Executive Summary

…..3.8

Introduction

Z. Kundzewicz, F.F. Hattermann and A. Becker

1.1 What is this document for?

The Water Framework Directive (WFD), which came into force in December 2000, arranges a framework for actions of the European Communities in the field of water policy, with the key objective of achieving a “good water status” for all waters in the European Union (EU) by 2015. The Directive imposes legal obligations for the authorities in EU Member States. In order to assist the EU Member States and, in particular, the water managers within river basin districts in implementing the Directive, a number of activities have been launched. The EU Member States, Norway, and the European Commission have developed a common strategy for supporting the coherent and harmonious implementation of WFD, which has led to the publication of practical guidance documents on various technical issues related to the Directive. In addition, a number of scientific efforts at the European level and associated with the WFD have been made with the aim of providing the scientific support in implementing the Directive.

The objective of the present document is to offer guidance to water managers on the model-supported implementation of the Water Framework Directive at the level of a river basin district and at other levels (such as sub-basins, national, or international scale in the case of international river basins). The document is a deliverable of the Concerted Action “HarmoniCA” (Harmonised Modelling Tools for Integrated Basin Management), and in particular, it’s Work Package 3 (WP3). The HarmoniCA Project has been launched within the Sixth EU Framework Programme with the overall objective of creating a forum for unambiguous communication, information exchange, and harmonisation of the use and development of Information and Communication tools (referred to as IC-tools or ICT) relevant to river management, and the implementation of the WFD (see http://www.harmoni-ca.info). Work Package 3 of the HarmoniCA Project was conceived to promote the development of a harmonised general methodology (planning framework) for water management along the lines of the Directive.

In addition to HarmoniCA, there are further EU projects providing scientific support in implementing the Directive, such as those belonging to the CatchMod (Integrated Catchment Water Modelling) Cluster with the objective of developing common harmonised modelling tools and methodologies for the integrated management of water at the river basin scale, cf. http://www.harmonit.org/links/catchmod.htm. The CatchMod Cluster groups a number of projects from the “Harmon[y]” family and some other projects. The HarmoniQuA (Quality Assurance) Project provides methodological framework for quality assurance in modelling and model applications, cf. http://harmoniqua.wau.nl. The HarmoniCOP (Harmonising Collaborative Planning) Project, aims to improve participation in water management, cf. http://harmonicop.info. The following projects also play useful roles and provide guidance: the Harmoni-RiB (Harmonised Techniques and Representative River Basin Data for Assessment and Use of Uncertainty Information in Integrated Water Management), cf. www.harmonirib.com, the HarmonIT (project aimed at the development and implementation of a European Open Modelling Interface and Environment), cf. http://www.harmonit.org, the EuroHarp (Towards European Harmonised Procedures for Quantification of Nutrient Losses from Diffuse Sources), cf. http://www.euroharp.org, TransCat (project dealing with Integrated Water Management of Transboundary Catchments), cf. http://www.transcat-project.net/index.php, and the BMW (Benchmark Models for the Water Framework Directive), cf. target=_blank>www.environment.fi/syke/bmw. for specific modelling related issues

The present document complements the material provided by working groups of water managers and other experts in the field of water in Europe, established to develop several guidance documents (GD 1 to 13) in the framework of the “Common Implementation Strategy” (CIS) for the implementation of the WFD. In particular, the present document refers to the Guidance Document No. 11 (GD 11) on Planning Process (EC, 2003), which presents a general overview of the whole planning cycle and provides recommendations for its successful implementation. As explicitly stated in the foreword to the latter publication, “[it] is a living document that will need continuous input and improvements…”. The present document reacts to this invitation and offers complementary contribution: guidance on model-supported implementation of the Directive; and in particular on the use of models for the planning measures and integrated river basin management. It should help water managers to better understand how models may be used for planning purposes, while special attention should be given to the problem of predicting an uncertain future, one very likely to differ from the present.

This could be useful, since the GD 11 does not focus on modelling. The considerable potential and challenges in modelling are neither explained nor discussed in that guidance document that only makes ten references to “model[s]” or “modelling” (EC, 2003). This corroborates the observation that, indeed, there is a niche for projects like HarmoniCA. It is worthwhile to follow in what context the GD 11 (EC, 2003) mentions models and modelling.

Modelling is mainly referred to in a rather short section (4.5: The appropriate toolbox). The GD 11 states that models exemplify the systems approach to water resources planning. It observes that models permit a definition and evaluation of numerous alternatives that represent various possible compromises among conflicting groups, values, and management objectives (such as options for engineering structures, operating and allocating policies, and different assumptions made in the analysis) in physical and economic contexts. In particular, a rigorous and objective analysis should help to identify the possible trade-offs between quantifiable objectives so that further debate and analysis can be more informed. As stated in (EC, 2003), “models can represent, in a fairly structured and ordered manner, the important interdependencies and interactions among the various control structures and users of a water resources system.” Models are viewed as tools capable of answering well-posed questions about the behaviour of the system being studied. However, it can be useful to feed answers derived from the models back into questions and to examine whether or not rephrasing the original questions is necessary. That is, models can be used a in two-way process: they produce information that may be channelled into decision making (formulation of plans) or they produce information that is fed back to aid in redefining the problem (EC, 2003).

The present document consists of five chapters. After the brief general introduction, the opening chapter explains the objective of the document and provides brief information on the Water Framework Directive. Chapter 2 explains the planning process of implementing the Directive and demonstrates how models can help in this context. This refers to identifying the current status and setup of monitoring programmes, and to designing, setting up, implementing, and evaluating programme measures. Chapter 3 introduces a primer on modelling, tailored to the needs of practitioners. It provides an easy introduction to modelling and then proceeds to explaining several specific technical issues related to the modelling process, such as model selection, models for impact analysis, integrated modelling, problems of model calibration and validation, special requirements for stakeholder participation, uncertainty assessment and quality assurance aspects. Here, references are made to relevant documents and guidelines prepared in the HarmoniCA project and other EU projects, particularly from the CatchMod Cluster. Finally, Chapter 5 breifly presents six case studies from different parts of Europe to illustrate the practical applicability of the planning framework in the WFD implementation. They are very important for illustrating how concepts from earlier parts of the document are applied to real-world situations. The case studies cover several aspects of mesoscale river basin management, water quantity and quality issues, and the role of modelling, with two case studies located in pilot river basins.

The primary targeted audience is authorities entrusted with implementing the WFD. First and foremost, these are water managers in river basin districts, but also other stakeholders with stakes and/or interest in the implementing: the Directive, such as landowners, decision makers at different levels (from central government, via provinces through to municipalities and counties), water users, scientists, educators, NGOs, nature protection activists and environmentalists, farmers, media, and the broader public. The overall aim is to fulfil the various and varying demands of all parties interested in and affected by the implementation of the Water Framework Directive. It is hoped that this document contributes to improving the understanding of modelling aspects and facilitates the planning process for implementing the Directive. The emphasis lies on practical aspects, hence the material refers to practice, and case studies illustrate implementing the planning process in practice.

The ideas is to keep the document simple in structure, content, presentation, and terminology to keep it readable for a broader, practice-oriented audience.We avoid using the original, formal and legal language of the Directive without dropping technical contents of the message. The aim is to keep the present material in a complementary and not in a repetition fashion vis-à-vis the Guidance Document 11 (EC, 2003) while providing additional material, with particular references to modelling, and seeking synergy. The superstructural question document aims to assist in answering is: How to manage water in river basin districts in order to meet the WFD objective.

1.2 The Water Framework Directive – a challenge

On 22 December 2000, the Water Framework Directive (WFD) was published in the official Journal of the European Communities [I.327/1-72], after it had been accepted as Directive 2000/60/EC by the European Parliament and the Council on 23 October 2000, hence coming into force (EC, 2000). The Directive arranges a framework for analysing, planning and managing water resources at river basin scales, with the major objective of achieving at least (par. 26, EC, 2000) “good water status” for all waters by the year 2015.

The purpose of the Directive (Art. 1, EC, 2000) is to establish a framework for the long-term protection of freshwaters, preventing future deterioration and protecting and enhancing the status of ecosystems (aquatic, terrestrial, and wetlands); to promote sustainable water use; to ensure reduction of pollutant loads; and to contribute to mitigating floods and droughts. As stated in paragraph (25) of the Directive, environmental objectives should be set to ensure that good status of surface water and groundwater is achieved throughout the Community and that deterioration of waters is prevented at Community level. The objective of achieving good water status should be pursued for each river basin (par. 33, EC 2000).

The Directive is a challenge to EU nations and their water sectors. In several areas the current quality of water is far below the WFD target, so long-lasting and costly efforts are required to reach good status of all waters by 2015. The principles of the Directive have been greeted by the increasingly environment-friendly societies of EU nations, yet the WFD is not easy to implement under a variety of national conditions, different legislations, standards, guidelines, traditions, etc.

The WFD creates a legal obligation for the authorities in EU Member States to organize water management within river basin districts (rather than within administrative units). The Directive provides relevant advice for planning processes. Preamble 13 of the Directive highlights the possible diversity of conditions and needs in the various river basins (and districts), which require different specific solutions. This diversity should be taken into account in implementing of the WFD. Further, Preamble 13 states that decisions should be made as close as possible to the locations where water is affected or used (need for competent, decentralized, grassroots decision making in accordance to the subsidiarity principle). Preamble 28 emphasises the time lag in the processes of renewing water resources (particularly long for groundwater) to be taken into account in the planning measures for achieving the good water status. It can take a very long time to reverse a trend towards increasing groundwater pollution. Article 4 tackles environmental objectives, Article 5 deals with assessing the current status of waters in a river basin district, while Article 8 regulates monitoring. The comprehensive Article 11 of the WFD defines some basic requirements concerning the programme measures, established in order to achieve the environmental objectives of the WFD. The WFD requires Member States to produce a management plan for each river basin district (Article 13) involving stakeholders in plan development. The plans should envisage reporting mechanisms to the Commission and public (Article 15).

Integration is indeed a key concept underlying the Water Framework Directive for regulating the management of water protection within the river basins and the river basin districts (EC, 2003). Integration is interpreted in a very broad sense in the WFD, much broader than in the classic integrated water management approach, where the scope was typically reduced to joint consideration of surface water and groundwater and/or of water quantity and quality aspects. The Directive jointly considers, quality, environmental, ecological, and quantity objectives for protecting valuable aquatic ecosystems and ensuring a general good status of waters. It embraces integrating all water resources (fresh surface water and groundwater), and all water uses, functions and values. The notion also integrates disciplines, analyses and expertise (hydrology, hydraulics, ecology, chemistry, soil sciences, agronomy, forestry, technology, engineering and economics) to aid in implementing the Directive in the most cost-effective manner. The Directive also requires integrating water legislation into a common and coherent framework and consideration of significant management and environmental aspects. It calls for using a wide range of measures, including economic and financial instruments, e.g. water pricing. Furthermore, integrating stakeholders and the civil society in decision making is required, as well as integrating different decision-making levels (local, regional or national), and water management from different Member States in the case of international basins.

The need for integrated river basin management has risen because managing environmental processes independently (without integration) may not be sufficient and may not lead to optimal decisions. Due to complexity of processes and systems involved, managers turn to integrated models and decision support systems.

The recent understanding of integrated water resources management (IWRM), as defined by the Technical Advisory Committee of the Global Water Partnership, reads (GWP 2000): “Integrated water resources management is a process, which promotes the coordinated development and management of water, land and related resources in order to maximize the resultant economic and social welfare in an equitable manner without compromising the sustainability of vital ecosystems.”

The Key words here are “process”, “coordinated development and management of water, land and related resources”, considering the relationship with “economic and social welfare” and equity, and “the sustainability of vital ecosystems”. It clearly points in the direction of “Integrated River Basin Management” (IRBM).

Mostert (1998) describes five aspects capable of causing problems in the application of IWRM:

This list can be extended by adding further elements, such as:

The distinction between “integrated” and “traditional” management of water resources or river basins has to do with the scope and sphere they operation in. The latter is typically sector-oriented (water supply, irrigation, hydropower, etc.) and focused on the satisfaction of the demands perceived within each sector. On the other hand, the former attempts to take a cross-sector approach and to focus on integrated management of the available water and land resources, accounting for their spatial and temporal variability and associated uncertainties .

This necessitates the decision-making procedure to include a coordinated inter- and cross-disciplinary dialogue between the social, natural and engineering sciences while retaining the distinctive depth of the individual disciplines. It should support a dialogue between stakeholders and decision makers on the one hand, and scientists, representatives of governmental Environment Protection Agencies (EPAs) and non-governmental organizations (NGOs) on the other to develop a process of social learning. What emerges is the most appropriate compromise alternative, which will then be implemented.

This general concept is accepted as the IWRM Paradigm, adopted by the European Water Framework Directive (EC, 2000), forming core element of any future planning methodology.

According to Article 11 and Annex VI of the WFD (EC, 2000), basic and supplementary measures should be implemented in order to achieve environmental objectives of the Directive. The list of basic measures, the minimum requirements to comply with, consists of the following EU regulations introduced in the past decades (in chronological order):

Further legislative acts, such as the recent Floods Directive, should be also considered.

In cases where these measures do not suffice to reach the environmental objectives defined under Article 4 of the Directive (EC, 2000), Member States are entitled to supplementary measures (cf. Annex VI part B of WFD). They comprise (EC, 2000):

The measures listed above should be considered pragmatic action alternatives, established with the purpose of reaching environmental objectives. Any other actions resulting from different sectoral policies can also be considered measures and implementations (falling under the item (xvii) above, i.e. other relevant measures). Actions such as urban development plans, best agricultural practices, and flood risk plans, can be considered “supplementary” measures in regard to meeting environmental objectives.

The Directive provides a long-term policy basis for water management with clearly defined interim objectives. It delineates a road map of objectives to be met in the future, drawing clear deadlines for each of the requirements to add up to an ambitious overall timetable. The period to 2015 refers to the Directive proper, while the second period (2015-2027) will be managed with due consideration of the experience gained in the first period. The key milestones and deadlines for achieving them are listed below (Table 1.1).

Table 1.1: Key milestones of the European Water Framework Directive (Issues foreseen to have been completed already are shown in italics).

2 How can models help implementing the Water Framework Directive?

F.F. Hattermann, A. Becker, Z.W. Kundzewicz, Y. Laurans, S. Muhar, R. Sonsini-Sessa, P. Stalnacke, P. Willems

2.1 Framework for model-supported participatory planning of measures in implementing the Water Framework Directive

2.1.1 The planning process of implementing the Water Framework Directive

The general structure of the planning process to implement the Water Framework Directive contains the following main components identified in the GD 11 (EC, 2003) (Table 2.1):

Table 2.1: General structure of the planning process to implement the Water Framework Directive.

A short explanation of the particular components is offered in the sequel.

(o) Setting the scene is an initial phase that aims at defining the system of interest (such as a river basin or a river basin district), and describing the problem to be solved and the goals of the plan. This includes the overall objective of “good water status” and sustainable development, as required by the WFD. For that purpose, the system boundaries and the time and space domains should be defined. In addition, the authorities, stakeholders, sectors involved, and the institutional and legal framework of the plan should be identified.

(i) Assessing the current status and preliminary gap analysis includes describing the “characteristics of each river basin district”, in particular those of the water bodies, including artificial and heavily modified water bodies, and establishing reference conditions. Here, protected areas must be designated and registered, and an economic analysis of water uses performed. It is necessary to identify significant current and future anthropogenic pressures, to assess their impacts, and to determine which water bodies will be at risk of failing to meet the environmental objectives. This also includes the putting the massive effort needed into preliminary analysis of information and knowledge gaps.

(ii) Setting up the environmental objectives aims at reaching “good ecological status”, “good ecological potential”, and specific objectives. When goals and targets are set, they serve as the foundation for the decision on programmes measures.

(iii) Establishing monitoring programmes is required for implementing the Directive in order to improve our knowledge and understanding of the particular river basin and the determining the risk of failing to meet the WFD objectives. Monitoring can serve as surveillance (for improving determining the water bodies at risk of failing to meet the Directive’s objectives), operation (for checking the progress made in establishing the Directive’s objectives) and investigative measures (to explain why the given body is at risk and to assess the risk).

(iv) Gap analysis means identifying “gaps” by comparing scenarios of future system performance with the “goal” status (e.g. good ecological status). The gap analysis builds on assessing the current status (component i) compared with the environmental objectives of the Directive (component ii). It also uses the future development scenario (Baseline Scenario, a business as usual projection into the future, often computed by with models, see Glossary), which serves as a reference for planning measures.

(v) Setting up programme of measures should identify a feasible set of alternatives for enabling overall goals to be met, while collecting a satisfactory level of possible consensus among the stakeholders. This part of the planning process includes two aspects: determining scenarios and alternative measures. Scenarios mainly deal with future external driving forces in the river basin such as global developments, EU policies, climate change etc. Projections (visions into the future) and assumptions (e.g. by experts) are required to estimate effects and impacts. Many “alternative measures” (action alternatives) can be defined before effects/impacts are investigated and compared. It should be emphasized that economic instruments and pricing policies must be considered as well. This part of the planning process takes into account the principles of sustainable development, of “polluter pays”, “risk taker pays”, “cost recovery”, and “source control rather than the end-of-the-pipe” approach. This is the phase to discuss possible alternatives in a broad participation process .

..

(vi) Developing River Basin Management Plans is an essential milestone in the planning process. A plan is a principal mechanism of implementing measures to achieve the WFD objectives by thoroughly evaluating and comparing different measures (alternatives), a task that experts should tackle. This can be essential as some measures are expensive, and the negotiation based on model results preceeding the implementation would facilitate informed decision making.

(vii) Implementing the programme of measures (v) includes preparing the interim report on the implementation.

(viii) Evaluating the success of the implemented measures is an indispensable part of the planning process. It should inform about the progress made to meet the WFD objectives. Indicators like water levels, discharges, water quality characteristics (e.g. N, P), other substances and solutes, BOD, biomass etc., are required to evaluate and compare how the system performs with not measures taken (“do-nothing” option) and what effects (impacts) of all the alternatives (measures) on the system are projected. Stakeholders must define a set of evaluation criteria reflecting the values that underlie their judgments in accordance to the WFD and existing national or regional regulations,. These are the “goal criteria”, e.g. setting the “threshold” values for water levels, discharges, water quality parameters, and ecological criteria not to be exceeded.

(ix) The public should be informed and consulted, and interested parties actively involved early on in the process, e.g. in steps (i), (ii), (iv), (v), (vi), (viii). The concept of measures and River Basin Management Plan, prepared by experts and managers, should be discussed with stakeholders and other experts to get a reaction – of approval or criticism, advice and other comments. Stakeholders can also define measures to be considered (action alternatives). “It’s crucial for the legitimacy of a planning process to start dialogue as early as in the phases of problem defining and setting the agenda” (EC, 2003).

The above components of the planning process are discussed in detail in GD 11 (EC, 2003). It is worth stating that these components do not necessarily have to be followed in rigid succession from (o) to (ix). Even though the planning process is described as a linear sequence of components, several feedbacks and iterative processes are possible and necessary indeed. The planning includes a number of components that depend on each other and should be jointly developed. For instance, informing and consulting the public and actively involving interested parties should already take place in much earlier stages, at best “as early as possible” (cf. discussion above).

The Directive aims at triggering positive changes in the future. The planning must consider future “boundary conditions”, such as development and variability (seasonal and inter-annual) in water availability conditions, quality and demand. Beyond that, changes in land use and wastewater release affecting water availability and quality and thus the ecological status of the water bodies need to be known. This also applies to water demand (water use), subject to change due to technological changes or economic development (including resulting land use and land-cover changes) or to weather conditions. The Baseline Scenario should also take into the projections of the climate change into account. In the planning interval of 2015, climate models are quite consistent in projections of both temperature and precipitation, unlike in later intervals (eg. 2050), when precipitation projections may largely differ between models disagreeing even on the direction of changes. The scenarios should include information about the normal (average) conditions as well as the extreme conditions (e.g. very dry summers). Projections into the future are required in form of scenarios based on model simulations, expert judgments or sometimes reasonable assumptions (e.g. trend extrapolations), on weather conditions (climate) and other external driving forces (e. g. globalization, population dynamics, economic growth, EU policies) over the planning period. The internally consistent scenarios should be interpreted as plausible alternative projections into the future, rather than being treated as crisp and accurate predictions.

The impacts corresponding to each scenario and measure (alternative) have to be estimated. This is generally a simulation step, using calibrated models. If more than one scenario exists, the estimation has to be made for each of them. At the end of this step an impact matrix is produced with elements as values of the indicators derived from each scenario and each alternative to help evaluate the results.

In order to facilitate model support for water managers dealing with the WFD implementation process, it is crucial to make the modelling process structure (see e.g. the five modelling steps as described in Scholten et al. 2004) applicable to the framework for water management suggested in GD 11 (EC 2003), and shown in Figure 2.1.

As it has been identified in the GD 11 (EC, 2003), models are a considerably aid in several components of the planning process to implement the Water Framework Directive. Models are indispensable for impact prediction and what-if analyses (in planning, designing, and managing water resources systems). They are useful in virtually all the steps in the planning process. They turn out to be beneficial and – in a variety of circumstances – indispensable. Models can help us understand and optimize the efficiency of water resources usage.

Figure 2.1: The management approach including the main stages of the WFD implementation process according to GD 11 (EC, 2003)

How to support the planning process outlined in Figure 2.1Figure 2.1 by use of models and in consultation with stakeholders is the core of this guide. The formal methodology is described in the following section.

2.1.2 Framework for model-supported implementation of the Directive

The challenges of the WFD are pointed out in Section 1.2. It is required to bring the water bodies in Europe into a “good chemical and ecological status”, but at the same time to account for cost-effectiveness and to honour the needs and demands of many different stakeholders, and environmental concerns. Thus, carefully planning the implementation is important, combining the nine steps of the WFD (as introduced in GD 11, see Section 2.1.1) and the five typical modelling tasks as described and explained by Scholten et al. (2004).

The framework for model-supported implementation of the Directive should integrate modelling and management to allow adaptation driven by participation, facilitating social learning and supporting decision making (Pahl-Wostl 2002a&b). Such a framework (flowchart for model-supported implementation of WFD) is presented in the sequel (Figure 2.2). Its Main components were developed in (Wenzel, 1999, Soncini-Sessa et al. 2004 and Becker et al. 2005).

The individual steps in the combined planning framework for model-supported river basin management, schematically illustrated in Figure 2.2, will be briefly explained in the following, focusing on “what” has to be done in each step. A general explanation of the “how” is much more difficult to give, because water-related problems where models are applied to support the planning process are very diverse, ranging from operating water storage reservoirs to water quality management in lakes, over salinisation problems to biodiversity loss, etc. Therefore, a selection of case studies will give guidance on how typical water-related problems have been investigated in different European countries using models to support the planning process (see Chapter 6). The case studies cover methodologies to set up a Baseline Scenario (Section 6.2), water supply related problems (see Section 6.3 and 6.4) and water quality related problems (see Section 6.5 and 6.6). Many additional examples show how and when model applications are useful to facilitate water management and also public participation.

Figure 2.2: Framework for model-supported participatory planning of measures and Integrated River Basin Management (planning framework).

The planning framework shown in Figure 2.2 describes the main tasks to be considered for integrating participatory river basin planning and modelling along the first main stages of the WFD implementation. It can be applied to manage and structure the entire implementation process, but also to organize individual modelling tasks. A very detailed methodology for structuring the modelling process as such has been compiled by the EU HarmoniQuA Project (MoST - Modelling Support Tool, see http://www.wise-rtd.info).

Step 1 – Problem description and goal definition: “Setting the scene”

This phase aims at defining the system of interest (river basin or river basin district), describes the problem to be solved and the goals of the plan, including the “good water status” and cost-effective sustainable development, as required by the WFD. In other words, the state of the system has to be characterized, including the relevant drivers and pressures. Looking at the entire WFD implementation process, this phase corresponds to the first three implementation steps according to GD 11, namely

• Assessing the current status and preliminary gap analysis,
• Setting up the environmental objectives, and
• Establishing monitoring programmes.

The preliminary gap analysis forms the background information to define and describe the modelling problem (“what to model”), the setup of the environmental objectives is the input to develop a model study plan (“how to model”). For that purpose the boundaries of the system (climate and management) and the time and space domains must be defined, as well as the authorities, stakeholders and sectors involved, (institutional and stakeholder analysis) and the institutional and legal framework of the plan.

The aim of this step is to develop the working concept of the project. Starting from the problem description and goal definition the “means of solution” must be conceived and designed. In principle, a kind of “project proposal” needs to be developed describing how the given task can be solved. This work needs careful thinking, is primarily expert and manager based, and largely is intellectual work. The final concept should be discussed with stakeholders and experts (individually or in a group discussion), to get their approval, advice and comments. “It’s crucial for the legitimacy of a planning process to start dialogue as early as in the phases of problem defining and setting the agenda” (GD 11, EC 2003d). Therefore all activities that may raise consensus must be promoted in this step.

Following Pahl-Wostl 2004 three types of stakeholder and public participation can be distinguished (see also Section 4):

• Information: The lowest form of participation is information.
• Consultation: In the case of consultation different stakeholder groups and the public at large are asked to give their opinion on a management plan / scenario.
• Active involvement: Stakeholders and the public are actively involved in river basin management.

The latter is preferable. It can lead to to a pro-active management regime via a co-production of knowledge and of co-decision making (viz. adaptive management, see Section 4.1), and, in the end, can help to prevent undesirable impacts of unsuitable management measures.

Step 2 – Conceptualisation

The previous step aimed at “setting the scene”, including preliminary gap analysis, defining environmental objectives and setting up a model study plan. This step helps conceptualize the management and modelling problem.

Step 2a – Identification of mIdentification of measures

In this step classes of measures (or actions) possibly allowing the environmental goals to be attained are identified , while accounting for the different stakeholders’ interests. Examples of possible measures are the implementation of reservoirs or sewage treatment plants, changes in water management general, but also changes in land use management (watershed management), establishment of wetlands, and riparian buffer zones. These measures should be sustainable and increase the buffering capacity and resilience of the hydrological system.

A coherent set of measures (following one story line) represents a management alternative. Part of the action identification should be to clearly state “who is doing what and when”. The final scope of the planning process is to identify the set of alternatives that allow the goals of the plan to be reached (the good status of the water bodies) while building consensus among the stakeholders. This step is necessary in order to design models that allow the assessment of the resulting effects/impacts in the simulations to be incorporated and easily activated. Moreover, there should be capacity to adapt the model to new management options.

Step 2b – Criteria and indicator identification

A very sensitive task is to identify criteria and indicators which can be used to describe the water-related problems and to measure the consequences of implementing new management measures.

In addition to the criteria defined by the WFD, and the existing legislation (including national or regional regulations), stakeholders must define a set of evaluation criteria reflecting the values that underlie their judgments. These are the “goal criteria”, such as “threshold” values of water levels, discharges, or water quality and ecological criteria not to be exceeded in order to reach a good status of the water body. It is appropriate to structure the criteria as a hierarchy, starting off with the goal of the plan and iteratively refining it. A comprehensive description of work with various criteria is given in the Verbano case study (see Section 6.4).

Indicators are required to evaluate and compare how the hydrological system behaves without any action taken and in effect of the management alternative(s) i.e. measure(s). Examples for such indicators are water levels, discharges, water quality characteristics, for example N, P, BOD, biomass etc. These indicators also specify the type of model to be applied to investigate the state of the hydrological system.

Identifying evaluation criteria and indicators should result from discussions with the relevant stakeholders and local experts.

If the system is affected by uncertain inputs or by the stochasticity of processes, the indicator values are uncertain making quantitative measures of risk and uncertainty necessary (see Spree river basin study, Section 6.6).

Step 2c – Model setup including calibration and validation

In order to simulate the system response to changes in management or climate and the trajectories needed to quantify the indicators, a site-specific model of the entire system is required. Three sub-steps can be distinguished in building the model (see also EU project HarmoniQuA - Scholten et al. 2004):

(a) model setup;
(b) model calibration;
(c) validation.

Clearly, they belong together and include the aspects of uncertainty assessment and quality assurance. First of all, the decision must be made whether and where models are to be applied and what type of models (e.g., detailed, parsimonious) could be used (see Section 3.2.1, and also the Scheldt river basin case study, see Section 6.3). The most important driver is the required accuracy of the results: if there is demand for very accurate and detailed model results, a more sophisticated model has to be applied, and the relevant data have to be collected accordingly (Højberg et al. 2006).

Model validation results (from comparing observed data and modelling results for periods / locations where the model has not been calibrated) are very important for raising the confidence of the stakeholders in the models and so, for justifying further modelling for the impact analyses. Here it is beneficial to have computer-based illustrations and summary interpretations of simulation results. In any case, a calibrated and validated model for the river basin in question (or hydrological system under investigation) is delivered after this step. The choice of the model types and their degree of detail strongly depends on the indicators defined in step 2b, regulated by the data available to operate the model and the actions selected in step 2a.

It is not mandatory for the model to be mathematical. Another option could be to use knowledge of an expert able to describe the effects a given alternative would induce (such a case is presented in this document for the Möll in Section 6.7). However, mathematical models are frequently used. The validation of the model including sensitivity and uncertainty analyses defines the reliability of the model results (see Section 3.3).

Sometimes two or more models are adopted; e.g., a simpler model for the design of alternatives (see step 3b) and a more complex one for a more accurate estimation of effects (see Section 3.2 and the Scheldt case study in Section 6.3). If this is the case, it is good practice for the first to be a parsimonious version of the second. Similarly, in multi-scale studies the larger-scale study could be based on a simpler (parsimonious) model while in the smaller-scale study a more detailed model is required.

Step 3 – Scenario definition and alternative design (two steps)

In order to set up the programme of measures aiming to reach the good status of the water bodies according to the WFD, projections into the future are required. The first inputs to the model are the driving forces unaffected by the alternatives and variables describing alternative external conditions (e.g. climate change, globalization etc.). Their trajectories are called scenarios, since they describe the background scene on which the alternatives act. Both the scenarios and the management alternatives (the second input to the models) have to be quantitative and the scope of this step is defined by determining them. It may not be necessary to specify just one scenario. Neither is the scenario necessarily deterministic (indicating certainty of the future); it can be stochastic (uncertain) (see also Spree/Havel case study in Section 6.6). For scenario designing see also Section 2.3.2.

Step 3a – Choice of scenario

The scenario(s) may be chosen by experts or obtained by running models that describe the processes producing the driving forces; e.g. the future scenario of rainfall can be produced by a climate model, while the future scenario of land use is often proposed by an expert.

The length of the time horizon defining the scenario must be sufficiently long to observe all the types of possible significant events in the system (e.g. 20 or 50 years). There may be different scenarios for designing alternatives (design scenarios) and for estimating impacts.

Step 3b – Design of the programme of measures (action alternatives)

The term “measure” or “action alternative” is used to emphasize that system modelling serves the investigations of “alternative measures”. Very often ,it is the case that the only measures (alternatives) considered during a real design process are those suggested by the stakeholders and the manager’s experience. This is a convenient starting point, but one should consider all alternative measures obtained by all combinations of actions identified in step 2 leading to achieving the environmental objectives (“good water status” proposed by the WFD). Generally, the resulting number of alternatives is fairly large. Therefore it is necessary to screen them in a way that optimal combinations in terms of benefit for the different stakeholders are selected (the so-called Pareto optimality, where subordinate alternatives are removed, see Soncini-Sessa, 2005). For details see the Verbano case study in Section 6.4.

Step 4 – Simulation and estimation of effects/impacts

The effects/impacts resulting from each scenario and measures (alternative) have to be estimated by computing the values the indicators take for each alternative. This is generally a simulation step where calibrated models are used. If more than one scenario exists, the estimation has to be made with respect to each one of them. At the end of this step an impact matrix is produced whose elements are the indicator values for each scenario and alternative. It serves to evaluate and compare results (see Figure 2.3).

Figure 2.53: Impact matrix illustrating the impacts of different planning alternatives on specific criteria (Wenzel 2005). Translation of German terms: Versorgungssicherheit – Supply reliability; Mindestabfluß – Least discharge; Wasserwerke – Water supply units; Kraftwerke – Water power units.

Step 5 - Evaluation of the alternatives

Given the Impact Matrix, the goal of this step is to determine the “value” that each sector assigns to each alternative (in general, this “value” may not be related to the indicator values in a simple way). A number of evaluation techniques exist for assisting the analyst in reaching this goal (e.g., multi-attribute value theory and analytic hierarchy process, see Soncini-Sessa, 2005).

When only one decision maker and one stakeholder exist, the optimal alternative can be easily found by ranking the alternatives with respect to their values compared with the adopted goal criteria and indicators, after which the procedure terminates. If either the number of decision makers, the number of stakeholders, or both, are greater than one, a different ranking of alternatives is found for each stakeholder or decision maker at the end of this step. For this reason, the decision process is not completed yet and proceeding with the following step is necessary.

Step 6 – Comparison and negotiation

The optimum result of this step is identifying a set of measures (action alternatives) that can be perceived as a fair trade-off between different stakeholders' interests without encounting anyone’s opposition and taking Articles 4.5-7 of the WFD into account. If such an alternative cannot be found, the step can be brought to an end by identifying the alternatives that live up to the environmental objectives (“good water status”) and gather a broad (yet imcomplete) consensus amongst the stakeholders. Supporters and opponents of each of these alternatives must be identified. The process begins by briefing each stakeholder on the other stakeholders' points of view. In case of relevance, this includes a briefing on the negative effects that the actions preferred by the stakeholder in question actually have on the other stakeholders and the environment (Article 14 of the WFD). Once this information has been shared, the core of this step is the negotiation amongst stakeholders to reach a compromise. The negotiation can be carried out (e.g., with the help of Pareto race, multi-criteria and equity analysis, see Soncini-Sessa, 2005). The result of this step is a set of compromise alternatives that form the programme of measures to be included in the River Basin Management Plan.

2.1.3 Checklist for effective model use

As stated by Loucks & van Beck (2005), the most important aspect of model use today is communication: planners and managers articulate their needs for information. Modellers pick them up and communicate assumptions and results to stakeholders. The next iteration can follow. Water managers and modellers should start cooperating in early stages of the project and, ideally, sustain into the project application phase. Involving stakeholders in model building creates a feeling of co-ownership and leads to a better understanding of everyone’s intentions, concerns, and priorities. The final HarmoniCOP handbook "Learning together to manage together – Improving participation in water management" (see http://www.harmonicop.uos.de/handbook.php) gives practical information about participation processes in river basin management and how to support the implementation of the public participation provisions of the European Water Framework Directive.

Due to the complex nature of management problems in river basins it is impossible to already design the ultimate plan at the beginning of the WFD implementation process. Unforeseen problems are likely to occur and new pressures make it necessary to adjust the programme of measures and to adapt to the emerging pressures and problems. It is therefore crucial that the approach is flexible enough to adapt to new ideas and solutions. A broad participation process including all relevant stakeholders helps identify alternative management measures in order to mitigate problems.

At the stages of identifying and characterizing the individual river basin, including an assessment of the current ecological status, impacts and pressures, preliminary gap analysis, and establishing environmental objectives, modelling may be useful to support the definition of the reference conditions and to assess the possible pressures. An Evaluation of the susceptibility of water status to the pressures can be established by using both monitoring and modelling. Modelling can be useful, because in this initial phase available data are often insufficient (e.g., the data records do not include all the necessary variables, or are incomplete in time and space, or simply erroneous). In combination with monitoring information (both ground-truth and remotely-sensed) and expert judgment, models can help the analysts: (i) assess the impacts of the various pressures, but also (ii) design optimal monitoring networks, and (iii) interpolate the available data. At the stage of designing River Basin Management Plans and programmes of measures, modelling is a useful tool in supporting the assessments and quantification of the effects and costs of various measures under consideration. Furthermore, on-line modelling is often used to support the operational decision making in the stage of implementation of the measures, for example in managing reservoirs (reservoir operation), flood protection and urban drainage systems. Finally, at the stage of evaluating the effects of the planned or implemented measures on the environment, modelling may support the monitoring in order to extract maximum information from the monitoring data, e.g. by indicating errors and inadequacies in the data and by filtering out the effects of climate variability.

The additional value of models lies in them representing the relevant hydrological and ecological processes in a cost-effective way, in due time, and without the need for an “active experiment”. Therefore, models can be used, along with expert judgments and stakeholder dialogues, to characterize water bodies, to evaluate impacts of planned measures in a river basin, to support the implementation of monitoring networks, and more generally for decision support in the planning process as well as in the operational mode.

In order to calibrate models, their parameters should be estimable from readily obtainable data. Operational model use requires data in adequate quantity and quality, and sufficiently good knowledge of the boundary conditions (climate, water and land use management). In practice, this is typically not the case. To overcome the problem of incomplete and inaccurate data, one may try to revise and update the model whenever new data become available. A possible solution in the case of changing and unknown boundary conditions is to integrate and describe possible changes in climate, water and land use management in the form of scenarios, where models since validated can be used to evaluate the possible impacts under scenario conditions on the water quantity and quality. Models can be used to indicate errors and inadequacies in the monitoring network.

From the discussion above it becomes clear that the model setup is often not straightforward, instead it has to be revised and updated in an iterative way. This is theoretically possible in each of the planning and implementation phases of the Directive. To minimize the modelling costs and to avoid the development of models that do not apply to the management problem, involving stakeholders and public participation is absolutely necessary in the initial model setup phase. Another crucial issue is the correct and accurate documentation of each modelling step in a modelling protocol in such a way that the simulation results can be reproduced and retraced. The technically and scientifically adequate execution of all tasks included in a modelling study is called quality assurance, the core of the HarmoniQuA Project of the EU (Scholten et al. 2004).

HarmoniQuA strongly focuses on modelling and quality assurance. It has defined the following five activities for the modelling process:

1. Description of the modelling problem and the development of a model study plan;

2. Compilation of data bases and conceptualisation of the model problem;

3. Setup of a model describing the relevant processes;

4. Calibration and validation of the model;

5. Simulation of the processes and evaluation of the results.

Computer models are in common use nowadays to support planning processes and water resources management. However, as pointed out in Section 1.2, the WFD forms a new challenge for water management and therefore also for hydrological modelling, because the approach of the WFD is to involve all relevant stakeholders on the river basin scale in a common striving to reach the goal of “good chemical/ecological status” of the hydrological system. The challenge is to integrate the interests of all stakeholders (including the environment, e.g., via environmental flows) having sometimes very diverse particular interests, bearing in mind an uncertain future, and considering complex feedbacks in the system structure and changing boundary conditions. Model descriptions may help facilitate this process.

Figure 2.4: WFD’s implementation timetable and the role of modelling structured by the identification phase (blue), the design phase (red), the implementation phase (reen) and the evaluation phase (purple).

Four main management tasks can be identified where models can play a particularly important role to support the implementation process of the WFD, by providing additional information about the chemical and/or ecological status of the specific water bodies and possible impacts of management interventions (Rekolainen et al. 2004, Refsgaard et al. 2005a), which are (see Figure 2.4):

• Identification - assessment of the current status and setup of monitoring programmes (according to the plan – to be finalized by 22 December 2006);
• Design – setup of programmes of measures and River Basin Management Plans (until 22 December 2009);
• Implementation – bringing the programmes of measures into action (until 22 December 2012);
• Evaluation – review of the programmes of measures and improvements achieved (22 December 2015).

Figure 2.4 illustrates the time-frame of the implementation process and the phases where modelling can play an important role.

2.1.4 Iterative process of arriving at the decision

When a compromise alternative can be identified, it means that all the stakeholders are satisfied. Thus the procedure ends and the alternative can be presented to the decision maker(s) as the optimal compromise among all the involved interests and environmental objectives (reply “yes” in Figure 2.3Figure 2.2). In the more common “no compromise” case, the analyst, the stakeholders and the decision maker(s) have to examine whether all possible ways to broaden the consensus have been explored. An iterative application of the planning process should be started in which, depending on the result, different alternatives for proceeding can be considered further.

The following alternatives may primarily be taken into account:

• mitigation of negative impacts by technical measures;
• compensation by external (political or economic) measures (e.g. by subsidies);
• introducing new alternatives in process control (e.g. reservoir building, extending wastewater treatment capacities etc.);
• adaptation through a social learning process with the stakeholders.

The last procedure is very challenging and promising. Fundamentals for this process are being developed within the HarmoniCA and HarmoniCOP projects (see Ridder et al. 2005). They are briefly introduced in Chapter 4.

Mitigation, compensation and proposing new alternatives

If an alternative achieves the consensus of the stakeholder majority, but not all of them, it is important to understand whether measures of mitigation and compensation would further broaden the consensus by satisfying some of the unsatisfied stakeholders. Then it is necessary to identify new kinds of actions to be included in the relevant alternative that specifically act upon the unsatisfied players. A next iteration (from step 2 in Figure 2.2) commences, and it ends when no more mitigation actions to broaden the consensus can be identified. If opponents still exist at this point, the possibility of compensating the disadvantage they perceive must be explored. Here, adaptive management comes into the picture (see Section 4.1).

Political choice and final decision making

Once all the possibilities have been explored and it is clear to the analyst, stakeholders and decision maker(s) that nothing else can be done to broaden the consensus, the procedure is terminated. At that point the set of compromise alternatives includes alternatives supported by the consensus of at least one sector. It is up to the political decision maker to select from this set the alternative of best compromise, i.e., the alternative that accounts best for the different interests and environmental functioning. Technically, the step takes the on form of an evaluation when there is only one decision maker.

The planning of measures

Stakeholders may define as many measures (action alternatives) as they find necessary. With each of them a “planning cycle” may be implemented. The same procedure applied for determining the Baseline Scenario is now repeated, i.e. different measures to overcome the observed gaps and problems are taken as input to the model (or model system) and the resulting effects/impacts (i.e. to simulate the system behavior). Responses capable of avoiding or overcoming the dangerous and undesired pressures on the river basin system are especially sought after.

The suggested framework (Figure 2.2) allows this task to be performed. However, it is recommended to begin with a few, most promising and attractive alternatives, and to present the result of these simulations to the stakeholders for review and discussion. Then they can develop a better understanding of the system and the range of possibilities for an efficient control. This is a first phase of social learning and a good opportunity for searching compromise alternatives.

In the best case, one of the investigated alternatives fulfils the expectations of all stakeholders or comes close to it. In the Spree river basin study for example (Section 2.3.5) the building of a new reservoir was such a solution, fulfilling the requirement of raising water availability to better satisfy water demand (of all stakeholders). Then the evaluation process (step 5) and also the negotiation (step 6) can be completed.

In other cases many more alternatives may be taken into account, for example, (a) other structures (such as levee systems, wastewater treatment plants, or the like, for controlling the processes), or (b) normative or (c) regulatory actions, as in the Verbano case study (Section 6.4). In such cases a set of alternatives is defined for the investigation, and by estimating effects and impacts a multiple “impact matrix” can be established to aid the evaluation (step 5). Optimization techniques and other advanced participation approaches may be included in the evaluation procedure to achieve consensus among the stakeholders in interviews, group discussions, hearings, workshops etc.

In the end, attractive alternatives can be found through negotiation (step 6), from which a set of compromise alternatives may be determined and delivered to the final decision maker (see Verbano case study in Section 6.4).

Iterative planning and conclusions

Whenever consensus cannot be achieved in step 6 (Figure 2.2), a new planning cycle must be started, normally with step 3 where new scenarios and/or alternative measures are designed or a change in paradigm is agreed on.

2.2 1st implementation phase: Assessment of current status, setup of monitoring programmes and evaluation of a Baseline Scenario (gap analysis)

The EU Member States have to undertake several key actions in the first phase of the implementation process of the Water Framework Directive by the end of 2006:

• the assessment of the current status of the water bodies, ready by 22 December 2004;
• the setup of monitoring programmes and implementation and maintenance of monitoring networks, which should be adapted regionally and be operational until 22 December 2006; and
• the evaluation of the water state under business-as-usual management (Baseline Scenario).

These initial works are of crucial importance for the further phases of the implementation as demanded in the WFD: to describe the water management problem (gap analysis), to design programmes of measures, and to evaluate the performance of the measures. Due to the lack of consistent data during the first implementation phase, models and other tools, especially Geographic Information Systems (GIS), in combination with monitoring information help to improve the description of the water bodies, to specify water-related problems and to optimize the monitoring network.

2.2.1 Identification and characterization of water bodies and optimization of the monitoring network

Annex II of the WFD describes a process to identify, categorize and typify water bodies (see also Guidance Document 2 “Identification of Water Bodies”, EC 2003). Type-specific reference conditions have to be identified for each water body type (e.g. lakes, rivers, reservoirs, aquifers). They will be compared with type-specific reference conditions for each water body type to assess the status of a specific water body or group of bodies (guidance on this topic can be found in Guidance Document 6 on “Intercalibration”, EC 2003). “Status classes” of numerical values for the biological quality elements of surface waters in Member State’s assessment systems have to be established, with the ecological status of rivers determined by the lower values (high / good / moderate / poor / bad) from the relevant biological and physico-chemical monitoring results. Groundwater will be classified only as being in either a “good” or “bad” status.

2.2.2 Joint use of monitoring and modellin

Water bodies are complex environmental systems with many unknowns and uncertainties incorporated due to the incomplete knowledge of the processes, to scaling aspects, and to the high spatial and temporal variability of the processes. A comprehensive understanding of the hydrological system, which can be formulated as a site-specific model description of the relevant hydrological processes, in combination with monitoring data, will be of great help in characterizing the water body under investigation and in identifying pressures. It is therefore obvious that the implementation of the European Water Framework Directive creates new challenges on the joint use of monitoring and modelling, and provides excellent opportunities to promote the interaction between monitoring and modelling (Jørgensen et al. 2007).

The models can serve as a tool to analyze, for example, different realizations of an unknown/uncertain hydrological environment (e.g. of the subsurface, Figure 2.5), to identify important flow and transport pathways, sources for contamination (point and diffuse), and impacts of buffers (reservoirs, riverine vegetation) and management measures etc. Based on the modelling information, it is then possible to find monitoring gaps and to improve the monitoring network.

However, due to the complexity of the monitoring/modelling problem, in most cases it is not possible to develop an optimal model or to design an optimal monitoring network. The reason for this is the amount of information and knowledge at hand in the first phase of the implementation process about the relevant processes in the specific river basins. A dynamic approach is therefore proposed (Jørgensen et al. 2007) to overcome this gap during the initial stage of the implementation. Models may indicate errors and inadequacies in the monitoring network. They can also aid in defining the reference conditions, interpolating observations and designing new monitoring networks, or optimizing existing ones in terms of location and number of stations, frequency of measurements, choice of (indicator) parameters etc. (Kampenhorst et al. 2005). The model is revised and updated as new data become available (about the use of models of different complexity and different data demand see Section 3.2.1).

Figure 2.85: Example of joint use of monitoring and modelling: 180 drill log information of the subsurface was used in combination with advanced geo-statistical tools to simulate a set of realizations of the subsurface corresponding to the input information (left: two realizations). Right: by overlaying the realizations, areas with a high reliability of the reconstruction are visible (in red, e.g. the locations of the drill logs), whilst areas with a poor reproduction also appear (blue). The blue areas are the locations where additional monitoring data are needed.

Steps to be taken following Figure 2.2:

Step 1: Problem description: The fundamental first step is to evaluate any existing information about the river basin, including relevant maps, articles and observations. There will be no data arising for the first assessment from the Article 8 monitoring programmes as they do not have to be operational until the end of 2006. These data should be available only later for subsequent assessments for future River Basin Management Plans. However, the countries do not start from scratch; they already have extensive and sometimes long-lasting monitoring programmes in place. These monitoring programmes have to be revised and likely upgraded, and the assessment of the ecological status has to be repeated in an iterative way to accommodate the new information. Problems arise if the data availability is insufficient to identify and characterize the hydrological system under investigation. Here, data assimilation also using expert judgment and modelling approaches in addition to ‘hard’ or traditional monitoring data will be needed to complete the data base.

Figure 2.6: Example of joint use of monitoring and modelling: a statistical methodology was applied for defining an optimal set of sampling intervals (for dissolved oxygen DO) for the operation of a river water quality model. Starting with an extensive set of measurements (left), it is the aim to reduce the number of observations to obtain just as much data as necessary for a calibration with an acceptable uncertainty and thereby to reduce the resulting range of uncertainty (see confidence interval 95 % high – 95 % low) in the parameters (right) (Vandenberghe et al. 2005).

Step 2: Conceptualisation (model setup): Water managers normally have a profound system understanding of the river basin under investigation (in form of a “mental model” resulting from their experience). The task is now to conceptualise the problem identified in the previous step and to set up a site-specific mathematical model of the hydrological system based on the information gained during the evaluation of the observations, whereby the model setup has to focus on the specific management problem. It is recommended to define indicators, such as water levels, water quality characteristics, fish species, and critical thresholds (criteria) of the indicators, to evaluate the state of the system. A modelling protocol should be designed to record the modelling steps (see Section 3.4 and Scholten et al 2004).

Table 3.1 (Section 3.2.1) introduces tools and methods applicable in addition to monitoring data to gain the necessary knowledge about the water bodies under investigation (for a more comprehensive general introduction to modelling and tools see Chapter 3). Evaluation of the data demand in order to operate these tools is included in the table. This is important, because the data availability, which is necessary to use the tools and models, determines their applicability to solve the management problem.

Simple approaches with low data demand, such as GIS-based analytical/statistical methods or simple conceptual models have to be the first choice in river basins where the compilation of the necessary data basis just started or is incomplete. More advanced approaches like physically-based distributed hydrological models integrating different aspects of the water cycle, the environment, and management sectors can be applied where the data support is adequate, or where models have already been applied and adjusted in earlier projects (for questions of model selecting see Section 3.2.1).

The choice of a tool is already very important in this initial stage because model applications in subsequent stages of the implementation of the WFD are likely to build on this first model setup and information gained during the first stage. It is therefore important to consider possible model applications in later stages of the implementation process while conducting the first model setup, compiling the data base and designing the monitoring network.

A good compromise would be to apply, due to the lack of more accurate data, a simplified description of the relevant hydrological processes first, to use it (in combination with monitoring data) to indicate knowledge gaps, and to determine what further investigations should focus on, and based on the first gap analysis to prepare a more complex and physically-based model which can represent, in an adequate way, the relevant hydrological processes. It may well be that the data support is not sufficient enough to fully identify the boundary conditions and parameter structure of a complex model, and the monitoring network has to be improved and designed accordingly.

Scenario definition: During this initial phase of the implementation process, there is no need to formulate a scenario (projection into the future), since the aim of the modelling exercise is to acquire additional knowledge about the current status of the hydrological system under investigation. Therefore, it might be sufficient to run the model in a steady-state mode (under assumption of no changes of the boundary conditions during the simulation process).

Simulation / estimation of impacts: The site-specific model should be applied with due consideration of the different possible descriptions of the basin environment and climate / management boundary conditions in order to check the agreement of modelling results with the observations and to identify knowledge gaps.

Evaluation of the results: The last step in the process is the evaluation of the results. The advantage of starting with a simple model description of the hydrological system (like a combination of GIS, statistical analysis and conceptual equations), is that they are easy to apply and give first (rough) results quickly and in a cost-effective way. For example, information on the long-term water and nutrient balance of the entire river basin can be obtained, although the spatial and temporal aggregation of the information will be high. Therefore, evaluating the results has to take into account that the uncertainty of the outcome of simple models is usually high (see Section 3.3). However, the achievable accuracy can be sufficient to characterize water bodies which are not in danger of failing to meet the WFD objective of reaching a “good chemical/ecological status”. Otherwise, in a second iteration, the work can focus on the areas of the basin where the water management problems could not be solved applying the simple model setup, with the help of a more complex and physically-based description of the hydrological system.

2.2.3 Identification of pressures and formulation of a Baseline Scenario

The next step recommended by the WFD is to collect and maintain databases containing information on the type and magnitude of the significant anthropogenic pressures the surface water bodies in each river basin district are subject to (Article 5 of the Directive). An assessment must be made of the vulnerability of the water bodies to the pressures identified and how likely water bodies within the river basin district will fail to meet the environmental quality objectives set under Article 4 of the WFD. This assessment will use any available data, while the extent of existing monitoring data will again vary greatly from country to country and also from river basin to river basin. Additional modelling information will be necessary (Figure 2.7) especially for the formulation of the Baseline Scenario.

Figure 2.117: Identification of pressures –the observed and simulated groundwater levels are shown (Hattermann et al. 2005).
(a) The observed trend in groundwater levels starting from summer 1983 cannot be reproduced taking only observed climate variability as driver.
(b) The groundwater hydrographs agree when the simulation is done assuming a management induced decline in the drainage basis of -0.35 m in the period 1983 - 1984 (d1) and of -0.80 m in the period 1990-1991 (d2).

Steps to be followed following Figure 2.2:

Problem description and goal definition: The conceptual problem is that solely applying monitoring information to identify the pressures on the system might not be sufficient. This is because the WFD explicitly demands to also include already planned management measures in the evaluation of the pressures (for example measures necessary due to national programmes or to implementation plans of the EU Urban Wastewater Directive etc.). These new measures might change the state of the hydrological system and result in additional pressures (see case study of the Marne Pilot River Basin and Seine-Normandie District in Section 6.2). One important method, besides data evaluation, that aids analysts in identify these pressures on the hydrological system is the formulation of a Baseline Scenario the anticipated “good status” (Article 4 of the Directive) can be compared to and any gaps in reaching it can be determined. Based on the understanding of these gaps, the programme of measures can then be planned according to the WFD to reach the “good status”.

A model description of the hydrological system taking additional management options into account, would be an option for analyzing the development of the hydrological system under investigation and for evaluating impacts of new management alternatives.

Conceptualisation (model setup): The model description of the Baseline Scenario has to build on the work which has been done to characterize the state of the hydrological system under investigation. Like in the previous step, it is recommended to consider data demand vs. data availability when selecting models to describe and evaluate the hydrological environment. While the modelling to support the characterization of individual water bodies can often focus on one component of the hydrological system (like groundwater), the modelling challenge to evaluate a Baseline Scenario is normally higher. This is so, because many different hydrological processes interact in a river basin, management has to be considered, and the approach has to be dynamical and unsteady-state (i.e., the boundary conditions may change during the modelling period, for example because of the new management measures).

Possible indicators to evaluate the state of the river basin are water levels and nutrient concentrations in groundwater, rivers and lakes. More challenging are ecological indicators (fish species, algae composition) and river morphology. The site-specific model has to prove that it is able to reproduce the relevant hydrological processes (calibration), also under changing boundary conditions (validation (see Section 3.2.4). The model setup and validation process should be recorded in a modelling protocol (see Section 3.4 and Scholten et al. 2005) to guarantee transparency in the modelling process.

It is strongly recommended to guarantee maintenance of the site-specific model setup, because it is very likely that modelling will also be needed to support the further steps of the implementation process, as listed below. Model maintenance will also enhance transparency, because it may be necessary to repeat the modelling exercise in order to investigate potential problems with the modelling results, or to include new management strategies.

Scenario definition: The Baseline Scenario is a business-as-usual scenario including already planned and formally agreed-on new management measures. External drivers may be population growth, globalization (socio-economic change), climate change etc. The optimal situation would be predictions being available. Otherwise, assumptions must be made on future developments (e.g. trend extrapolations) with stakeholders and experts participating (see Marne case study in Section 6.2). Another option here is to use stochastic models for time series generation into the future. For instance, one can generate precipitation by a method which represents a mixture of objective simulation techniques and stakeholder advice (see Spree river basin case study, Section 6.6).

Simulation / estimation of impacts: The simulation should be done first without representing the planned management measures in the model framework, and afterwards including them, to evaluate the impacts of the new measurements and the sensitivity of the hydrological system to the anticipated changes. Like in any modelling exercise, it is important to analyse the robustness of the simulation results as well or, in other words, the reliability / uncertainty of the results (see Section 3.4).

Evaluation of the results and alternatives: The evaluation of the long-term impact of the business-as-usual scenario on the ecological status of the river basin will be the key information to identify obstacles in reaching the objectives of the WFD. There cannot be a real negotiation of the alternatives here, since there is consensus on the new management measures, They will be taken into account according to the Baseline Scenario (by definition of the Baseline Scenario, they are already in construction or at least planned). Nevertheless, evaluating their impacts will give additional information about possible management options for the future, and the setup of the programme of measures will have to build on the results gained during the evaluation of the Baseline Scenario.

2.2.4 Example 1: Marne/Seine/Normandie – modelling and the Baseline Scenario

(For the full description of the case study see Section 6.2)

The basin characterisation according to Article 5 of the WFD intends to provide input to the decision-making process and the public participation from 2005 to 2009. This is a prepares a programme of measures that should be started by 2009 and aims at achieving the overall WFD objectives by 2015. Then integrating the current dynamics of the water status and policy appears necessary. Therefore, it is important to anticipate results from the implementation of existing European water directives before they have been completed. Some progress is expected in the near future (e.g. from “terminating” the implementation of the Urban Wastewater Directive and of the Nitrogen Directive), while at the same time some environmental factors may worsen (e.g. related to pesticides). Hence evaluating the business-as-usual (BAU) trends is unavoidable in deriving a Baseline Scenario (BLS).

Step 1 - Problem description and definition of the environmental objectives

The general methodology that was used for designing the modelling tools was indeed strongly dependent on the precise goals derived from the WFD context. The final goal is to reach the “good ecological status” in all water bodies until 2015. In order to reach this goal the following questions need to be answered:

Step 2 – Conceptualisation Based on this goal definition, the model-supported planning system can be described as follows, starting from the anticipated “end-results” and going up to the institutional and legal analysis. See the figure to the left for a schematic organisation of the system. Figure 2.8: Structure of the model-supported management planning in the Seine-Normandie basin.

This drivers-pressures model is linked to ecological quality models (see Figure 2.13Figure 2.8) by simulated GIS-localised net pressures. The modelling methodology used for the drivers-pressures model does not demand a specific technology. It is based on linear programming with current office software (such as Access).

The modelling methodology used for the status simulation tools is by far more complex. Although their basic “engine” stems classical ecological equations, they are developed through specific applications for the Seine basin, linking GIS with calculation engines.

Step 3 - Definition of scenarios

External drivers Context scenarios were built by referring to existing sector analyses, EU and national agricultural projections, national interpolated demographic data, etc. Adaptations of large-scale scenarios to the basin or to specific sectors, when needed, were based on expert forums and stakeholder knowledge.

Internal drivers In Baseline Scenario, internal drivers comprise the technical translation of the business-as-usual water policy into an “equipment programme” of implementing existing water-related directives and national provisions and orientations.

Step 4 and 5 – Analysis of effects and evaluation

The quantification of effects and impacts of scenarios relates to 2 linked important outputs:

• The evolution of pressures, impacts on rivers, and the corresponding ecological status.
• The assessment of economic cost of the equipment programme, and how it relates to the current expenses.

Figure 2.149: Risk assessment (likelihood of failing to reach the “good status”) map for Seine-Normandie.

The evaluation deals with policy-relevant issues of the Baseline Scenario results. Results are meaningful with respect to general assessment of the likelihood of failing to reach a good ecological status in Seine-Normandie water bodies and the cost of business-as-usual policy, comparing to current expenses, and evaluating of required changes in investment rates and volumes (see Figure 2.14Figure 2.9).

2.3 2nd implementation phase: Support for design and setup of programmes of measures and of the River Basin Management Plan

2.3.1 The River Basin Management Plan

Perhaps the most important and central implementation task according to the WFD is the establishment of programmes of measures (WFD Article 11) to produce and publish River Basin Management Plans (RBMPs) for each River Basin District (RBD), including the designation of heavily modified water bodies, by 22 December 2009 (Article 13, Article 4.3) . The RBMP serves as the main reporting mechanism for river basin district authorities to the EC and summarizes the results of the two first implementation phases. The functions of the plan are (after GD 11, cf. EC 2003):

Item (1) of the RBMP, the inventory and documentation of the water bodies, of the environmental objectives and of human impacts, was the main task during the first implementation phase and should be finished by December 2006. The main task for the second implementation phase is to set up programmes of measures describing how to reach the environmental objectives identified in the intercalibration process of the first phase. For each river basin district, the programmes of measures to be established by the end of 2009 are to describe the regulatory provisions or basic measures to be implemented in order to achieve the objectives of what the management plan defines for 2015. This includes taking pricing measures to provide users with incentives to manage water more efficiently. Measures may be decided on a national level in accordance with community and/or national laws (e.g. indicating responsible authorities, a reporting system, and defining protected areas, discharge control etc.).

2.3.2 Planning for an uncertain future – scenario definition and model support

The basic problem in designing the River Basin Management Plan lies in making it for an unknown or at least uncertain future and at the same time integrating water availability, quality and demand in a cost-effective way. The plan must take into account the development and variability (seasonal and inter-annual) of water availability, quality and demand. Meanwhile water quality is often strongly dependent on water quantity, i.e. stream flow and/or water volume (storage). This problem is illustrated in Figure 2.10.

The left side in the figure represents the past, which is known and can be evaluated on the basis of available monitoring and observation data. The future, on the right side, is unknown or uncertain, at least. As far as water availability is concerned, it is unknown whether and when a drought, or a sequence of dry years (with prolonged droughts), or an extreme flood period are to be expected. This is the case although the information on future vulnerability has to be considered in the programme of measures in order to avoid undesired impacts for the public under such conditions.

Figure 2.10: Typical situation in a planning study (Kaden et al. 2002).

Concerning water quality, the future is more certain since water quality can be influenced by local water and land use management, even though political decision making has its own level of uncertainty. Projections into the future are required, in the form of scenarios (possibly assumptions) related to future weather conditions (climate) and other external driving forces over the planning period of, for example, 20 years, calling for taking climate change and socio-economic growth into account. The scenario technique is an appropriate tool here. The scenarios should be internally consistent and plausible alternative projections into the future (rather than being misleadingly promoted as accurate and reliable predictions).<:p>

In summary, information on land use change and related scenarios of climate change is required in the form of:

(a) projections of the driving forces (especially external drivers) into the future (over the planning horizon of e.g. 20 to 50 years). These driving forces can be climate change or variability and socio-economy (agriculture, industrial development);

(b) a design of the measures that will definitely occurring (e.g. basic measures), and have already been planned or confirmed for implementation (the same as for the Baseline Scenario), and additional measures potentially leading to an improvement of the water status.

Figure 2.11: Example of a comprehensive model system to investigate land use and climate scenario impacts on water resources (example taken from the GLOWA-Elbe project, see Wechsung et al. 2005).

Figure 2.11 illustrates information flow (model system) investigating consistent land use and climate change scenario impacts on water availability and water quality. A similar information flow and model setup can be applied for investigating water demand (water management), susceptible to change due to technology changes or economic development, or in dependence on weather conditions. Considering global (external) drivers must be considered, since they cannot be influenced locally but can have a significant impact on the hydrological system (water quantity and quality). Regionalization of the global changes is important, Future data may not exist, especially for socio-economic boundary conditions, so expert knowledge has to be put to use.

During the scenario setup it is to describe water availability and quality under normal (average) conditions as well as under extreme conditions (low flows and droughts in summer and other seasons, but also high flow conditions and floods). In extreme conditions there may be too little water to fulfill water needs and to dilute pollution, or too much water, so that the capacity of sewage systems is exceeded. Often one must take the real sequence of the varying hydrological conditions and events into account reaching from normal” (average) years and seasons to wet and dry conditions, including extreme events (floods and droughts). A continuous simulation in the future (monthly or even daily) by using Monte Carlo techniques may be required for a longer future periods (for example 20 or 50 years).

Here a fundamental decision must be taken by the responsible modelling group (in agreement with decision makers and stakeholders): Should a stochastic generation technique be applied to generate long time series of water availability in terms of river flow (discharge), water quality, or precipitation? And what is the adequate time step in the time series generation (a month, ten days, a day)? Time series generation is the only appropriate solution in river basins with existing reservoirs, since intra- and inter-annual water storage plays an important role for demand periods. This case is discussed in the Spree river basin case study (see Section 6.6). Only such a technique makes it is possible to link wet / water surplus periods (seasons or years), where water can be stored with dry seasons and years, where stream flow becomes small or may even disappear, as for example in many ephemeral Mediterranean rivers. Similar conditions may also occur in other zones and environments, including the continental climate zone in Eastern Europe and in the Eastern part of Central Europe. It is advisable to try to compensate the deficits by water stored during the surplus periods, for example in winter or in flood seasons.

Similar statements can be made on water quality as it directly depends on water quantity. Planning under such conditions is a rather difficult and challenging task. The stochastic character and variability of hydrological processes must be considered. as do long time series of river flow, or any other variable of interest in the planning process, and exceedance probabilities.

Simpler cases may exist with planning done with defined reference values, for example with long-term average flow or any extreme flow. In turn, simplifying assumptions can be accepted in the planning process. However, in the more complex cases, as mentioned before, the troublesome efforts are required, some of which are explained in the case studies.

2.3.3 Model-supported design of the programme of measures

Models will certainly play a key role in facilitating the selection of measures to achieve the environmental objectives according to the WFD, and to develop the River Basin Management Plan. The reasons for this that planning has to advance several projections into future development of external drivers and management options. An appropriate model description of relevant processes in a river basin allows for assessing the impact of many combinations of the external drivers and different management strategies and facilitates decision making by eliminating unfavourable solutions,.

Steps to be done following Figure 2.2:

Step 1 - Problem description: The problems can be defined on the basis of experiences in the first implementation phase, the characterization of the water bodies, and the preliminary gap identification according to Article 5 of the Directive (see Baseline Scenario). The questions to be answered in this phase are:

• which management alternatives are appropriate to achieve the environmental objectives, and
• which of them have the highest cost-effectiveness and are supported by the pertaining stakeholders.

It is absolutely necessary to establish a good communication platform of water experts, stakeholders and modellers for broadly discussing the problem definition and goal identification from the start. Later on, these discussions help to set up a site-specific model addressing water related problems. They also aid in increasing the acceptance and understanding model results.

Step 2 – Conceptualisation and model setup: The model setup for supporting the design of the programme of measures is usually the most advanced component of the planning process, because it has to integrate many different hydrological variables interacting in the river basin, and it has to consider different management options. The approach has to be dynamic and unsteady-state, i.e. for covering changing boundary conditions during the modelling period to correspond with the designated management measures. The challenge in the river basin model lies in reproducing a variety of different possible management options (boundary conditions) and their impacts, unlike in the model setup for the Baseline Scenario. For that purpose, the boundaries of the hydrological system and the time and space domains must be defined carefully. The responsible authorities, stakeholders, and sectors involved should be determined (institutional and stakeholder analysis). Furthermore, the institutional and legal framework of the plan should be defined and considered in the model setup.

Following the DPSIR (Drivers, Pressures, State, Impacts, Responses) scheme, the relevant Drivers, Pressures, and State of the hydrological system, Impacts and possible Responses have to be defined. Identifying potential management options is primarily expert- and manager-based and is intellectual work. The final concept should be discussed with stakeholders and experts either individually or in group discussions to receive their approval, advice and comments. Therefore all activities that may raise consensus must be promoted in this step.

The potential management alternatives, environmental objectives and related indicators determine the tools and model setup for investigating pressures on the state of the system and for evaluating their impacts. The case studies Witte, Nete, and Dender in the Scheldt Pilot River Basin, and Verbano summarized at the end of this chapter give an insight into how the goal definition and stakeholder identification was performed in two example river basins.

Characterizing of the water bodies including preliminary gap identification was the goal of the previous implementation phase finalized by 2005. It can be of use here. The Site-specific model setup for the Baseline Scenario is further useful work from the previous phase. It can be applied and improved in this one to estimate the ecological status of the river basin while considering additional management interventions. The model setup benefits if the data needed for modelling was considered while the monitoring network was implemented according to Article 8 of the WFD. It should build on the existing modelling activities undertaken during the identification of the river basin districts and the investigation of the Baseline Scenario.

However, if no model description of the river basin exists, starting off by setting up a simple, e.g. GIS-based model description of the hydrological system is recommended. Similar to the investigation of the Baseline Scenario, the simple model setup may already suffice to compare different management options with adequate accuracy and credibility and to design the programme of measures. If not, an evaluation of the existing data bases compiled during the first phase of implementation until the end of 2006 is recommeded in order to assess the data support for a more complex site-specific model setup.

Step 3 - Scenario definition: Scenario definition plays a key role in designing the programme of management measures for meeting the environmental objectives that are pivotal to the RBMP. The scenarios determine the boundary conditions water management in a certain river basin will be based in future (see Section 2.3.2). Part of the scenario(s), especially the external drivers, will have to be chosen by experts or obtained by running models that describe the relevant processes producing the driving forces; e.g. the future scenario of rainfall. However, the WFD explicitly requests the involvement of the public in the planning process. The development of different water management scenarios offers an excellent opportunity to discuss and integrate different visions (alternatives) of future water and land use management in a given river basin by consulting various stakeholders and the public.

The challenge of scenario building is to identify a set of scenarios which combine the external driving forces (climate, socio-economy) and possible water management alternatives agreed upon during the participation process. The scenarios should also include extremes (droughts and floods) in order to determine the response of the hydrological status in a river basin under such conditions and to investigate the vulnerability of the system, and possibly climate change and its impacts.

Step 4 – Simulation and estimation of effects/impacts: The effects/impacts produced by each scenario and management measure (alternative) have to be estimated by computing the indicator values for each alternative. If there is more than one, An estimation has to be made for each scenario and realization of one. At the end of this step an impact matrix comprising the values of the indicators obtained for each scenario and alternative (or a number of alternatives applied jointly) can be produced. It aids in evaluating and discussing the results in support of negotiations.

Step 5 - Evaluation of the alternatives: Given the simulation results (impact matrix, see Section 1) the goal of this step is to determine the “value” that each sector assigns to each management alternative (as a rule, this “value” may not be linearly related to the values of the indicators). This impact matrix serves as a tool to communicate and discuss the impact of different scenarios of boundary conditions and management options and to evaluate the best compromise. Again, this evaluation has to be performed in consultation with the responsible water managers and relevant stakeholders. It may be necessary to repeat part of the modelling exercise and to include new management alternatives which were not included in the original set of scenario trajectories (cf. the interplay of adaptation in management and modelling in Section 1.1). This is why it is so important to ensure a flexible model setup, including the capacity to repeat the modelling exercise for a new set of boundary conditions and new management alternatives.

2.3.4 Example 2: Water quality modelling in the Scheldt Pilot River Basin

(For the full description of the case study see Section 6.3)

The case study describes the integration of a larger diversity of water management problems for the Dender tributary of the Scheldt Pilot River Basin, whereby this brief introduction focuses on water quality modelling.

Step 1 - Problem description and definition of the environmental objectives

According to the WFD, the Dender basin is classified as a ‘heavily modified water body’. The main water quality problems are high pollution levels, low DO (Dissolved Oxygen) concentrations, and high water temperature. Industrial pollution sources are mostly responsible for the water quality deterioration, followed by agricultural and untreated domestic pollution. The concentrations of heavy metals are high in the rivers due to the prominent industrial contribution.

Step 2 – Conceptualisation

Figure 2.12 gives a plan view of the Dender River basin. The geometry of the catchment, the river Dender and the 6 major tributaries are shown for the Flemish part of the basin. The figure also zooms into the urban drainage system connected to the Molenbeek tributary in sub-basin no. 431, an area responsible for a large part of the urban pollution of that river.

Figure 2.12: Plan view of the 12 subcatchments in the River Dender basin, including the locations of the urban drainage system (with 13 outfalls towards the Molenbeek tributary) in sub-catchment area No. 431. The basin has a total area of 708 km2 and encloses 12 sub-catchments areas.

A variety of physico-chemical water quality variables were identified to serve as indicators for the water quality improvement in the basin (for organic pollution: DO (Dissolved Oxygen), BOD (Biological Oxygen Demand); for nitrogen nutrients: NH4-N and NO3-N; for phosphorous nutrients: dissolved phosphorous, organic phosphorous, and PO4-P; and heavy metals), while the criteria include thresholds for physico-chemical river water quality according to the Flemish percentile-based standard.

To describe the impact of the various sources of water and pollution on the receiving water, models for the different subsystems in

Figure 2.12 have to be set up and linked (Figure 2.13). Based on the principles discussed in Section 3.2.1 of this document, a model type is chosen for each subsystem with the level of detail in the model structure expected to be the most appropriate. This is done for each pollution source and/or subsystem involved. Figure 2.13 illustrates the complementary use of detailed physically based and parsimonious conceptual models.

Figure 2.13: Integration of the submodels in the Dender River basin case study at the conceptual level, and complementary use of the detailed submodels (outer circle) and the conceptual submodels (inner circle).

Step 3 – Scenario definition

Three water management actions (forming the so called MAP action plan) were considered to improve the water quality, namely a) stricter emission standards for the industry, b) improvement of wastewater treatment infrastructure, and c) limitation of the agricultural pollution.

Step 4 and 5 – Analysis of effects and evaluation

By means of the parsimonious integrated model, a simulation was performed for all of 1993. Figure 2.14 shows the simulated time series of DO (Dissolved Oxygen), BOD (Biological Oxygen Demand) and NH4-N at km 35 for both the current water quality state and the sanitation state. The percentage of time, during which Flemish immission standards are exceeded, is shown. The figure shows that for the current water quality state the DO concentration remains below the Flemish percentile-based standard in 70 % of the time. In the worst case, the percentage decreases to 40 %. The effect is higher for BOD and NH4-N: the percentage of standard’s exceedence time decreases from 85 % to 5 % for BOD and from 9 % to 0 % for NH4-N. The concentration of NO3-N and P components seem to exceed the standard with the same frequency after sanitation. The effect of reduction in the nutrient losses due to agricultural activities is very low during heavy winter storms. The concentration of NO3-N and P components still exceed the threshold of the standard during that period.

Figure 2.14: Evaluation of two water quality sanitation scenarios on DO and BOD concentration changes (time series at 35 km distance along the Dender), and comparison with the Flemish percentile-based standard.

2.3.5 Example 3: Water Management in the Spree river basin

(For the full description of the case study see Section 6.6)

Step 1 - Problem description and definition of the objectives for the environment

Since 1980s much scientific and technical work has been done on water management planning in the Spree river basin because for two major reasons: The Lusatian lignite mining district within the basin had a serious impact on water availability and management in the basin, while at the same time a decrease in annual precipitation has been observed. The urban agglomeration of Berlin and the Spreewald biosphere reserve, two important sub-regions in the basin, have been directly affected by those impacts. The management objective for the future is to have a balanced and sustainable water management in the basin, protecting the biosphere reserve and guaranteeing a minimum water inflow to Berlin concerning impacts of climate change.

Step 2 – Conceptualisation

In order to investigate how the existing hydrological system can fulfil the water demands and how the situation can be improved by appropriate measures (action alternatives), a large scale long-term simulation into the future of the hydrological conditions in the river basin was required . For this purpose an earlier developed hydrological and water balance modelling system of the Spree river basin was used to describe the system as indicated in Figure 2.15.

Figure 2.15: Structure of the hydrological (rainfall-runoff) model system for the Spree river basin upstream of Berlin (left). It also shows the River system, water users, balance points and reservoirs in the Spree river system upstream of Berlin and in the connected Schwarze Elster system (right).

The model system describes, in a coupled way, the processes of runoff generation, runoff routing in the river system, water availability and water balance at various user points and reliabilities of water supply at these points (river cross-sections), as indicated in Figure 2.15. The structure of the hydrological modelling system (rainfall-runoff model) for the 22 selected sub-basins of the Spree river basin upstream of Berlin is shown. The structure of the river system model embraces about 400 water users, 170 water balance points, 14 reservoirs, and 50 dynamical elements.

Step 3 - Scenario Definition and Alternative Design

The reference climate scenario time for the investigation was the recent past. The general assumption for the future climate was that temperature was the driving force of the observed climate trend. A statistical regional climate model was used to generate 100 time series of the climate of the next 50 years, representing the observed climate variability and driven by the temperature trend from a Global Circulation Model.

In addition, the trend of coal lignite mining was extrapolated into the future (50 years), and the pumping of groundwater into the rivers in the remaining mining area was projected accordingly, based on numerical groundwater modelling. These were the two principal scenarios for future development.

Step 4 and 5 – Analysis of effects and evaluation

In order to investigate how the existing water management system can fulfil the demands and how the situation can be improved or changed with the two action alternatives above, a large-scale long-term simulation of the hydrological conditions in the river basin was performed. For this purpose the hydrological and water balance modelling system WBalMo® was used.

Figure 2.16: Reliabilities of water inflow to Berlin for different management alternatives.

The main result of the investigation is presented in Figure 2.16. The lowest curve illustrates the reliability of water supply at the inflow gauge to Berlin for the existing system (without additional action). It is discernible that with the system currently in place the demand can only be fulfilled in the autumn and winter months (October to March). Deficiencies may occur in all other months. They may already start in April and increase in May. Meanwhile the reliability drops below 50 % in the long-term average in summer months (June, July and August), i.e. the system is absolutely insufficient. The figure also shows how the situation can be improved if a new reservoir (Lohsa II) is additionally included in the system and / or if water is transfered from the Odra River. However, in the meantime, the idea of water transfer was abandoned, mainly because of insufficient water quality. Instead, other alternatives or measures were considered and numerous scenario analyses were made.

2.3.6 Example 4: The MANTRA-East approach

(For the full description of the case study see Section 6.5)

Introduction

The main focus in this case study is the scenario methodology. The qualitative scenarios based on the stakeholder opinions and policy scientists were translated into quantitative scenarios. Finally, they were used as input for computer models (i.e. ICT-tools) to determine the future riverine loads into a trans-boundary lake as well as the ecological impact in the lake.

Step 1 - Problem description and goal definition

Lake Peipsi is one of the largest European lakes (surface area of 3531 km2). The lake and its basin are located in the Baltic Sea drainage basin: The lake discharges into the Narva River, which in turn flows into the Gulf of Finland. The basin is shared by Russia (59 % of the basin area), Estonia (33 %) Latvia (8 %) and Belarus (0.3%).

The major environmental problems in the Peipsi Lake Basin evolve around water eutrophication and reduced fish stocks. Eutrophication due to significant nutrient loads in Lake Peipsi represents a lake water quality problem.

Step 2 - Conceptualisation

The MANTRA-East project had three main objectives. The first was to evaluate the applicability of the EU Water Framework Directive in the EU border regions. This included an assessment of the state of eutrophication (e.g. ecological status) in lakes and river basins, as well as the development of strategic lake and river basin tools for source apportionment, storage, and time-trends in nutrient loads. The second objective was to develop methods for improving communication and utilisation of scientific information. The third objective was to develop institutional mechanisms and policy instruments for decision making in planning of measures under conditions of transition and uncertainty.

To assess changes in emissions and their effects on nutrient runoff in a quantitative manner, a straightforward long-term nutrient emission and runoff model was adopted. Since the nutrient emissions and transport rates are variable in space and time, it was a logical inference to use a geographical information system (GIS) to model the past and future changes in nutrient emissions, and simulate the resulting river loads to Lake Peipsi.

Step 3 - Scenario definition

The approach decided on was following the various steps in the Planning Framework with scenarios. The main stages towards achieving outcomes were to identify the key variables, ask key questions for the future, and determine the most probable scenarios. In a regional perspective the key factors considered were: economic development, and trans-boundary cooperation, resulting in four different alternatives (Figure 2.17):

• high economic development / high level of cooperation (‘fast development scenario’);
• high economic development / low level of cooperation (‘isolation scenario’);
• low economic development / high level of cooperation (‘business as usual scenario’); and
• low economic development / low level of cooperation (‘crisis scenario’).

Figure 2.17: Basic scenarios for the Estonian-Russian border.

Steps 4 and 5 - Estimation of the impact and evaluation of the alternatives

Loads of total Nitrogen and total Phosphorus into the lake have decreased during the 1990s, in particular the former. This decrease is spread evenly over the Russian/Latvian (Velikaya River) and the Estonian part of the drainage basin. Total Nitrogen loads are expected to decrease in all scenarios except for in scenario II (Target/Fast Development), mainly influenced by an increase in the River Velikaya load (Russia/Latvia). Total Phosphorus loads are projected to decrease under all scenarios.

Results from an ecosystem lake model revealed that reduced riverine loading of Nitrogen enhances the blue-green algae growth. More precisely, the highest N/P ratio and the less favourable conditions for cyanobacteria can be expected in case of the ‘Fast development’ scenario (Figure 2.18).

Figure 2.18: Reaction of phytoplankton to different loading scenarios in 2015-2019. Source: Noges et al. (2003).

2.3.7 Example 5: The Möll study - Integrative river management in the Alpine region

(For the full description of the case study see Section 6.7)

The Austrian river Möll was used as a pilot river in order to (1) apply the “Leitbild-concept” for specifying the characteristics of the good ecological status according to the WFD, (2) evaluate the current status of the river, and (3) develop and evaluate management alternatives (necessary measures) both from the ecological (hydromorphology, fish, terrestrial vegetation) and economic (cost-effectiveness) perspective.

Step 1 - Problem description and definition of the environmental objectives

The Möll River is a characteristic Alpine river that has suffered a series of anthropogenic impacts and uses that are typical for the entire Alpine region. Today the Möll River is subject to many human interferences. In the 1960s and 70s the river was systematically channelised and stabilized. This totally changed the morphological characteristics. Moreover, a series of power stations were constructed, causing two large impoundments, long river stretches with water abstraction as well as a peak flow regime typical for hydropower and reservoir flushing. These factors, along with the intensive land use, explain why only fragments of the natural riverine landscape remain today and underline the urgency for measures to improve the ecological status.

Step 2 – Conceptualisation

A Cost-effectiveness analysis (CEA), as indicated in the WFD (Appendix IIIb), was used to compare measure implementation costs and ecological improvements. A version of the “fixed-effectiveness” approach was selected in which the degree of target fulfilment (“good status”) was already specified at the beginning of the project. The costs of achieving the target in various scenarios are investigated. Based on this method, cost effects of individual measures on such economic sectors as agriculture, forestry, hydropower, trade and industry, were investigated. The intangible effects arising from the project (e.g. improved nature protection) were also incorporated and qualitatively described.

Step 3 – Scenario and alternative definition

Nine scenarios of measures were defined: five uncombined alternatives for each type of impact (water diversion, hydropower-related flow peaking regime, channelization and bank stabilization, clear-cutting of floodplain forests, reservoir flushing), and four combined alternatives representing combinations of these.

Step 4 and 5 – Analysis of effects and evaluation

The Leitbild concept contained a map from the period prior to river modifications (1830), historical photos and near-natural reference sites of comparable rivers. A five-level scale was used for the ecological evaluation: “high status” (according to the Leitbild), “good”, “moderate”, “poor” and “bad status”. A quantitative analysis of the ecological quality elements was performed for 19 subsections of the river. Quality elements were hydromorphology for habitat conditions, fish fauna for aquatic biocenoses, and terrestrial vegetation for the floodplain character.

An expert judgement served to predict the ecological status for each subsection and scenario. The predicted status was compared to the present one (status quo), lying in the categories “moderate” and “poor”. Figure 2.20 shows the visualisation of one of the restoration scenarios (“restoration of river and floodplain forest”).

Figure 2.20: Current situation of the channelised river bed (left). Visualized restoration measures such as the reinitialisation of braiding river sections with floodplain forests (right)

Within this assessment process the effects of the different alternatives on hydromorphology, fish fauna and vegetation were predicted. River channel restoration clearly represents a key measure. Nonetheless, morphological restoration measures alone were insufficient to attain good ecological status. Only in combination with additional measures such as low-flow augmentation and / or attenuation of hydropower-related peaking or management of reservoir flushing, the values fell below 2.5, reaching the “good status” range (see Table 2.2).

Table 2.2: Ecological evaluation of the different scenario alternatives.

2.4 3rd implementation phase: Support for implementing the programmes of measures

The implementation period from 2010 to 2012 (WFD Article 11-7) is the phase the management measures of the River Basin Management Plan come into action in. While implementing the measures, it is likely that implications will lead to deviations from the original management plan. In addition, water agencies have a demand for models which describe more sectoral components of the water management system to manage such systems as storage reservoirs and wastewater treatment plants, for example. Therefore, two possible model applications are relevant during this period:

• Support to adapt the River Basin Management Plan and the programmes of measures:
• Support of operational management of water resources and hydrological networks.

2.4.1 Adjustment of River Basin Management Plans

The implementation of the measures agreed upon while designing the River Basin Management Plans faces planning (and modelling) with reality. It is very likely that unforeseen problems will arise when the measures are put into practice. It is recommended that an evaluation of the new measures and the improvements achieved is performed at regular intervals to sort out implications as early as possible. From a technical point of view, the model setup to support this process might be the same as the one used in the previous phase, but it is possible that model adaptations are necessary.

As indicated before, it is strongly recommended to store the model setup which was used for the design of the programme of measures (Section 2.3) in order to be able to reproduce the simulation results having led to the selection of specific measures of the River Basin Management Plan. In case one of the originally planned measures proves to be problematic or ineffective, this model setup can be used again in implementing the measures to adjust specific management tasks and to find alternative measures. During the adjustment of the programme of measures, it is also important and necessary to update the model setup to new monitoring data and other experience/information gathered during the implementation exercise. This gives an accurate model description of the hydrological processes allowing analysis of the implementation problems. The continuous evaluation of the success of the measures and of the model setup will also facilitate the major revision of the River Basin Management Plan as demanded in the WFD for the period 2013-2015 (Article 15, see Section 1.1).

Generally, the model implementation steps which have to be undertaken here are the same as described in Section 2.3. Again, the model setup should be maintained in order to reactivate it for the major evaluation of the River Basin Management Plan.

2.4.2 Operational modelling

One of the major fields hydrological models are applied in today is operational modelling of specific hydrological components to support short-term (day by day) water management. Examples are simulation and management of water storage reservoirs, flood control, irrigation, groundwater management (pumping and treatment), wastewater treatment, and discharge of urban channel networks.

The restriction to simulate only specific components of the entire hydrological system, often under well defined (man-made and controlled) boundary conditions (channel architecture, reservoirs), allows to achieve a detailed description of the relevant processes and hence the high modelling accuracy needed in operational modelling. In many cases, models for operational water management have to be linked to other hydrological models (e.g., precipitation-runoff models) to investigate the long-term impacts of specific pressures on the system. In any case, they are important in the process of adjusting the short-term management of several technical measures which will be implemented according to the River Basin Management Plan.

Steps, following Figure 2.2:

Step 1 - Problem description: The environmental objectives and possible gaps resulting in the programme of measures to overcome these gaps have been described in the River Basin Management Plan. They can be updated if new monitoring data are available.

Step 2 - Conceptualisation and model setup: The selection of operational models is determined by the implemented technical measures, such as hydraulic models for water level control, a water quality model for management of water treatment plants etc. Another relatively simple task is to define indicators (nutrient concentrations, water levels, reservoir storage volumes), criteria (thresholds of concentrations, water levels) and boundary conditions (channel architecture, in- and outputs). Some of the boundary conditions are taken from online monitoring networks for making it possible to react to the ongoing changes in the boundary conditions and to adjust the model results to the actual conditions. The challenge lies in calibrating the model using historical time series to an accuracy satisfying the management demands. This is crucial, because the simulation accuracy demanded in operational modelling is much higher than in other model applications. For example, a small underestimation of nutrient concentrations in sewage discharge might lead to algae blooming and oxygen deficit in surface water bodies. An inaccurate simulation of the water level in rivers may lead to a failure of the flood preparedness system, or to problems in river navigation. It is therefore recommended to perform a comprehensive sensitivity and uncertainty analysis to identify problematic management conditions. It may be necessary to implement the operational model in a broader model system in order to define the necessary boundary conditions, which may change over time. Such a model environment could be e.g. the model setup described in Section 2.3.3.

Step 3 - Scenario definition: The scenarios for testing the robustness of the management solutions can be defined on the basis of scenarios for designing the River Basin Management Plan. Pressures on the state of the system may come from external drivers (climate variability or climate change) and internal changes in land use, urban sprawl and water management (due to population growth, increasing standards of living, industrial development etc.). They can be defined by experts by applying empirical methods or generated by using model outputs.

Step 4 – Simulation and estimation of effects/impacts: Operational simulation and forecast of short-term impacts have to be performed in order to aid in near-future water management. Long-term simulations have to be carried out under scenario conditions for each of the possible scenarios to produce a matrix of the indicator values (nutrient concentrations, water levels, etc.) for each scenario, both for average and for extreme conditions.

Step 5 - Evaluation of the alternatives: Short-term responses have to be implemented to adjust the daily water management in case the indicator values lie beyond the satisfactory range (determined by threshold values). The likelihood for such events to happen can be reduced by evaluating the long-term simulations under scenario conditions, by identifying problematic boundary conditions and by adjusting the implementation of the measures (for example building water treatment plans) to extreme events of a certain likelihood. As for the other modelling exercises, this has to be done in consultation with the relevant experts, and it might be necessary to repeat the modelling campaign.

2.5 4th implementation phase: Support for evaluating the programmes of measures and improvements achieved

The WFD demands a major evaluation of the programme of measures and improvements achieved (Article 15 of the Directive). The time frame for this activity is the period 2013-2015. Gaps to achieve the environmental objectives have to be identified and the River Basin Management Plan has to be updated accordingly. Just like in the design of the programme of measures, this will include projections into the future for evaluating the success of the first implementation period under long-term conditions. A continuous evaluation of improvements achieved by the new measures should be done during the implementation of the measures, as described in Section 0. The major evaluation of the programme of measures starting in 2013 will have to build on these experiences and, in addition, include long-term projections into the future in form of changing boundary conditions (climate, land use, and water management). The simplest solution in terms of model support would be to apply the site-specific model setup used for designing the River Basin Management Plan and to adapt it taking the updated monitoring data and improved management experiences into account.

The modelling steps can be structured in the way described in Section 2.3.3. The problem description has to be based on the gap analysis demanded in Article 15, the conceptualisation (model setup) has to be upgraded if additional indicators need to be investigated. This includes another calibration and validation of the model for the hydrological processes under investigation with due consideration of the updated data records (of the period 2010-2012). The scenario definition should be adapted to the newest information on the long-term trajectories of the relevant boundary conditions. The simulation and estimation of impacts will render the indicator values under changed boundary conditions, bearing in mind the updated information on the state of the hydrological system. The WFD demands for public participation and the evaluation of the results has to be met in consultation with relevant water experts and stakeholders. It may be necessary to repeat the modelling exercise several times, including additional management and adaptation strategies, and considering an adjusted programme of measures accepted by all stakeholders.

3 Modelling – a primer for practitioners

P. Willems, F.F. Hattermann, and Z. Kundzewicz

3.1 Introduction

Mathematical models implemented on computers are of paramount importance to support water managers and decision makers in implementing River Basin Management Plans as required in the WFD. The aim is to do everything possible to find and design appropriate model solutions and implement these in practice, based on the needs and demands of potential clients and customers (Abbott & Refsgaard 1996). The common use of mathematical models to support decision making in water management makes it indispensable for water managers to have a certain understanding of modelling including model types, model selection, model setup, but also limitations. This is important, because water managers are often those responsible for commissioning a modelling project. Hence, their basic understanding of modelling issues is really necessary to formulate the project framework and to settle the modelling tasks. Moreover, it is in order to be able to evaluate the model results in terms of robustness and reliability.

Several guidance documents recently developed in different EU projects deal with various aspects relevant in the modelling process. These aspects include model benchmarking and selection (BMW), model linking (HarmonIT, Gregersen and Blind 2005), monitoring and uncertainty assessment throughout the whole planning process (HarmoniRIB, Refsgaard et al. 2005a), and Quality Assurance (QA) (HarmoniQuA, Refsgaard et al. 2005b). Access to all this information is provided by means of a dedicated web-site that contains links and references to ICT-tools, benchmarking reports, guidance documents, and a tools-selection manual that guides the user to the tools which are adequate for the issue at hand, the characteristics of the river basin, the data availability etc. (see http://www.wise-rtd.info and Chapter 5).

3.1.1 What is a model?

The term “modelling” is commonly interpreted as replacing an object under consideration by another object - a model. One of the reasons of doing so is the ability to draw information about the real object of interest from examining the model. The model imitates (mimics) selected aspects of the (possibly very complex) real object of interest, which are deemed important in the study at hand, while for economical computation and for clarity, the model should omit those aspects which are deemed unessential in the specific study.

The formalized models belong to two general categories (see Figure 3.1): material models and symbolic models. The former represent a real system by another real system. This category can be subdivided into a class of physical, also called iconic (look-alike), models (e.g. hydraulic laboratory models of a dam or a channel, built in an appropriate scale) and analogue models, e.g. electrical analogues - electrical circuits whose behaviour resembles (in terms of some variables) the performance of another physical object beyond the class of electrical circuits. Performance of material models should be similar, in the relevant aspects, to the object under consideration and should be easier and cheaper to study. Experiments on material models can be conducted under more favourable and observable conditions, while experiments (i.e. feeding the system with controlled input signal and observation of the resulting output signal) on the object may be difficult, if not impossible.

Today, most formalized models fall into the class of symbolic models, whose main sub-class, mathematical models, are by far most commonly used. The triumphal march of mathematical models, which are transformed to a computer code, is mainly a result of the emerging computational capabilities offered by ever more powerful and affordable digital computers and increasingly comprehensive, user-friendly, and flexible computer software.

Mathematical modelling can be understood as the use of mathematical constructs (formulae, relations, graphs, equations, systems of coupled equations) to describe features of systems or processes. Virtually, every use of a mathematical equation to represent links between variables, or to mimic a temporal or spatial structure of variable(s), can be called mathematical modelling.

Figure 3.1: A real world catchment (top) and two possible formalizations: a conceptual scheme of a computer model (left) and a material model (right).

There are many classifications of mathematical models, often based on the division of the models into the following dichotomies: static – dynamic, linear – nonlinear, stationary – nonstationary, lumped – distributed, deterministic – stochastic, continuous – discrete. Most of these categories are self-explanatory and their interpretation can be easily understood, so it is not discussed here - for extensive explanation, see Dooge (1973), Singh (1988).

One criterion of classification of models is the degree of their physical justification. Here one can distinguish at least three classes:

• Mechanistic, process-based (physically-based) techniques involving (possibly complex) systems of equations expressing physical laws and theoretical concepts (e.g. rigorous nonlinear partial differential equations of water flow and substances in irregular open channels with floodplains).
• Purely empirical, black-box (or data-based) techniques, which match the input and output signals of the system, without mimicking the internal structure. The instantaneous unit hydrograph (IUH) or artificial neuron networks (ANN) belong to this class.
• Conceptual models, consisting of simple elements, which may simulate, in an approximate way, processes occurring in the basin. There have been several attempts to render conceptual models closer to physically-based ones by physically interpreting conceptual parameters (cf. Kundzewicz, 1986), making it possible to apply a conceptual model to ungauged basins.

Combinations of the above mentioned classes are conceivable, for example the combination of process-based techniques and conceptual models. They are often referred to as process-oriented models: The process of interest is formulated in a physically-based way, while the boundary conditions may correspond to simplified conceptual approaches.

3.2 Specifics of the modelling process – techniques and tools

In the following, a brief introduction is given to modelling strategies for integrated water resources management in river basins, with several references to activities and reports of ongoing EU projects. The preliminary items, such as the problem definition and scenario development are not discussed here, because they were tackled in Chapter 2. The following sections will focus on techniques and tools and give a general introduction into the model setup and modelling strategy for integrated water resources management in river basins, with references to relevant guidance documents and reports. The first part focuses on model selection. The next one introduces component models for impact analysis, discussing their complexity and data needs. It is followed by a section describing problems in model integration when component models, designed to simulate specific processes in a river basin, are linked together in order to build a holistic model system for multidisciplinary integrated impact analysis for water management in river basins. Special requirements with regard to models and model systems in the light of stakeholder participation and decision-making processes are discussed, as are the reliability of modelling results under unavoidable uncertainties in input data, model parameters and model structure. The last section is devoted to quality assurance in river basin modelling.

3.2.1 Model selection

Selecting an appropriate model to investigate a certain problem related to water management is a very important task, because it determines the character of the simulation results in terms of reliability, robustness, and temporal and spatial resolution. For several reasons, it is not always possible to apply a model representing a large number of hydrological processes and having the finest temporal and spatial resolution. The model complexity has to correspond to the problem under investigation and should include only the essential relevant processes, while the low-order effects and links have to be ignored. In addition, more complex models or model systems usually have a far higher data demand to identify the model parameters (transmissivity, residence times, turn-over etc.). Without accurate data input, such a model will fail to produce very precise outcomes. The data support is indeed the most important factor for determining the accuracy of modelling results. If high accuracy results are required, this may mean that additional monitoring campaigns have to be initiated.

In principle however, modelling is possible with all levels of data quantity and quality (Højberg et al. 2006). A complex physically-based model can also be applied without having all the site-specific data, taking instead information from literature and expert assessment. The modeller then has to take into account that the quality and quantity of available data will affect the uncertainty of the model outputs and as such the usefulness of the model simulations (Refsgaard et al. 2005b, Refsgaard & Henriksen 2004). The relevant question is therefore what the requirements to the accuracy of the modelling are. Only when the answer to this question is given can the data requirements be assessed in a meaningful manner (Højberg et al. 2006).

Figure 3.2 illustrates the selection process: The first step is to define the modelling target (e.g. modelling of surface water quality) including the relevant indicators (e.g. nitrogen and phosphorous concentrations) and criteria (e.g. maximum concentrations). The next step is to define the necessary level of problem complexity and accuracy of the results, which are possibly (i) high (many interacting processes, unsteady-state boundary conditions, high risk for humans if wrong decision is are taken) or (ii) low (only one component of the water cycle has to be investigated, boundary conditions do not change, low risk). It is also important to estimate the available resources in terms of manpower, data and experience. The problem complexity and the available resources determine the complexity of the site-specific model . It can be applied to support management in the basin under investigation. Again, whenever a high accuracy of the modelling results is needed, one has to make sure that the resources to prepare and process the data and to operate a more sophisticated model are available.

Figure 3.2: The model selection process. It is important to have a clear vision of the modelling target, available resources, and complexity of the problem including the accuracy required.

Four combinations are possible:

Table 3.1 lists the most important model types classified by their complexity. Also included are descriptions of the domain of their applicability and characteristics of their basic data demand.

Table 3.1: Models to simulate different domains with different complexity.

There are many mathematical models on the market encapsulated in software packages and products that are neither products of research institutions nor commercial software companies. It is often difficult to evaluate the relative advantages and disadvantages of proposed models in operational use. There are a number of factors and criteria involved in the selection of a model, as compiled in Table 3.2 (after WMO, 1994, modified):

There has been a broad consensus on the statement that the range of model complexity should agree with the nature of the problem and the amount of data available to build and validate the model (Singh 1995). The most appropriate model type may exist for each particular application, depending on the project objectives, data requirements, and the data availability. Whenever different model types with different detail are compared (only those that meet the modelling objectives, see also Section 3.2.1), the most appropriate model has the optimal balance between uncertainties resulting from the model-structure on the one hand, and model-input and parameter uncertainties on the other hand (see Section 3.3 and Figure 3.3). The problem still under discussion is: Which solutions apply to the particular problem at hand, and what is the appropriate complexity of models and model systems in order to fulfil the tasks in question (Beven 1997).

Figure 3.3: Balancing different types of uncertainties to determine the optimal model structure detail for a specific application (Willems et al. 2003).

Because of the fact that the amount of data available for model applications is not the same for all subsystems and emission sources in a watershed, the most appropriate model types may differ for the various submodels. More data are available for some subsystems, allowing for a more detailed model to be implemented. Yet, harmonization with respect to the submodel accuracy is needed (coupling a very good and a very poor model is not likely to provide optimal performance). Matching appropriate complexity of models and tasks at hand is not easy (see Beven 1997).

Models with a higher degree of detail than the optimal model detail (cf. Figure 3.5Figure 3.3) can also be called ‘over-parameterized models’. Preferably, they should not be included in a model system. The number of parameters is too large for an accurate calibration on the basis of the limited data available. Due to the over-parameterization, this means that different sets of model parameters with equally good fits to the observed model output may exist. So there is no unique optimal solution of the parameter identification problem. Further, parameters of submodels are considered poorly identifiable as they cannot be identified from the available data accurately and in a unique way. However, some submodels may be needed conceptually in the model structure. These submodels cannot be excluded from the model, even if they are poorly identifiable. The parameters of these submodels are, however, difficult to calibrate and will remain uncertain. It is important to be aware of how these uncertainties may affect model outcomes and the decisions based on them (see also Section 3.4). Instead, additional effort should be made to improve the data support of the model.

3.2.2 Component models

Component models in this context are models which are designed to investigate specific processes in river basins, such as soil water processes or features of urban drainage systems. Typical examples are groundwater models, hydraulic models to simulate discharge in rivers and channels, lake models, and reservoir models (see Table 3.1Table 3.1). The advantage of using a component model is that it is possible to focus on a single hydrological process that might be of certain interest. This may allow an advanced model description of this single process in terms of the physical background to be used to represent the basic flow and transport equations. Component models are mostly applied on small spatial scales (1-1000 km2) allowing to compile an adequate data base with sufficiently high spatial and temporal resolution. The opportunity to focus on a single process in combination with the high data support in such small-scale applications, if properly implemented, leads to modelling results of high reliability. However, the basic assumptions made in Section 3.2.1 concerning model complexity, parameter identification, and data support also apply component models, of course.

The boundary conditions for applying component models are often of the steady-state type (assuming that inflow/input is equal to outflow/output), at least in the calibration phase. However, if scenario impacts of changing climate or management drivers are the very topic of the model exercise, then the model cannot be applied in a steady-state mode, and the changing boundary conditions have to be defined in a data base coupled to the model, or the component model has to be linked to other models delivering the changing boundary conditions.

In summary, component models are:

• process specific;
• often advanced in terms of the mathematical background to solve the basic physical equations;
• often calibrated in a steady-state mode;
• mostly applied on a spatial scale up to 1000 km2.

One important sub-species of component models are tools for operational modelling (often in real time). Examples of such tools are models to simulate and manage water storage reservoirs, flood control, irrigation, groundwater pumping and treatment, wastewater treatment and discharge of urban channel networks. Operational models have some specific common characteristics, for example:

• they are mostly component/sector specific;
• built for short-term forecast (of water quality, river water level, reservoir storage), often including an updating option (adjustment component, depending on the stochastic properties of the time series of so-far forecast errors);
• often linked to an on-line monitoring network;
• advanced in terms of management and visualization interfaces.

Operational modelling has been one of the first and most important model applications in hydrology, because this type of model has been used in quasi-industrial sectors like sewage treatment, freshwater supply etc. since the beginning of model applications. The common use and economic importance of these models has led to the development of several high-end products. However often they are not freely available. The restriction to simulate only specific components of the entire hydrological system, often under well defined (man-made and controlled) boundary conditions (channel characteristics, storages), permitting a detailed description of the relevant processes and therefore the high modelling accuracy needed in operational modelling (e.g. for flood protection, oxygen control).

Models for operational water management as well as other component models can be linked with other hydrological models such as precipitation-runoff models to investigate the long-term impacts of specific pressures on the system.

3.2.3 Model systems for integrative and multidisciplinary analysis

Scientific improvements in process knowledge as well as improved computation power enable integrating models describing different hydrological processes in the river basin into one model system (see Figure 3.4). Depending on the river basin, specific objectives and management options, such an integrated model may consist of several component models dealing with

• eco-hydrological aspects (water balance, diffuse and point pollution sources, transport of nutrients, groundwater depletion, irrigation);
• urban water management (urban drainage systems, wastewater treatment);
• reservoir control, dams, municipal water supply;
• socio-economy;
• regional climate change;
• land use change;
• agriculture and forest management;
• multi-criteria analysis of the results.

Multi-scale (in temporal and spatial sense) integrated models or model systems of water resources are the preferred tools to support water management planning for implementing the WFD. These models deal with processes on different time scales (for example short-term processes like generating overland flow and long-term processes like retention of nutrients in groundwater) and on different spatial scales (for example point-source pollution and diffuse emissions of nutrients).

Figure 3.4: Different model components linked for multidisciplinary integrated impact analysis (example taken from the German GLOWA-Elbe project).

Before starting the integrate the models into one model system, some basic choices have to be made. These choices are about the following issues:

• relevant processes which have to be included;
• the manner of integration (dynamic or “cascade” approach, the latter being a combination of linear reservoirs in series, for instance);
• how the integrated model will be available after it is completed;
• the running time;
• the degree of interaction with the end-user;
• will individual models retain their original code?

Based on these choices, a software architecture for the integrated model incorporating and linking the individual models should be designed. Some components can be integrated directly, while others have to be adapted or rebuilt. After connecting all the individual models a first sensitivity analysis can be carried out to eliminate non-crucial model elements in the given application.

The selection of the individual models to be linked in the model system is important, because a large number of different simulation tools already exist for the detailed modelling of individual sub-components of the river basin system, such as the river network, urban drainage systems, wastewater treatment plants, groundwater resources (aquifers), etc. These models are most often based on (full or simplified) equations of mathematical physics, describing the laws of conservation of mass (continuity) and momentum, plus transport equation. They have high data and computation needs. Normally, the integration of these model types in one model system for integrated catchment management is only feasible at relatively small spatial scales, for the simulation of short time periods, for well defined tasks, and for investigating single components of the water and nutrient cycle (e.g. groundwater simulations or representation of pollutant transport in rivers). Their application has to be supported by a large number of observations and input data (see Section 3.2.1).

The spatial and temporal scales are, however, key elements of an integrated modelling approach at the river-basin scale and at the level where integrated strategic planning and water management is conducted. Long-term effects (such as accumulation and retention of nutrients in water bodies, including aquifers) have to be considered in the modelling and decision-making processes. By the integrating different subsystems at these levels, physical phenomena that interact at largely different spatial and temporal scales have to be described in a combined way. Two examples are:

• The large spatial scale of the natural surface, subsurface, and groundwater runoff in a hydrographic catchment has to be combined with the small spatial scale of the man-built urban drainage systems covering a part of the catchment.
• The short-term impact of combined sewer overflows (relevant time scale in the duration of minutes to hours) has to be combined with the long-term impact of accumulative pollution from wastewater treatment plants, diffuse domestic and industrial wastewater and groundwater leaching.

Therefore, an integrated modelling tool at the river-basin scale has to consider these large ranges of temporal and spatial scales. It also means that all subsystems with significant contributions to the water management problems under consideration have to be modelled (see Figure 3.9Figure 3.5). This may lead to a large number of systems and a large spatial scale to be modelled, so that simplifications in model complexity and spatial representation of the landscape characteristics have to be found and implemented.

Figure 3.5: Integration of models at 2 different levels (detailed models at small scale level, and conceptual models at large scale level) (Willems et al. 2003).

The application of semi-distributed process-oriented models (Wade et al. 2002, Krysanova et al. 1998, Bicknell et al. 1997, Schulze 1994, Arnold et al. 1994) is often a good compromise between the conditions of data availability and model complexity. In such models, processes of interest are represented in physically-based mathematical equations, and fluxes between them are expressed by simplified or conceptual formulations. The spatial heterogeneity of the landscape pattern is aggregated on the hydrotope scale (hydrotopes are unique for example in their broad soil and land use characteristics) so that the spatial variations in land use and water management can be taken into account. These models normally operate on a daily time step and are designed for meso- to macroscale river basins and long-term simulations, which are the spatial scales and time horizons relevant for the implementation of the WFD. Some of these models are public domain and their source code is available free of charge, making it possible to exchange individual modules or to extend the model by using new modules representing further processes, according to the needs of the user. These computer software packages are well documented and have been validated in many model applications. Nevertheless, the basic principles mentioned in Section 3.2.1 regarding the model complexity and the data requirements have to be considered when applying these tools.

A subclass of semi-distributed process-oriented models are the so called eco-hydrological models. Eco-hydrology links ecology, i.e. science of interrelationships of organisms and their environments, and hydrology, i.e. science of water cycle in the nature, dealing with the properties, distribution, and circulation of water.

Apart from the abiotic climatic / meteorological input (precipitation, temperature, etc.) and water-related components (water quantity: level of surface or groundwater, water flow, soil moisture, snow cover and snow water equivalent; and water quality, in a chemical, biological, and physical context, including sediment generation and transport), a mathematical model of an eco-hydrological system should deal with biotic elements (e.g. links between water and vegetation, calculation of net primary production, crop yield, and biodiversity, related to both flora and fauna) and possibly with socio-economy issues (such as assessment and evaluation of ecosystem services).

For integrated modelling purposes, a limited number of interfaces are currently set up for the “user-friendly” linking of some of these existing modelling systems in a number of EU projects. One example is the open modelling environment to be set up in the ‘HarmonIT’ Project. As indicated before, the BMW Project provides guidelines to and tools for the user-friendly selection of benchmarked models. The EUROHARP Project gives guidance towards harmonized procedures for the integrated quantification of catchment-scale nutrient losses from European drainage basins. The primary objective of the CLIME Project is to develop a suite of benchmarked models that can be used to simulate the responses of lakes to changes in climate. Integrated water management of transboundary catchments and integrated Decision Support Systems are the goal of the TRANSCAT Project. For links to information relevant to all of these and other related projects, see http://www.wise-rtd.info.

3.2.4 Model calibration and validation

The procedure for building the optimal model structure (free of unnecessary poorly identifiable sub-models) for a certain application is called ‘model structure identification’. In the data-based modelling methodologies, this identification is only based on observed time series for the input and output variables. Generally, it is important to highlight the purpose of the model. If, for example, land use changes need to be evaluated, there is no other choice but to use a distributed model etc.

Model parameter values can be derived by an identification (estimation or calibration) process (see Figure 3.6). Whenever the parameters represent physical properties, their values can be derived from measurements in a direct or indirect way (e.g. by using correlations between the physical properties and the measurements). This can be done in a more direct (and thus more accurate) way for models with a higher physically-based level. If no measurements are available, parameter values can be roughly estimated on the basis of the modeller’s experience (with applications to other systems). Wherever observations (e.g. measurements) are available for some model output variables, the parameter values can be calibrated by matching the simulated with the observed model output.

Figure 3.6: The calibration process.

Parameter calibration is called ‘optimal’ when it derives ‘optimal parameter values’. Three conditions are embedded in the definition of optimal model parameters:

The two latter conditions on accuracy and physical realism can be partly tested by model validation with data which were not used for calibration.

Generally, calibration depends highly on the purpose of the model. Different requirements are set for flood routing and flood damage evaluation models than for modelling water quality or aquifer recharge.

Figure 3.137: Observed data (a) of river discharge and different stages of model calibration (b).

The term validation describes the process when the adequacy of a site-specific model is tested. This is mostly done using data not applied earlier during the calibration process. The procedure is also called split-sampling test, since the available data set was split into a part of data for calibration and a part of data for validation. In course of validations the model is run for a time-period which does not overlap with the calibration period, or it is applied in a neighbouring, comparable region/basin without calibration (e.g. to model an ungauged basin). More advanced methods to validate a model are described in Madsen et al. (2007). Combined with an uncertainty analysis (see next section), the validation indicates the predictive capacity of the site-specific model.

3.3 Reliability of model results

Throughout the whole modelling process it is crucial to bear in mind that the input data, model parameters and model structure have an inherent uncertainty, which will affect model outcomes and hence also the decision-making process (Refsgaard et al. 2005b). The uncertainties can be quantified by carrying out an uncertainty analysis.

Complementary to the model structure identification and model calibration procedures discussed in Section 3.2.4, an uncertainty analysis provides the means to evaluate the effects of the different sources of uncertainty involved in the integrated modelling. It supplies decision makers with important additional information on the uncertainty in the data and the information that they use as a basis for their decisions. It allows one to evaluate the ‘decision-support power’ of the integrated model. This way, water policies can be set up for which efficiency can be achieved up to specified acceptable risk levels. The uncertainty analysis also provides a modeller and a decision maker with information on the importance of various types of model limitations and uncertainty sources. Part of uncertainty analysis should ideally be a comprehensive sensitivity analysis assessing the contribution of individual sources of uncertainty (see Refsgaard et al. 2006). Combined with the uncertainty analysis, a sensitivity analysis allows defining priority directions for model improvement, and removable model components (e.g. replaced by a constant value of some parameter in the broader system).

In each mathematical modelling application, different sources of uncertainty may be involved. These sources can be classified into three categories:

• uncertainties in the model input variables: ‘input uncertainties’ (see Section 3.3.1);
• uncertainties in the model parameter values: ‘parameter uncertainties’ (see Section 3.3.1);
• uncertainties originating from the imperfect description of the physical reality by a limited number of mathematical relationships or due to an incomplete knowledge of the physical system: ‘model structure uncertainties’ (see Section 3.3.2).

Collectively, the three types of uncertainty sources determine the total uncertainty in the model output (and furthermore in the indicator variables and the water management decisions). The uncertainty in model output may be used in a decision-making process by means of a risk analysis in terms of a ‘water management risk’.

Harmonized techniques and representative river basin data for assessment and use of uncertainty information in integrated water management are the core of the HarmoniRIB Project (Refsgaard et al. 2005a), and Refsgaard et al. 2006 gives a comprehensive guidance to uncertainty assessment in hydrological modelling.

3.3.1 Uncertainties in input data and model parameters

Model input uncertainties may result from measurement errors (if the model input is directly measured) or from errors in estimation (in case the input is estimated indirectly from existing historical observations or in scenario analysis with external driving forces). An example is rainfall, where the measurements normally underestimate the real precipitation amount. It has to be interpolated to have input data at sites without observations. Also, for those model parameter values (e.g. storage parameters) that are measured or estimated (on the basis of experience), the parameter uncertainties may be caused by measurement or estimation errors (see Figure 3.8). For parameters that are calibrated on the basis of observations of the model output and model input variables (i.e., via inverse modelling), parameter uncertainties arise whenever erroneous, incomplete data (with significant data gaps in “critical” range and/or limited time series of measurements) are used for calibration or whenever the calibration procedure is not adequately rigorous. Parameter uncertainties decrease with the increase of the length of a time series of observations used for the calibration.

Uncertainties in input data and model parameters can be represented by ‘stochastic terms’ taking the form of random variables. The randomness in the magnitude of these variables is described by probability distributions. The stochastic component of an uncertain temporal variable may be correlated in time. Similarly, the stochastic component of an uncertain spatial variable may be correlated in space. The degree of temporal and spatial autocorrelation may be quantified by an autocorrelation function or a semivariogram. Multiple uncertain inputs may also be cross-correlated. The stochastic terms can be added to the variables treated as uncertain (the model input variables x, the model parameters p, and the model structure, to come up with the model output Y; see Figure 3.8).

Figure 3.8: Representation of three different types of uncertainties in probabilistic modelling by means of stochastic terms (Willems et al. 2003).

To calculate the propagation of uncertainties and their contribution to the total uncertainty of the model results, different techniques can be used (see Refsgaard et al. 2006). The most popular technique is the Monte-Carlo method (either using Ordinary Sampling, Importance Sampling or Latin Hypercube methods). These techniques also allow to calculate and rank the sensitivity of the model results to single data inputs and model parameters, where the sensitivity can be expressed as the correlation of changes in the data or parameters to changes in the model output. Alternatives to the Monte Carlo technique are based on an analytical variance calculation. These usually involve some degree of simplifying the model response behavior.

Example: Equation 1 shows a simple linear equation to calculate fluctuations in groundwater table depth (Smedema & Rycroft (1983)), assuming that the variation in water table height h [m] at time step t is linearly related to the rate of change in groundwater recharge Rc in mm per day: Equation 1 Here, S is the specific yield [m3 m-3] or flow active porosity. The reaction factor a is a function of the hydraulic transmissivity T [m2 d-1] and the slope length L [m]: Equation 2 At the macroscale, the basic geo-hydrological input data (transmissivity, specific yield) are usually not available. Therefore the value of a has to be calibrated. Figure 3.9 shows the fluctuations in groundwater table simulated by Equation 1 for an experimental plot (Figure 3.9a) and the possible range in simulation results for varying the model parameters transmissivity (Figure 3.9b) by 25 % and the specific yield (Figure 3.9c) by ± 1.5 %. The results show the sensitivity of the modelled groundwater fluctuations (or of Equation 1) to changes in input parameters. This is important considering that geo-hydraulic parameters are normally difficult to estimate at a regional scale. They also show that the sensitivity of the results varies with changes in the input data (groundwater recharge), which are normally also uncertain. Not shown in this example are uncertainties due to deficits in model structure. They can only be revealed by comparing them with monitoring data or by a model comparison (see Section 3.3.2). Figure 3.9: Groundwater recharge and groundwater table as calculated by Equation 1 (a) with S = 0.02 m3 m-3, T = 80.0 m2 d-1 and L = 800 m. (b) shows the range of the results for varying T between 60.0 m2 d-1 and 100.0 m2 d-1, and (c) for varying S between 0.005 m3 m-3 and 0.035 m3 m-3.

3.3.2 Uncertainties in model structure

Model structure uncertainties are often dominant. They are caused by the modeller’s limitations to describe physical reality and by complex scaling issues. They can be considered the uncertainties even remaining in the idealized case of error-free input (measurements of perfect accuracy) and optimal calibration. Theoretically, if all input data and parameter uncertainties were known in the same way as the input and parameter uncertainties, even the model structure uncertainties could be described for the different submodels (different mathematical relations or subsets of relations) separately. However, some authors indicate that this is not necessary (Willems 2003). Also, a lumped representation could be used by considering the total uncertainty of all model structure uncertainties of the model output variables. The separate description for the different submodels (or the model structure ‘uncertainty decomposition’) is only possible for models with a transparent model structure and a calibration procedure. Uncertainty decomposition, also derives a transparent uncertainty structure. This has important advantages. The contributions of uncertainty sources to the total uncertainty in the model output can be quantified and compared. The comparison of the uncertainties resulting from the data on the one hand, and the uncertainties resulting from the model structure on the other hand is particularly interesting. Ragas et al. (1997) called these two types of uncertainties ‘operational’ and ‘fundamental’ uncertainties, respectively. Seeking an appropriate balance between them is recommended. Whenever the operational uncertainties dominate, much more attention should be given to data collection than to model structure improvement in an attempt to improve the model results. Whenever the fundamental uncertainties dominate, the reverse is recommended. By comparing the contributions of the different uncertainty sources, efficient measures to reduce the total uncertainty in the model results can be determined.

3.3.3 Quantification of uncertainties and how to cope with them

The different sources of uncertainty can be quantified in a stepwise manner. The model input uncertainties are estimated independently first, since they do not depend on the other uncertainty sources. They can be represented by means of stochastic terms that are added to the input variables of the model. A large number of random simulations are performed to quantify the contribution of the input uncertainties to the total uncertainty in the model output variables. Parameter uncertainties are quantified independently on the basis of the data used for calibration (e.g. with the help of regression) or they are estimated on the basis of experience. They are represented by probability distributions and correlations for all model parameters. After the measurement error in the observations has been compensated, the total uncertainty of the model output variables is quantified on the basis of the observed model residuals (difference between modelled and observed values of the model output variables for historical periods). If all input data and parameter uncertainties are known (idealistic rather than realistic what, in practice, is seldomly the case), the remaining uncertainty (total uncertainty minus the contribution of the input and parameter uncertainties and the measurement errors) can be considered the model structure uncertainty. Whenever additional measurements are available for some internal state variables, uncertainties can be decomposed further for different submodels (Willems et al. 2003).

During the procedure of quantifying uncertainties, the correlations between the different model output or state variables should be taken into account. Part of the remaining uncertainty in the output variables (after subtracting the input and parameter uncertainties and the measurement errors) may be explained by the model structure uncertainties of other output variables. For instance, this applies to the water quality variables that are the output of a combined hydrodynamic - water quality model. In order to calculate the model structure uncertainties in the water quality model, the contribution of of those in the hydrodynamic submodel should be subtracted from the remaining uncertainty in the (output) water quality variables. So the different uncertainty sources have to be quantified for the most upstream submodels first. Based on the same concept, whenever different sub-catchments are modelled, the uncertainties of the most upstream catchments have to be quantified first.

In risk management, the estimation of the uncertainty in the model results is used for limiting the probability (or the risk) of inconsistent decisions or ‘wrong’ designs. These risks exist whenever water management decisions (e.g. flood control strategies, more severe pollution emission standards) or engineering designs (e.g. hydraulic control structures, water or wastewater treatment infrastructure) are based on uncertain information (in this context model results). More specifically, risk is defined as:

Risk = probability of occurrence of failure * measure of consequence risk

In this formula, the multiplier should be viewed as a ‘convolution’ operator because different types of occurrences (e.g. at different time scales and with different magnitude) and consequences can contribute to the risk. With respect to this definition, the risk of inconsistent decision making and design can be assessed after probabilistic modelling and after the consequences of inconsistent decisions/designs have been quantified. Contradictory to probabilistic modelling just involving technical aspects, social and economic aspects should also be taken into the equation of quantifying the consequences. More specifically, the consequences are determined by the damaged elements, their ‘vulnerability’, and the ‘values’ attributed to social and economic aspects. Therefore, if the vulnerability of the element under investigation is high, the risk can only be minimized by minimizing the data, parameter and/or model structure uncertainty.

For instance, flooding , comprises material and human damages (the latter may include fatalities), and vulnerability is a function of exposure, sensitivity to damages, and adaptive capacity. For water quality problems, the damaged elements mainly are linked to the aquatic life. In comparison with material damage of flooding, it is much more difficult to quantify these ‘values’. This is also the case for valuing human lives lost during severe floods.

3.4 Quality Assurance

Over the past three decades, computer based model systems have been increasingly used to analyse pressures on water resources and to support water management in general. A rapid development of hydrological models was observed, mainly due to the vast increase in computational power and data availability along with the development of powerful Geographical Information Systems (GIS). The trend to base more water management decisions on model studies is likely to be reinforced in the course of implementing the EU Water Framework Directive (Refsgaard et al. 2005b).

However, what is often lacking is (a) an adequate documentation of the entire modelling process and (b) the quantification of the predictive capability of the site-specific model. The former (a) includes documenting the goal definition, the data pre-processing, the model set-up, the calibration and validation and the single simulation steps. The aim is to be able to reproduce each modelling step leading to the final result. The latter (b) includes a comprehensive validation and uncertainty analysis of the model

The procedural and operational framework to assure technically and scientifically adequate performance of a modelling study is called Quality Assurance (QA) (NRC 1990). According to Scholten et al. 2005, Quality Assurance for water management can be defined as protocols and guidelines to support a good application of models in water management with two main components: QA in development of model codes; and QA in relation to applying models for water management. Refsgaard (2002) classifies QA guidelines according to how much focus is put on the consensus-building process between modeller and water resources manager. He distinguishes three classes:

The most recent set of QA guidelines was developed by the HarmoniQuA Project (2002-2005) (Scholten et al. 2005). Within the HarmoniQuA toolbox, the complex process of modelling is decomposed into tasks defined at levels consistent with the conceptual understanding and action repertoire of the associated key-players (e.g. modellers, water resource managers, auditors, other stakeholders, and the general public). The body of knowledge describing these tasks has been compiled by experts within various domains of water management, discussed until a consensus was reached and stored in a Knowledge Base (KB) with an ontology-based structure, also including a glossary of terms and concepts used in model-based water management.

Figure 3.10: The five steps and 45 tasks of the modelling process in the HarmoniQuA knowledge base (Refsgaard et al. 2005b).

The following set of software tools provides functionality in the HarmoniQuA system (Refsgaard et al. 2005b):

• guideline tool: generates guidelines from the KB;
• monitoring tool: monitors all activities within a modelling job and stores these activities as a single model journal in a model archive;
• report tool: generates reports from a model journal;
• advisor tool: advises modellers in new modelling jobs based on decisions and choices of previous jobs and associated model journals in the model archive.

The idea of the tools is that each single modelling step is guided, documented and stored. Emphasis is placed on transparency and reproducibility, accuracy criteria, model validation, and uncertainty assessment (beyond the technical, modelling related uncertainties). Thereby, the five main modelling steps as outlined in Section 2.1.3 and illustrated in Figure 3.10 are decomposed into 45 tasks considering the order and possible feedbacks. The tools and related publications can be downloaded under http://harmoniqua.wau.nl, and Refsgaard et al. 2005b give a comprehensive introduction into QA.

3.5 Synthesis

There is a plethora of models and it is not easy to get an overview of what is available and which is the optimal model for a particular problem. Inflation of models and the lack of comparison exercises between models make it difficult to select the proper model for a given application. The issue of selecting the criterion for comparing models is a critical one. It cannot be a „beauty contest”. Models should be judged by how well they do their job in a particular application. From a practical viewpoint, a model should work well by rendering results comparable to observations. In doing this, it should have an acceptable value of the goodness-of-fit criterion measuring the difference between the observed and the modelled values of the variable in concern. Scientists also seek internal consistency, non-ambiguity, and a consideration of all the most important contributing processes. Summarily, they want to be certain that the model works right for right reasons.

Typically, there is a trade-off between accuracy and complexity. Simple models usually produce less accurate results than those from complex (more costly and time-consuming) and physically-based models (provided that appropriate data are available). However, at times, simple models may produce reasonable results which are comparable to those of the complex models, providing a clue that a simpler model should be chosen. Models with balanced components are advocated. Since the weakest link drives the quality of the whole, making the already good component even better may not influence the overall quality, while elimination of a “bottleneck” may lead to a visible improvement.

When selecting models and using them, it is always essential to remember the range of permissible (legitimate) model applicability and the assumptions taken in model development. Further, it is important to examine whether a model has been thoroughly tested in a variety of environments, including conditions similar to those in the study at hand. Some reports on model use (e.g. scientific articles) may be misleading, since they may only offer scarce illustrations referring only to one or a few test cases in situations the model work reasonably well in. Whether or not the model could work in much different conditions remains open.

Distributed or semi-distributed models are becoming increasingly common. According to Abbott & Refsgaard (1996), they help us „do everything that is reasonably possible … for the purpose of analyzing problems and of designing and implementing remedial measures”. They allow a user to benefit from the distributed data fields made available through remote sensing (providing highly distributed observation, including in inaccessible areas), and the GIS revolution (that enhanced the capacity of spatial studies). However, although remote sensing has the potential to provide masses of spatially-distributed records, these remotely-sensed data are only indirect representations of physical variables, and converting this information into values of the variable in concern is not free of problems. The GIS techniques, which used to be applied in hydrological mapping are now increasingly being used in hydrological modelling. Distributed models are conceived to use all kinds of data and information related to hydrology (including extreme events), topography, soils, land use, meteorology, climatology, geology and hydrogeology, erosion and sediment transport, plant physiology, ecology, economy, etc.

A search engine finds numerous entries related to the term „distributed hydrological model” (in the case of Google: 451,000 entries were found on 23 December 2005). However, despite the potential usefulness, distributed hydrological models are not applied commonly. For instance, the distribution of the pioneering and „iconic” SHE model has been „far below the initial expectations of its developers” (Abbott & Refsgaard, 1996). Among the reasons of this state-of-the-are: tradition, fear of unnecessary complications, but also lack of sufficient data (such models require the existence and availability of large amounts of data with the appropriate accuracy), lack of adequate understanding and knowledge of the processes. For a discussion about the scientific foundation of physically-based models at different scales see Beven (1996).

4 The role of participation, social learning, and adaptation in complex systems

Claudia Pahl-Wostl

4.1 Paradigms of adaptive management and the role of participation

In recent years stakeholder and public participation has received increasing attention in water resource management with it’s strong tradition in the engineering and technical sciences. The increased awareness of the human dimension is related to the insight that improved governance and integrated solutions are required to deal with the complexity of today’s water related problems. This is also emphasized in European legislation. The European Water Framework Directive (EC 2000) prescribes the involvement of interest groups and the public at large in river basin management. A Guidance Document was developed on this issue (GD 8, EC 2003c). Some authorities may prefer to fulfil the minimum requirements to comply with the legal regulations; others may perceive a chance to fundamentally change the practice of river basin management here, for example from a technical control paradigm to an adaptive management paradigm.

The role of participation is tremendous, in particular in the discussion on scenario definition and alternative design, action, criteria and indicator identification, and in the evaluation and negotiation steps. These participation processes are often performed in a straightforward manner and with simple means, i.e. without introducing advanced experiences and recent progress in communication techniques, and without any adaptive management practices whatsoever. These practices come increasingly into the picture and need more attention.

In any case it is necessary to provide arguments and collect experiences on the potential, practices and limitations of participation.

As an introduction one can distinguish the following types of participation:

• Information: The lowest form of participation is information. It is not truly participation, yet information is a prerequisite for any empowerment of citizens. Access to information is crucial to allow different interest groups and the broad public to participate in environmental decision making.
• Consultation: In the case of consultation different stakeholder groups and the public at large are asked to give their opinion on a management plan / scenario.
• Active involvement: Stakeholders and the public are actively involved in river basin management. In this case one may call it a co-production of knowledge, a co-decision-making. Active involvement can range from just having interviews, discussions with the authorities and experts, to actively contributing to policy development (co-designing), influencing decisions (co-decision-making), or even full responsibility for (parts of) river basin management.

Active involvement of stakeholders and the public at large is the most interesting and challenging. It can result in social learning, and this is essential for achieving integrated water resources management. The methodology to be proposed must take care of this fact.

Moreover, Pahl-Wostl et al. (2004) define adaptive management as “a systematic process for continually improving management policies and practices by learning from the outcomes of implemented management strategies. The most effective form of adaptive management employs management programs that are designed to experimentally compare selected policies or practices by evaluating alternative hypotheses about the system being managed. As it is defined in the approach promoted here, adaptive management has as another target - to increase the adaptive capacity of the (water) system. It is aimed at integrated system design. The problem to be solved is how to increase the ability of the whole human-technology-environment system to respond to change rather than to react to undesirable impacts of change. Hence it is a pro-active management style. Increasing the ability for change includes, for example, increasing the use of small-scale technology or combining formal regulations with informal institutional settings“.

Introducing these ways of thinking into the methodology of river basin management would open new horizons for further progress, especially in cases where the traditional control-based paradigm ends in “no-consensus”.

4.2 What is social learning?

Social learning is a dialogue-based process through which different stakeholders interact to achieve an inclusive, systemic, shared understanding of a given system, set of issues and ways to manage them. It increases the capacity of different authorities, experts, interest groups, and the public to manage their river basins effectively. Collective action and the resolution of conflicts require people to recognize their interdependence and their differences, and learn to deal with them constructively. The different groups need to learn and increase their awareness about their biophysical environment and about the complexity of social interactions

Processes of social learning should contain the following elements (Pahl-Wostl et al. 2004).

• Building up a shared problem perception in a group of actors, in particular when the problem is not very well-defined (this does not imply consensus building).
• Building trust as a base for a critical self-reflection, which implies recognition of individual mental frames and images, and how they pertain to decision making.
• Recognizing mutual dependencies and interactions in the social environment.
• Reflecting on assumptions about the cause-effect relationships in the system to be managed.
• Reflecting on subjective valuation schemes.
• Engaging in collective decision and learning processes (this may include the development of new management strategies and the introduction of new formal and informal rules, including adaptive management).

The nature of the processes of social learning will determine the types of decisions to be taken and will particularly promote innovation. Many intricate problems in water management require social groups to start communicating and processes of social learning . Collaborative governance to overcome the fragmentation in responsibilities characterizing wide areas of water management.

4.3 Role of models to support public participation

The tradition of science and the use of models are based on the assumption that the scientific observer is outside the system and produces objective knowledge that can be tested against empirical/experimental investigation. The predictive capacity of a model is generally assumed to reflect the quality of understanding the processes characterizing system behaviour. However, the situation is different for social systems. Human beings may change the rules under which they operate when they are faced with new knowledge and/or change their mental frames and expectations on an issue – hence the tradition of “hard” systems analysis does not apply anymore and new tools have to be developed (Checkland 1993, Pahl-Wostl 1995). Mental models about system behaviour are crucial for the analysis of environmental problems and the choice of measures and strategies. The social sciences acknowledge that reality is partly constructed. The importance of a socially constructed reality also suggests that models may serve as communication tools in processes of social learning.

Pahl-Wostl et al. (2004) have summarized this development and characterized the present situation, ongoing developments, and challenges. These challenges need to be taken into account in further research, and the best practices should be further developed and introduced into practical application.

For example, Pahl-Wostl et al. (2004) explain she explains that “mental models are shaped by the role of actors in a social system, their previous experience and cognitive biases resulting from heuristics that allow human beings to survive and act in a very complex and partly unpredictable world. Mental models determine the selective processing of information. Experience may help to construct a context from few pieces of information, to draw analogies to previous situations and select a type of response and behaviour that is deemed to be appropriate based on previous experience. Sometimes selective information processing may prevent learning and the adaptation to a changing environment – this applies for individuals, for enterprises or for scientific organizations”.

Mental models may be proven to be factually false. Then they should be corrected in agreement with the other actors.

4.4 Rules of good practice

Participation is a resource-intensive process. So preparing it well is important since it may be better to have no participation at all than an ill-prepared and badly managed participatory process. A first checklist for issues to be considered is as follows:

Different stakeholder groups may have different conceptions of what constitutes a ‘good’ participatory process within a given decision area – who should be involved when, for what purpose, and how? How should the success of participation be measured?

A range of views have been disclosed and the existence of conflicting views on what constitutes good practice was recognised as a significant challenge to implementation (Webler & Tuler 2001). Processes put in place must not only acknowledge differences of opinion regarding objective and legitimacy. They also must incorporate them somehow.

The final HarmoniCOP handbook "Learning together to manage together – Improving participation in input and parameter uncertainties water management" (see http://www.harmonicop.uos.de/handbook.php) gives practical information about participation processes in river basin management and about supporting the implementation of the public participation provisions of the European Water Framework Directive. It thereby will provide a basis for the development of improved integrated models and decision support tools.

5 The WISE-RTD webportal: Experiences, guidance and tools

Wim deLange, Patrick Willems, Fred Hattermann

It is one of the core objectives of HarmoniCA project to support the transfer of knowledge between the world of WFD-implementers/policy-makers and the research/technology world. One of the means through which HarmoniCA aims to reach that objective is the ‘WISE-RTD web-portal’. This web portal will contain information on the experiences gained by pilot projects involved in the implementation of the WFD as well as information on tools, technologies and methodologies, developed by research projects and projects on the WFD implementation. The web portal has different user-entries, such as for WFD implementers, policy-makers and water managers, which guide different kinds of users to available knowledge by using pictures and terminology, generally known by that particular user-group (see Figure 5.1).

Three types of information sources

(i) Guidance

The use of RTD results in policy implementations requires different types of guidance, depending on the user and the specific tasks. They can be classified broadly in guidance for scientists and policy implementers.

(i.a) Guidance for scientists

Scientists are mainly interested in receiving technical guidance on the use of a specific RTD result, which most often takes the form of a methodology, technique, ICT tool or mathematical model. Because the technical documentation on these techniques or tools can be interpreted as an information source of type (iii), the technical guidances cover more general topics such as guidance on the use of tools (e.g. on model selection, model linking, model calibration and validation, model sensitivity and uncertainty analysis, etc.), the monitoring process, the stakeholder participation, etc. Such technical guidance is set up by various RTD projects (e.g. projects of the CatchMod cluster of the EC, complemented and generalized with technical guidance offered by HarmoniCA; e.g. Refsgaard et al. 2005b).

Figure 5.1: The WISE-RTD web-portal supports the transfer of knowledge between the world of WFD-implementers/policy-makers and the research/technology world.

(i.b) Guidance for policy implementers

Policy implementers are more interested in information about techniques and tools that can be used for the various policy implementation tasks, in information on the data needs and other costs/benefits. The 14 CIS Guidance Documents are examples of such guidance pertaining to the WFD and its implementation tasks. These GDs are classified as per type of WFD tasks or per water-related discipline (economic analysis, analysis of pressures and impacts, monitoring, etc.). One of the CIS Guidance Documents (GD 11) focuses on the planning in the WFD implementation. A more general planning process scheme towards RBMP, which gives the application of ICT tools a more prominent role, is the IMA-PIP framework being developed by HarmoniCA (Becker et al. 2005).

Figure 5.32: The WISE-RTD webportal.

In both types of guidance for policy implementers, guidance is given for specific policy implementation tasks. For example, a complete list of WFD tasks can be found in GD 11. In the present study they were cross-checked against the tasks reported in the other CIS GDs, completed and classified according to the water-related disciplines as covered by the different CIS GDs. The WFD tasks were also classified in three levels according to the main tasks and subtasks, as will be discussed in more detail in next sections.

(ii) Experiences and cases

It is useful for both scientists and policy implementers to consult examples of existing policy implementations and the results of the use of RTD tools. These examples can complement the guidance and enhance the dissemination of experience among different policy implementers and tool users. This dissemination step needs to contribute to a large part in closing the science-policy gap.

For the WFD implementation, examples are mainly available for the PRBs, but also in demonstration projects such as in the INTERREG III programme (e.g. Scaldit project covering the Scheldt river basin district).

(iii) RTD results and ICT tools

The list of RTD results and related tools and instruments (hereafter shortly called “tools”) that have been produced and developed by the scientific community in support by software developers, is lengthy. To enable analysis and reporting on their applicability to policy implementation, classifications are needed, such as by water-related discipline, by policy implementation task, by input and output variables and physical processes considered (e.g. Rekolainen, 2003).

Most relevant RTD projects for the WFD are the projects funded in the framework of EU FP5 under the EESD (Energy, Environment and Sustainable Development) thematic activity and in the framework of EU FP6 under the Priority 6.3 (Global Change and Ecosystems), and ERA-NET (networking within the European Research Area), but also the projects funded by the LIFE programme and through structural funds (e.g. INTERREG III).

Information sources in support of policy implementation tasks

It is clear that each of the above-listed types of information sources can provide different and complementary types of information to support the policy implementation tasks:

• By means of links between the experiences and the policy implementation tasks: for every case study or implementation project, it can be specified which policy implementation tasks were covered together with details on the specific implementation. These links allow for each policy implementation task examples to be provided on the practical implementation under various conditions, i.e. local circumstances.
• By means of links between the tools and the policy implementation tasks: every tool can be used to support one or more of specific policy implementation tasks. These links allow for each policy implementation task lists to be given on the available tools together with details on the specific support provided by the tool, on the required input data and the type of tool output or results.
• By means of links between the policy implementation tasks and the guidance: guidance can be given on how the tasks can be carried out technically, as well as explanations on the role ICT tools and models can play in the implementation of the tasks.

6 Practical experiences from existing case studies and pilot river basins

6.1 Brief characterization of the case studies

6.2 Case study 1: Marne/Seine/Normandie - Using models for the WFD Baseline Scenario

6.3 Case study 2: Witte Nete and Dender tributaries in the Scheldt Pilot River Basin

6.4 Case study 3: Participatory Integrated Planning (PIP) in the Verbano case study

6.5 Case study 4: The MANTRA-East approach to apply the Planning Framework

6.6 Case study 5: The Spree river basin case study

6.7 Case study 6: The Möll study - Integrative river management in the alpine region

6.8 Conclusion: lessons learned from the case studies

7 Glossary

(to be corrected and extended)

Adaptation Initiatives and measures to reduce the vulnerability of natural and human systems against actual or expected climate change effects. Various types of adaptation exist, e.g. anticipatory and reactive, private and public, and autonomous and planned. Examples are raising river or coastal dikes, building reservoirs, renaturation of wetlands, insurance systems, etc.

Adaptive capacity The whole of capabilities, resources and institutions of a country or region to implement effective adaptation measures.

Baseline Scenario The reference for measurable quantities from which an alternative outcome can be measured, e.g. a non-intervention scenario is used as a reference in the analysis of intervention scenarios. In the context of the WFD, it is a projection into the future assuming the development of drivers and pressures in the reference period after basic measures are implemented. The Baseline Scenario serves as a business as usual scenario to investigate the long-term development of water resources under pressure. Additional management options to overcome possible water related problems have to be compared against this reference.

Calibration (Model calibration) The procedure of adjusting parameter values of a model to reproduce the response of reality within the range of accuracy specified in the performance criteria.

CIS The Common Implementation Strategy of the European Commission.

Data Assimilation A continuous update of the model by utilizing new monitoring data, e.g. of flood forecast models by remote sensing data.

DPSIR A useful insight into issues of management of complex freshwater systems can be achieved if the Driving forces – Pressures – States – Impacts - Responses (DPSIR) scheme of the Euro-pean Environment Agency (EEA) is applied. The Driving forces generate a Pressure upon the system, thus altering its State. This alteration represents an Impact, i.e., an effect upon the environment and society. When the society is affected in an unfavourable way, it reacts by devising and implementing Responses that can target either the Drivers, the Pressures, the State, or directly the undesired or threatening Impacts, so that the latter are avoided, reduced, or compensated.

GIS Geographical Information Systems are computer-based tools to process and analyse spatial information.

IWRM The recent understanding of integrated water resources management (IWRM), as defined by the Technical Advisory Committee of the Global Water Partnership, reads (GWP 2000): “Integrated water resources management is a process, which promotes the co-ordinated development and management of water, land and related resources, in order to maximize the resultant economic and social welfare in an equitable manner without compromising the sustainability of vital ecosystems.”

Integrated assessment An interdisciplinary process of combining, interpreting and communicating knowledge from diverse scientific disciplines so that all relevant aspects of a complex societal issue can be evaluated and considered for the benefit of decision-making.

Model A site-specific model established for a particular study area, including input data and parameter values.

Model code A mathematical formulation in the form of a computer program that is so generic that it, without program changes, can be used to establish a model with the same basic type of equations (but allowing different input variables and parameter values) for different study areas.

Model setup Establishment of a site-specific model using a model code. This requires, among other things, the definition of boundary and initial conditions and parameter assessment from field and laboratory data.

NGO A non-governmental organization, e.g. nature conservation organizations, which can form a legal body in the course of the implementation of Water Framework Directive.

Projection A projection is a potential future evolution of a quantity or set of quantities, often computed with the aid of a model. Projections are distinguished from predictions in order to emphasize that projections involve assumptions concerning, for example, future socioeconomic and technological developments that may or may not be realised, and are therefore subject to substantial uncertainty. See also Climate projection; Climate prediction.

Quality Assurance (QA) Quality assurance is defined by NRC (1990) as the procedural and operational framework used by an organisation managing the modelling study to assure technically and scientifically adequate execution of all tasks included in the study, and to assure that all modelling-based analysis is reproducible and defensible.

RBD River Basin District, the spatial unit where the implementation of the WFD is organized in form of a River Basin Management Plan.

RBMP The River Basin Management Plan summarizes the results of the two first implementation phases (characterisation of water bodies, identification of pressures, and impacts and the programme of measures). The functions of the plan are (after GD 11, cf. EC 2003) to serve as a fundamental inventory and documentation mechanism for information gathered according to the Directive and to co-ordinate the programmes of measures and other relevant programmes concerning the area of the river basin district.

Scenario A plausible and often simplified description of how the future may develop, based on a coherent and internally consistent set of assumptions about driving forces and key relationships. Scenarios may be derived from projections, but are often based on additional information from other sources, sometimes combined with a narrative storyline. See also SRES scenarios; Climate scenario; Emission scenario.

Uncertainty An expression of the degree to which a value (e.g., the future state of a hydrological system) is unknown. Uncertainty can result from lack of information or from disagreement about what is known or even knowable. It may have many types of sources, from quantifiable errors in the data to ambiguously defined concepts or terminology, or uncertain projections of human behaviour. Uncertainty can therefore be represented by quantitative measures, for example, a range of values calculated by various models, or by qualitative statements, for example, reflecting the judgement of a team of experts (see Moss and Schneider, 2000; Manning et al., 2004). See also Likelihood; Confidence.

Validation (Model validation) Substantiation that a model within its domain of applicability possesses a satisfactory range of accuracy consistent with the intended application of the model.

Vulnerability The degree to which a system is susceptible to, and unable to cope with, adverse effects of external pressures, including impacts of land use, water management and climate variability and extremes. Vulnerability is a function of the character, magnitude, and rate of pressures to which a system is exposed, its sensitivity, and its adaptive capacity.

WFD The European Water Framework Directive, which came into force in December 2000, establishes a framework for actions of the European Communities in the field of water policy, with the key objective to achieve a “good water status” for all waters of the European Union (EU) by 2015.

WISE The Water Information System for Europe, see http://www.wise.eu.

WISE-RTD The Research-Technology-Development web portal of WISE (Water Information System for Europe, see http://www.wise-rtd.info) supporting the transfer of knowledge between the world of WFD-implementers and policy-makers and the research and technology world.

8 References

• Abbott, MB, Refsgaard, JC (1996) Distributed Hydrological Modeling. Kluwer Academic Publishers, Dordrecht / Bosten / London.
• Anderson, MP and Woessner, WW (1992) The role of postaudit in model validation. Advances in Water Resources, 15, 167-173.
• Arnold, JG, Williams, JR, Srinivasan, R, King, KW, Griggs, RH (1994) SWAT, Soil and Water Assessment Tool.USDA, Agriculture Research Service, Grassland, Soil and Water Research Laboratory, Temple, TX.
• Becker, A et al. (2002) Zusammenfassung der erzielten Ergebnisse aus den integrativen Forschungen. http://www.glowa-elbe.de/summ_becker.html
• Bicknell, BR, Imhoff, JC, Kittle, JL, Jr., Donigian, AS, Jr., and Johanson, RC (1997) Hydrological Simulation Program-Fortran. User's manual for version 11: U.S. Environmental Protection Agency. National Exposure Research Laboratory. Athens, Ga., EPA/600/R-97/080, 755 p.
• Dooge, J.C.I. (1973). Linear Theory of Hydrologic Systems. Techn. Bull. No. 1468, Agric. Res. Stat. USDA, Washington, USA.
• Beven, K (1997) ‘TOPMODEL: a critique’, Hydrological Processes, 11, 1069-1085.
• Checkland, P (1993). System Thinking. System Practice. Wiley Chichester, 350 pp.
• EC (2000) Establishing a framework for community action in the field of water policy. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000, Official Journal of the European Communities, Brussels.
• EC (2003a) Guidance document No 1: The implementation challenge of the Water Framework Directive (2000/60/EC). Produced by Working Group 2.6 - WATECO (Economics and the environment). Luxembourg: Office for Official Publications of the European Communities.
• EC (2003b) Guidance Document No 11: Common implementation strategy for the water framework directive (2000/60/EC). Produced by Working Group 2.9 – Planning Processes. Luxembourg: Office for Official Publications of the European Communities.
• EUW (2004): European Workshop "From initial characterisation to River Basin Management Plan". Paris 21-22 Oct. 2004.
• Forkel, C (1996) Das numerische Modell - ein schmaler Grat zwischen vertrauenswürdigem Werkzeug und gefährlichem Spielzeug. Presented at the 26. IWASA, RWTH Aachen, 4-5 January 1996.
• FWR (1998) Urban pollution management (UPM). Second edition. Foundation for Water Research, UK.
• Global Water Partnership Technical Advisory Committee (TAC) (2000). Integrated Water Resources Management. Global Water Partnership. Stockholm, Sweden.
• Giupponi, C, Mysiak, J, Fassio, A, and Cogan, V (2000) MULINO: Multi-sectoral, Integrated and Operational Decision Support System for Sustainable Use of Water Resources at the Catchment Scale, Ghassemi, E.; M. McAller, F. Oxley, and Scoccimarro (eds), Proceeding from MODSIM 2001, Volume 3, Canberra, Australia, 2000.
• GWP - Global Water Partnership (2000). Towards water security: Framework for action, 75–76.
• Hare, M and Pahl-Wostl, C (2002). Stakeholder categorization in participatory integrated assessment processes. Integrated Assessment 3, 50–62
• Harremoës, P (2003) The role of uncertainty in application of integrated urban water modeling. Proceedings of the IMUG Conference, Tilburg, 23-25 April 2003.
• Heuvelink, GBM. (1998a): Error Propagation in Environmental Modelling with GIS. London: Taylor & Francis, 127.
• Heuvelink, GBM (1998b): Uncertainty analysis in environmental modelling under a change of spatial scale. Nutrient Cycling in Agro-ecosystems 50, 255-264.
• Højberg AL, Refsgaard JC & Jørgensen LF (2006). Do the WFD monitoring programmes provide sufficient data for modellling? In: Refsgaard JC & Højberg AL (eds) 2006: Nordic Water 2006. Proceedings. XXIV Nordic Hydrological Conference 2006 – Experiences and Challenges in implementation of the EU Water Framework Directive. NHP Report No. 49.
• Huxham, C (2000). The challenge of collaborative governance. Public Management, 2, 337-357. 2000.
• Jørgensen, LF, Refsgaard JC, Højberg, AL J. (2007) The inadequacy of monitoring without modelling support. Environ. Monit. 9 (931)
• Kaden, S et al. (2002) Großräumige Wasserbewirtschaftungsmodelle für gebietsbezogene Analysen. http://www.glowa-elbe.de/status_praes/wasy/gesamt.html
• Klemes, V (1986) Operational testing of hydrological simulation models. Hydrological Sciences Journal, 31, 13-24.
• Kundzewicz, ZW (1986) Physically-based hydrological flood routing. Hydrol. Sci. J. 31(2) 237-261.
• Kundzewicz, ZW (2007) Mathematical modelling of ecohydrological systems in the changing world (in print).
• Kundzewicz, ZW, Afouda, A & Szolgay, J (1987) Mathematical modelling, in: Kundzewicz, Z.W., Gottschalk, L. & Webb, B. (eds) Hydrology 2000, IAHS Publ. No. 171, 71-78.
• Loucks, DP & van Beck, E (2005) Water Resources Systems Planning and Management. An Introduction to Methods, Models and Applications, UNESCO Publishing.
• Madsen, H, van Griensven, A, Højbjerg, AL, Doherty, J, Hill, MC, Poeter, E, Solomatine, D, Kuczera, G (2007) HarmoniCA Guidance Model Calibration.
• Mahmood, K & Yevjevich, V (eds) 1975. Unsteady Flow in Open Channels. Water Resources. Publ., Fort Collins, CO, USA (in three volumes).
• Melching, CS, Yen, BC, Wenzel, HG (1990). A reliability estimation in modeling watershed runoff with uncertainties, Water Resources Research 26(10), 2275-2286.
• Middlemis, H (2000) Murray-Darling Basin Commission. Groundwater flow modelling guideline. Aquaterra Consulting Pty Ltd., South Perth. Western Australia. Project no. 125.
• Mostert, E (1998) A Framework for Conflict Resolution. Water International 23, Nr. 4
• Nilsson, S (2003) The role and use of information in Transboundary Water Management. PhD Thesis, Department of Land and Water Resources Engineering, Royal Institute of Technology, Stockholm, Sweden.
• Nooteboom, B (2002) Trust: Forms, Foundations, Functions, Failures and Figures. Edward Elgar, Cheltenham, UK.
• Nooteboom, B and Six, F (eds)(2003). The Trust Process in Organizations. Edward Elgar, Cheltenham, UK.
• NRC (1990) Ground Water Models: Scientific and Regulatory Applications. National Research Council, National Academy Press, Washington, D.C.
• OECD (1994). Environmental Indicators. Indicateurs d'environnement. OECD Core Set. Corps central de l'OCDE. Organisation for Economic Co-operation and Development, Paris
• Pahl-Wostl, C (1995) The Dynamic Nature of Ecosystems: Chaos and Order Entwined. Wiley, Chichester, 288.
• Pahl-Wostl, C (2002a) Participative and Stakeholder-based policy design, evaluation and modeling processes. Integrated Assessment 3, 3–14.
• Pahl-Wostl, C (2002b). Towards sustainability in the water sector: The importance of human actors and processes of social learning. Aquatic Sciences 64, 394-411. 2002b.
• Pahl-Wostl, C, Schmidt, S and Jakeman, T (eds) (2004) The Implications of Complexity for Integrated Resources Management, iEMSs 2004 International Congress: "Complexity and Integrated Resources Management". International Environmental Modelling and Software Society, Osnabrück, Germany, June 2004
• Radwan, M (2002) River water quality modelling as water resources management tool at catchment scale, Ph.D thesis, Faculty of Engineering, K.U.Leuven, Belgium.
• Radwan, M, Willems, P, Berlamont, J (2003) Sensitivity and uncertainty analysis for river water quality modelling, Journal of Hydroinformatics, [in press]
• Ragas, AMJ, Haans, JLM, Leuven, RSWE (1997) Selecting water quality models for discharge permitting, European Water Pollution Control 7(5), 59-67.
• Refsgaard, JC, van der Sluijs, JP, Højberg, AL, Vanrolleghem, P (2006) HarmoniCA Guidance Uncertainty Analysis. http://www.harmoni-ca.info/toolbox/docs/Harmoni-ca_Guidance_1_Uncertainty_Analysis.pdf
• Refsgaard, JC, Nilsson, B, Brown, J, Klauer, B, Moore, R, Bech, T, Vurro, M, Blind, M, Castilla, G, Tsanis I and Biza P (2005a) Harmonised techniques and representative river basin data for assessment and use of uncertainty information in integrated water management (HarmoniRiB), Environmental Science & Policy 8(3), 267-277.
• Refsgaard, JC, Henriksen, HJ, Harrar, WG, Scholten, H and Kassahun, A (2005b) Quality assurance in model based water management – review of existing practice and outline of new approaches, Environmental Modelling & Software, 20(10), 1201-1215.
• Refsgaard JC, Henriksen HJ (2004) Modelling guidelines - terminology and guiding principles. Advances in Water Resources, 27, 71-82.
• Refsgaard, JC (ed.) (2002) State-of-the-art report on quality assurance in modelling related to river basin management. HarmoniQuA Report, D-WP1-1, Copenhagen, Denmark. http://HarmoniQuA.org/deliverables.htm
• Rykiel, ER (1996) Testing ecological models: the meaning of validation. Ecological Modelling 90, 229-244.
• Scholten, H, Refsgaard, JC, Kassahun, A (2004) Structuring multidisciplinary knowledge for model based water management: the HarmoniQuA approach. Pahl, C, Schmidt, S, Jakeman, T (eds), iEMSs 2004 International Congress: Complexity and Integrated Resources Management". International Environmental Modelling and Software Society, Osnabrück, Germany.
• Scholten, H, Van Waveren, RH, Groot, S, Van Geer, F, Wösten, H, Koeze, RD and Noort, J J (2000) Good Modelling Practice in water management. In: Proceedings HydroInformatics2000. International Association for Hydraulic Research, Cedar Rapids, Iowa, USA.
• Schulze, RE (ed.) (1994) Hydrology and agrohydrology: a text to accompany the ACRU-300 agrohydrological modelling system. Agricultural Catchments Research Unit, Department of Agricultural Engineering, University of Natal.
• Singh, VP (1988) Hydrologic Systems: Rainfall-Runoff Modelling, Vol. I, Prentice Hall, Englewood Cliffs, NJ, USA.
• Singh, PS (ed.) (1995) Computer Models of Watershed Hydrology. Water Resources Publications, Littleton, Colorado, USA.
• SIW (2005): Second International Workshop on Implementing Economic Analysis in the WFD. "Opening the black box". Proceedings of the workshop, Paris 17-18 February.
• Soncini-Sessa, R (2004) Integrated modelling and participatory decision making in practice: The Verbano project. (in Italian), Mc Graw Hill, Milano (English forthcoming with Elsevier).
• UNCSD (1996) Indicators of Sustainable Development Framework and Methodologies. Commission on Sustainable Development. United Nations, New York
• Wade, AJ, Durand, P, Beaujouan, V, Wessel, WW, Raat, KJ, Wjitehead, PG, Butterfield, D, Rankinen, K and Lepisto, A (2002) A nitrogen model for European catchments: INCA, new model structure an equations. ISSN: 1027-5606. Hydrology and Earth System Sciences 6(3), 559 – 582.
• Webler, T and Tuler, S (2001) Public Participation in Watershed Management Planning, Human Ecology Review 8(2), 29-39
• Wechsung, F, Becker, A, Gräfe, P (eds) (2005): Auswirkungen des glo-balen Wandels auf Wasser, Umwelt und Gesellschaft im Elbegebiet. Weißensee Ver-lag Ökologie, Potsdam, Germany
• Wenger, E (1998) Communities of Practice; Learning, Meaning, and Identity. Cambridge: University Press.
• Willems, P (2000) Probabilistic immission modelling of receiving surface waters. Ph.D thesis, Faculty of Engineering, K.U.Leuven, Belgium.
• Willems, P, Berlamont, J (2002) Probabilistic emission and immission modelling: Case-study of the combined sewer - WWTP - receiving water system at Dessel (Belgium). Water Science and Technology 45(3), 117-124.
• Willems, P (2003) Methodology for integrated catchment modelling, Integrated Modelling User Group (IMUG) International Conference, Tilburg, 23-25 april 2003, proceedings (Holapress, Valkenswaard): 8 p.; see also: http://www.wrcplc.co.uk/imug/imug2003/15_paper_PWillems_Belgium.pdf
• Willems, P, Radwan, M, Popa, D, El-Sadek, A, Berlamont, J (2003) Surface water quantity and quality modelling by the complementary use of detailed and simplified models: Case-studies of Dender and Witte Nete watersheds – Belgium, In: Monitoring and modelling catchment water quantity and quality, Verhoest, N, Hudson, J, Hoeben, R, De Troch, F (eds), IHP-VI: Technical Documents in Hydrology, No. 66, International Hydrological Programme UNESCO, 105-112.
• WMO (World Meteorological Organization) (1994) Guide to Hydrological Practices, WMO-No. 168, Geneva, Switzerland, Fifth Edition.

List of Figures:

• Figure 2.1: The management approach including the main stages of the WFD implementation process according to GD 11 (EC, 2003). 20
• Figure 2.2: Framework for model-supported participatory planning of measures and Integrated River Basin Management (planning framework). 21
• Figure 2.3: Impact matrix illustrating the impacts of different planning alternatives on specific criteria (Wenzel 2005). Translation of German terms: Versorgungssicherheit – Supply reliability; Mindestabfluß – Least discharge; Wasserwerke – Water supply units; Kraftwerke – Water power units. 27
• Figure 2.4: WFD’s implementation timetable and the role of modelling structured by the identification phase (blue), the design phase (red), the implementation phase (reen) and the evaluation phase (purple). 32
• Figure 2.5: Example of joint use of monitoring and modelling: 180 drill log information of the subsurface was used in combination with advanced geo-statistical tools to simulate a set of realizations of the subsurface corresponding to the input information (left: two realizations). Right: by overlaying the realizations, areas with a high reliability of the reconstruction are visible (in red, e.g. the locations of the drill logs), whilst areas with a poor reproduction also appear (blue). The blue areas are the locations where additional monitoring data are needed. 38
• Figure 2.6: Example of joint use of monitoring and modelling: a statistical methodology was applied for defining an optimal set of sampling intervals (for dissolved oxygen DO) for the operation of a river water quality model. Starting with an extensive set of measurements (left), it is the aim to reduce the number of observations to obtain just as much data as necessary for a calibration with an acceptable uncertainty and thereby to reduce the resulting range of uncertainty (see confidence interval 95 % high – 95 % low) in the parameters (right) (Vandenberghe et al. 2005). 39
• Figure 2.7: Identification of pressures –the observed and simulated groundwater levels are shown (Hattermann et al. 2005). 43
• Figure 2.8: Structure of the model-supported management planning in the Seine-Normandie basin. 47
• Figure 2.9: Risk assessment (likelihood of failing to reach the “good status”) map for Seine-Normandie. 48
• Figure 2.10: Typical situation in a planning study (Kaden et al. 2002). 51
• Figure 2.11: Example of a comprehensive model system to investigate land use and climate scenario impacts on water resources (example taken from the GLOWA-Elbe project, see Wechsung et al. 2005). 52
• Figure 2.12: Plan view of the 12 subcatchments in the River Dender basin, including the locations of the urban drainage system (with 13 outfalls towards the Molenbeek tributary) in sub-catchment area No. 431. The basin has a total area of 708 km2 and encloses 12 sub-catchments areas. 58
• Figure 2.13: Integration of the submodels in the Dender River basin case study at the conceptual level, and complementary use of the detailed submodels (outer circle) and the conceptual submodels (inner circle). 59
• Figure 2.14: Evaluation of two water quality sanitation scenarios on DO and BOD concentration changes (time series at 35 km distance along the Dender), and comparison with the Flemish percentile-based standard. 60
• Figure 2.16: Reliabilities of water inflow to Berlin for different management alternatives. 62
• Figure 2.17: Basic scenarios for the Estonian-Russian border. 64
• Figure 2.18: Reaction of phytoplankton to different loading scenarios in 2015-2019. Source: Noges et al. (2003). 64
• Figure 2.19: Stakeholder workshops to discuss possible management options. 65
• Figure 2.20: Current situation of the channelised river bed (left). Visualized restoration measures such as the reinitialisation of braiding river sections with floodplain forests (right) 66
• Figure 3.1: A real world catchment (top) and two possible formalizations: a conceptual scheme of a computer model (left) and a material model (right). 73
• Figure 3.2: The model selection process. It is important to have a clear vision of the modelling target, available resources, and complexity of the problem including the accuracy required. 76
• Figure 3.3: Balancing different types of uncertainties to determine the optimal model structure detail for a specific application (Willems et al. 2003). 80
• Figure 3.4: Different model components linked for multidisciplinary integrated impact analysis (example taken from the German GLOWA-Elbe project). 84
• Figure 3.5: Integration of models at 2 different levels (detailed models at small scale level, and conceptual models at large scale level) (Willems et al. 2003). 86
• Figure 3.6: The calibration process. 89
• Figure 3.7: Observed data (a) of river discharge and different stages of model calibration (b). 90
• Figure 3.8: Representation of three different types of uncertainties in probabilistic modelling by means of stochastic terms (Willems et al. 2003). 93
• Figure 3.9: Groundwater recharge and groundwater table as calculated by Equation 1 (a) with S = 0.02 m3 m-3, T = 80.0 m2 d-1 and L = 800 m. (b) shows the range of the results for varying T between 60.0 m2 d-1 and 100.0 m2 d-1, and (c) for varying S between 0.005 m3 m-3 and 0.035 m3 m-3. 95
• Figure 3.10: The five steps and 45 tasks of the modelling process in the HarmoniQuA knowledge base (Refsgaard et al. 2005b). 99
• Figure 5.1: The WISE-RTD web-portal supports the transfer of knowledge between the world of WFD-implementers/policy-makers and the research/technology world. 108
• Figure 5.2: The WISE-RTD webportal. 109