performers of projects;
• The main purpose behind the development of 4^th generation models is to make computational hydraulics more useful to a wider range of end users; and
• Research to support the development of this generation of model is technological, unlike earlier generations which were supported by scientific research.

The main developments being undertaken on these 4th generation models is in the user interfaces. As customers require more and more features as well as technical support, a large organisation is needed, dedicated to sustaining and developing the software. With end users undertaking a considerable amount of the modelling work, the specialist developer-modeller is less prominent. This is one of the challenges the industry faces – that of ensuring operators of modelling software have the necessary background and experience in computational hydraulics and numerical methods.

The 4^th generation modeller these days is typically not dedicated solely to the modelling task alone. Within public organizations, at least in New Zealand, few have dedicated modelling teams. More typically modelling is seen as a tool that needs to be used from time to time to support risk assessments of existing infrastructure, or as a basis for capital works planning and design. Several engineering consultants have managed to set up permanent and skilled modelling teams to support the multidisciplinary activities of their companies.


5^th Generation – Intelligent Modelling System. Numerical modelling is today a well accepted technology and is part of most consulting engineers’ toolbox. Indeed it would be unusual to find any substantial hydraulic investigation being undertaken without any modelling. We have arrived at this point due to the rapid dissemination of 4^th generation models into mainstream engineering consultancies which began more than 20 years ago. We are now entering the next generation: 5^th generation modelling. The main characteristics of the 5^th generation as hypothesised by Abbot et al (1991) are:

• Fifth generation models will be dependent upon expert advice serving facilities and equivalent intelligent interrogation facilities;
• While 4^th generation tools provide the modeller with a set of tools, these are “closed”, as the engineer only parameterizes these tools. Fifth generation modelling will provide the engineer, chemist, biologist or sedimentologist with a more open working environment, where the user is able to assemble his or her own tool sets;
• 5^th generation models will link together many diverse elements, including object oriented and structured databases, graphics facilities in addition to the core computational engines; and
• 5^th generation models will be used in a much broader sense, for diagnostic systems, planning and scheduling systems, on-line control systems and more.

The term hydroinformatics has been accepted as meaning the application of information technology to hydraulics and water resources, of which 4^th and 5^th generation modelling systems form a core part. Written more than 15 years ago, the prediction of the 5^th generation model is beginning to be seen today. The porting of common software to state of the art software platforms (e.g. MOUSE to MIKE URBAN) resulting in the development of close integration of GIS, asset management databases and openness to decision support frameworks is an excellent example.

While the use of expert systems to guide the modeller may be still some way off, there is definitely a move to expose models to wider audiences than has been seen before. An example is the use of models in Decision Support Systems (DSS) as described by Shipton et al (2008). Up to now model outputs have required interpretation and presentation by the modeller to the end client, who therefore only sees what is presented. A DSS can provide a “shell” through which non-expert users can dynamically interact with a model to take advantage of the enormous amounts of information that can be generated by a well validated model. Hence non-expert users may gain an improved understanding of existing systems or to try “what if” scenarios of proposed interventions.



The 5th Generation - Decision Support Systems Built Around Models and Data Sources


4^th generation models tend to have been designed for a single purpose only, e.g. MIKE 11 for open channels, MOUSE for sewers, and MIKE 21 for inland flooding and coastal waters. In the transition from 4^th to 5^th generation modelling, the past 5 years has seen the successful integration of different types of model which leverage the power of each type to produce a hybrid which offers the advantages of the individual components. Hence MIKE FLOOD incorporates MIKE 11, MIKE 21 and MIKE URBAN. Other commercial vendors have followed the trend, and with the continued increase in computing power, use of these models for undertaking 2D urban or rural flood studies has now been commonplace for the past few years.

Today the linking of different components from different vendors’ software is also possible through the Open Modelling Interface (OpenMI), a joint EU funded development which includes DHI, Deltares (formerly WL Delft) and Wallingford Software. Hence the “tool assembly” technology Abbott refers to is in some way already available. The OpenMI environment also allows modellers to take on aspects of the developer role again e.g. with OpenMI one can dynamically couple a propriety model such as MIKE SHE with one’s own code as a DLL.



OpenMI Founding Organisations


Other 5^th generation modelling developments beginning to be implemented are:

Agent based modelling – whereby the behavioural characteristics of aquatic species are included in the model (e.g. ECOLAB) enabling an assessment of the effects on behavioural patterns of such species resulting from changes in the eco-hydraulic regimes to be undertaken.



Agent Based Modelling – River and flood plain showing habitat conditions and species movement (white line)


Online forecasting – models of urban water supply, wastewater collection and river networks are being established to run in a forecasting model which provide operators and system managers with continuous predictions of the performance of the system. 5^th generation modelling also allows operators to carry out “what if” scenarios and update the prototype system configuration in real time.

For further information of 5^th generation modelling please contact DHI (info.nz@dhigroup.com)


References

• Abbott, M.B., Havnø, K. and Lindberg, S. (1991); “The Fourth Generation of Modelling in Hydraulics”, Journal of Hydraulics Research, 29 (5), 581-600.
• Shipton, S., Breen, S., van der Linde, J. and Wood, M (2008).; “Use of Models in Decision Support – Case Studies in Urban Reticulation and Flood Forecasting”, NZWWA 50th Anniversary Conference, Christchurch, 2008.