Conclusions and Recommendations

Results from all Use Cases

• Legacy and in-house models were successfully migrated in OpenMI
• All models were set up, calibrated and validated in the study area
• In most cases, the selected models were successfully linked and run using the Configuration Editor provided by the OpenMI Association Technical Committee
  • Visual Modflow (Use Case C) was limited by its migration to exchange data in OpenMI only with MIKE 11.
  • New compliance rules demand that developers provide an .xml file with exchange parameters and information so that interested researchers can see whether a specific model satisfies their needs
• Missing exchange variables and numerical error/instabilities at selected nodes were addressed by modifying the available model codes and adjusting internal timestep lengths

Building an OpenMI Community

• Improve the training of students/ future modellers in OpenMI
  • Introduce the use of OpenMI in courses to support teaching of integrated modelling (across disciplines)
  • Set up research focusing new avenues enabled by OpenMI (sustainability assessment, multi-modelling etc)
  • Use the OpenMI E-learning tool (upload new training material and download existing courses) ; expand into a collaborative learning environment.
• Increase the availability of experts capable of migrating models in OpenMI
  • Provide seminars and training for new developers
• Increase the availability of OpenMI compliant models
  • Migrate legacy and in-house models
  • Built the new models to be OpenMI compliant
  • Request partners and cooperating companies to migrate their models in OpenMI
  • Approach departments from different scientific fields, such as economics, public health, computer science and suggest the use of OpenMI
• Use OpenMI in research work, studies and papers
• Present insights/thoughts/results on OpenMI in conferences

Use Case A:
The effect of advection-dispersion on sewage effluent discharged daily in Pinios River (Thessaly, Greece)

The upper Pinios watershed

Use Case A involves the linking of a Rainfall Runoff to a Hydraulic to a Quality model component, in order to assess the impact of human activities (agricultural, industrial and municipal releases) on the water quality of the Pinios River. The current study focused only on BOD5 and DO concentrations. The simulated concentration values varied significantly in time and space depending on rainfall, river flow and pollutant discharge rates.

Two independent study groups from the National Technical University of Athens (NTUA) set up models from different developers to evaluate water quality along the upper 75 kilometers of Pinios, during different time periods and following their own model assumptions/limitations. At the end of the study, both groups ran their individual linked model schemes (1 & 2) for the same month in order to also compare their results.

	MIKE 11 (DHI)		RISH-1D (NTUA)	RISQ-1D (NTUA)	OTIS (USGS)
Type of model	NAM module (rainfall-runoff)	Hydraulic (including rainfall-runoff input)	Hydraulic	Water Quality	Water Quality
Compliance status	OpenMI compliant		Migrated by NTUA	Migrated by NTUA	Migrated by NTUA
Linked Scheme 1	X		X	X
Linked Scheme 2		X			X

The data sets supporting Pinios Use Case A were obtained from the records of Public Power Corporation, the Ministry of Environment, the Ministry of Development as well as from field surveys and discussion with local population. GIS, Matlab and Excel supported the data pre-processing and visualization steps that preceded the linking phase.

Linked Scheme 1 as represented in the Configuration Editor provided by the OpenMI Association Technical Committee

The main objectives/milestones of the study were the following:

• Migrate in-house and legacy models in OpenMI
• Set up, calibrate and validate the selected models in the study area
• Run the models separately and linked in OpenMI and evaluate the results (accuracy, time)
• Evaluate the impact of different plausible pollution scenarios at specific locations
• Compare the two linked scheme results

Testing and Pollution Scenarios
Separate and OpenMI linked runs of Schemes 1 and 2
WWTP failure upstream (14.5 km)
Untreated industrial releases at a downstream point
Increased diffused pollution due to increased agricultural and livestock production (and/or illegal landfills and malpractices)
Comparative analysis linked Scheme 1 and 2

Study Results

• Separate and OpenMI linked runs matched successfully each other in both Linked Schemes
• There was an additional time delay related to linked model runs, especially in the case of Linked Scheme 2. Extra calculations were added to the wrapper of the water quality model (OTIS) to overcome limitations from missing critical MIKE 11 variables (those variables were not provided as exchange parameters in OpenMI). That coding approach may have contributed to the significant time overhead of the linked runs
• The different initial assumptions, spatial model representations and model limitations led to different BOD estimations between Linked Scheme 1 and Linked Scheme 2, as observed during the Comparative model runs. Swapping models between different research groups needs careful thinking and cooperation
• The impact of the Pollution Scenarios was significant only during low flow conditions. At selected locations close to the incidents, the Pollution Scenarios could be responsible for changes in the Water Class category and require further actions in order to allow specific water uses

Use Case B: The impact of climate change on the reliability of a reservoir

Use Case B links the rainfall-runoff NAM module of MIKE-11 to a reservoir management model (RMM-NTUA) in order to assess the impact of climate change on the reliability of Smokovo reservoir (situated on the southwest part of the Thessaly Water District, Figure 6). During the second phase of this study, the benefits of a possible restoration scenario of drained lake Xiniada were considered by linking two reservoir management models, RMM-NTUA and RMM-NTUA (2), to MIKE 11 and to an additional "Rule" component that calculates the total available volume of the system and returns it back to each reservoir. Xiniada (~36.6 km²) is located approximately 20km to the east of the Smokovo reservoir and today serves mainly local irrigation needs.

The Smokovo reservoir watershed

The different water demands to be satisfied by Smokovo reservoir releases and water abstractions are shown in the following figure:

The Smokovo reservoir functions

The data supporting the analysis were obtained from the historical records of the Public Power Corporation, the Ministry of Environment and the Ministry of Agriculture.

The main objectives/ miliestones of the Smokovo study were the following:

The multi-reservoir scheme as represented in the Configuration Editor

Phase 1

• Modify and migrate the RMM-NTUA model and Rule component in OpenMI
• Set up, calibrate and validate MIKE 11 and RMM-NTUA in the study area
• Run the models separately and linked in OpenMI and evaluate their results (accuracy, time)
• Generate and evaluate simplified climate change scenarios assuming 0% (normal), +10% (high) and -10% (low) change on historical precipitation depths.

Phase 2

• Define the Lake Xiniada reservoir properties and input parameters and set up the RMM-NTUA (2) model
• Set up the MIKE 11 model for the additional watershed area (upstream of Lake Xiniada)
• Run the simulations of the multi-reservoir system for the same time period as the single reservoir runs
• Investigate future development scenarios where water demand increases downstream, water allocation changes and rainfall rates vary along the watershed

Study Results

• The two model components (RMM and Rule) were successfully migrated in OpenMI
• A multiple reservoir system was set up from the existing models, taking into account the interactions between reservoirs and catchments, and managing reservoir operation to balance supply and demands.
• It was suggested by the end users that the set up of such scenarios using OpenMI saved substantial time, by allowing the models to exchange data at run time, instead of running the models consecutively and separately.
• Even a relatively small-scale change on precipitation affected notably the Smokovo reservoir yield, as denoted through the 5-year simulation. During continuous dry seasons (selected from historical data), Lake Xiniada's contribution is too small to cover an increased future demand

Use Case C: The lake Karla watershed: Linking a rainfall runoff model to a groundwater model

The aim of Use Case C was to apply integrated modelling for the estimation of the surface water and groundwater resource potential of the restored Lake Karla watershed in Thessaly (see Figure 8). The water balance of the study area was evaluated for existing water management conditions and for different strategies dealing with the development and operation of proposed hydro-technical projects and other water demand management measures. The main partner of Use Case C was the Department of Civil Engineering of the University of Thessaly (UTH), Greece.

Lake Karla occupied until 1962 most of the eastern part of the Thessaly plain. It was one of the most important Greek wetlands and a natural reservoir which provided significant water storage and recharge to groundwater. Incomplete technical studies with intention to protect the surrounding area from flooding and gain some agriculture land had led to the drainage of the lake. The decision to restore part of Lake Karla was made in the early 1980's but the construction work started few years ago. The project today is near completion. The restored lake will occupy approximately 38 km² with a drainage area of 1171 km².

The lake Karla watershed in Thessaly

Use Case C participants chose to link UTHBAL, a water balance model developed and migrated in OpenMI by UTH, to Visual Modflow. During the testing phase, UTHBAL exchanged data in OpenMI with various dummy models provided by the OpenMI Association Technical Committee to explore the different linking options (id based and spatial linking). Then, UTHBAL and Visual Modflow were set up for the study area, calibrated and validated. After several trial and error efforts to link the two models and an additional communication with Visual Modflow developers, it was understood that Visual Modflow was limited to exchange data in OpenMI only with MIKE 11. For that reason, the final coupling part between UTHBAL and Visual Modflow was conducted in a semi-automatic way.

Linking UTHBAL to Visual Modflow
Type of Exchange and Spatial Characteristics	Number of nodes/XYZ polygons	Simulation Period
Unidirectional - Lumped	1 mapped to ~ 40000 polygons	1987-1997
Unidirectional - Semi-distributed	200 mapped to ~40000 polygons	1987-1997
Unidirectional - Fully Distributed	195000 mapped to ~40000 polygons	1987-1997
Operational
Scenario 1: An up to 60% pumping reduction is evenly distributed but only at the wells within the southern area, which covers part of the study area	Scenario 2: The "do nothing" for indicative reasons: no reduction in the pumping rate of the wells	Scenario 3: An up to 60% pumping reduction accomplished by closing down selected wells due to their high pumping rate and their proximity to others

Some study results and key issues

• The simulations using larger groundwater grid produced abrupt changes in hydraulic heads. The most effective and cost efficient simulation for this Use Case was the coupling of the semi-distributed UTHBAL with 200 m grid Modflow.
• Due to the inability of Visual Modflow to be fully OpenMI compliant, a lot of effort was put into developing semi-automatic approaches for the coupling process.

Modellers

National Technical University of Athens (GR)	University of Thessaly (GR)
End Users Prof. Maria Mimikou Prof. A. Stamou Dr. Christos Makropoulos Dr. Eleftheria Safiolea Dr. Andreas Efstratiades Ms. Sandra Baki Ms. Soso Douka Ms. Paraskevi Economidou	End Users Prof. Antonis Liakopoulos Prof. Athanasios Loukas Prof. Nikitas Mylopoulos Dr. Lampros Vasiliades Mr. Pantelis Sidiropoulos
Developers Mr. Vasilis Kaffes Mr. John Liagouris	Developer Dr. Konstantinos Kokkinos
Web Designer Mr. Elias Sotiropoulos
Contact Person Dr. E. Safiolea Phone: +30-210-772-2885 Email: safiolea@chi.civil.ntua.gr	Contact Person Dr. K. Kokkinos Phone: +30-2410-233-528 Email: k_kokkinos@teilar.gr