Conclusions

The OpenMI-Life project was successful in demonstrating that:

• In general, improvements in results could be obtained compared to the current standalone models.
• Integrated modelling provides a better insight to and understanding of the interactions between different water domains and the necessary data exchange.
• Modellers obtain a better understanding of the potential and limitations of their own models and the associated models in establishing linked simulations.

Although some specific integrated modelling frameworks may currently have better performances, the OpenMI offers a unique technology standard that allows the linking of legacy models from different organisations in their own software and hardware environments. Partners are committed to participating in the further implementation and improvements of the OpenMI in future projects, in the short, medium or long term. A small number of practical (technical, organisational and legal) issues still need to be overcome, but the partners are looking at the possibilities of future applications, such as extensions of the above use cases and new applications. The project provided a good opportunity for the different partners (competent authorities, model users, model developers and the OpenMI Association) to collaborate and learn from each other's practical organisation.

The following recommendations can be made for a successful further implementation of the OpenMI:

• Before embarking on a practical application, it is important that all partners involved make a thorough investigation of the specific needs for linking and the potential improvement that is expected from an integrated approach. This will avoid the situation where scarce resources are spent on less interesting cases.
• A number of technical issues remain to be solved or improved, e.g. user interfaces (link configuration and temporal data operations), performance (multi-threading), remote linking.
• Where possible, the models and the interface should be prepared consistently from an early stage in order to minimise the need for modification and the cost when linking them up.
• Guidance documents and best practice manuals should be composed from as broad a user community as possible, e.g. to gain experience on numeric stability issues and the use of more efficient geospatial linking.
• Partners need to strive for high-level collaboration agreements in order to minimise or overcome practical institutional and legal barriers; agreements must be sought with software providers to guarantee technical support and to resolve licensing issues.

Use Case A

Use Case A demonstrates the added value of the OpenMI when linking hydraulic sewer and river models with a view to optimising investments and the operational strategies of the sewerage and river managers.
Flows in the sewers and rivers are often interdependent; changes to the sewer network (or to its operating policy) can impact river flows, and hence the flood risk. Further on, the sewer network changes can alter river water quality. Similarly, changes to the controls or operating policy of the rivers, usually for better flood management, change the water level in the rivers, which can in turn affect the levels in the sewers and possibly change the frequency of overflow spills.

Figure 3: River Dijle and tributary Voer around drainage area of Leuven

The use case is situated in the Dijle river basin and more specifically around the town of Leuven. The study area comprises the larger drainage area of the town of Leuven (situated approximately 20 km east of Brussels). The central wastewater treatment plant of Leuven has a capacity of about 120,000 population equivalents. In this area there is a very strong interaction between the sewer system and the watercourses, mainly as a result of the complex branching of the river Dijle on its passage through the town centre of Leuven. In an area of about 75 km² no less than 180 sewer discharge points were identified along 50 km of watercourses. The majority of the sewer system is still combined, with a high number of combined sewer overflows (CSOs), although in some areas quite a number of separate (storm) sewers have been constructed during the last five to ten years.

The partners involved are Aquafin and the Division Operational Water Management of the Flemish Environment Agency (VMM-AOW). Both partners use InfoWorks (CS and RS respectively) as their modelling software, but as a result of different objectives and working procedures, the former standalone models were not really compatible or geared to one another.

Three scenarios were investigated in Use Case A:

• The first (or base) scenario linked the two models as they were available at the start of the demonstration phase. These models represent the situation as at mid-2008. This scenario was used to analyse differences between standalone and integrated modelling.
• The second scenario involved a different version of the CS sewer model. In this scenario, additional storm water storage/buffering volumes were introduced at major outfalls of the sewer system. This scenario was compared to the base scenario (with both in linked mode).
• The third scenario involved an update of the CS sewer model to reflect the situation as at mid-2009. No changes took place to the river system in the same period, but some new boundary nodes had to be introduced to allow linking with the new CS model. This scenario was also compared to the base scenario (with both in linked mode).

The first scenario showed significant differences in the behaviour of both the sewer and river systems. Especially in the smaller tributaries, the impact of discharges from the sewer system can be very significantly different between simulations from the linked models and those of the standalone sewer models. The simulations from the linked models gave a better insight to the validity of predefined boundary conditions that are commonly used with standalone sewer models. The second and third scenarios confirmed the impact that changes in one system (sewer) may have on the other system (river) and this stresses the need for a more integrated evaluation of strategic investments.

Overall the OpenMI proved to be a valid and useful technology for the linking of sewer and river models. Even though some technical and conceptual issues (integrated calibration, consistency of input data, performance and stability) remain to be resolved or improved, it was felt that a real-life continuation of this use case could be initiated in a very short time.

Use Case B

Use Case B demonstrates the added value of the OpenMI in linking an upstream non-tidal river model with a downstream tidal river model that were set up using different modelling software. The partners involved are the Division Operational Water Management of the Flemish Environment Agency (VMM-AOW), the competent authority for groundwater and the category 1 un-navigable watercourses, and Flanders Hydraulics Research (FH), the competent authority for the navigable watercourses. The two river managers are responsible for different parts of the water system, which are interconnected. Each has their own models using different software relevant to their needs and goals. It is often not desirable or practical to combine the models into one large model. Linking provides the opportunity for integrated management of the whole river system, because the models are able to simulate the impact of the management of the other part of the river system.

Figure 4: Location and extent of the models in Use Case B within Flanders

The use case is situated on the Dijle river. The downstream model is a tidal river model of the Dijle-Scheldt running from Werchter up to Terneuzen (the upstream Scheldt is cut off in Temse); the model is from Flanders Hydraulics and is built using MIKE11 (DHI software). The upstream non-tidal river models are the Dijle model that runs up to Werchter (at the confluence of the rivers Dijle and Demer) and the Barebeek-Weesbeek model running up to the confluence with the Dijle river; these models from VMM-AOW are built using Infoworks RS (Wallingford Software).

A first interaction point is situated in Werchter, at the confluence of the main river Dijle and its tributary Demer. At this point, the river Dijle becomes a navigable watercourse and comes under the authority of FH. The upstream Dijle constitutes a relatively large watershed and contains two large retention basins and multiple control structures. All these objects are modelled in detail in the InfoWorks RS model but are not present in the upstream rainfall run-off boundary of the standalone MIKE11 model. Therefore linking provides a more detailed upstream boundary for the downstream MIKE11 Dijle-Scheldt model. On the other hand, linking provides the upstream model with a more correct, tidal downstream boundary.
Because the influence of the tides is very low at Werchter, a second InfoWorks RS model was linked to the Dijle-Scheldt model. The InfoWorks RS Barebeek-Weesbeek model contains the whole of the Barebeek and Weesbeek watershed and includes two retention basins as well as the non-return valve control structures at the confluence with the downstream Dijle. This model benefits more from the linking with the tidal river model because it is more under the influence of the tides. Three extra interaction points are located at the confluence with the river Dijle.

Three scenarios were investigated in Use Case B:

• The first scenario was performed during the trial phase and constituted a set of simulations representing different situations and combinations relating to inland discharge and tidal levels.
• In the second scenario, a single event from December 2007 was run to evaluate the results compared with measurements along the river.
• Finally, in the third scenario, a forecast simulation was performed to demonstrate the added value of the OpenMI in forecasting.

From the first scenario, it was clear that the OpenMI is a workable method for modelling interactions between models built using different software. The coupling of the models showed an impact on the results of all models. These first coupled runs also revealed some shortcomings in the boundary definitions of the standalone models. The second scenario showed that the input received from a linked model is an improvement on the standalone boundary conditions, resulting in improved results for both models. In the third scenario, the OpenMI was used to link models that are currently used for flood forecasting. The definition of boundaries for forecasting is not straightforward and OpenMI linking provided the models with better boundary conditions for the forecast period. The simulations showed a significant improvement in the forecasting results.

All in all, the OpenMI proved to be a usable method for linking the models, providing each one with improved boundary conditions and resulting in improved results for each model. It is feasible that the OpenMI can serve as an operational tool for integrated river modelling, making it possible to asses the effects of new operational management and new infrastructure on another part of the river.

Use Case C

Use Case C demonstrates the added value of the OpenMI in linking two river flow (hydraulic) models to a river quality model in the Dijle river basin with a view to optimising river flow and quality calculations and to assessing the impact of the flow regulation on water quality and the impact of water quality during flooding. The use case is situated in the river Dijle basin.
The partners involved are the Unit River Basin Management of the Flemish Environment Agency (VMM-Staf AD-IWB), the Division Operational Water Management of the Flemish Environment Agency (VMM-AOW), Flanders Hydraulics Research (FH) and the University of Liege - Aquapole (ULG). The two river flow models are the models mentioned in Use Case B; the river quality model is the Pegase model developed by the ULG and used by VMM-Staf AD-IWB.

In order to cope with the spatial differences between the different models, Use Case C has been limited to the Dijle river basin (see Figure 5 The basin of the rivers Dijle and Demer and the location of the common linked reaches on the river Dijle) with an area of 1.276 km² and a population of 560,000 inhabitants. The catchment of the river Demer, the most important tributary of the river Dijle, and all non-hydraulically calculated tributaries have been modelled by the Pegase model without linking to river flow models, in order to avoid the increase in the number of linked rivers. In this way, a new Pegase database has been built, geographically limited to the Dijle and Demer rivers basin; the MIKE11 model has been limited to the part of the river Dijle upstream of the city of Mechelen and a small part of the river Demer. The linking of the InfoWorks RS and Pegase models occurred only on the main river Dijle upstream the city of Leuven (rivers in olive green in the olive green box on Figure 5). MIKE11 has been linked to Pegase downstream of the city of Leuven on the main river Dijle (river in dark blue in the blue box on Figure 5).

Figure 5: The basin of the rivers Dijle and Demer and the location of the common linked reaches on the river Dijle

Several scenarios were investigated in Use Case C.

Link between InfoWorks RS and PEGASE

Type of exchange	Number of links	Simulation period
Unidirectionnal	6	Winter 2002-2003
Unidirectionnal	6	Summer 2003
Unidirectionnal	296	Year 2000

Link between MIKE11 and PEGASE

Type of exchange	Number of links	Simulation period
Unidirectionnal	6	Winter 2002-2003
Unidirectionnal	6	Summer 2003
Unidirectionnal	6	Year 2006
Unidirectionnal	360	Year 2000
Unidirectionnal	360	Year 2006
Bidirectionnal	6 + 17	Year 2006

The scenarios with unidirectional links at 6 points were studied in the demonstration phase; the other scenarios were studied in the test phase.

From these scenarios we learned:

• The river flow models, if calibrated, could be useful for optimising the river flow during high water conditions.
• The river flow models still have to be adapted to simulate low flow conditions.
• Due to unreasonably high run times, linking the models at all nodes is not feasible and doesn't give large improvements in results.
• Bidirectional linking, providing releases flow from Pegase to the river flow model, is only interesting when simulating river flow upstream in little rivers and with time series values of releases flow.
• River quality is not very sensitive to small changes in river flow.

Use Case D

Use Case D demonstrates the added value of the OpenMI in linking an upstream one-dimensional tidal river model and a downstream two-dimensional tidal estuary model, which were set up with different modelling software. The linking provides one or more downstream boundaries for the 1D model and one or more upstream boundaries for the 2D tidal models. Two different 2D models, Kustzuid (Waqua) and Zeekennis (Delft 3D), and their link with a 1D model (MIKE11) of the rivers Leie and Bovenscheldt have been investigated in this use case. The partners involved are Flanders Hydraulics Research and Deltares.

Figure 6 shows the interaction between the Leie and Bovenscheldt 1D model and the Kustzuid 2D model. The models overlap between Melle and Dendermonde. This is the upstream part of the Sea Scheldt. This interaction is studied in detail at operational level for flood forecasting.

Figure 6: Coupling of the Leie and Bovenscheldt model with the Kustzuid model

In addition, in Spring 2009, the participants of Use Case D saw an opportunity to investigate the coupling of two 2D models. By linking the 2D WAQUA tidal model with the 2D SWAN wave model, predictions can be made of the impact of waves on coast- or flood-protection works such as dikes and dams. This is especially interesting in storm and high tide situations.

The work of this additional project is on-going at the time of printing of this brochure. Each use case went through three phases :

• Definition phase: During this phase the existing models were screened and the properties and characteristics of the links to be achieved were technically described and analysed.
• Test phase: During this phase the various issues and potential problems described during the definition phase were tested and corrections and improvements to the implementation of the OpenMI were made.
• Demonstration phase: The goal of this phase was to use the linked models in the closest possible approximation to current operational modelling practice.

Four scenarios were investigated in Use Case D:

• Scenarios 1 and 2 were studied in the test phase, to assess the impact of the use of the OpenMI and to test whether the OpenMI was able to handle large models.
• Scenarios 3 and 4 related to the demonstration phase. In these scenarios the OpenMI was used on a semi-operational level, meaning that predictions were repeated for a historical period with different conditions for tides at the coast and inland flow.

Scenarios 1 and 2 showed that the OpenMI is capable of handling integrated modelling not only for simplified models but also for more complex and realistic physical models. In addition, the OpenMI clearly demonstrated key benefits: the OpenMI solely exchanges data between models and doesn't intervene in the simulation process itself; and the results from both models are only influenced by the setting of common boundary conditions.

For scenarios 3 and 4, standalone predictions using two models were done by running the simulations in an iterative way and the results were compared to those of the integrated method using the OpenMI. In the case of high inland discharges (scenario 3), the use of the OpenMI has a great advantage. The standalone simulations had to be repeated several times to obtain an accurate result. When using the OpenMI, only one simulation had to be performed to get a solution that was comparable to the final result of the iterative simulations.
In the case of a storm at sea (scenario 4), the use of the OpenMI has a smaller advantage, since the stable solution is reached from the first standalone simulation. Although, in general, one can state that the use of the OpenMI has an advantage, no separate runs have to be carried out and no time series has to be exchanged.

The difference in the scenarios lies in the level of interaction. With high inland discharges, the systems (the Upper Scheldt and the Leie, and the Sea Scheldt) have a significant interaction with each other, since the lock weir in Melle is down; in the scenario of a storm at sea, there is no real interaction between the systems, since the lock weir separates the Sea Scheldt river from the Upper Scheldt and the Leie rivers.