MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

Miljø- og Fødevareudvalget 2017-18
MOF Alm.del Bilag 35
Offentligt

International evaluation of the

Danish marine models

Perf ormed by t he Panel of internati onal experts

10. oktober 2017

Implement Consulting Group

Strandvejen 54

2900 Hellerup

Tel +45 4586 7900

Email [email protected]

Implementconsultinggroup.com

CVR

Bank

SWIFT

Iban

32767788

4845-3450018236

DABADKKK

DK3030003450018236

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

Table of contents

Introduction .......................................................................................................... 1

1.1 Aim and focus of the evaluation ................................................................... 2

1.2 The basis and process of the evaluation ...................................................... 3

1.3 Content and structure of the evaluation report ............................................. 6

1.4 Future development following the evaluation ............................................... 6

Compliance with the Water Framework Directive ................................................ 8

2.1 Reference conditions and boundary setting ................................................. 8

2.2 Choice of indicators...................................................................................... 9

2.3 Intercalibration ............................................................................................ 10

2.4 “One out, all out” ......................................................................................... 11

2.5 Other stressors on the ecosystem............................................................... 11

Coastal water typology ........................................................................................ 13

3.1 Basic idea behind typology ......................................................................... 13

3.2 The Danish typology ................................................................................... 13

3.3 Suitability of the Danish typology ................................................................ 14

3.4 Suitability of the Danish monitoring programme .......................................... 14

3.5 Suggestions towards a modified approach ................................................. 15

3.6 The look abroad .......................................................................................... 15

The use of seagrass and Kd as environmental indicators ................................... 16

4.1 Kd as an indicator for the biological element “benthic vegetation, macroalgae

and angiosperms” ............................................................................................... 16

4.2 Other indicators used in the statistical modelling ........................................ 19

Emphasis on nitrogen versus phosphorus .......................................................... 21

5.1 Phosphorus limitation .................................................................................. 21

5.2 Treatment of nitrogen and phosphorus in the Scientific Documentation Report

.................................................................................................................... 21

5.3 Possible implications for management ........................................................ 22

5.4 Seasonality ................................................................................................. 23

Statistical modelling ............................................................................................ 24

6.1 Setup........................................................................................................... 24

6.2 Panel evaluation of basic model setup ........................................................ 25

6.3 Panel evaluation of statistical model results................................................ 26

Mechanistic modelling ......................................................................................... 27

7.1 The models ................................................................................................. 27

7.2 Model setup, calibration and validation ....................................................... 28

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

7.3 Validation .................................................................................................... 28

7.4 Reference conditions simulation ................................................................. 29

7.5 Scenarios and establishment of cause-effect relationships......................... 30

7.6 Conclusion on the mechanistic models ....................................................... 31

Calculation procedures to estimate Maximum Allowable Inputs from model results

............................................................................................................................ 32

8.1 Steps in the calculation of targets and MAI ................................................. 32

8.2 Averaging and “ensemble modelling” aspects in the procedure.................. 33

8.3 Conceptual differences between modelling approaches ............................. 35

8.4 Meta-modelling ........................................................................................... 36

Evaluation of Maximum Allowable Inputs results ................................................ 37

9.1 The overall Danish MAI in an international framework ................................ 37

9.2 Historic conditions as basis for target setting .............................................. 37

9.3 Effects of climate change on targets and MAI ............................................. 38

9.4 Relevance of typology on MAI .................................................................... 39

9.5 Relevance of indicator choice on MAI ......................................................... 39

9.6 Relevance of model quality and approach for MAI ...................................... 39

9.7 Conclusion and perspectives ...................................................................... 40

10. Overall assessment and conclusions .................................................................. 41

11. Recommendations for going further .................................................................... 43

12. List of references ................................................................................................ 45

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

Introduction

This report presents a scientific review of the Danish management approach regarding

coastal waters in relation to the implementation of the European Water Framework

Directive (WFD) in Denmark. The parties to the Agreement on Food and Agriculture

Package (22 December 2015) have decided to evaluate the modelling tools (pressure-

impact models) used to calculate the mitigation demands for nitrogen (N) runoff from

land in the Danish River Basin Management Plans. The results of the evaluation will be

utilised towards the development and application of models in the 3

generation water

plans valid for 2021-2027.

Task description by t he M inistry of Food and Agriculture

In agreement with the EU Water Framework Directive, Denmark has produced the River

Basin Management Plans devising a strategy for improving and securing that coastal

waters, lakes, streams and ground waters fulfil the demand for Good Ecological Status

as stated in the directive. For Danish coastal waters, it has been estimated that

reductions in N runoff from land are the primary concern if goals of Good Ecological

Status in coastal waters are to be fulfilled. On this background, mitigation measures have

been implemented in the 2015-2021 River Basin Management Plans to additionally

reduce the N runoff to coastal waters, corresponding to roughly half the total estimated

reduction needs.

The task of the evaluation panel is to perform a thorough evaluation of the marine

modelling tools that form the basis for the mitigation demands for land-based nitrogen

(N) runoff in the Danish River Basin Management Plans with regards to the importance

of N as well as other relevant pressures such as phosphorous, fisheries etc. In particular,

the evaluation panel has to:

Evaluate the use of models for determination of type-specific reference values

(according to the Water Framework Directive, Annex 2) for the water quality

element phytoplankton (chlorophyll).

Evaluate the use of models to determine environmental targets (Maximum

Allowable Inputs (MAI) of nitrogen)) and mitigation needs to achieve good

environmental status and evaluate differences and similarities between the use

of different methods and model types for coastal waters with different typology.

Evaluate the estimated nitrogen target loads and mitigation needs in the Danish

River Basin Management Plans and evaluate the method for determining the

Danish proportion of total mitigation needs. How is the current environmental

status in Danish coastal waters determined by N runoff from Danish land areas

in relation to other pressures such as N released from sediments and N loads

from catchments in neighbouring countries and airborne N deposition (the

Danish share of the total mitigation needs related N)?

ii.

iii.

Further, the Panel is expected to address the technical questions and comments from

the stakeholders.

Recruitment of experts

The Danish Ministry of Environment and Food has been responsible for the recruitment

of an international panel of five experts to carry out the evaluation. The recruitment of

experts has been conducted by a nomination process where the Danish Ministry of

Environment and Food has requested water management authorities in other countries

(Sweden, Finland, Poland, Germany, The Netherlands and England) and the European

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

Environment Agency, Joint Research Centre (JRC) and the European Commission (DG

Environment) to nominate experts to conduct the evaluation. It has been stated in the

request that the nominees should have expert knowledge in the following areas: marine

ecology, marine ecosystem models, statistical methods and experience in marine water

management in relation to the Water Framework Directive.

The request by the Ministry resulted in the nomination of 14 experts of which 9 experts

subsequently indicated that they were interested in being part of an expert panel. Of

these, the Ministry has selected the following five experts to conduct the evaluation:

•

Professor Peter Herman, Deltares, Institute for applied research in the field of

water and subsurface, the Netherlands.

Professor Alice Newton, NILU – Norwegian Institute for Air Research

Professor Gerald Schernewski, Leibniz Institute for Baltic Sea Research,

Warnemunde

Director Bo Gustafsson, Baltic Nest Institute (BNI), Stockholm University,

Sweden

Senior Researcher Olli Malve, Finnish Environment Institute SYKE

Professor Peter Herman was chosen as chairman of the Panel

•

The five experts were chosen according to an assessment of their qualifications with

regards to experience with and competences in the following fields of study:

marine

ecology/coastal ecology,

coastal

ecosystem modelling, use of statistics in environmental

science

and

marine management experience related to the implementation of the Water

Framework Directive.

1.1

Aim and focus of the evaluation

This section presents the aim and focus of the evaluation according to the international

panel (hereafter referred to as the Panel) and should therefore be seen as the Panel’s

further operationalisation of the task description in section 1.1.

The main aim of the evaluation

The main aim of the evaluation is to review whether the marine models – as

presented in the Scientific Documentation Report and as commented by the

researches and stakeholders –

provide solid and robust scientific evidence that the

proposed reductions in land-based N runoff will be both necessary and sufficient to

reach Good Ecological Status as defined in the Water Framework Directive.

•

By “solid”, the Panel means well based in international scientific literature, well

performed, credible

By “robust”, the Panel means not unduly dependent on arbitrary details, reliable

with acceptable precision

By “necessary”, the Panel means that by doing less the goals would not be

reached

By “sufficient”, the Panel means that by executing the plans, there is a high

probability of reaching the goals

•

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

The evaluation concerns the modelling tools (pressure/impact models) forming the basis

for the mitigation demands for land-based nitrogen (N) runoff in the Danish River Basin

Management Plans. The evaluation results will enter into the calculation of N mitigation

demands for coastal marine areas in the 3

generation water plans valid in 2021-2027.

The evaluation will answer questions related to points (i)-(iii) in the task description

above and is therefore focused on the scientific underpinning of the plans, in particular

the modelling tools. The evaluation must take into account the internationally agreed

goals of achieving Good Ecological Status in the Water Framework Directive. One that

basis, the Panel has defined the aim and focus of the evaluation as stated in the Text

Box shown above.

The scope of the evaluation does not include other models than the marine model and

other environmental targets than those applying to coastal areas. The scope of the

evaluation does not include the societal costs and benefits of the measures that would

be needed to fulfil the environmental targets.

1.2

The basis and process of the evaluation

The basis for the ev al uation

The basis on which the Panel has made the final evaluation consists of the following

materials:

•

The Scientific Documentation Report written by Aarhus University (DCE) and

DHI in June 2017, which documents the model tools and calculated MAI that

were developed for the Ministry over the period 2013-2015.

Questions and comments from the stakeholders to the Scientific Documentation

Report (see Annex 1 of the evaluation report)

Answers from the researchers to questions and comments which were

formulated by the Panel after the members of the Panel read and considered

the report as well as the questions and comments from the stakeholders (see

Annex 2a and 2b of the evaluation report).

Answers from the Panel on how they took into account each of the technical

questions and comments from the stakeholders (see Annex 3 of the evaluation

report).

Selected background materials cited by the researchers, the stakeholders and

the Panel

•

The means for ens uri ng independence during t he proc ess

It is considered crucial that the evaluation of the Danish marine models be performed by

independent scientists. In order to guarantee independence, it was decided that the

Ministry of Environment and Food, the scientists from AU and DHI and the stakeholders

should keep arm’s length to the Panel throughout the process of the evaluation.

Implement Consulting Group (Implement) was engaged by the Ministry to facilitate the

process.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

Figure 1. Communication model

As illustrated above, the communication model was designed to facilitate a dialogue that

ensured arm’s length between the involved parties and to promote a transparent flow of

communication. Implement has been the link between the Panel, the stakeholders and

the scientists. Besides facilitating the final writing workshop and the preparations leading

up to that, the main role of Implement has therefore been to ensure timely

communication and convey relevant material and information between the parties.

The evaluati on process

The evaluation process started in June 2017. It resulted in an evaluation report on 19

September, which was finalised after a writing workshop in Helsingør which took place

between 11-15 September. After the hearing process between 19. September and 2

October some minor corrections were made to the final report which was completed on

10 October.

The text and the activity plan below provide a more detailed overview of the evaluation

process.

Initially, the stakeholders from Blåt Fremdriftsforum, the scientists from AU and DHI and

the Panel were invited to participate in separate meetings where Implement explained

the process of the evaluation. At the meetings, the activity plan and a communication

model were presented to make sure that all parties were properly informed about the

practical aspects, important deadlines and rules of communication. The process leading

up to the final evaluation workshop was thereafter as follows with respect to each of the

parties:

•

The stakeholders received the Scientific Documentation Report written by the

scientists from AU and DHI on 6 June and had until 4 July to formulate

questions and comments to the report. The comments and questions had to be

submitted in a table – made specifically for that purpose – that followed the

structure of the report. A “hotline” for questions regarding the practical aspects

of formulating and submitting the questions and comments was established by

Implement. The stakeholders submitted their questions and comments on 4

July, and they were all forwarded by Implement to the Panel on 6 July. In order

to sum up their main points of view in front of the Panel, the stakeholders were

invited to participate in a physical meeting with the Panel hosted by the Ministry

in Helsingør on 11 September during the writing workshop.

The Panel received the Scientific Documentation Report at the same time as

the stakeholders – on 6 June. Implement held a couple of status meetings with

the Panel in June and the beginning of July when the comments and questions

from the stakeholders were forwarded to the Panel on 6 July. Based on the

reading of the Scientific Documentation Report and the questions and

comments from the stakeholders, the Panel has jointly formulated questions to

•

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

the Danish scientists that were forwarded by Implement on 15 August. The

scientists from AU and DHI answered these questions on 4 September in order

for the Panel to take it into account in the evaluation. Throughout August,

Implement held some status meetings with the Panel to monitor the progression

and to prepare the writing workshop in Helsingør.

•

The scientists from AU and DHI worked out the Scientific Documentation

Report that was distributed by Implement to the stakeholders and the Panel.

The scientists received the comments and questions by the stakeholders on 6

July as an orientation. As stated above, the scientists received questions from

the Panel and replied to these on 4 September. During the writing workshop

from 11-15 September, the scientists have answered a limited amount of

additional questions from the Panel.

The Ministry of Environment and Food was not directly involved in the

evaluation process due to the principle of arm’s length. Implement has

occasionally informed the Ministry about the progress in the evaluation, and the

parties had a dialogue about the practicalities of the writing seminar in

Helsingør. Representatives from the Ministry were present at the meeting with

the stakeholders and the Panel in Helsingør on 11 September.

•

The writing of the evaluation report was thereafter carried out by the Panel and facilitated

by Implement in a final writing workshop in Helsingør between 11-15 September. The

evaluation report was edited and submitted for hearing on 19 September.

The hearing of the evaluation report among stakeholders from Blåt Fremdriftsforum and

the scientists from AU and DHI took place between 19 September and 2 October.

Figure 2. Activity plan of the evaluation process

After the hearing process, the evaluation report will be published by the Ministry of

Environment and Food along with annexes containing the hearing comments and

answers by the Panel. The activity plan above illustrates the entire process of the

evaluation.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

1.3

Content and structure of the evaluation report

The evaluation report is divided into a number of themes which the evaluation panel

found to be the most important in order to cover the topics in the terms of reference and

pursue the aim of the evaluation. According to the Panel, the main themes are those

covered in Chapters 2-9 in the evaluation report which has the following structure:

•

Introduction (Chapter 1)

Compliance with the Water Framework Directive (Chapter 2)

Coastal water typology (Chapter 3)

The use of seagrass and Kd as environmental indicators (Chapter 4)

Emphasis on nitrogen versus phosphorus (Chapter 5)

Statistical modelling (Chapter 6)

Mechanistic modelling (Chapter 7)

Calculation procedures to estimate Maximum Allowable Inputs from model

results (Chapter 8)

Evaluation of Maximum Allowable Input results (Chapter 9)

Overall assessment and conclusions (Chapter 10)

Recommendations for going further (chapter 11)

•

By going through the most important themes and discussing the main problems within

each theme, the review by the Panel focuses on whether these problems have been

adequately solved in the Scientific Documentation Report – rather than going through

the details in the report chapter by chapter.

This means that the review text by the Panel mainly concentrates on investigating

possible weaknesses in the overall modelling approach followed by the researchers from

Aarhus University (DCE) and DHI. However, the review contains conclusions with

respect to both the strengths and weaknesses of the approach, and critical remarks

should be viewed in the context of the overall assessment as presented in Chapter 10.

After the thematic chapters, the evaluation contains an overall assessment of the marine

modelling approach and report. The final assessment will provide an answer to the

central question on whether the modelling approach and report provides solid and robust

scientific evidence that the proposed reductions in land-based N runoff will be both

necessary and sufficient to reach Good Ecological Status as defined in the Water

Framework Directive. Moreover, the assessment will answer other related questions to

cover the terms of reference.

Finally, recommendations are given as to how the Danish marine models might be

improved in the future. Focus is on improvement that can be made within a reasonable

time frame and without investing excessive resources.

1.4 Future development following the evaluation

Once the researchers have made the adjustments to the modelling, they are encouraged

to publish their work in peer-reviewed journals to showcase Danish leadership in this

field.

The work of the researchers has been performed over decades and several

administrations. This “organic” process has given rise to numerous interactions between

the scientists and authorities. In order to avoid confusion and misunderstandings, terms

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

of reference, the scope of the missions set by the Ministry and agreements on choices,

e.g. indicators to be used, should be well-defined. This can be important for the further

political process, but has not been subject to examination by the Panel.

The Panel hopes that the attention given to the views of the stakeholders and the

responses of the researchers during the scientific scrutiny of the Scientific

Documentation Report will help to build trust between the parties and contribute to a

successful outcome.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

2. Compliance with the Water Framework

Directive

This chapter examines whether the Scientific Documentation Report complies with

Directive 2000/60/EC, commonly referred to as the Water Framework Directive (WFD). It

also addresses some of the concerns and questions of the stakeholders. In this chapter,

we focus on the following general questions relating to the compliance with the Water

Framework Directive:

•

Is the procedure for setting the type-specific reference condition WFD

compliant?

Is the choice of indicators selected WFD compliant?

Have the indicators been intercalibrated?

Has the “one-out, all-out” principle been respected?

•

The questions are the basis for the subsections of the chapter. In addition, the chapter

devotes attention to the question whether all relevant stressors have sufficiently been

taken into account.

The community policy on water was adopted by the European Parliament and Council on

23 October 2000 as an integrated Community Directive 2000/60/EC, commonly referred

to as the Water Framework Directive (WFD). It was published in the Official Journal (OJ

L 327) on 22 December 2000 and was also adopted by member states (MS). In

Denmark, it was adopted as national legislation in 2003.

Article 1 states: “The purpose of this Directive is to establish a framework for the

protection of inland surface waters, transitional waters, coastal waters and groundwater”.

Preamble 26 of the WFD states that “member states should aim to achieve the objective

of at least good water status by defining and implementing the necessary measures

within integrated programmes of measures, taking into account existing community

requirements”. Article 4 introduces the concept of the River Basin Management Plans

(RBMP) as fundamental to “making operational the programmes of measures”, and

these are detailed in article 13. RBMP are a single system of water management by river

basin, which are the natural geographical and hydrological units, instead of according to

administrative or political boundaries.

The report “Development of models and methods to support the establishment of the

Danish River Management Plans”, which we refer to as the Scientific Documentation

Report, contributes to the implementation of the WFD to maintain or achieve Good

Ecological Status in Danish Coastal Waters (CW). Therefore, an evaluation of the WFD

compliance of the methodology and results is valuable and important.

2.1 Reference conditions and boundary setting

Annex 2 of the WFD (section 1.1) addresses the characterisation of surface water body

types. First, the water bodies must be placed in one of the surface water categories:

rivers, lakes, transitional waters or coastal waters. Another possible category is artificial

and heavily modified bodies of water.

The WFD then specifies that the water bodies are to be differentiated according to their

type. Denmark merged transitional waters with coastal waters, therefore a Fish BQE is

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

not necessary. Denmark has 119 marine water bodies

. These are categorised into six

open water body types and 12 estuarine water body types, all included in coastal waters,

according to a report by Dahl et al (2005). In Chapter 3, we discuss the further

implications and consequences of this typology. We note that in their answers to the

Panel, the researchers state that there is a project proposal on an “update of the

typology applied towards the RBMP 2021-2027”.

Annex 2 of the WFD (section 1.3) specifies the procedure for the “establishment of type-

specific reference conditions for surface water body types”. Type-specific reference

conditions (RC) may be either spatially based or based on modelling or may be derived

using a combination of these methods. Where it is not possible to use these methods,

member states may use expert judgement to establish such conditions. The Danish

approach relies on modelling and a 1900 baseline, since there are no pristine systems

that can be used as a reference. This approach is appropriate, WFD compliant and

better than only using expert judgement.

Good Ecological Status (GES) falls between the “High/Good” boundary and the “Good-

Moderate” status. The relationship between reference condition, the boundaries and

GES is shown in Figure 3. The setting of the reference conditions and the boundaries,

especially the G/M boundary, is important. This determines whether management

measures are necessary. Classification below the G/M boundary requires management

measures to be adopted.

Figure 3. Relationship between reference condition, the boundaries and GES

Target values must fall in the green (GES) range.

2.2 Choice of indicators

Annex 5 of the WFD specifies the quality elements (QE) for the classification of

ecological status of coastal waters (1.1.4). Good Ecological Status is an assessment

based on a combination of biological quality elements (e.g. phytoplankton, other aquatic

flora and benthic invertebrates); Hydro-morphological elements (e.g. structure and

substrate of the seabed, tidal regime); chemical and physico-chemical elements (e.g.

transparency, oxygenation conditions, nutrient conditions).

The indicators chosen in the Danish RBMP report are Chlorophyll a, Kd and a benthic

index as well as some secondary indicators in statistical modelling approaches (see

Chapter 4 of this evaluation report). Denmark has not adopted transitional waters as a

separate category, and therefore there is no need to include a fish biological quality

element. Chlorophyll a is a proxy for phytoplankton biomass and has been

intercalibrated (see 2.3 below). Kd is a measure of attenuation, hence an indirect

measure of growth conditions for benthic plants and algae. Thus, it is not a direct

indicator of aquatic flora (eelgrass), but rather a light control on the distribution of

eelgrass. Furthermore, Kd is not independent of Chlorophyll a, since phytoplankton cells

contribute to light attenuation and a loss in transparency. Kd has not been

Scientific Documentation Report, section 3.1, p. 14, Bek.nr. 837 2016

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

intercalibrated, although eelgrass depth limit, for which it is a proxy, has (see 2.3 below).

The Danish Benthic index addresses the benthic invertebrates’ biological quality

element. Chlorophyll a was the indicator chosen for the intercalibration of the WFD,

which Denmark participated in. The marine models under evaluation only considered

indicators for the physico-chemical elements’ oxygenation condition and nutrient

condition in the statistical modelling.

•

Most of the calculations in the modelling are based on only Chlorophyll a and

Kd to derive the targets for N

Furthermore, the choice of Kd as an indicator for submerged aquatic vegetation

(eelgrass) may be insufficient (Chapter 4)

The inclusion of other indicators throughout the process and modelling

(oxygenation condition and nutrient limitation) are also discussed in Chapter 4

•

2.3 Intercalibration

A Common Implementation Strategy (CIS) has been in operation since 2001, bringing

together national experts, stakeholders and the Commission involved in the

implementation of the WFD. During the process, a series of Guidance Documents and

CIS Thematic Information Sheets were produced. These are not legally binding but give

mainly technical advice about the implementation process.

The WFD requires the national classifications of Good Ecological Status to be

harmonised through an intercalibration exercise, Birk et al (2013). This is to avoid

adjacent water bodies being classified in a different way. Intercalibration was carried out

by member states that share typologies and transboundary water bodies. In the case of

Denmark, the shared water types were NEA 1/26C: NEA 8B and BC 6. These are

explained in Table 1.

Table 1. Common typologies intercalibrated for Chlorophyll a with Germany and

Sweden

Code

NEA

1/26C

NEA 8B

BC 6

Water type

The North-East Atlantic, enclosed seas, exposed

or sheltered, partly stratified

The North-East Atlantic Kattegat coastal waters

Baltic Coast (SW)

Shared

with

Intercalibration

Intercalibrated

As mentioned in 2.2, Chlorophyll a was the indicator chosen for the intercalibration of the

WFD, which Denmark participated in. The overall status with respect to intercalibration of

the indicators used in the Danish marine models is as follows:

•

Chlorophyll a has been successfully intercalibrated with SE and DE

Kd has not been intercalibrated (as confirmed by the researchers from Aarhus

University (DCE) and DHI and the European Commission’s Joint Research

Centre).

Eelgrass depth limit has been intercalibrated

•

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

2.4 “One out, all out”

The WFD preamble 11 specifies that it is “based on the precautionary principle and on

the principles that preventive action should be taken”. The “one-out, all-out” principle is a

key principle that reflects the WFD integrated approach for the protection of water

resources and associated aquatic ecosystems. Quality elements comprised in the

definition of ecological status provide a holistic picture of the health of the aquatic

environment. The overall status would only be “good” if all the elements comprised are at

least considered “good”. This ensures that all pressures capable of degrading the water

status are addressed and are a guarantee of the environmental integrity of the objectives

of the directive.

Progress achieved towards “good” status of water bodies can be reported using

indicators at individual quality element level. However, this does not preclude the “one-

out, all-out” principle. The WFD will be reviewed by 2019, taking into account the results

of the second RBMPs. The proper implementation of the Nitrates Directive, which is a

basic measure under the WFD, is necessary for the achievement of the WFD objectives.

However, in many cases, this will not be sufficient, and additional measures will have to

be taken by member states to ensure that the WFD objectives are reached.

Based on the “one-out, all-out” principle, indicators for different quality elements should

be considered individually. If one is classified as below the G/M boundary, then

management measures must be applied. This was not applied in the Scientific

Documentation Report as confirmed by the researchers in their answers to the panel

questions. The different methods of aggregation and their implications in both the WFD

and MSFD are discussed in Borja et al (2014). We further discuss the “one-out, all-out”

principle in relation to indicators in Chapter 4 and in relation to the calculation

procedures in Chapter 8.

2.5 Other stressors on the ecosystem

In his classic paper on eutrophication problems, Cloern (2001) describes how the vision

on eutrophication problems has evolved from viewing nutrient enrichment as a single

isolated issue, towards a vision that emphasises the interactions between multiple

stressors, the physics and hydrography of the systems, and eutrophication. He makes a

plea for integrated models and tools that describe how nutrient enrichment modulates

the response of ecosystems to other stressors, such as chemical pollution, introduction

of invasive species, habitat modifications, fishing pressure and others, in the physical

setting of a water body. The different stressors should not be viewed as additive factors

with – from a management perspective – the option to choose reduction of any of these

stressors to obtain a similar percentage of improvement in the ecosystem response.

Restoring physical habitat quality, as an example, will have very little effect if

eutrophication leads to oxygen problems, low transparency of the water or low

phytoplankton quality due to the interactions of the causal factor “physical structure” with

eutrophication. Reversely, remediation of eutrophication problems may not suffice to

improve ecological quality, if additional action on other stressors is needed.

In many questions and comments of the stakeholders, reference was made to the report

by Andersen et al (2017) that lists many stressors on the marine ecosystem and, using a

particular weighting, concludes on an overall percentage of stress due to nutrient

loading. One could try and argue that this provides evidence that similar improvements

of ecological status could be obtained by working on other stressors than nutrient

loading, but in so doing would miss the essential point that the effect of the different

stressors is not additive and that the final ecosystem response is modulated by the

interaction between the stressors, not their individual additive effect. The Panel endorses

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

the fundamental view on interaction between stressors, and on the key role of nutrient

loading and eutrophication in modulating the ecological response of Danish coastal

waters, that is expressed in the Scientific Documentation Report and in the models

(especially the mechanistic models) underlying the analyses.

This fundamental view on the importance of water quality as the main modulator in

promoting Good Ecological Status is fully in line with the Water Framework Directive

implementation and with the use of intercalibrated indicators such as Chlorophyll a and

measures of chemical pollution as the prime measures of ecological status. Inclusion in

the WFD was based on extensive and in-depth reviews of ample scientific evidence.

Other legal instruments, e.g. the Marine Strategy Framework Directive, take a broader

view and also include more explicitly other stressors such as invasive species, shipping,

fishing and physical modification. The Panel is convinced that these aspects fully merit

inclusion in a holistic view on restoring Good Ecological Status but in no way decrease

the importance that has to be attached to controlling nutrient loadings as a necessary

condition for restoration of Good Ecological Status.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

3. Coastal water typology

The Scientific Documentation Report uses a modified Danish coastal water typology as

the basis to calculate reference conditions and targets for coastal waters as well as

Maximum Allowable Inputs (MAI). The typology is a crucial element for all following

steps. Therefore, in this chapter, the Panel evaluates its suitability, analyses its

shortcomings and provides suggestions.

3.1 Basic idea behind typology

Annex 2 of the WFD gives instructions on how typology should be carried out and lists

the obligatory and optional factors that can be used (see Chapter 2 of this evaluation).

Most European Union member states applied the most specific system B. In this

approach, the physical and chemical factors that determine the characteristics of coastal

and transitional waters are latitude, longitude, tidal range and salinity as obligatory

factors. Optional factors are current velocity, wave exposure, mean water temperature,

mixing characteristics, turbidity, retention time, mean substratum composition and water

temperature range.

The Common Implementation Strategy (see Chapter 2) for the WFD (2000/60/EC):

“Guidance Document No 5” on “Transitional and Coastal Waters – Typology, Reference

Conditions and Classification Systems” provides a detailed guideline for carrying out a

characterisation of all water bodies, referred to as typology. The aim is to produce a

simple physical typology that is both ecologically relevant and practical to implement. It

aims at linking similar water bodies under one type to enable the establishment of type-

specific reference conditions. Guidance Document No 5 suggests ranges for several

factors that could be used in the typology.

Once the water has been characterised as transitional (TW) or coastal (CW), a typology

for each is developed by the member states. Denmark, like Germany, chose to include

estuarine waters within coastal waters because all the parameters, except depth, are the

same. Whether a typology separates transitional and coastal waters or combines both

under coastal waters does not make a difference for the calculation of reference

conditions, targets and Maximum Allowable Inputs (MAI). It does not necessarily affect

the number of types nor the number of water bodies in a country. Because of the narrow

guidelines, most countries considered the typology development as a largely technical

task.

3.2 The Danish typology

According to the Common Implementation Strategy for the WFD (2000/60/EC), Dahl et al

(2005) divided the Danish coastal waters into 15 different types: 5 open water types and

10 estuary types. Transitional waters were included in the typology. Dahl et al (2005)

state that “the large number of types reflects the strong salinity gradient present in the

Danish coastal waters, but also that the physical factors that are relevant for defining a

type, vary greatly among the Danish estuaries”. The national typologies in the Baltic

Region show many similarities, and in several cases coastal and transitional waters were

merged into one system to reduce complexity. In the Scientific Documentation Report,

the Danish typology is further simplified and types are merged. The aim of this

simplification is that less Chlorophyll a reference and threshold values for a Good

Ecological Status (target values that fall between the High-Good and Good-Moderate

boundaries) have to be defined.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

The Common Implementation Strategy for the WFD (2000/60/EC) reminds member

states that, when developing a typology, they should keep the major objective of the

Directive in mind, namely to establish a framework for the protection of both water quality

and water resources preventing further deterioration and protecting and enhancing

ecosystems. It is pointed out that typology is a tool to assist this process, and it is

recognised “that a simple typology system needs to be complemented by more complex

reference conditions that cover ranges of biological conditions” (p. 28). It means that

every country has the freedom to adjust the typology to its own needs and to refine it to

the required degree.

3.3 Suitability of the Danish typology

The major question is whether the typology in the Scientific Documentation Report is

sufficiently detailed to allow the definition of reliable reference and target values for

Chlorophyll a and the other indicators in all coastal waters. Reliable means that these

values well reflect the ecological conditions and properties of all coastal waters. This is a

precondition for defining target values that allow to derive to derive reliable MAI for each

water body. The general impression is that the typology allows the derivation of suitable

target values and MAI for water bodies in the sea with strong water mixing. An indication

is that the inter-calibrated values for Chlorophyll a with Germany and Sweden for the sea

and outer coastal waters are well in agreement with the results of the Scientific

Documentation Report (Schernewski et al, 2015). In general, the comparable German

Chlorophyll a target values for the open sea are slightly lower, but would allow a cross-

border harmonisation.

Many fjords and coastal bays share a similar Chlorophyll a target concentration of 3.6

mg/m³, namely Norsminde Fjord, Mariager Fjord (outer), Nissum Bredning, Randers

Fjord (outer), Horsens Fjord, Kolding Fjord, Vejle Fjord, Odense Fjord, Nyborg Fjord,

Kerteminde Fjord, Holckenhavn Fjord, Bredningen, Emtekær Nor, Nærå Strand,

Nakkebølle Fjord, Dalby Bugt, Karrebæk Fjord and Roskilde Fjord.

The Scientific Documentation Report and the additional data tables provided by the

authors of the report show that water bodies with diverse properties are represented by

only one target value. The typology is too simplified to reflect the specific characteristics

of the individual fjordic water bodies. The consequence is a large and not sufficiently

justified variation in the required load reduction for each water body. In the

understanding of the Panel, the Danish typology does not sufficiently reflect the

individual properties of the many Danish fjords and inner coastal waters. The solution

could be either to subdivide the typology for these systems, taking into account

especially water exchange rate and fresh water discharge, or to develop individual

Chlorophyll a target values for every single water body. The statistical modelling,

especially when carried out across water bodies, could be an excellent basis for this.

3.4 Suitability of the Danish monitoring programme

A precondition for a refined typology for fjords and inner coastal waters is the existence

of a suitable and comprehensive monitoring programme. The present Danish national

monitoring programme includes more than 90 stations along the coast and in the sea. It

is very comprehensive and seems to be well-adjusted to the WFD requirements.

Altogether, 119 water bodies are separated in Denmark. In some cases, fjord systems

are divided into two or more water bodies and are represented only by one monitoring

station. Examples are Mariager Fjord, Randers Fjord, Vejle Fjord and Flensborg Fjord. It

means that practically every water body or spatially linked group of water bodies (like a

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

fjord) are represented by one monitoring station. This is important, because only the

existence of a monitoring station and regular data collection allows assessing whether

the target is reached or not.

3.5 Suggestions towards a modified approach

Such a comprehensive monitoring programme not only allows a refinement of the

typology, but would allow the definition of individual Chlorophyll a reference and target

values for every water body, respectively some spatially linked group of water bodies.

We strongly suggest considering this approach, especially when the aim is to calculate

as precise and water-body specific MAI as possible. Denmark is one of the few countries

in Europe, where the necessary data, expertise and models are available for such a

comprehensive approach. In detail, it has to be assessed if additional monitoring

stations, temporary data collection at some locations or complementation of the

monitoring programme with remote sensing might be necessary. Neither the application

of the meta-modelling nor monitoring of the success of the proposed measures are

possible without a minimal set of follow-up actions in the field.

3.6 The look abroad

Similar discussions took place in Germany as well, and the results are reflected in a

national report and an international publication (Schernewski et al, 2015). Germany

carried out a comprehensive revision of all German Baltic reference and target values for

nutrients and Chlorophyll a. The discussion process within the accompanying official

national working group came to the conclusion that especially the different estuaries and

lagoons have so specific properties and behaviours, and that type-specific Chlorophyll a

and nutrient reference and target values would be too general. As a consequence,

specific Chlorophyll a and nutrient reference and target values were developed for every

single water body, resulting in 35 major Chlorophyll a reference and target values for the

German Baltic waters alone.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

4. The use of seagrass and Kd as

environmental indicators

The use of Chlorophyll a as an indicator for phytoplankton is widespread and accepted in

the WFD. The indicator is intercalibrated between Denmark and neighbouring countries

and is fully endorsed by the Panel. In this chapter, we will discuss the appropriateness of

other indicators. We discuss the use of Kd as an indicator for aquatic macrophytes and

angiosperms, and subsequently devote a discussion to the indicators for hypoxia and

nutrient limitation, used in the statistical modelling approach. The main questions are

whether these indicators are appropriate for demonstrating important ecological quality

aspects, whether they can be related to nutrient inputs, and whether they should be

maintained in the scientific modelling approaches.

4.1 Kd as an indicator for the biological element “benthic

vegetation, macroalgae and angiosperms”

One of the three main indicators used in the WFD as measures of Good Ecological

Status is the condition for aquatic macrophytes and angiosperms. In most Danish

estuarine and marine waters, this concerns eelgrass (Zostera

marina),

even though this

is not the only species of angiosperm that occurs. Pondweed (Potamogeton species)

and

Ruppia

may cover extensive parts of some systems and should be taken into

account as “angiosperm vegetation”. However, in systems where this occurs (e.g.

Odense Fjord), it is still eelgrass that dominates the deeper (>1.5 m) parts and thus

remains the most critical indicator. The Panel has no complete overview of the situation

in all the different water bodies, but stresses the generality of the required “angiosperm

vegetation” indicator, so that it may occasionally differ from the single “Zostera maximum

depth” indicator, at least in principle.

In the scientific documentation report and the underlying model work, water

transparency, expressed as the light extinction coefficient Kd (m

-1

) is used as a proxy for

the depth limit for eelgrass. This is based on solid scientific evidence that eelgrass needs

a light intensity at the bottom of between 10-20% of the incident light. The choice of 14%

is based on this literature, and on area-specific experiments for Danish waters, and is

well justified. However, the Panel points out that, due to the non-linear interaction

between light intensity and Kd and in the presence of temporal variability in Kd, the

average light intensity reaching the bottom in a water system at a particular depth may

differ significantly from the light intensity calculated at this depth using the average Kd.

As clearly stated in the report, water transparency is a necessary but insufficient

condition for seagrass to re-establish in these estuarine systems. Recent studies of

seagrass reproduction, as well as adult seagrass survival, have pointed to factors such

as disturbance by floating algae, resuspension of fine material, disturbance by lugworms,

herbicides and others to influence recovery (Flindt et al, 2016; Kuusemäe et al, 2016;

Canal-Verges et al, 2016). It is likely that the presence of eelgrass itself plays a role in

these circumstances, not only as a seed source but also by collecting and fixing fine

sediment material. It has been shown in general (van der Heide et al, 2011) and in a

specific restoration case in North America (Orth et al, 2012) that this may lead to

alternative stable states and strong non-linear behaviour: once extensive seagrass

As an example, if Kd has a lognormal distribution with a logarithmic mean of -1.0 (mean approx.

0.4 m

-1

), and a logarithmic standard deviation of 0.5, the depth was on average 14% of incident light

reaches the bottom is 6.15 m, whereas the depth limit calculated from the average Kd is 4.85 m.

Note that the difference depends on the statistical distribution of Kd in the field, which the Panel

cannot evaluate.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

meadows are present, they contribute to keeping the water clear and extend their range

to deeper waters, but in the absence of meadows the water remains turbid and prevents

the development of meadows. As a consequence, restoration of seagrass meadows (i.e.

transition from unvegetated to vegetated state) may require more stringent conditions to

be fulfilled than what is needed to maintain an existing vegetation. These more stringent

conditions may relate to lower nutrient loadings, but may also relate to the exclusion of

other disturbing factors. Therefore, it is unlikely that Kd as a sole indicator covers the

entire range of conditions needed for eelgrass restoration, but it is even more unlikely

that restoration will succeed without at least restoring Kd to the levels needed for the

Good-Moderate boundary conditions.

The time course of Kd in the water bodies studied by statistical modelling is shown in the

Annexes to this evaluation report. In most cases, it is very difficult or impossible to detect

a significant downward trend in the values. Even though a significant correlation of

summer (June to August) averages of Kd with N load is reported for 16 out of 22 stations

[p 94], the slopes of these relations are very low [p 94], and no material changes in

yearly averages are observed over time despite changes in N loading. Similarly, in the

mechanistic modelling, slopes for change of Kd as a function of N load are usually small,

and the model is not able to reproduce the reference (observed around 1900) Kd values

by modelling reference loads of 1900. The total range in Kd across all systems at

reference loadings is approximately 0.15 m

-1

, whereas the total range in the observations

is approximately 0.35 m

-1

. The model is calibrated to reproduce the mean reasonably,

but is not able to capture the full (temporal and cross-system) variation very well.

Based on a statistical analysis of all Danish systems since the 1980s, Riemann et al

(2016) report a significant increase in Secchi depth between 1980 and 1990, when many

systems made the transition from hypertrophic to eutrophic state, but absence of any

systematic trend afterwards. In addition, they remark that Secchi disc depth data even

overestimate the effect on Kd, because of a shift from scattering to absorption as the

main light extinction mechanism between 1980 and 2010. In summary, none of the

within-system statistical analyses or models seem to be able to demonstrate a strong

dependence of Kd on nutrient loading in the period 1990-2013.

However, when viewed across systems, the data shown in annex B of the Scientific

Documentation Report for Chlorophyll a and Kd in the systems studied with the statistical

modelling strongly suggest a close correlation between average Chlorophyll a

concentration and average Kd over the study period (see Figure 4 in Chapter 8). It is

likely that a common cause – most probably the relative influence of the freshwater end

member in the water of the estuary – determines both. However, within each of the

systems, we do not observe a correlated evolution in time of the two indicators over the

1990-2012 period. A further interesting observation is that the targets for Kd and

Chlorophyll a, as derived in the Scientific Documentation report, show exactly the same

correlation but in a narrower range of both indicators. As these reference values

represent historic conditions, the suggestion is that on a centennial time scale, Kd and

Chlorophyll a covary in time. Therefore, it is possible that Kd does respond to nutrient

loading, but with significant delay and only on long time scales.

From these considerations, the Panel concludes that both indicators represent

eutrophication effects, but that the estimation of the effect of nutrient reduction on

Chlorophyll a is more reliable than the estimation of this effect on Kd.

The Scientific Documentation Report suggests that other causes, in particular the influx

of dissolved and particulate organic matter from freshwater, as well as long-term storage

of fine and fluffy sediment material, influences the transparency of the water. Trends in

benthic filter feeders (that have decreased significantly in biomass between 1990 and

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

2017 – see Riemann et al, 2016) may also be a causal factor. It can be assumed that

filter feeders decrease in biomass as a consequence of the decrease in phytoplankton

primary production, which may result in less filtration and fixation of fine particles in the

sediment. It is difficult to evaluate each of these hypotheses, but the very good

correlation between average chlorophyll and average Kd across systems suggests that

the influx of some substance with the freshwater, that may be higher nowadays than in

1900, plays a dominant role. In their answer to the questions of the Panel, the

researchers have thoroughly analysed and dismissed the possibility that herbicides play

a significant role in this freshwater influence. The most probable hypothesis is that the

influx of coloured organic substances has increased between 1900 and now.

The consequences of incorporating Kd as an important indicator for water quality, in the

absence of strong slopes between nutrient loading and Kd, are different for the

mechanistic modelling and the statistical modelling exercises. The mechanistic modelling

estimates which part of the distance between target and status can be bridged by

reducing Danish land-based N sources. It corrects for this fraction in the calculation of

the effort required. The Panel finds this approach appropriate and does not think it leads

to unjustifiable overestimation of efforts needed.

The statistical modelling approach does not follow the same reasoning as the

mechanistic modelling. For some water bodies, N load reductions of well above 100%

are calculated to be needed in order to bring Kd down to target levels. This is of course

physically impossible. The problem is solved by “translating” the required very high

efforts into realisable efforts [25% when the calculation is 25%-100%, 50% for calculation

100%-200%, 75% for calculation >200%). Despite questions to the researchers, the

Panel has not been able to discover the logic behind this translation. The researchers

argue that this is basically expert judgement and further argue that 25% is the order of

magnitude of interannual variation of the N load, therefore when an effort is estimated to

be “large”, it should be above this level but not too much. In the opinion of the Panel, the

“translation” introduces an unnecessary element of arbitrariness into the whole

procedure that is in contrast with the general evidence-based approach and that

therefore exposes the entire procedure to unproductive criticism. The Panel furthermore

observes that the situation here is analogous to the situation treated in the chapter on

mechanistic modelling, where very often the target value cannot entirely be reached with

reduction of Danish land-based N. Therefore, the Panel suggests harmonising the

approach across the two modelling lines and adopting the approach of the mechanistic

modelling also in the statistical modelling.

In further work, the Panel recommends reviewing the approach for this WFD indicator by

starting from the basic observation that not Kd, but survival and restoration of aquatic

angiosperm vegetation is the real criterion. In some systems, this criterion may actually

be fulfilled by other species than eelgrass (e.g.

Ruppia

Potamogeton

species), in

which case the criterion could also be considered as generally fulfilled. However, in most

cases, eelgrass will be the species of interest. As mentioned above, recent modelling

work of Kuusemäe et al (2016) and Flindt et al (2016) has taken a more comprehensive

view on restoration of eelgrass, and the influence of nutrient loading on the process. This

work is actually built into the mechanistic models used in the present study, but the

results have not been directly used in order to estimate the influence of nutrient

reduction on seagrass restoration. The Panel proposes to make better use of these

models, probably after more extensive validation, to more directly estimate the effect of

nutrient reductions on seagrass development possibilities.

In view of the apparent difficulties in estimating the effect of nutrient reductions on Kd at

short time scales, the insufficiency of Kd as a representation of all factors needed for

restoration of seagrass, and the high correlation between Kd and Chlorophyll a both in

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

status and targets at longer time scales, the Panel suggests to relatively downweigh the

importance of Kd in the final calculations of reductions needed. It further recommends

pursuing studies attempting to estimate conditions for seagrass restoration based on

already developed more comprehensive models. In the absence of the latter, and given

the correlation between Kd and Chlorophyll a, the Panel is of the opinion that adherence

to the “one-out, all-out” principle with respect to Kd and Chlorophyll a, is not imperative.

A weighted average of reduction needs for both indicators might be preferable.

4.2 Other indicators used in the statistical modelling

In contrast to the mechanistic modelling, the statistical modelling bases its conclusions

on three other indicators: (1) the occurrence of hypoxia, (2) ecological signs of hypoxia

from nutrients and (3) chlorophyll and (4) the number of days of N limitation of

phytoplankton growth. Indicator (2) and (3) are given half weight as they estimate one

element together. Compared to Chlorophyll a and Kd, the combination (2) - (3) and the

indicators (1) and (4) are given half the weight.

The Panel is surprised by the inclusion of these indicators in only one line of modelling,

as it could also have been done in the mechanistic modelling. The latter contains all the

variables needed to estimate hypoxic/anoxic conditions as well as direct estimates of

nutrient limitation of phytoplankton growth. The asymmetric situation leads to a decrease

of comparability of the two models and decreases credibility of the procedure of

averaging both approaches, e.g. in meta-modelling. The Panel further notes that in

meta-modelling based on the statistical approach, the ancillary indicators are sometimes

included and sometimes not, depending on data availability.

With respect to the occurrence of hypoxia, the researchers note in the Scientific

Documentation Report that:

“There is direct evidence for a relationship between nutrient loadings and oxygen

concentrations in bottom water (Markager et al, 2006) and the size of hypoxic/anoxic

areas (Scavia et al, 2003; Christensen et al submitted). However, these relationships are

complicated by a considerable time lag and a high sensitivity to climate variables like

water temperature and wind stress.”

With respect to the number of days with nutrient limitation, figure 8.7 of the Scientific

Documentation Report shows a direct correlation with Chlorophyll a concentrations, but

with considerable scatter (considering the log scale of the y-axis). Two questions can

thus be posed: Do the additional indicators measure a significantly different indicator

compared to Chlorophyll a and Kd, and can the effects of nutrient reduction on both

indicators be estimated reliably?

The Panel is of the opinion that both criteria lead to doubt about the usefulness of these

indicators. It is clear that phytoplankton production and biomass is related to the amount

of organic matter sinking to the bottom and fuelling oxygen consumption. This could be a

reason to include the spring in the Chlorophyll a indicator, but anyhow a correlation

between Chlorophyll a and the probability of occurrence of hypoxia can be expected.

Note, moreover, that excessive summer Chlorophyll a is the ecological sign of hypoxia

used. The high dependence of occurrence of hypoxia on weather conditions induces

considerable variability that obscures effects of nutrient reductions on the indicators.

There is essentially only one (on/off) observation per year. In addition to these

difficulties, a rather arbitrary look-up table approach has to be used in order to estimate

the required nutrient reduction for improvement in the hypoxia indicators.

For the indicator “days with nutrient limitation”, it can be expected that nutrient reduction,

if effective on Chlorophyll a at all, can only be effective through the increase of the time

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

duration of nutrient limitation. It is hard to see how the response of this indicator could

differ from the response of Chlorophyll a concentration. The latter, however, can be

measured more easily and more reliably. We note that there is considerable

disagreement in the literature on the correct value of Km, the Monod limitation

parameter, and that it differs considerably between different phytoplankton species and

groups. Also for this indicator, a look-up table has to be used to estimate the required

load reductions.

In summary, even though the ancillary indicators aim at describing important ecological

phenomena, it is not easy to translate them into required load reductions (expert

judgment and look-up tables are needed) and their added value compared to Chlorophyll

a and Kd is limited. Therefore, the Panel is of the opinion that these indicators do not

bring a substantial improvement of the approach. The Panel recommends using the

mechanistic models to better study how the important phenomenon of oxygen depletion

can be linked directly to required nutrient reductions before using it in practice to

estimate required nutrient reduction. If, based on these studies, it can be decided to use

these additional indicators, they should be introduced in both statistical and mechanistic

modelling approaches for consistency of the approach.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

5. Emphasis on nitrogen versus phosphorus

In this chapter, it is evaluated to what extent the Scientific Documentation Report

a priori

focused exclusively on nitrogen (N) reductions as measures to reach Good Ecological

Status, or if evidence is given that rules out positive effects from phosphorus (P) load

reductions. We thus address the question whether management options were

unnecessarily limited by focus on reduction of the yearly N load only.

5.1 Phosphorus limitation

The nutrient emissions from large point sources were dramatically reduced already

during the 1980s, causing a large and relatively sudden decrease in the P loads. After

that first effort, focus has been directed to mitigation of N loads, primarily through

measures in agriculture. This has resulted in a smoother, but substantial, decline in the

nutrient inputs (mostly N inputs) thereafter. Currently, the overall inputs of N and P are

roughly about 4.2 and 3.4 times higher, respectively, than estimated reference inputs for

the year 1900 (Riemann et al, 2016). This indicates that the N/P ratio of nutrient inputs is

not exceptionally deviating from the historic inputs.

Previous studies have shown substantial and significant reductions of primary production

(e.g. Timmermann et al, 2014) and Chlorophyll a concentrations (e.g., Riemann et al,

2016) in response to the early P load reductions. Thus, there is no doubt that reduction

of P loads can, in principle, lead to improvement of water quality in terms of WFD

indicators. However, it is uncertain to what degree these historical responses are

transferrable to present-day conditions, because emissions from point sources did not

have the annual cycle of the diffuse sources and, moreover, they were observed in

generally hypertrophic situations that are not comparable to the present state.

Traditionally, marine coastal waters have been regarded as N limited, but in the past

decades, scientists have become increasingly aware of complicated co-limitation

patterns and intricate nutrient dynamics. Processes such as N fixation and sediment P

release can modify long-term response compared to the direct response of

phytoplankton to nutrient additions on short time scales. A number of studies from

Danish waters confirm that N is in general limiting algal production during summer time,

and P is often limiting in spring, but there are seasonal and spatial variations of nutrient

limitation. These field studies suggest that at least in a number of systems, regulation of

annual primary production by P load reduction could be feasible.

5.2 Treatment of nitrogen and phosphorus in the Scientific

Documentation Report

There are several elements that have contributed to the large emphasis on N load

reduction in the Scientific Documentation Report. In particular, we discuss the nature of

the indicators used, the selection of the study period, the procedures of the statistical

modelling and the characteristics of the mechanistic model.

The basis for all calculations are the indicators Chlorophyll a and Kd during summer.

This has potential implications for the exclusive focus on nitrogen load reductions.

Summer phytoplankton in most Danish water bodies is predominantly nitrogen limited.

The choice of summer Chlorophyll a as an indicator may have focused the attention

primarily on processes that are dominant in summer and on nitrogen loads as a primary

factor responsible for eutrophication. This is also pointed out in the Scientific

Documentation Report, where it is suggested that developing new indicators focusing on

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

other parts of the season would give a more diverse focus on both nitrogen and

phosphorus.

In general, the Panel is of the opinion that the selection of indicators only representing

summer conditions could be too restrictive. In waters with some degree of stratification,

the spring bloom has the highest contribution to export production, fuelling the organic

matter on the sediment and largely determining the oxygen demand in the rest of the

season that could lead to P release from the sediments. The Scientific Documentation

Report suggests that limitation of the spring bloom by P occurs in a number of water

bodies, thus suggesting that the effectiveness of P load reduction on an indicator

representing the full growing season could be significant.

Another factor potentially excluding possible influence of P in the analysis is the selected

period for the statistical model (1990-2013). This period excludes most of the period of

major development of efficient sewage treatment in the 1980s that caused a major

decrease in point source P loads. For most water bodies, the P load trends that are now

dominated by diffuse sources are less significant than N load trends, and thereby it is

naturally more difficult to find significant effects. However, the Panel endorses the choice

of period, because the seasonality and mechanisms of P limitation in current situations

may differ from the historical, point-source dominated situation, as argued above.

In the statistical modelling approach, the variable selection procedure may have masked

the potential role of phosphorus load reduction. There is a bias in the selection of

variables towards regressions with N. This occurs first through the automated variable

selection process. Whenever N load is selected as the dominant variable, possible P

dependence is disregarded because P load is no longer considered as a secondary

independent variable. If, on the other hand, P is selected as the dominant controlling

variable, that regression model is not used. Thus, potential influence of P load

reductions, or combinations of N and P load reductions, are not investigated further.

The mechanistic models include all relevant processes for modelling effects from both N

and P and combinations of them both. However, major focus in the formulations of

scenarios is on N, and the few scenarios, including also P reductions, are not detailed

and perhaps not optimal for exploring the influence from P load reduction. In addition, we

observe that for a significant portion of the water bodies, the models seem to

overestimate P concentrations during summer. This can have eliminated the potential

impact from P load reductions on the indicators.

5.3 Possible implications for management

Based on the different factors leading to a focus on N load reduction, the Panel

concludes that the study does not demonstrate significant contributions from P loads on

the summer indicators, but the evidence is not strong enough to exclude that P

reductions or combined N and P reductions could be effective in reducing year-averaged

chlorophyll levels as well as sediment oxygen demand.

Keeping the option for combined N/P reduction open may have significant management

implications in regions where very large N load reductions are demanded. Focused

studies resulting in an envelope of combinations of Maximum Allowable Inputs of N and

P would probably lead to greater flexibility and more cost-efficient nutrient reduction

management in these areas. In making this recommendation, the Panel acknowledges

that great efforts have already been made to reduce the P load from urban waste waters,

and that little gain has to be expected from intensifying those efforts, following the law of

diminishing returns. However, any innovative approach to reducing remaining P loads,

including the P load from agriculture, could significantly enlarge the portfolio of potential

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

measures. The Panel recommends using basin load models in combination with the

mechanistic models used in the Scientific Documentation Report to investigate these

possibilities.

5.4 Seasonality

The exclusive focus on summer indicators in combination with water bodies with short

residence times implies a direct link between summer loads and the indicator. Typical

residence times in Danish estuaries are short in many cases, ranging from a few days to

about 3 months (Rasmussen and Josefsson, 2002). Even if the indicators would include

the spring phytoplankton bloom, regulation by N loads would mostly focus on the

summer period in water bodies where P limits the spring bloom. There seems to be a

possibility to regulate Good Ecological Status by focusing on the summer loads, rather

than on the yearly integrated loads. The Panel recognises that the problem is

complicated by N retention in the system in the form of organic N stocks accumulating

over the season and even years, so that the calculation is not straightforward. Moreover,

spatial displacement of problems to other systems as a consequence of flushing winter

nutrient loads has to be taken into account. Even so, the Panel estimates that the

modelling tools developed, especially the mechanistic modelling, are able to investigate

scenarios with seasonal regulation of the N (and P) input into the system. Therefore,

nutrient load management could be focused on optimising the effect in the coastal

estuaries. The Panel does not have a complete overview of the potential, in agricultural

practice, to focus in particular on summer N load. However, it recommends exploring the

possibilities to do so and use the mechanistic models to estimate how this would affect

the GES indicators.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

6. Statistical modelling

In this chapter, the Panel reviews and sums up the objectives and basic setup of

statistical modelling and their usage in defining Maximum Allowable Inputs (MAI). The

main issues are averaging of statistical and mechanistic models, the analysis of within

and cross-system variability in Chlorophyll a and Kd responses, collinearity of

phosphorus (P) and nitrogen (N) loading, filtering out the effect of flushing and the

uncertainty and resulting risk of over- and under-dimensioning of MAI.

6.1 Setup

The statistical model approach as presented in the Scientific Documentation Report aims

at demonstrating the dependence of the indicators Chlorophyll a, Kd, hypoxia, anoxia,

number of days with nutrient limitation on the N and P loading of the system as well as

on some other physical and chemical characteristics of the system. The statistical

models (there is one model per sufficiently monitored water body) also estimate how

concentrations of Total N and Total P depend on the nutrient loadings and physico-

chemical characteristics. These latter analyses are informative on the functioning of the

systems but are not really used any further in the overall modelling procedure.

A few basic choices for the setup of the statistical modelling have been made at the start

of the study. The most important choices were:

•

Restrict the database analysed by the statistical modelling to the period 1991-

2012. This implies that the major decrease in P input, as well as the ecological

consequences of this decrease, in general are not part of the analysed

database.

Restrict the construction of statistical models to those systems where sufficient

data are available. What is “sufficient” is always open to discussion, but the

Panel is of the opinion that the choices are reasonable and have been well

justified.

Use annual averages of nutrient loads, concentrations and other variables as

the basis for modelling.

Construct one statistical model per water body without cross-system model

building.

Perform a variable selection method for significant independent variables,

where (due to collinearity problems) only one type of nutrient loading (either N

or P) was entered into the set of independent variables.

•

The most important results of the statistical models are the slopes of the relation

between N loads and the indicators Chlorophyll a and Kd. These slopes are only

determined, if N load was selected as the most important independent variable and thus

entered into the statistical model. When this was not the case, substitute solutions have

been used. The statistical model that was used to estimate the slopes (Partial Least

Squares) is different from the model used to select the variables (Multiple Linear

Regression).

The construction and use of the statistical model are well explained in the Scientific

Documentation Report. Measures of goodness-of-fit are given at different stages in the

description. No formal uncertainty analysis of the model as a whole, nor variance

estimation of the estimated parameters (in particular the N load – indicator slopes) have

been given.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

6.2 Panel evaluation of basic model setup

The Panel distinguishes three major uses of the results of the statistical modelling:

•

Estimation of the relation between nutrient loading and indicators at relatively

long time scales (5-10 years), as a basis for estimation of reference conditions

of Chlorophyll a, and of the effectiveness of load reductions in reaching the

target conditions.

Provide insight into the water body characteristics that explain the differences

between water bodies in status or slopes.

Provide an independent, evidence-based check on the accuracy of the

mechanistic modelling approach.

•

The Panel remarks that the statistical models are not needed to ascertain that nutrients,

both N and P, are important for phytoplankton. This point was also made by the

researchers, stressing that the body of scientific evidence showing these relations is

massive.

In contrast to the researchers, however, the Panel questions if the step of variable

selection was needed at all. It involves mixing of two methods (MLR and PLS). It also

leads to the suggestion that in some systems, N load was not involved at all in

determining Chlorophyll a and Kd. In systems where N load was selected as the most

important determining factor, possible secondary effects of P load cannot be shown and

are obliterated. The most important consequence of this option, however, is that it may

lead to biased estimates of the slopes and MAI. If, in a particular water body, the slope is

very small (close to zero), it is very likely that N load will not be selected as the most

important independent variable in the variable selection procedure. Subsequently, for

this system, the slope will be estimated as the average type-specific slope, almost

inevitably leading to a higher slope than shown by the data. This will then lead to a lower

reference and target value for the system than the one suggested by the data. As these

reference values will enter into a type-specific averaging afterwards, the final

consequences of these choices become difficult to assess, but likely affect the targets for

all systems in the type.

Moreover, the Panel is of the opinion that there is no real reason for estimating the short-

term response of the indicators on year-to-year variations in nutrient loads, with or

without time lags of a few months. Both nutrient loads and concentrations of nutrients

and chlorophyll are known to vary considerably with freshwater discharge, which is

variable from year to year. Short-term (i.e. year-to-year) responses of indicators to short-

term variations in nutrient loads will not necessarily be the same as the decadal-scale

responses that the study really wants to estimate. For instance, high discharge will not

only increase the total load of nutrients to a system, but simultaneously also decrease

the freshwater residence time and thus the ability of the ecosystem to take up and use

these nutrients. This may contrast with a decadal-scale increase in nutrient load, where

clearer and possibly also different ecological responses might be expected.

Therefore, the Panel is of the opinion that a clearer focus on the long-term slopes and

the cross-system variability is needed. Through the use of mixed or Bayesian

hierarchical models, short-term and long-term variations can be separated and

collinearity between variables can be built in as part of the model (Malve & Qian, 2006).

Danish water systems differ in a number of morphological and hydrographical

characteristics, leading to a diversity of systems that is not very well captured by the few

types used in the typology (see Chapter 3 in this evaluation report). However, there are a

few characteristics that presumably dominate the differences in nutrient, chlorophyll and

Kd status between systems. The relative influence of freshwater in the water, dependent

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

on discharge rates, flushing rates and exchange rates with the coastal system, will most

probably be a key parameter. Nutrient concentrations in seawater are relatively stable

and do not differ very much between the reference conditions and now. In contrast,

nutrient concentrations in freshwater are much higher and are obviously much more

directly influenced by nutrient loads. As a consequence, it may be expected that much of

the variation in status and slopes between systems may be explained with a cross-

system statistical model as suggested above. The main purpose of this setup is to

improve within-system estimates of slopes with information coming from similar systems

elsewhere, and to improve meta-modelling applications. It should lead to a model that

estimates the slopes (which are the results of primary importance) based on independent

variables summarising the water body characteristics, while simultaneously estimating

(and evaluating) system-specific deviations. Such an approach could constitute an

improvement with respect to the current within-system modelling approach.

In order for the statistical model to provide an independent, evidence-based check on

the results of the mechanistic modelling, two requirements must be fulfilled. First, the

procedures of the statistical and mechanistic modelling should not be unduly mixed at

early stages (see comments in Chapter 8 in this evaluation report). Second, the

statistical model should contain a formal estimation of variances of the estimated

parameters. Statistical modelling techniques have much better formal methods to

estimate uncertainty than mechanistic models, and this opportunity should be taken in

order to better formalise both uncertainty resulting from modelling and from data

uncertainty. For this evaluation to be effective, the setup of a single cross-system

statistical model is better suited than the current set of separate within-system models.

6.3 Panel evaluation of statistical model results

Even though the simultaneous development of two model lines, statistical and

mechanistic, may seem redundant at first sight, the Panel endorses the continuation of

this approach. The richness of the Danish database is an internationally exceptional

asset that provides the opportunity of an evidence-based check on mechanistic model

outcomes. This asset should be used, and the two modelling lines are a very good way

to do so. However, the Panel recommends strengthening this aspect, e.g. by keeping the

two model lines more separate and independent throughout the modelling procedure, so

that the check becomes clearer and more explicit in the final stages of result

interpretation. Furthermore, as specified above, the Panel is of the opinion that a cross-

system approach in the statistical modelling would strengthen the possibilities of

obtaining insight into possible causes for model divergence and would assist better in

choosing final management strategies based on the model comparison. In addition, a

formal uncertainty analysis of the statistical model would contribute to this goal.

With respect to the present outcomes of the statistical modelling, the Panel sees reasons

to suspect bias in estimated slopes and reference values due to the variable selection

procedure, as specified above. The Panel suspects that the slope estimates, being a

mixture of short-term and long-term ecological responses, might be biased as estimators

of the long-term response. However, the Panel does not consider these remarks as a

reason to entirely dismiss the statistical model results as unreliable. The mentioned

discrepancies are probably minor in comparison with the overall range of the results and

in comparison with the inevitable variability in the observations. The within-systems PLS

regression approach used is robust and not expected to be overly influenced by the

mixture of short and long time scales. The variable selection procedures have led to the

replacement of slopes with type-averaged slopes, but mostly in types with small slopes.

Nonetheless, there is enough reason to improve the statistical model and the slope

estimates that follow from it.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

7. Mechanistic modelling

The mechanistic models are evaluated in this chapter with respect to included processes

and technical implementation, performance and the different scenarios that are used.

7.1 The models

The mechanistic modelling is based on the DHI systems MIKE 3 combined with

ECOLAB. Four models are set up: a large-scale model encompassing the whole Baltic

Sea up to Skagerrak (IDW model) and three models of specific estuaries; Limfjorden,

Roskilde Fjord and Odense Fjord (estuary models). In all, 45 of the 119 Danish water

bodies are covered by the mechanistic models. The IDW and estuary models differ in

some specific ways, adapting them to the circumstances. However, the three

implementations of the estuary model are identical in terms of processes, but needed

somewhat different calibration.

The pelagic dynamics in both models follow classic NPZ concepts similar to other

models, and the bacterial loop is not explicitly resolved. An addition to many other similar

models is that internal nutrient pools are explicitly modelled using the Droop equations

(Droop, 1968). Both models also feature explicit benthic vegetation state-variables, but

not benthic fauna.

The estuary models are quite comprehensive in terms of processes, including

sophisticated representation of benthic vegetation and elaborate description of

resuspension coupled to dynamic wave-shear processes from the hydrodynamic model.

Spatial sediment characteristics are taken into account both for sediment-water

interaction and as controlling the benthic vegetation.

Specifically, the IDW includes three autotrophic groups to take into account the seasonal

succession and nitrogen fixation typical for the open sea areas of the Baltic Sea. Further,

the representation of the sediments does not include explicit representation of inorganic

particles and instead an empirical direct relationship between shear stress and turbidity

is used. Simplification of the sediment module was necessary because of lack of detailed

information from the wider area and because of computational constraints.

In many biogeochemical models, Chlorophyll a is estimated from the autotroph biomass

in retrospect using a specific ratio. The models used in the Scientific Documentation

Report are more advanced in this aspect in that Chlorophyll a is dynamically calculated

based on fitness of the autotrophs and light conditions. In the IDW model, where there

are three autotrophic functional groups, the weighted average contributions from all

groups are taken into account in calculating the production and removal of Chlorophyll a.

The water transparency, Kd, is computed from a relationship that includes Chlorophyll a

concentration, detritus carbon, (coloured) dissolved organic carbon and inorganic matter.

All these components are explicitly modelled, although the inorganic matter

representation in the IDW is a less complicated empirical relationship than in the estuary

model.

In summary, the models are quite comprehensive and include all processes that we think

are relevant for the problem at hand. The MIKE system, with its sub-components, is a

mature system, although it is not so frequently used by research scientists, and,

therefore, there are not that many peer-reviewed articles with applications as there are

for some open access model systems. Despite this, we have no reason to question the

model system capabilities.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

7.2 Model setup, calibration and validation

All model setups have high resolution, both in horizontal and vertical. The IDW resolution

is sufficient to resolve the internal physical dynamics, both of the narrow straits and

geostrophically balanced Kattegat-Skagerrak front. The computing cost of the high

resolution and high degree of complexity is significant, leading to a trade-off in the

execution of calibration and experiment simulations. All relevant forcing functions are

taken into account in a sensible way. The time period of simulation was 2002-2011. A

critical part of the riverine inputs is the division of whatever carbon data available into the

different categories of organic carbon in the model, especially the CDOC (coloured

dissolved organic carbon) that influences Kd and the refractory and labile fractions of

organic nutrients. This has been handled to the extent possible according to the

Scientific Documentation Report.

At least the hydrodynamics of the IDW model have been used previously and were set

up as a part of the EIA for the Fehmarnbelt fixed link project. The models for Odense

Fjord and Roskilde Fjord are applied to vegetation modelling applications in Kuusemäe

et al (2016) and Flindt et al (2016). Only the Limfjord model is newly developed. Thus,

there is some history behind three out of four implementations.

All four model implementations are calibrated independently. That resulted in somewhat

different parameter setting, also of the structurally identical estuary models. According to

the researchers, there are only about 10 parameters that differ, and all of these are

within the sediment module. The actual calibration procedure is not described in detail,

but for the three models that have a past history, it can be expected that this has been

an iterative process over some time.

7.3 Validation

The hydrodynamics are evaluated quantitatively with respect to salinity and temperature.

Salinity is important since it indicates whether circulation is correct and gives the right

mixing between the riverine water and open sea water in the estuaries of different sizes.

Temperature is of less importance for the circulation, but of imperative importance for the

biogeochemical processes. The quantitative comparison shows that the model results

are well within the criteria. Upon request from the Panel, the researchers supplied direct

time-series comparisons between observations and model results for all four models,

and inspection of these shows excellent agreement between model and data for both

salinity and temperature. The Panel is convinced that the models give a quite accurate

representation of the physical processes.

The validation of the biogeochemical models is done primarily through comparison with

observations of Chlorophyll a, Kd and nutrient concentrations. To simplify presentation of

the validation of the biogeochemical processes in the models, results are aggregated per

water type and month. This presentation may hinder interpretation of the magnitude of

the difference between modelled and observed annual cycles. Quantitative skill

assessment was performed by computing a cost function (measuring mean deviation

scaled by variation of the variable) and correlation from simultaneous model results and

observations. Upon request from the Panel, the researchers also supplied example time-

series of concurrent observations and model results from selected locations for the

standard measured variables.

The model separates well the differences in Chlorophyll a, Kd and nutrients between the

water types, and mean values are well captured for all variables.

The seasonal cycle of Chlorophyll a is well captured, although levels are somewhat low

during late spring – early summer in type 2 and 3 water bodies, and the autumn bloom

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

seems to be underestimated in type 1 and 2 water bodies. The seasonal cycle of Kd is

quite weak in especially type 1 and 2 water bodies, so it is difficult to value the accuracy

from the seasonal averages in these water bodies. The tendency for all types 1-3 is,

however, that Kd in summer is less than during winter, indicating some influence from an

early spring bloom, but probably more from winter river runoff and turbidity from

resuspended material. The time-series plots of Kd supplied by the researcher confirm the

complications. The two open sea stations show seemingly random variations in time of

observed Kd due to short-term variability, and no visual seasonal cycle or trend can be

identified. There is no annual cycle (and only small variation) to be seen in the time-

series supplied for Odense Fjord and central Limfjorden, neither in observations nor in

model results. In Roskilde Fjord, there is significant variation in Kd with the seasons, but

from visual inspection the pattern is irregular and not very well captured by the model,

and it is not obvious what causes the variations. The time-series with clear seasonal

cycle are from the inner part of Skive Fjord, and here the model accurately simulates the

low Kd in winter time and high Kd in summer time.

The seasonal TN is modelled accurately for all water types. Winter DIN is somewhat

overestimated in type 1 and 2 water bodies, and DIN is somewhat overestimated in late

spring – early summer in type 3 waters. The overestimate of winter DIN in type 1 waters

is confirmed for the time-series examples supplied by the researchers. However, overall,

the model performs well on the nitrogen cycles.

There seems to be a consistent overestimation of DIP in the summer in type 1 and 2

waters, although somewhat later in type 1 than in type 2 waters. From inspection of the

time-series, it seems that the problem is larger in the IDW, smaller in the Limfjorden

model, while in the Odense Fjord and Roskilde Fjord, the seasonal cycle is quite correct.

Winter DIP and TP concentrations are accurately modelled for all water types.

The quantitative validation in terms of cost function and correlation confirms the

qualitative validation discussed above. Overall, the model is low in bias (cost function)

indicating that the levels are modelled accurately, with exception of DIP in type 3 and to

some extent type 1 waters and Kd in the type 5 waters. However, correlation is absent

for type 1 water bodies and weak for type 3 water bodies for Kd.

The comparison between modelled and observed primary production indicates that the

model performs well in this respect.

7.4 Reference conditions simulation

A hindcast simulation representing conditions around 1900 was performed. Forcing in

general was kept as for the 2002-2011 period, but loads and nutrient boundary

conditions needed adjustments. Appropriate waterborne and airborne loads were

obtained from existing well-established data sets, and boundary concentrations in

Skagerrak were adjusted according to previously published methodology. To overcome

the computational challenge of running the whole of the Baltic Sea to steady state, initial

conditions were adjusted in the IDW model according to literature values. It is the Panel’s

opinion that the setup of the simulation of reference conditions with the mechanistic

model is sound and based on current published scientific knowledge on the nutrient

loads around 1900.

Reference Chlorophyll a concentrations for all water bodies were extracted as average of

the last 5 years of the simulation. In a few cases, simulations were repeated in order to

be sure that average conditions were in equilibrium with the reference loads. It is unclear

whether Chlorophyll a concentrations were spatially averaged over the water bodies or

not.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

7.5 Scenarios and establishment of cause-effect

relationships

A prerequisite in construction of load reduction scenarios is implementation of BSAP for

other countries than Denmark. That implies major reductions of primarily phosphorus to

Baltic Proper, Gulf of Finland and Gulf of Riga, but also nitrogen to Baltic Proper,

Kattegat and Gulf of Finland. The response time to the load reductions to Baltic Proper

and the Gulfs is very long. Estimations show that during the first decade after

implementation, all changes are within natural variability, but significant reduction in

winter nutrient (primarily phosphorus) concentrations will be seen between one and two

decades after implementation (HELCOM, 2013). The response time-scale has been

shown to vary between models (Eilola et al, 2011), but is long in all cases. This means

that the influence from load reductions to the Baltic Proper and the Gulf is limited during

the decade considered here. It could be noted that in the longer perspective, nutrient

concentrations would continue to decrease, and according to the underlying calculations

in the BSAP, load reductions to the Baltic Proper was a prerequisite for obtaining GES in

the Danish straits.

Three nitrogen reduction scenarios for the Danish loads were constructed by reducing

proportionally all waterborne loads by 15, 30 and 60%, respectively.

There is also a set of scenarios where the three nitrogen scenarios are combined with a

spatially distributed phosphorus reduction scenario according to reductions specified by

the Danish EPA. It is mentioned that no significant effect could be detected from the P

load scenarios, but there is no further elaboration on these scenarios. If the distribution is

such that most of the reduction occurs to relatively few water bodies, there could

potentially be an effect in these that would not be seen overall.

Indicator values from model results are calculated as water body spatial means, and

these are corrected to match the mean observational value at the measurement station.

The scenarios without phosphorus load reduction are used to estimate parameters for a

simplified surrogate model, built on temporal averages over 2007-2011. The three

scenarios are used to establish the linear response function. The extrapolated value at

present day Danish loads will represent the indicator value, given only reductions by

other countries. In the Scientific Documentation Report, also an average indicator value

from the reference scenario is included to indicate how much higher the value will be

because of higher loads from other countries. For most water bodies and indicators, the

linear approximation is appropriate. It should be remembered though that only a

proportion of the full effect from BSAP reductions has had time to develop in the

scenarios, and one would expect that for open sea water bodies, especially in the south,

water quality will continue to improve as time goes by.

There are implications from the approach of running scenarios with a constant

proportional load decrease in the scenarios. Some water bodies will be subject to

change due to load reductions to adjacent water bodies. Therefore, one cannot directly

sub-divide MAI to individual water bodies, if there is a risk that reduction is necessary

also in adjacent basins to obtain GES. To fully disentangle the individual contribution

spatially between all water bodies, one would need to test sensitivity to load reductions

to each individual water body by itself, and perhaps, if the effect is non-linear, even

combinations of water bodies. This would be a major computational challenge, and the

improvement in the results would most probably be minor. The reason for the latter is

that the problem mostly applies to open sea water bodies that would in any case

integrate the load reduction for a relatively large region, while enclosed water bodies still

are mostly dominated by local reductions.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

7.6 Conclusion on the mechanistic models

Having evaluated the mechanistic models, the Panel comes to the following conclusions:

•

The models are clearly state-of-the-art, both in terms of numerical techniques

and processes included. The quality of the results follows a high standard and

is as good as, or better than, other similar coupled physical-biogeochemical

model systems.

The hydrodynamics seem to perform excellently.

Levels of Chlorophyll a, Kd and nutrients are accurately modelled across water

body types.

The biogeochemistry seems to perform overall somewhat better for nitrogen

than for phosphorus, although in the models for Roskilde Fjord and Odense

Fjord, also phosphorus performs excellently. Weakest is the performance of

nutrients in the IDW model, where relatively frequently DIP seems to be

overestimated during summer or early autumn and nitrogen during winter.

Observed short-term variability in Kd in open waters is such that it seems

impossible to model.

Long-term response to large changes in nutrient loads has not been validated.

The nitrogen reduction scenarios are appropriately set up and relevant.

The scenario for P reduction is not extensively described, and it cannot be

judged whether it forms sufficient basis for exclusive focus on N.

It should be noted that in a longer time perspective, >10-20 years, the effect

from BSAP load reductions will influence the open sea water bodies, especially

in the southern part of the region.

It would be extremely valuable to extend the mechanistic modelling system to

as many water bodies as possible.

•

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

8. Calculation procedures to estimate Maximum

Allowable Inputs from model results

In this chapter, we discuss the general build-up of the procedure to estimate reference

conditions, Good-Moderate boundary targets and the required N load reductions to

reach the target conditions. These procedures are based on the statistical and

mechanistic model results, but use and interpret them in a diversity of ways. In our

discussion, we focus on how the different models interact and on the different steps

taken to arrive at the final MAI per water body.

8.1 Steps in the calculation of targets and MAI

Despite the general logical nature of the procedure, and even though the Scientific

Documentation Report gives extensive explanations of the detailed procedures followed,

it is not easy to follow and weight the different steps used in deriving the Maximum

Allowable Inputs (MAI) for the water bodies. The essential steps, as the Panel

understands them, are summarised in the diagram shown in Table 2. The left column

refers to the procedures followed in the statistical modelling, and the right column to the

mechanistic modelling. Joint cells point to steps where both approaches are joined.

Table 2. Essential steps in the calculation procedure of targets and MAI in the Scientific Documentation.

Steps with averaging have red boxes.

Statistical modelling

Estimate the slope of the Chlorophyll a/N load relation

for those systems where N load was selected as a

significant independent variable in the regressions. For

8 water bodies where this was not the case, the

average type-specific slope was used.

Estimate the slope of the Kd/N load relationship, and

substitute with average type-specific slopes where no

significant relations could be found (6 water bodies).

Estimate 1900 reference Chlorophyll a levels, using

1900 N loads and the slopes as input.

Estimate 1900 reference Chlorophyll a levels, using a

1900 scenario with adjusted nutrient inputs (N, P),

adjusted benthic stocks etc.

Use the same historic data on Kd as 1900 reference

as the statistical modelling.

Mechanistic modelling

Do not estimate 1900 Kd from the models; use historic

observations instead. Where no direct observations

were available, use observations from nearby similar

water bodies.

Estimate Chlorophyll a reference levels per water body type by averaging the reference levels coming from the

statistical and the mechanistic models of all water bodies in the type. Notes: For type 1, some subtypes are

defined; a few systems have a

status aparte.

The slopes are not averaged, but kept per water body and model type.

The same procedure is NOT followed for Kd. Historic references are used in both approaches.

Estimate the required N load reduction to reach the

target values for Chlorophyll a, Kd, hypoxia, anoxia,

days with N limitation. Where logical inconsistencies

may exist (reductions >100%), use a look-up table to

substitute calculations. Unclear how this is done if

Estimate the required N load reduction to reach the

target values for Chlorophyll a and Kd, taking into

account the fraction due to Danish land-based

sources. Based on scenarios with varying degree of

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

>100% is needed for Chlorophyll a.

Calculate the required load reduction as a weighted

average of the results in previous step.

overall reductions of N input.

Calculate the required load reduction as a simple

average of the results in previous step.

Smooth the variability in required N load reduction by

regional averaging. Unclear what was the basis for

delineating the regions.

Meta-model systems without a model.

IF status information is available, use type-averaged

slopes for N Chla, N Kd and N other indicators (latter

only if their status is known).

Calculate weighted average required N reduction.

ELSE use type-averaged required reduction.

Meta-model systems without a model.

IF status information is available, use type-averaged

slopes for N Chla, N Kd. Calculate average required N

reduction.

ELSE use regionally averaged required reduction.

Average required N reduction for meta-modelled systems across statistical and mechanistic approaches.

IF mechanistic model exists for system: Drop information from statistical model and only use mechanistic model

result.

ELSE use statistical model result.

Apply upstream-downstream rules.

All done!

8.2 Averaging and “ensemble modelling” aspects in the

procedure

In this procedure, both modelling approaches are largely independent and focused on

individual water bodies. However, four critical averaging steps intervene:

•

The Chlorophyll a targets are averaged per type, over the statistical and

mechanistic models. As far as the Panel understands it, this does not apply to

the slopes, which remain water body-specific. In the Scientific Documentation

Report, the averaging is justified as a means of reducing variability. As a

consequence of the averaging, there is a possibility of incompatibility between

slopes and targets: More than 100% of N load should be reduced. It remains

unclear how this is solved. Presumably the reductions >100% enter the

weighted average over indicators as such and are corrected by the averaging

over indicators.

The required load reduction for the different indicators (most importantly

Chlorophyll a and Kd) is averaged quite early in the procedure. This averaging

violates the “one-out-all-out” principle, as clearly stated by the researchers, but

is justified in the Scientific Documentation Report based on arguments of

reducing random variation in the required load reductions.

The mechanistic model results, in terms of required percentage load reduction,

are spatially averaged before the final meta-modelling step is applied to

systems without monitoring data. This step obscures differences between water

systems in regions. In the Scientific Documentation Report, it is very shortly

discussed, and justified based on observed variability between systems in the

regions that is – without proper argumentation – attributed to variability in the

•

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

data on system status. It is not clear to the Panel why this would be the

justification for a relatively drastic step in reducing the spatial differentiation in

the model results. The Panel feels that this is an important yet poorly

argumented step in the entire procedure.

•

For meta-modelled systems, statistical and mechanistic meta-model results are

averaged. This is not the case, however, for individually modelled systems

where the mechanistic model has prevalence in the cases where both models

are available.

In the opinion of the Panel, the most problematic aspect of the procedure is the

averaging of Chlorophyll a reference (and GM boundary target) values across model

types and within water body types. As the types are relatively broad and contain water

bodies with quite dissimilar characteristics, this loss of local detail can easily lead to

situations where too much effort is spent in one system and too little in another. In that

case, there will be both economic and ecological loss of efficiency. In this respect, it is

informative to compare the targets used for Kd and for Chlorophyll a across all systems,

as is shown in Figure 4.

Figure 4. Comparison of target values for Kd and Chlorophyll a across all water

bodies. Target values are shown as orange crosses. For comparison, the average

status values 1990-2012 are shown as blue diamonds. Regression lines of the two

sets are remarkably similar.

The figure shows that the target values of Kd, based on historic observations, are quite

variable within each of the water body types and largely overlapping between types.

There is much more discrimination between water bodies based on Kd than based on

the (uniformed) Chlorophyll a targets, implying that the latter are not optimised for the

water bodies.

The averaging of the required load reductions across the indicators (primarily Chlorophyll

a and Kd) early on in the procedure renders it impossible to judge whether, and how

much, the final results in terms of MAI depend on this violation of the one-out-all-out

principle. If, for reasons of compliance to the WFD procedures, it would be decided that

this is unacceptable, the present results cannot be used to make the recalculation. Also,

no elements are offered to evaluate the importance of this decision.

The adoption of common Chlorophyll a target values across the statistical and

mechanistic models also leads to a loss of independence of the two modelling

approaches. As a consequence, comparison of the required N load reduction obtained

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

by the two different methods, as a check on the methodology, is not really possible

anymore. The Panel is of the opinion that it would be better to keep both methods

separated up to the last stage and then do an in-depth comparison, taking into account

water body characteristics to explain or understand any discrepancies. Averaging two

independent model results as “ensemble modelling” is an option that can be taken in

order to reduce variation in results, but it does not necessarily lead to a better solution. If

one of the methods is biased (e.g. is clearly unable to make reliable estimates in

particular types of systems), averaging is a worse solution than dropping the bad

prediction. The availability of extensive databases on almost all systems should allow the

model comparison to be evidence-based (e.g. model comparisons with data can be

made for systems where predictions differ substantially), so that well-justified choices

can be made.

Very little justification is given for the choice to give prevalence to the mechanistic

models where both models are available. Even if the choice could be well justified (which

is questionable since an independent comparison is impossible), it contrasts with the

meta-modelling approach where both are averaged. Consistency in the choice would

improve the overall approach.

In the opinion of the Panel, the spatial averaging of the mechanistic model results is not

necessary nor justified. There can be good reasons, from a management point of view,

to smoothen required load reductions regionally so that no abrupt changes in

requirements occur at too small scales where control would be virtually impossible.

However, such decisions could better be made on the basis of a map showing the

original model results, allowing one to judge whether a management problem is posed or

not. As it is performed now, it remains unclear to what degree the spatial averaging leads

to under- or overdoing in particular watersheds, leading to lack of transparency in the

management rules.

Summarising, the Panel recommends postponing the averaging operations to the very

last stages of the procedure. This will keep the two modelling approaches independent, it

will allow estimating the consequences of violating the one-out-all-out principle and avoid

confusion due to regional averaging. As a consequence, the effects of different

modelling strategies and different indicators will remain clear in the different results on

nutrient reduction. A close examination and comparison of these differences will allow

making informed decisions on the choice of strategy.

8.3 Conceptual differences between modelling approaches

There are a few points where the statistical and mechanistic modelling approaches are

not conceptually consistent. The most important point is that the statistical modelling

takes into account additional indicators, while the mechanistic models (although capable

of doing the same, as all relevant variables are calculated in the model) only focus on

Chlorophyll a and Kd. Even if this would not lead to large differences (this cannot be

controlled based on the report), the Panel feels that it leads to a different treatment of

water bodies, depending on the model applied to it, that is difficult to justify. Based on

the observation that the ancillary indicators are strongly correlated with Chlorophyll a and

can hardly be justified as probing independent characteristics of the ecosystem (see

Chapter 4), the Panel suggests dropping these ancillary indicators from the procedure. It

will make the two modelling approaches more comparable without apparent loss of

information on the ecosystem. The ancillary variables could better be used as

corroborating evidence for the need to take action (or not) in the water bodies concerned

and as documentation of the range of ecological results to be expected from sanitation of

the nutrient input.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

A second potential inconsistency between the two models is that the mechanistic

modelling explicitly separates how much of the distance to target can be reached by

reducing Danish land-based N loading alone, while such separation is not done for the

statistical model. However, the latter bases its regression approach on Danish (in fact:

local) land-based N loads, so that, implicitly, the reasoning is probably more similar than

it may appear when written out. Overall, the consequences of this difference are

extremely difficult to trace, but intuitively the Panel does not estimate this to be a major

conceptual difficulty. It may, however, have consequences in practice. A comparison of

independent model results of the two approaches would also inform better about this

aspect.

8.4 Meta-modelling

With respect to meta-modelling, the Panel remarks that the coarseness of the typology

also has potential impact on the meta-modelling. It is clear that for the meta-modelling,

some knowledge on the (un-sampled or under-sampled) systems must be used in order

to set the best targets and use the appropriate slopes. As the typology used is rather

coarse, the current choices may not be optimal for these systems. The Panel sees a role

for statistical modelling here, provided that the statistical modelling would also focus on

understanding and modelling cross-system differences (in slopes and consequently in

targets) as a function of hydrographic and morphological characteristics of the systems.

Particularly the importance of freshwater influence in the systems and the flushing rates

may be overarching determining characteristics. The Panel estimates that a regression-

based approach could be better than a classification approach.

In the opinion of the Panel, the meta-modelling of the North Sea water bodies is less

reliable than that of the other water bodies. For the North Sea coastal systems, the

background modelling, which has focused on the Baltic systems instead, is not very

strong, and the meta-modelling is based on daring extrapolations from systems with

quite different ecological characteristics. The Panel recommends that more study is

made of the North Sea estuaries in order to improve the estimates of the nutrient load

reduction requirements, based on their very different physical and ecological

characteristics as well as on the very different basis (OSPAR intercalibration) for the

references and targets.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

9. Evaluation of Maximum Allowable Inputs

results

Maximum Allowable Inputs (MAI) define the annual load of nutrients, in this context of

nitrogen, that are acceptable to keep a coastal water body in a Good Ecological Status

according to the WFD or allow a water body to return to this status. Since nutrient load

management is a complex task and nutrient load reductions are associated with high

costs, reliable overall and water body-specific MAI are of outstanding importance. In this

chapter, the Panel reflects to what extent the suggested MAI can be regarded as reliable

enough to form the basis for policy and management actions.

9.1 The overall Danish MAI in an international framework

The nutrient reduction scheme of the HELCOM Baltic Sea Action Plan was revised in the

2013 HELCOM Ministerial Meeting, based on a new and more complete dataset as well

as an improved modelling approach. The new MAI, compared to the reference inputs of

1997-2003 for the Baltic Sea sub-basins Kattegat and Danish straits, demands only a

minor load reduction requirement of about 3%. In this revision, Denmark agreed to

reduce N loads to the Baltic Sea (from both land and air) by 2,890 t/a and P loads by 38

t/a. The Scientific Documentation Report suggests low N load reductions (>10%) for

Western Jutland and most parts of Zealand as well as Lolland and Falster (Figure 8.23,

p. 127). This seems reasonable and is well in agreement with international requirements.

However, to meet the targets for a good status, the Scientific Documentation Report

demands much higher load reduction, especially on Funen and Jutland. Here, Denmark

faces a situation similar to the Baltic coastal waters of Germany. Especially the inner

coastal waters, estuaries and bays in Germany require higher N load reduction than

demanded in the HELCOM Baltic Sea Action Plan to reach the GES. According to the

German plans, the N load from German Baltic river basins has to be reduced by 21,500

tTN/a, with an average maximum allowed total N concentration in rivers of 2.5 mg/l,

resulting in an overall reduction of 34%.

For Denmark, depending on the model approach, an average overall reduction between

29% and 34% is suggested. There are many similarities with respect to geomorphology,

land use pattern and intensity as well as population and state of sewage purification

between the German and Danish Baltic catchments, and the coastal waters share many

similarities too. Therefore, the very good agreement in the assumed relative reduction

requirements between both countries indicates that the values meet the right order of

magnitude and seem reasonable.

However, the reliability of water body-specific MAI depends on the approach for

calculating reference conditions and subsequent target conditions, the typology and

type-specific targets, the considered indicators, the applied weighting, the model and

meta-model approach as well as the data processing and aggregation. The major

question is if all these aspects are sufficiently taken into account and if the application

has a sufficient quality to determine reliable water body-specific MAI and mitigation

needs to achieve the GES in Danish coastal waters.

9.2 Historic conditions as basis for target setting

The process in the Scientific Documentation Report follows the implementation

guidelines of the WFD. It means that it is based on historic reference conditions and

assumes that these conditions can serve as a basis for the definition of present and

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

future targets. The reference conditions describe the status of biological quality elements

that would exist in a situation with no or very minor disturbance from human activities.

Reference conditions are therefore not pristine conditions. The WFD allows different

methods to calculate reference conditions. In countries with long monitoring data records

and the availability of suitable models, historic conditions are usually used as reference

state. Because of data availability, this period often refers to a period around 1900, being

aware that this period not always reflects a state with very minor disturbance from

human activities. Similar to Germany, the Scientific Documentation Report uses the

years around 1900 as reference. The Panel finds this approach well justified and the

data basis sufficient and suitable.

However, it is obvious that between 1900 and today, land-use pattern and population

densities have changed and different regions in Denmark developed differently until

today. Further, the year 1900 is well suitable to reflect a high ecological status in rural

areas, while cities already at that time emitted significant amounts of untreated sewage

and caused pollution in their surroundings beyond the thresholds for a high ecological

status.

For the definition of reliable targets, the question is less how did it look like in 1900, but

rather how would reference conditions in a region look like, assuming present land-use

and population pattern. This means that targets and water body-specific MAI based on

historic conditions around 1900 bear uncertainties and for some water bodies may

require a deeper analysis. This is especially true for areas with known strong changes

between 1900 and today. However, the Panel agrees that this approach is the best

choice that still ensures full compliance with technical WFD implementation guidelines.

In Germany, the official national working group on targets and MAI discussed if reference

conditions should be calculated for and translated into the present situation. The

approach was to use combined river basin and marine models and present population

density and land-use pattern as well as the historic specific emissions per hectare and

capita to calculate resulting regionalised Chlorophyll a and nutrients concentrations. The

idea was to use the values as reference conditions to account for the fact that different

regions developed differently during the last 120 years and to be able to provide even

more reliable water body-specific targets. However, the majority of the working group

declined this approach for containing too many assumptions and for not fully following

the technical WFD implementation guidelines. Denmark would face similar problems with

this alternative approach.

9.3 Effects of climate change on targets and MAI

Climate change shows its effects only gradually on a time horizon of decades, while the

implementation of the WFD and measures to reach GES must take place within a

decade. Further, depending on the emission scenario, climate change effects on

countries and on regions within a country are uncertain, they show a large variability and

are hard to predict. The Scientific Documentation Report addresses this topic and, in our

opinion, provides sufficient evidence and reasons why climate change has not been

taken into account in the definition of targets and in calculating MAI in Denmark.

However, several nutrient load reduction measures in river basins show the full effect

only after decades. Major effects of climate change on Danish coastal waters, very likely,

will result from changed nutrient loads as a result of altered spatial and seasonal

precipitation and discharge patterns. Therefore, linked river basin – coastal water – sea

models used for the assessment of the effectiveness of measures in the river basin

should take into account climate change effects on river basin loads and shifts between

nutrients. However, climate change can affect internal processes in coastal waters as

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

well. Riemann et al (2016), for example, point out that more frequent stratification and

higher water temperatures presumably hampered the improvement of bottom water

oxygen conditions and counteracted the expected positive effects of reduced nutrient

inputs in Denmark.

9.4 Relevance of typology on MAI

As indicated in Chapter 3, the Panel has the opinion that the Danish typology used in the

Scientific Documentation Report does not sufficiently reflect the individual properties of

the many Danish fjords and inner coastal waters. This is also true for the typology

reported in Dahl et al (2005). Type-specific targets for the indicators, especially

Chlorophyll a, that are applied to a wide range of significantly different water bodies do

not sufficiently reflect their properties and behaviour to loads reductions. Consequences

are less reliable water body-specific MAI. This may cause an underestimation of the

required load reduction for some water bodies and an overestimation for others.

9.5 Relevance of indicator choice on MAI

The Panel agrees that Chlorophyll a is a core indicator, and coastal water body-specific

Chlorophyll a concentrations are a sound basis for calculating water body-specific MAI.

Further, the Panel agrees that water transparency has to be restored as one necessary

condition to enable the recovery of eelgrass in coastal waters. Potentially, Kd can serve

as an indicator for describing suitable growing conditions for eelgrass. Eelgrass can

serve to indicate the status of macrophytes, a biological element in the WFD. Therefore,

Kd has the potential to be an important parameter for calculating MAI.

However, as pointed out in Chapter 4, the relationship between Kd in coastal waters and

external nutrient loading is sometimes very weak. Further, Kd and the insufficient

relationship have different consequences for and are differently treated in the

mechanistic and the statistical modelling exercises. In the statistical modelling approach,

for example, the use of Kd in some cases causes impossible N load reduction

requirements of above 100%. Further, Kd shows only a slow response to load reduction,

the data are subject to high variability, and it shows a correlation to Chlorophyll a.

Altogether, we consider Kd as a less suitable indicator in many Danish coastal water

bodies. A strong weight of Kd in the calculation of MAI should be avoided and would add

uncertainty to water body-specific MAI. Chapter 4 outlines possible solutions to

overcome or at least to deal with some of these problems. In the Scientific

Documentation Report, other indicators are sometimes mentioned and used in the

statistical model. We do not see a major advantage of these indicators for the calculation

of MAI, because they do not provide significant new information or show correlations to

the exiting indicators.

9.6 Relevance of model quality and approach for MAI

In general, the mechanic model has a very good potential for calculating water body-

specific MAI, but in the present state it does not cover all water bodies. The statistical

modelling is based on real monitoring data, and in most coastal water bodies it can serve

as a valuable tool to assess long-term trends as well as the mechanistic model

performance. As indicated in Chapter 8, the model application and the process of

calculating water body-specific MAI are complex and not entirely convincing. Most

problematic is the averaging of Chlorophyll a reference values across both models and

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

within coastal water types. This has negative consequences for the meta-modelled water

bodies as well.

9.7 Conclusion and perspectives

Many of these aspects and shortcomings were mentioned and pointed out by several

stakeholders as well. The Panel picked up the stakeholder comments and examined in

some detail the MAI for specific areas with very high nutrient load reduction demands.

Altogether, the Panel largely shares the stakeholder concerns.

The calculation of water body-specific MAI is a challenging task, but potentially has one

major advantage: It allows the development of water body-specific management options

and solutions. For this purpose, the coastal water and sea models should be combined

with river basin models providing information about the quantitative potential and

efficiency of single (or sets of) measures and providing load reduction scenarios for

coastal models. If river basin models are able to provide nutrient load data on a monthly

basis, this would allow the development of scenarios that take into account the

seasonality of emissions. Assessing how seasonally differentiated emissions affect the

status of coastal water bodies could lead to optimised, cost-effective management.

Taking into account all aspects and associated problems, the Panel has the impression

that the water body-specific MAI are not sufficiently reliable to serve as a basis for

decision-making and planning of load reduction measures. Further, the MAI are only

addressing nitrogen load reductions and leaving out the possibility of potentially

managing water bodies via phosphorus load reduction. However, models, competences

and data are available in Denmark to meet the challenge to calculate water body-specific

MAI. Even a modified processing of the existing model results might lead to much more

reliable MAI.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

10. Overall assessment and conclusions

The Water Framework Directive aims at restoring Good Ecological Status in surface

waters in Europe. The Scientific Documentation Report proposes measures of nutrient

load reduction to reach this Good Ecological Status in Danish transitional and coastal

waters. The Panel fully endorses the importance attached to nutrient reductions as a

necessary requirement to reach this Good Ecological Status and stresses the

importance of nutrient conditions as a modulating factor interacting with any additional

measures taken to improve the state of the system.

In comparison with many other European countries, Denmark has excellent databases,

models and scientific expertise as a basis for the implementation of the Water

Framework Directive. The Panel was delighted to see that these resources have been

mobilised to achieve a leading position at the European scale. The Panel was impressed

by the openness and transparency of the interaction between government, researchers

and stakeholders as well as by the high intellectual level of the discussions. This open

exchange of ideas and opinions is a perfect basis for a further improvement of the

scientific basis for the WFD implementation.

The Panel has reviewed the choice of indicators and procedures, in the context of the

WFD requirements and specifications, and found that the indicators, the methods to

determine reference conditions and the methods to determine required actions were

WFD compliant. The Danish implementation is based on either direct historical

observation or model determination of reference conditions. Little or no uncontrollable

“expert judgement” is involved. In that respect, the Danish models are attaining the

highest possible standard of WFD implementation.

The Panel has analysed the consequences of using a relatively coarse typology of

coastal waters for calculating reference conditions, targets and Maximum Allowable

Inputs of nitrogen. The Panel concludes that the use of a coarse typology has led to

reduction requirements that are not optimal for each of the individual water bodies. The

Panel is convinced that the full use of available data and models would allow Denmark to

forego the typology and develop advanced, specific reduction targets for each water

body. The Panel recommends focusing on the water body scale of resolution throughout

the scientific process. The regional grouping of reduction measures should be decided

upon only at the stage of translating scientific advice into management action plans.

The Panel has analysed the indicators used and concluded that Chlorophyll a is a useful

intercalibrated indicator of phytoplankton, while Kd is less optimal as an indicator of

benthic angiosperms and macrophytes. The other indicators, used in the statistical

modelling only, currently present methodological problems and are not yet mature

enough for inclusion in the management plans. The Panel has identified promising

developments in the modelling with respect to angiosperm and macrophyte indicators

and made recommendations on how to extend and develop the indicator set in the

future.

In view of the large efforts in the past to remove P load from point sources, the Panel

endorses the emphasis placed in the Scientific Documentation Report on reducing N

loads from diffuse sources. However, at least in principle, there could be an additional

role for P load reduction and for seasonal regulation of the N load. The Panel is of the

opinion that these options merit further scientific exploration, especially in watersheds

where high efforts for N load reduction are required.

Although the maintenance of two parallel modelling lines (statistical and mechanistic)

may seem redundant at first sight, the Panel strongly endorses maintaining these lines.

Given the wealth of data available, it provides unique possibilities for evidence-based

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

checking of mechanistic model results. The Panel assesses the mechanistic model as a

state-of-the-art, very comprehensive tool, but emphasises that independent checking on

data as well as uncertainty analysis remain necessary and can be performed by the

statistical approach. This coherence can be optimised by improving the approach and

methods of the statistical modelling.

The Panel endorses the general logic of the methodology to derive reference and target

values from the models and to calculate the required N load reduction to reach the

targets. The Panel has identified several points in the workflow where averaging is

performed. This results in interdependence of model types, loss of indicator resolution

and loss of spatial resolution. It also adds complexity to the procedure and makes it very

difficult to understand. None of these losses are necessary since the model results and

database do permit a fully transparent derivation of water body-specific required nutrient

reduction.

Summing up these different aspects of the work, the Panel positively evaluates that

nutrient load reductions are based on

solid

scientific evidence and generally high-level

modelling approaches. The Panel is very positive about the near lack of expert judgment

in the work and is of the opinion that in the few places where it does occur, it is not

necessary and can be removed. The general (country-averaged) level of required

nutrient load reduction compares favourably with independent efforts in similar areas and

seems a

robust

measure of what is needed. At the same time, the Panel assesses the

spatial resolution of the required efforts as

unnecessarily coarse.

The Panel is

convinced that the rich database, combined with an

improved statistical approach

and

the high-resolution mechanistic modelling tools, are able to derive improved, water body-

specific MAI values. Current scientific insight endorses the view that the overall

reductions proposed are

necessary,

but cannot guarantee that they will be

sufficient.

Especially for benthic angiosperms and macrophytes, additional measures may be

needed.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

11. Recommendations for going further

Monitoring:

The Danish national monitoring programme used in the Scientific

Documentation Report includes more than 90 stations along the coast and in the sea. It

is very comprehensive and is generally well adjusted to the WFD requirements. It forms

the basis for the further development of models, for most calculations and is required to

evaluate the success of measures and whether the targets of the WFD are met. The

Panel recommends maintaining this monitoring system at full strength and assessing if

additional monitoring stations will be required for a water body-specific management.

Typology:

The typology has weaknesses in reflecting the individual properties of fjordic

water bodies. Instead of suggesting a refinement of the existing typology, we

recommend calculating reference conditions and targets for each of the 119 water

bodies in Denmark. Denmark is one of the few countries in Europe, where the necessary

data, expertise and models are available for such a comprehensive approach. By taking

specific conditions and individuality of every water body into account, the calculated

targets and water body-specific Maximum Allowable Inputs will be optimised and lead to

minimal waste of resources. For purposes of intercalibration, a robust typology can be

based on the results of the water body-specific analyses.

Choice of indicators:

Chlorophyll a is a generally accepted and intercalibrated indicator

of phytoplankton. Kd, as a measure for macrophytes and angiosperms, has certain

limitations. The Panel recommends building on recent efforts towards comprehensive

modelling of eelgrass in order to derive a better indicator of macrophytes, but to keep Kd

as a proxy meanwhile. The other indicators used in the statistical modelling address

important ecological questions, but are not mature in the sense that they lack a clear

quantitative relation with nutrient loading. The Panel recommends leaving them out of

the present modelling and developing targeted modelling directed at their incorporation

into the indicator system.

Statistical modelling:

The Panel sees great merit in the strategy to maintain two

independent lines of modelling, one based on statistical data analysis and the other

based on mechanistic modelling. The Panel recommends reorienting the statistical

modelling towards optimal estimation of the long-term slopes of the indicators on nutrient

loading in a cross-systems analysis way and keeping in principle both N and P loading

as explanatory variables. The Panel recommends elaborating the uncertainty analysis in

the statistical modelling and suggests that this will be facilitated when a single cross-

system advanced modelling approach is chosen.

Mechanistic models:

The mechanistic models are state-of-the-art, both in terms of

numerical technique and included processes. They are powerful tools for providing a

sound scientific basis for the implementation of the WFD in Denmark. A shortcoming is

that they do not cover all water bodies. As a consequence, different approaches were

used for the definition of reference conditions, targets and MAI in different water bodies.

We recommend extending a mechanistic modelling approach to as many water bodies

as possible to ensure that, in future, a uniform methodology can be used for the

definition of water body-specific MAI.

Methods to derive targets and MAI from the models:

The Panel recommends

simplifying the calculation procedure by removing the averaging steps between models,

between indicators, between water bodies within types and between water bodies on a

regional basis. In this way, the differences and correspondences between modelling

approaches, indicators and water bodies will become clear and can be further analysed.

Cross-checking of results of the statistical and mechanistic model approaches in

systems, where both are available, will form a basis for extrapolation to all systems. The

Panel recommends deriving one MAI per water body in this way and only deciding in a

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

later phase on regional averaging or lumping, when scientific results are translated into

management actions.

River basin interactions:

River basin models allow calculating the load reduction

potential of nitrogen and phosphorus for each river basin, the development of water

body-specific nitrogen and phosphorus load reduction scenarios and cost estimates.

Further, they allow addressing seasonal load and limitation patterns. The Panel

recommends a combination of river basin and coastal water models to enable the

development of water body-specific optimised management concepts that consider both

nitrogen and phosphorus.

International approach:

The technical WFD implementation guidelines force similar

approaches in all member states. As a consequence, requirements, modelling and

challenges are similar in different countries. Further, the WFD asks for an intercalibration

and harmonisation of targets with neighbouring countries. Therefore, the Panel

recommends a co-ordinated joint scientific approach, especially between Denmark,

Germany and Sweden.

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

12. List of references

Andersen, J.H., Harvey, T., Kallenbach, E., Murray, C., Al-Hamdani, Z., Stock, A. (2017).

Under the Surface. Report L.NR. 7128-2017 DK6 by NIVA Denmark.

Birk, S., Willby, N.J. , Kelly, M.G., Bonne, W., Borja, A., Poikane, S. , van de Bund, W.

(2013). Intercalibrating Classifications of Ecological Status: Europe's Quest for

Common Management Objectives for Aquatic Ecosystems. Sci Total Environ

454-455, 490-499.

Borja, A., Prins, T.C., Simboura, N., Andersen, J.H., Berg, T., Marques, J.C., Neto, J.M.,

Papadopoulou, N., Reker, J., Teixeira, H. and Uusitalo, L., 2014. Tales from a

thousand and one ways to integrate marine ecosystem components when

assessing the environmental status. Frontiers in Marine Science, 1, p.72.

Canal-Vergés P, Petersen JK, Rasmussen EK, Erichsen AC, Flindt MR (2016) Validating

GIS tool to assess eelgrass potential recovery in the Limfjorden (Denmark).

Ecological Modelling 338:135-148.

http://dx.doi.org/10.1016/j.ecolmodel.2016.04.023.

Christensen et al. (subm.)

Referred to in the Scientific Documentation Report, but not

found back

Cloern, J.E. (2001). Our evolving conceptual model of the coastal eutrophication

problem. Marine Ecology Progress Series, 210, 223-253.

Dahl K (red.), Andersen JH (red.), Riemann B (red.), Carstensen J, Christiansen

T, Krause-Jensen D, Josefson AB, Larsen MM, Petersen JK, Rasmussen

MB, Stand J (2005) Redskaber til vurdering af miljø- og naturkvalitet i de

danske farvande. Typeinddeling, udvalgte indikatorer og eksempler på

klassifikation. Danmarks Miljøundersøgelser. 158 s. - Faglig rapport fra DMU

nr. 535.

Droop, M.R. (1968). Vitamin B12 and marine ecology IV: the kinetics of uptake, growth

and inhibition in Monochrysis lutheri. J. Mar Biol. Ass UK, 48, 689-733.

Eilola, K., Gustafsson, B. G., Kuznetsov, I., Meier, H. M., Neumann, T., & Savchuk, O. P.

(2011). Evaluation of biogeochemical cycles in an ensemble of three state-of-the-

art numerical models of the Baltic Sea. Journal of Marine Systems, 88, 267-284.

Flindt, M. R., Rasmussen, E. K., Valdemarsen, T., Erichsen, A., Kaas, H., & Canal-

Vergés, P. (2016). Using a GIS-tool to evaluate potential eelgrass

reestablishment in estuaries. Ecological Modelling, 338, 122-134.

http://doi.org/10.1016/j.ecolmodel.2016.07.005

HELCOM. (2013). Summary report on the development of revised Maximum Allowable

Inputs (MAI) and updated Country Allocated Reduction Targets (CART) of the

Baltic Sea Action Plan. Background Document for the 2013 HELCOM Ministerial

Meeting, 1-23.

Kuusemäe, K., Rasmussen, E. K., Canal-Vergés, P., & Flindt, M. R. (2016). Modelling

stressors on the eelgrass recovery process in two Danish estuaries. Ecological

Modelling, 333, 11-42.

http://doi.org/10.1016/j.ecolmodel.2016.04.008

Malve, O., Qian, S. 2006. Estimating nutrients and chlorophyll a relationships in Finnish

Lakes. Environmental Science & Technology, 40(24),pp. 7848-7853. DOI:

10.1021/es061359b.

Markager S, Storm L, Stedmon C (2006) Limfjordens miljøtilstand 1985 til 2003.

Sammenhæng mellem næringsstoftilførsler, klima og hydrografi belyst ved hjælp

MOF, Alm.del - 2017-18 - Bilag 35: Rapporten fra det internationale ekspertpanel om evaluering af de danske marine modeller

af emperiske modeller. Danmarks Miljøundersøgelser. 219 s. - Faglig rapport

fra DMU, nr. 577.

Orth RJ, Moore KA, Marion SR, Wilcox DJ, Parrish DB (2012) Seed addition facilitates

eelgrass recovery in a coastal bay system. Marine Ecology Progress Series

448:177-195.

Rasmussen, B., & Josefson, A. B. (2002). Consistent Estimates for the Residence Time

of Micro-tidal Estuaries.

Estuarine, Coastal and Shelf Science, 54(1),

65-73.

http://doi.org/10.1006/ecss.2001.0836

Riemann, B., Carstensen, J., Dahl, K., Fossing, H., Hansen, J. W., Jakobsen, H. H., et al

(2016). Recovery of Danish Coastal Ecosystems After Reductions in Nutrient

Loading: A Holistic Ecosystem Approach.

Estuaries and Coasts, 39(1),

82-97.

http://doi.org/10.1007/s12237-015-9980-0

Scavia D, Rabalais NN, Turner RE, Justi

D, Wiseman WJ (2003) Predicting the

response of Gulf of Mexico hypoxia to variations in Mississippi River

nitrogen load. Limnology and Oceanography 48:951-956.

doi:10.4319/lo.2003.48.3.0951

Schernewski, G., Friedland, R., Carstens, M., Hirt, U., Leujak, W., Nausch, G.,

Neumann, T., Petenati, T., Sagert, S., Wasmund, N., von Weber, M. (2015).

Implementation of European marine policy: New water quality targets for German

Baltic waters.

Marine Policy,

51, 305-321.

https://doi.org/10.1016/j.marpol.2014.09.002.

Timmermann, K., Dinesen, G. E., Markager, S., Ravn-Jonsen, L., Bassompierre, M.,

Roth, E., & Støttrup, J. G. (2014). Development and Use of a Bioeconomic Model

for Management of Mussel Fisheries under Different Nutrient Regimes in the

Temperate Estuary of the Limfjord, Denmark.

Ecology and Society, 19(1),

art 14-

11.

http://doi.org/10.5751/ES-06041-190114

van der Heide T, van Nes E., van Katwijk MM, Olff H, Smolders AJ (2011) Positive

Feedbacks in Seagrass Ecosystems – Evidence from Large-Scale Empirical

Data. PLoS ONE 6(1): e16504.