Marine Board – ESFESF Marine Board Position Paper 7

Modelling in Coastal and ShelfSeas – European Challenges

MarineBoard

EUROPEAN SCIENCE FOUNDATION

EUROPEAN SCIENCE FOUNDATION

The European Science Foundation (ESF) acts as a catalyst for the development ofscience by bringing together leading scientists and funding agencies to debate,

plan and implement pan-European scientific and science policy initiatives. It isalso responsible for the management of COST (European Cooperation in the fieldof Scientific and Technical Research).ESF is the European association of 77 major national funding agencies devoted toscientific research in 30 countries. It represents all scientific disciplines: physicaland engineering sciences, life, earth and environmental sciences, medicalsciences, humanities and social sciences. The Foundation assists its MemberOrganisations in two main ways. It brings scientists together in its ScientificForward Looks, Exploratory Workshops, Programmes, Networks, EUROCORES,and ESF Research Conferences, to work on topics of common concern includingResearch Infrastructures. It also conducts the joint studies of issues of strategicimportance in European science policy and manages, on behalf of its MemberOrganisations, grant schemes, such as EURYI (European Young InvestigatorAwards).It maintains close relations with other scientific institutions within and outsideEurope. By its activities, the ESF adds value by cooperation and coordinationacross national frontiers and endeavours, offers expert scientific advice onstrategic issues, and provides the European forum for science.

Marine Board – ESFThe Marine Board operating within ESF is a non-governmental body created inOctober 1995. Its institutional membership is composed of organisations whichare major national marine scientific institutes and funding organisations withintheir country in Europe. The ESF Marine Board was formed in order to improveco-ordination between European marine science organisations and to developstrategies for marine science in Europe.Presently, with its membership of 25 marine research organisations from 17 European countries, the Marine Board has the appropriate representation tobe a unique forum for marine science in Europe and world-wide.In developing its activities, the Marine Board is addressing four main objectives:creating a forum for its member organisations; identifying scientific strategicissues; providing a voice for European marine science; and promoting synergyamong national programmes and research facilities.

Cover picture: Meteosat – artificially coloured area Europe, 23rd July 1994, 11.55 UTC – received and processed at ESOC (Darmstadt) © ESA/EUMETSAT

Modelling in Coastal and Shelf Seas – European Challenges

EditorsDavid Prandle, Hans Los, Thomas Pohlmann, Yann-Hervé de Roeck, Tapani Stipa

ContributorsWolfgang Fennel, João Gomes Ferreira, Michael Hartnett, Peter Herman, Michiel Knaapen,Morten Pejrup, Roger Proctor, Karline Soetaert,Takvor Soukissian, Georg Umgiesser, Waldemar Walczowski,Niamh Connolly (Marine Board Secretariat)

The Marine Board of the European Science Foundation regularly establishesWorking Groups of experts to address marine science and technology topicswhich need to be elaborated on. These Working Groups facilitate scientists to gettogether, reinforce their relations, create new opportunities and establish commonapproaches on projects, while also heightening awareness and visibility. The expected output of such a Working Group is, in principle, a position paper to be used subsequently at national or European levels.

The issue of Hydrodynamic Modelling of Coastal and Shelf Seas was identified bythe Marine Board as a subject appropriate for the establishment of a WorkingGroup. This Working Goup, chaired by David Prandle, concentrated its analysison operational oceanography and the implications of this in terms of modellingand data assimilation.

The analysis by this Working Group does not cover the entire breadth of the subject; aspects such as mathematically innovative modelling, new types ofecosystems models, coupling of physical to fishery ecosystem models,the approach to open source models, quality standards and skill scores were notconsidered within the scope of this report. However, this report does illustrate the development effort needed to transform research tools into services for the many users of ocean space and resources.

The Marine Board thanks the Working Group for its work on a subject crucial to the future of coastal oceanography.

Jean-François Minster

Chairman of the Marine Board

Foreword

Contents

5 Executive Summary

6 Background and Introduction

7 End-user Requirements

7 Challenges

8 Scope and Development of Models

10 Operation of Models – Hardware and Software Requirements

11 Challenges

12 Data Requirements from Observations and Coupled Models

12 Observational data

13 Data from coupled models

14 Challenges

15 European Collaborations and Initiatives

15 Challenges

18 Appendices

18 1. Model Codes

21 2. Hydrodynamic Modelling

22 3. Ecological Modelling

25 4. Operational Oceanography

26 5. Terms of Reference and Membership

5

Executive Summary

Well-recognised requirements to supportthe diverse needs of the research

community include:• Access to supercomputers and ancillary

services for data management, visualisation,and analysis.

• Teams with adequate resources to bothdevelop existing modelling systems andintroduce new innovative technologies, withattendant programmes for visiting fellows,workshops, training, capacity building, etc.

• Long-term programmes to match the time-scales of technology development andinternational scientific programmesconcerned with Global Climate Change andholistic sustainability.

• Enhanced provision of and links to:(i) observational technologies and test-bedsites,(ii) permanent monitoring networks,(iii) meteorological and climate data(attendant assimilation from futuresatellites),(iv) data centres providing quality-controlledinformation for coastal seas, and(v) enhanced methodologies for dataassimilation, using the expertise withinmeteorological agencies.

Diverse applications of models range from nowcasting of waves,tides and storm surges to coupled ocean-atmosphere-sea-riverscenario forecasting of the effects of Global Climate Change onterrestrial, fluvial and marine ecology over millennia. The validityof models is limited by the degree to which the equation oralgorithms synthesise the governing processes. The accuracy ofmodel simulations depends further on the availability andsuitability (accuracy, resolution and duration) of both observationaland linked meteorological, oceanic and hydrological model datato set-up, force and assess calculations. Modelling is at a stagewhere major investments are required in infrastructure andorganisation: e.g. access to supercomputers, softwaremaintenance and data exchange. Europe needs to develop astrategic vision and translate this into internationally-competitivemodelling capabilities to address issues of both local and globalgovernance of the marine environment. A few major Europeanmarine modelling centres are likely to emerge in the next fiveyears, collaborating closely with existing meteorologicalinstitutions, with an associated network of centres addressinglocal applications.

6

Background and Introduction

The Marine Board recognised three primarydrivers underlying the development of

hydrodynamic models:1. Understanding and predicting impacts of,

and feedbacks from, ocean climate change.2. Establishing scientific and socio-economic

bases for sustainable development of shelfseas and their resources.

3. Advancing marine science and technology.

This development involves two discrete, butinter-related, pathways: (i) scenario testing (pre-operational); (ii) real-time forecasting(operational). Maintaining associated state-of-the-art capabilities in Europe is essential bothto underpin EU policies for marinegovernance of its coastal seas and to inform itsapproach to international issues such as GlobalClimate Change.

Coastal and marine resource managementrequires linking of science and decision making,using theory, models, and measurements of the physical, chemical, and biological marineenvironment. Models synthesise theory intoalgorithms and use observations to set-up,initialise, force, assimilate, and evaluatesimulations in hindcast, nowcast, and forecastmodes. Assimilation involves the combinationof information provided by observing networkswith the systematic temporal and spatialresolution of holistic knowledge incorporatedwithin numerical models. In operationalforecasting, assimilation involves structuredincorporation of near real-time observations toimprove nowcasts and forecasts. In non-operational modes, assimilation may be used incalibrating parameters (boundary conditions,surface roughness, etc.) to improve theaccuracy of simulations.

Over the past 40 years, numerical modellinghas developed rapidly in scope (fromhydrodynamics to ecology) and resolution(from one-dimensional, 102 elements to 3-D,108 elements) exploiting the contemporaneousdevelopment of computing power.Unfortunately, concurrent development inobservational capabilities has not matched thisresolution, despite exciting advances in areassuch as in remote sensing and sensortechnologies.

The following sections of the report seek toarticulate these capabilities and limitations,indicating past and present approaches adoptedin Europe. Associated challenges and futureoptions to sustain the science and technologyto meet the requirements of the end-user areidentified. In reviewing future strategies for thedevelopment of modelling, subsequent sectionsexamine sub-components of this system,namely: the requirements of the end-user;scope and development of modelling;operation of models; data requirements fromobservations and coupled models; andEuropean collaboration (Figure 1).

Objectives of the Marine Board – ESF include promoting thescience needed for effective management of coastal and marineresources. Related scientific and technical challenges aresummarised in the Marine Board publication Integrating MarineScience in Europe (Marine Board – ESF, 2002). The objective of theWorking Group on Hydrodynamic Modelling of Coastal and ShelfSeas, whose work is presented in this report, was to identifyinitiatives to foster scientific and technical excellence in themodelling of coastal and shelf seas.

OperationalModel

Ass

imila

tion

EndUserApplication

MonitoringNetworkSatelliteShips

Buoys etc.

Set-up DataBathymetry

Initial Conditions

Coupled ModelsAtmosphere

OceanCoast

Figure 1: Components of a modelling simulation system (EuroGOOS)

7

for local coastal models. Links betweenmeteorological and global operational modelsneed to be incorporated from the initial designphase. Awareness of evolving end-userrequirements by developers of these oceanmodels is essential.

To deliver the full range of benefits from ourmodels over the diverse scope of habitats inEurope, interfaces with socio-economic-politicalconcerns must be established. This may requiresimplification of our complex models, oraggregation of their results, in a form that can beaccommodated in related total-system models.

End-users, especially from developing countries,will be most interested in solving problemsdirectly related to their environment, primarily inthe coastal area. Capacity building must includetraining of skilled young scientists for the creation of know-how in modelling andforecasting, which will enable these countries tosolve problems locally. Two-way mobility mustbe encouraged.

Challenges

Introduction of the EU Water FrameworkDirective (WFD) emphasises the need fordevelopment of well-validated, reliable modelsfor simulating water quality-ecology-fisheries in European coastal waters. To enhance our understanding of the threat of Global ClimateChange, we need whole-system models toindicate related impacts both from and on coastalseas (including the impacts on marine biota andtheir potential biogeographic consequences). For both pre-operational and operational (pre-operational models – tested and validated codesused for real-life applications; operational models– routinely used for marine forecasting; seeAppendix 4) hazard forecasting, improvementsare required in accuracy, reliability, and resolutiontogether with extended warning periods; suchimprovements also provide enhanced designstatistics for coastal development.

Awide variety of modelling and monitoringapproaches exists, reflecting the diverse

range of interests and end-user concerns. Appliedinterests include surface ice in the Arctic,ecosystem dynamics for fish recruitment in theBay of Biscay, eutrophication in the Baltic Sea(coupled sea-hydrological model includingpredictions of river flows), and pollutanttransport in the Mediterranean. Appendices 2and 3 summarise the development and pertinentissues relating to hydrodynamic and ecologicalmodels. Alongside such applied interests, modelsare used to address many generic issues anddevelop scientific and technical capabilitiesincluding development of numerical algorithmsand validation procedures, optimal design ofmonitoring networks, and assimilationtechniques.

Models are widely used for management andpolicy strategies, such as assessment ofabsorptive capacity for licensing of discharges,evaluating environmental impacts of intervention(reclamation, dredging, etc.), and in both hindcastand forecast modes for climate change scenarios.

The coastal – marine area is generally the focusof end-user interests. Improved forecastingcapabilities for storm surges, sediment transport,and wave action are important to address userneeds in relation to flood protection, fisheries,coastal erosion, and prevention of pollution. An accurate description of the state of the offshoreocean is required to define boundary conditions

Models are used for:1. Improving weather forecasting, climate prediction, and to warn

of hazards, e.g. storm surges, oil or chemical spill movement,search and rescue, eutrophication, toxic algal blooms, and the consequences of future changes.

2. Assessing and understanding the current state of health ofmarine ecosystems and resources – their likely sensitivity tochanging conditions.

3. Developing environmental management policies which accountfor both anthropogenic influences and natural trends.

4. Advancing underpinning science and technology.

End-user Requirements

8

Models encompass: (i) non-dimensional conceptual modules

encapsulated into whole-system simulations,(ii) one dimensional (1-D), single point verticalprocess studies or cross- sectionally averagedrepresentations for rivers and estuaries, (iii) two dimensional (2-D), representations ofhorizontal circulation, and (iv) fully three dimensional (3-D).

In shelf seas, model applications haveprogressed from the 2-D barotropic models ofthe 1960s to the 3-D baroclinic (incorporatingtemperature and salinity induced densityvariations) of the 1970s. Initially, these 3-Dbaroclinic models used prescribed densityfields, calculating resultant circulations in adiagnostic mode. These were superseded inthe 1980s by prognostic models, whichcalculate evolving temperature and salinityfields. Such 3-D models are now used widelyfor applications from limnology, estuaries,

harbours, coastal bays to shelf seas andoceans.

Parameters of interest include tides, surges,waves, currents, temperature, salinity,turbidity, ice, sediment transport, and an ever-expanding range of biological and chemicalcomponents. Table1, extracted from theMarine Board publication Integrating MarineScience in Europe (2002), shows a comprehensive (but incomplete and notprioritorised) set of such parameters.

The scope of the models involves simulationsacross ocean-atmosphere-seas-coasts (Figure 2)and between physics-chemistry-biology-geology-hydrology extending over hours tocenturies and even millennia. This connectivityspans meteorological agencies, satellitemissions, international scientific, and surveyprogrammes (IGBP, CLIVAR, GOOS etc.)that also introduce specific coupling issues.

Coastal sea models are influencedimmediately and directly by meteorologicalforcing. Likewise, though generally lessimmediately and directly, they are impactedby conditions along the ocean-shelfboundaries. Hence, coastal modellers need tomaintain close links with developments in

Scope & Development of Models

The diverse applications of models range from short-termnowcasting of waves, tides and storm surges to coupled ocean-atmosphere-sea-river simulations of the effects of Global ClimateChange on terrestrial, fluvial, and marine ecology over millennia.Associated practitioners range from scientists and engineers tocoastal managers.

Table 1: Key parameters in the coastal area and shelf seas (Integrating Marine Science in Europe, Marine Board – ESF, 2002)

Physical Chemical Biological Geophysical

Temperature NO3, PO4, Si, NH4 Phytoplankton Seismicitybiomass (chlorophyll) and diversity

Salinity Trace metals, pH, Bacterial Bathymetryradionucleides and phytoplankton

cytometry

Density, pressure Dissolved gases Viral particles Gravimetry(O2, CO2, DMS)

Light, bioptics Volatile organic RNS, DNA, proteins Magnetismpollutants key enzymes

Turbidity, particle Pesticides Pelagic animals Acoustic signalssize distribution

Velocity, turbulence PCB, PAH, CFC Benthic communities Seafloor characteristics

9

ocean modelling and awareness of issuesproducing medium and longer-termvariability, such as the ThermohalineCirculation of the Atlantic, the North AtlanticOscillation, El Niño, etc. We anticipateextension of coupled ocean-atmospheremodels to incorporate sea-surface exchangefluxes within coastal-marine areas, with the latter utilising unstructured grids.

In the horizontal, rectangular grids are widelyused and are suitably adjusted to polarcoordinates (latitude and longitude) inregional seas. Irregular grids, generallytriangular or curvi-linear, are used for variableresolution. In Computational Fluid Dynamics,continuously adaptive grids provide a widespectrum of temporal and spatial resolution in multi-phase processes. This facility is nowused in ocean models to address localisedanomalies such as non-hydrostatic conditionsor eddy shedding. Immediate improvements in the accuracy of simulations can be achievedwith adaptable and flexible grids alongsidemore sophisticated numerical methods.The vertical resolution may be adjusted fordetailed descriptions near bed, near surface or

at the thermocline. For example, the sigmacoordinate system accommodates bottom-following with a uniform number of coordinatesurfaces occupying the water column.Understanding and enhanced representation ofturbulence effects in models is a central issuefor future marine studies. Development ofturbulence models is proceeding viainternational collaborations (see GOTM inAppendix 1). This work is supported by newmeasuring techniques like the microstructureprofiler, providing a direct comparison ofsimulated dissipation rates with in situmeasurements. Presently, efforts are focusedon applications of 1-D (vertical) models; there is still no clear consensus on the bestturbulence scheme to be implemented into 3-Dmodels. Resolution of horizontal turbulence isless advanced; values specified often relate tonumerical stability requirements or to observedvalues from dye dispersion experiments. In shallow coastal waters, the influence ofturbulence on the interacting dynamics ofcurrents and waves remains to be clearlyunderstood – this is especially true for near-bed processes.

Figure 2: Four embedded models run simultaneously, forced by tidal constituents and meteorological forecasts (IFREMER)

10 The evolution of models can be usefullycategorised as:

Generation 1: development of algorithms tosynthesise representations of processes;Generation 2: quantitative simulations ofspecific environments (pre-operational); andGeneration 3: fully operational systems with nowcasting and forecasting capabilities (see Appendix 4).

For pre-operational shallow water engineeringapplications, licensed codes are usedinternationally, with three EC countries as the major providers thereof. For scientificapplications, open-code community modelsare more commonly adopted; these generallyoriginate from the USA. These distinctionscan blur because open-codes require supportand licensed codes are often free for academicresearch. Commendably, the USA’s Office ofNaval Research (ONR) has supported theconversion of previously commercial ECcodes to the public-domain. Europeannetworks (fostered by EC Frameworkfunding) continue to support world-leadingspecialist code modules in areas from wavesto turbulence to ecology (for a list of widelyused model codes see Appendix 1.)

Model codes are becoming ever-increasinglycomplex. Assimilation techniques may requiremultiple simulations. Specialist technologiesare required to provide requisite speed andsophistication of inputting and outputting ofdata. Formalised approaches for model

validation and verification are necessary,including procedures for quality control ofmodules and assembly of a range of bench-testobservational data sets. Specialist software isrequired for diagnostic analyses, visualisation,and communication. Both the proprietary andpublic-domain model codes mentioned abovetypically involve investment of tens of yearsin software development and continuedmaintenance by sizeable teams. Such effort isincreasingly beyond the scope of mostEuropean modelling groups.

Existing operational forecasting systems inEuropean waters provide real-time and nearreal-time products describing wind field, waveheight spectra, temperature, salinity, floatingsea ice, chlorophyll, tides, currents, and stormsurges. Movements of oil slicks and algalblooms are also predicted on an emergencyoperational basis. Effective operation of real-time forecasts requires the resources of a meteorological agency for communications,processing and dissemination of forcing data,alongside oceanographic data centresresponsible for dissemination of quality-controlled marine data. Such agencies alsoprovide access to data required for

Operation of Models – Hardware & Software Requirements

Effective operation of both ocean and coupled shelf-sea modelsrequires access to supercomputers and continuous maintenance ofsoftware. Major infrastructural investment is needed if Europeanmodellers are to remain competitive. A small number of majorEuropean marine modelling centres are likely to emerge over the next five years with links to existing meteorological institution.These will support an associated network of centres addressinglocal applications.

Figure 3: Spatial and temporal resolution of oceanographic data (Proudman Oceanographic Laboratory)

rem

ote

sens

ing

Spat

ial c

over

age

(poi

nts

per 1

00 k

m2 )

Frequency (per year)1

1

10

102

103

104

105

106

10 102 103 104 105 106

ship/

labsa

mpling

models

fixednetwork

opera

tiona

l

11

assimilation in hindcast or what-if projectionsby the dispersed academic community.

Computing capacity restricts the optimumresolution in many simulations. The applicationof ecosystem models, often involvingcombinations of Lagrangian and Eulerianmethods for simulation of sediment andplankton movement, is severely limited bytheir computing requirements. New computingsystems, such as massively parallel or parallel-vector machines, are extensively used.

This problem makes it vital for software to beadaptable for running on different hardwareplatforms. Exploitation of future hardwaredevelopments will pose challenges for the optimisation of software architectures tocombine scalar and vector capabilities. In addition, development of algorithms torepresent processes over varying temporal andspatial scales, and ranges of complexity willcontinue – especially for ecologicalapplications.

Model resolution is also effectively limited bythe corresponding paucity of resolution inobservational data (especially bathymetry)used for setting-up, initialising, forcing(meteorological and along model boundaries),assimilation and validation (Figure 3). This paucity of data is a critical constraint inenvironmental applications.

Challenges

Effective operation of both ocean and coupledshelf-sea models requires access tosupercomputers and the software requirescontinuous maintenance. The mechanisms bywhich such integration can be facilitated needto be explored along with the needs forinfrastructure investment in very highperformance computers, high performancedata networks, new numerical algorithms, etc.

The organisational efficiency developed bymeteorological agencies must be used toattract the investment for observationalnetworks, data services, computationalfacilities, training, etc. needed to stimulateparallel developments of marine science. The success of the European Centre for Mid-range Weather Forecasts (ECMWF) instimulating European research intometeorology, climate, and oceanography isnoted – some (virtual) analogues in the marinecommunity might be conceived. An expedientcollaboration might involve separate institutesassuming delegated responsibilities forsupport and development of specific modules.A range of these is shown in Figure 4.

Visualisation, data banking and high-performance computing

Fish larvaemodelling

Contaminantmodelling

Climatology and extreme statistics

Ocean and atmospheric operational

forecasting

Sediment transportand resuspension

Coupled 3D baroclinichydrodynamics and WAM wave model ERSEM* biology

Tidalforcing

Meteorologicalforcing

SSTassimilation

Met Office Oceanmodel forcing

Figure 4: Ecological simulation system-component modules (Proudman Oceanographic Laboratory)

12

Data Requirements from Observations & Coupled Models

The accuracy of model simulations depends on: (i) the accuracyand resolution of the observational data used to set-up, initialise,force, assimilate, assess, and fine-tune the simulations; and (ii)the adequacy of forcing specified from coupled atmospheric andocean models.

Use is made of observations from on-shore (radar, tide gaugesetc.), off-shore moorings, ships, moored and drifting buoys,aircraft, and satellites. More and better observational data,extending over longer periods are essential if modelling accuracyand capabilities are to be enhanced. International collaborationis an obvious and valuable means of achieving this goal. Whileinternational funding supports satellite programmes, synergistic insitu monitoring presently relies on national funding.

Observational data

Formulation of coastal models requiresaccurate fine-resolution bathymetry, andideally, corresponding descriptions ofsurficial sediments. Subsequent operationsrequire river flows and their associatedtemperature, sediment, and ecologicalsignatures. Similar requirements apply towind and irradiance data for model forcingtogether with related data for open-sea/oceanboundary conditions. Real-timeobservational data are needed both forassimilation into operational models and forparameterisation-validation in pre-operational models.

Development of model simulations for tides,surges, and waves is constrained by limitedaccuracy and resolution of both bathymetryand wind forcing (data assimilations may beused to circumvent these limitations).Simulations of temperature, salinity,suspended sediment, water quality, andecological parameters are constrained by the availability of: (i) initialisation andforcing data, and (ii) subsequent assimilationdata being absent or restricted to surfacevalues.

Observational data can be obtained fromsatellites, aircraft, radar, buoys, floats,(cabled) moorings, gliders, AUVs (AutomatedUnderwater Vehicles), instrumented ferries,and VOS (Voluntary Observing Ships)together with meteorological and oceanmodels (Figure 5). Over the past two decades,remote sensing techniques have matured toprovide useful products of ocean wind,waves, temperature, ice conditions, suspendedsediments, chlorophyll, eddy, and frontallocations. Unfortunately, these techniquesprovide only sea-surface values and in situobservations are often necessary both forvertical profiles and calibration. For coastalapplications, improved spatial resolution, as provided from aircraft surveillance isespecially valuable. High frequency radarscan also provide synoptic surface fields ofcurrents, waves, and winds on scalesappropriate to the validation of coastalmodels.

Despite these advances, the range of marineparameters that can be accurately measured is severely restricted – especially in operationalmode (Figure 3). Moreover, the cost of theseobservations is orders of magnitude greaterthan that associated with the development orthe operation of models. Consequently, the effectiveness of simulations is severelylimited by shortcomings in the accuracy,spatial and temporal extent, and resolution ofsuch data.

Instrumentation is already lagging seriouslybehind model development and application,and this gap is expected to widen. Newsensors are needed, in particular sensorssuitable for installation on ferries and ships of opportunity and through-flow sensors for moorings. A new generation ofinstrumentation is needed for the validation of multi-species, size-class and species-resolving ecosystem models.

13

Assimilating in situ observations with remotesensing data, alongside rapid data processingand appropriate communications is essentialfor operational modelling. Particular attentionis required for assimilation in models ofcoastal seas – because of their rapid responsetimes and (often) large tidal excursions.

New cost-effective instrumentation (gliders,drifting buoys, and yoyo quasi-Eulerianprofilers for the shelves) is developing rapidly.However, permanent in situ observations arelikely to be the most expensive component ofan operational system, and it is important tooptimise the observational network in relationto the modelling system for the requisiteforecasts.

Data from coupled models

Accuracy, resolution, and extent (in time ahead)of wind forecasts are the primary limitingfactors for sea-state and surge forecasting.Likewise, sea surface heat exchange is clearly a determining factor in forecasting ocean

mixed-layer depth and ice formation. In bothcases, the need for dynamically coupled ocean-wave-ice-atmosphere models is an essentialelement to improve atmospheric forcing. Oceanbasin modelling requires better understandingof the processes associated with fluctuations inthe Gulf Stream and North Atlantic Current, theformation of Atlantic bottom water, ventilation,convection, and inter-annual changes in thestate of the Atlantic oscillation.

Coupling of regional sea and ocean models is a pre-requisite for longer term simulations(especially hindcasting) in shelf seas. Suchcoupling requires the resolving of differingrepresentations of specific processes – forexample the omission or exclusion of tides.For accurate simulation of European seas, we need improved understanding of the shelfedge and slope processes along the Atlanticmargin. This includes non-hydrostatic codes toresolve critical mixing processes. At the landboundary, coupling with hydrological modelswill complete the water cycle – although this is similarly dependent on development ofrelated monitoring systems.

Figure 5: Monitoring system(LEO–15 coastal observatory – Rutgers University)

Data

management

Centre–data

archive

Modelling

and product

development

14

In addition to the continual demands toenhance model performance in terms ofaccuracy, reliability, finer resolution, andextended forecast periods, there is an increasein the requirements to extend their scope fromphysics-chemistry-biology to ecology in fullycoupled ocean-atmosphere-terrestrialsimulations.

Human intervention in the marineenvironment continues to expand beyond thecoastal margins to shelf-wide activitiesincluding: fisheries; oil, gas and aggregateextraction; offshore energy installations andother industrial and commercial offshoredevelopments. Since associated regulatoryregimes must encompass operation of these activities alongside their environmentalimpacts, we need to link our marine modelswith their socio-economic counterparts. In such cases, coupling might be limited tosub-set representations (statistical emulators)encapsulating integrated parameters such asstratification levels or flushing times. To overcome the limitations of individualmodules in such total-system-simulations,methodologies are required both to quantifyand to incorporate the range of uncertaintiesassociated with model set-up, parameterisationand (future scenario) forcing. This requirementcan be achieved by ensemble simulationsproviding relative probabilities of variousoutcomes linked to specific estimates of risk.

Challenges

The design of new comprehensive networks,exploiting synergistic aspects of the completerange of instruments and platforms, integrallylinked to modelling requirements orcapabilities is a prospect as exciting as it isdaunting. Furthermore, specialist skills andsystems are required to assimilate suchobservational data in real-time. Enhancing andlinking investment in these network designsand associated assimilation techniques is a toppriority.

Lead-times between proof-of-concept,laboratory tests, and availability of commercialmarine packages have traditionally been in theorder of one or two decades. Hence, the paceof development of coastal modelling will begoverned by the foresight of scientists andtechnologists in responding to challenges(such as the EC Water Framework Directive)and prioritising areas of investment to providelonger-term observations with enhancedaccuracy and resolution.

15Extensive European collaboration has been

fostered via initiatives such as the ESFGrand Challenges, EC Framework Projects,EuroGOOS Regional Task Teams and Panelsetc. These collaborations have stimulatedprogrammes aimed at providing accurate fine-resolution bathymetry, routinestandardised sampling along ferry routes,effective exchange of marine andmeteorological data, specifications for futuresatellite missions, interaction between ocean-sea-coastal scientists. However, longerterm continuity remains a problem.

Support of community model codes (e.g.,GOTM, COHERENS, SWAN, in Appendix 1)involves quality assurance, documentation andversion control, training, user workshops, etc.While Europe can only support a limitednumber of such systems, the growingimportance of ensemble forecasts (foruncertainty estimates) emphasises theimportance of maintaining diversity andretaining expertise in international codes.Future accommodation of a diverse range ofmodules (model sub-systems) may befacilitated via couplers such as OASIS orPRISM. Taking into account the implementationplans for the Water Framework Directive, an adoption of standardised modules can allowindividual modelling groups to concentrate onmore specialised sub-modules. Appropriatevalidation benchmarks and protocols formodel outputs will be required. (Note, theCATCHMOD development towards agreedstandards for implementation of the WFD.)

The EC FP6 project Marine Environment andSecurity for the European Area (MERSEA),directly related to the Global Monitoring forEnvironment and Security (GMES) initiative,serves as an example of the value ofcollaboration at the programme level. Theoverall objective of MERSEA is to facilitatethe visibility, understanding and exchange ofthe ocean modelling data, output products forusers, and evaluate the strengths andweaknesses of the European capacity forocean monitoring and forecasting. Thiscollaboration integrates the followingexisting modelling-monitoring systems:FOAM (Met. Office, UK), MERCATOR(MERCATOR-OCEAN, France), MFS(INGV, Italy), and TOPAZ (NERSC,Norway). The aims of MERSEA are toembed a range of modelling applications(e.g., oil spill, ecological and regional) intoocean-scale systems. The Global Ocean DataAssimilation Experiment (GODAE) isanother initiative that might eventually leadto a stronger cooperation between Europeanand world-wide partners.

Challenges for Europe

Long-term leadership in science requires therecruitment of the most original, talented andtrained staff supported by state-of-the-arttechnologies linked to active globalcommunication networks. Hence, an obvioushigh priority and readily achievable initiativeis to link existing European funding andnetworks spanning: post-graduate trainingcourses and fellowships, specialist summerschools, workshops, conferences, journals,and international or national scienceprogrammes, such as the International GlobalBiosphere Programme (IGBP), etc. This willfacilitate the exchange of skills andexperience, software and data to enhancemodelling capabilities and guide strategicplanning. Institutionally, Europe needs toconsolidate the successful but occasionally

European Collaborations & Initiatives

Modelling has moved into an era that requires major investmentsin infrastructure and organisation (as in meteorology). Havingdeveloped a strategic vision, Europe needs to translate this intoeffective modelling capabilities to address both long-term globalissues and more immediate national and local concerns about themarine environment. The specific challenge to scientists is todevelop firstly the vision and then secondly the implementation ofthe framework to exploit these new opportunities created by an integrated European approach.

16

transient collaboration achieved through ECFramework Programmes and elsewhere.

The diversity of marine systems makes itunlikely that a single integrated model willevolve, as is the case for weather forecastingin the national meteorological agencies.Moreover, there is a continuing need for a wide range of types of models with differentcharacteristics to provide genuine ensembleenvelopes and cater for a range ofenvironments (such diversity does not obviatethe requirement that all models be validatedand robust). A systems approach is needed,capable of integrating marine modules and linking these into holistic simulators(geological, socio-economic etc.).Rationalisation of modules to ensureconsistency with the latter is an importantgoal, together with standardisation ofprescribed inputs such as bathymetry, tidalboundary conditions, etc. Finally, there will bea continuing need for a limited number ofglobal ocean models.

18

Appendix 1: Model Codes

The following figures indicate material fromEC Community Model web sites:

(a) Coherens, (b) SWAN, and (c) GOTM (list ofCommunity and Commercial models).

19

20

List of Community and Commercial modelsfrom GOTM web site.

21

Appendix 2: Hydrodynamic modelling

Previously, we noted that shelf sea modelapplications have progressed from the 2-D

barotropic models of the 1960s to the density-evolving 3-D baroclinic models of the 1980s.Development continues in areas such as:cross-spectral coupling of tides and waves,incorporation of non-hydrostatic internal wavesand utilisation of adaptive non-structured gridsalongside parallel computer architecture.Moreover, such models are widely applied inareas such as limnology, estuaries, harbours,coastal bays to shelf seas and oceans.

Shelf sea hydrodynamic modelling generallyfocuses on tides, surges and waves, since theserepresent the most energetic processes andprovide the background conditions for non-linear interactions with other dynamicalprocesses. Explicit solutions for associatedgravity-wave propagation introduce severerestrictions on the size of the allowable time-step. Two approaches are commonly used tosolve this problem. First, a time step splittingmethod is implemented in which thegravitational waves and the vertical viscosityand diffusivity, i.e. the most time-step limitingprocesses, are resolved explicitly with smallertime steps than all the other terms. Second, a complex alternative is the use of synchronoussemi-implicit time step in which the gravity-wave-producing term, the surface pressuregradients and their associated terms in the continuity equation, and normally also the vertical viscosity and diffusivity are treatedimplicitly whereas all other terms are solvedexplicitly. While this semi-implicit treatmentallows for a uniformly large time step, thenumerical calculation of this treatment is morecomplex to implement. However, in terms ofactual computational costs, there can bebenefits in using this algorithm. On the otherhand, the time-splitting method producesdisturbances due to the fact that the barotropicand baroclinic modes are not always directlycoupled, this can be overcome by a shortiterative loop achieving the convergence. Whileapplications involving parallel computing can

introduce additional complications, they maymake explicit schemes more attractive. To incorporate the effect of turbulence intothree-dimensional models a large number ofdifferent parameterisations have been applied.At the beginning, purely empirical formulas oralgebraic expressions were used. However,nowadays so-called two equation models havebeen proven to be a good compromise betweenaccuracy and efficiency, because these modelsstill assume a local equilibrium. Most well-known are the Mellor-Yamada and k-e-models,which have been shown to be equivalent. In theturbulent closure approaches the vertical eddyviscosity depends on either the turbulent kineticenergy with the length scale of the turbulentmotion or the turbulent kinetic energy with thedissipation rate. Two major factors that are usedto infer these quantities are the vertical velocityshear, which increases the vertical viscosity andthe buoyancy which in contrast suppresses it.

Two other water column mixing processes,where further work is necessary, are internalwave dynamics and convective mixing. It is known that breaking internal wavessignificantly contribute to vertical mixing but an adequate parameterisation of this process, for use in regional scale models, has yet to beachieved. Convection in the ocean is a downward sinking processes caused byinstability of the water column. This process canbe described by a non-hydrostatic formulationwhich introduces a new level of complexity, or alternatively by a special treatment in the turbulent closure scheme. The use of a localised scheme is problematic becauseconvection can also produce a counter-gradientflow and overshooting, neither phenomena canbe described in terms of a local equilibrium.Thus, a reasonable description of the fullconvective process would require a non-localturbulence model, such as the KPP model; herefurther research is needed. Fortunately, forapplications over scales of kilometres, muchsimpler convective adjustment schemesproduce satisfactory results.

22

Appendix 3: Ecological modelling

The development of conceptual andmathematical models has played a central

role in marine biology and ecology as a tool forsynthesis, prediction, and understanding. It isclear that significant progress has been madeover the past 20 years in the development ofnumerical process models of the marineenvironment. The development of large-scalemodels started in the early eighties andincluded elements of phytoplankton kinetics.In the first generation models, processes weredescribed by rather simple equations. In theseearly models, the transport was based on 2-Dhydrodynamic calculations for averageconditions (i.e. one representative day). Somemodels were comparatively refined and usedcomputational elements in the area of about

20x20 km. This allowed for some gradients incoastal areas. Most ecosystems models werelumped together in so called “box models” ofwhich only about 10 to 20 were considered forthe whole of the North Sea.

In recent years, important progress has beenmade on the scale of ocean basins, overallcirculation patterns and the distribution, andabundance of algal groups (e.g., dinoflagellates,flagellates etc.) as a function of water massdistributions, large scale current regimes, andfrontal systems. Examples of large-scale 3-Decosystem models applied to the North Sea andparts of the Northwest European ContinentalShelf are ERSEM (Figure 6), NORWECOM,POLCOMS-ERSEM, ECOHAM,

Figure 6: Ecosystem model (Proudman Oceanographic Laboratory)

23

COHERENS, Delft3D-ECO and ELISE. The models differ in the way they interactwith the physical models, the complexity ofthe trophic structure, and the way theinteraction with the sea bed is included.

With respect to the interaction with the physicalmodel, two modes exist: on-line and off-linecoupling. Some models can optionally be runeither way. The main advantage of the on-linecoupling method is the possibility to includefeedback between all relevant processes. On-line interfacing may be the only feasible optionfor detailed hydrodynamic models run for a long simulation period because the amount ofinformation to be stored for an off-linesimulation model may be impractically large. In contrast, the off-line method offers moreoptions to differentiate because of the level ofaggregation and the time-scale. So runningoffline may be (much) faster because of the useof a somewhat coarser grid and a longer time-step, but still being accurate enough for manymanagement and research questions. As a compromise some modelling systems allowfor a distinction in the level of detail by usingcurvilinear and/or nested grids.

Within the water phase, the models show greatvariation in the level of detail. Most, but not allrecent models include Nitrogen, Phosphorus,Silicate and Oxygen as parameters. Variation in the number of individual processes also hasto be taken into account. Some models includeonly one group of phytoplankton, othersdifferentiate between functional groups basedupon eco-physiological characteristics such asnutrient and light requirements or edibility,whereas some models focus in particular onharmful algal species. Few models includeexplicit equations for grazing, nor for the trophicinteractions with fish, and existing models paysurprisingly little attention to suspendedparticulate material (SPM), although highconcentrations in several coastal watersconsiderably affect the light climate, andhence, the local rate of primary production.Furthermore, existing SPM models are notvery accurate; hence, an improvement isnecessary.

Large discrepancies exist in the way the water-sea bed interface is simulated. A number ofmodels do not include a functional sea bed atall or they just consider a pool of sedimenteddead organic material, whereas other modelsinclude several bed layers and different formsof nutrients, and some even includeformulations for the benthic community.These differences may partially, but notcompletely, be explained by differences inscope of models. Obviously, the interactionbetween the water and sea bed is much moreintense in shallow coastal areas or estuariescompared with off-shore, deep areas. Inaddition, the important interactions betweenthe biological organisms at or in the sea bed,and physical factors such as stability, erosion,morphology or local hydrodynamic conditionsare only now being introduced in themainstream operational models. A revision ofthe appropriate level of details and theimportance of bio-geomorphologic processesseems necessary.

Many models are, and will in the future be,developed for specific regions. To apply thesemodels, they need adequate approved data forboundary conditions. In many cases, it isimpossible to obtain a complete set ofconditions from measurements; thus, the mostappropriate source for these data is large-scalemodels. These models should be set-up andmaintained at a supra-national level.Meteorological data, including modellingresults, should also be available at a centralEuropean level. Lack of approved data ondischarges is another factor hampering thedevelopment of regional models. To solve thisproblem, databases of loadings should bepublicly available at a European rather thannational level. Monitoring data are urgentlyneeded to validate model results, for nutrientsand total phytoplankton biomass quite anumber of data are available, although mainlyon a national level and mainly data from belowthe surface interface. There is a lack of data onvertical profiles and data on grazers are muchless abundant and irregular. There is a lack ofrelevant data for both formulation, as well asvalidation, of the water-sea bed interface.

24

Current operational models for fish stocks andfisheries do not take these implications intoaccount, nor do they interact directly withprimary and secondary production models.Hence, it is unclear to what extent changes infish stocks relate to changes in the plankton

Developing operational models for theseinteractions is, therefore, of major significanceand necessary to implement an ecosystem-based approach to marine management. OneGrand Challenge is the exploration of global-climate-change induced latitudinal migrationof species; this will require incorporation ofspecies-specific behaviour.

Present operational models pay little or noattention to toxic substances because problemsrelated to organic pesticides and heavy metalshave declined considerably in the Europeanmarine waters since the 1990s. Recently,concern about the impact of new types of toxicsubstances such as hormones is increasing.Thus a new generation of toxic substancemodels might have to be developed in the nearfuture to cope with these impacts.

Fisheries are an important economic activityin marine waters with potential implicationsfor the ecosystems (Figure 7).

Figure 7: Fisheries model (Proudman Oceanographic Laboratory)

community or to changes in the fishery itself.

25

Appendix 4: Operational Oceanography

Satelliteobservations

Model

In situobservations

Forecasts

Hindcasting

Observational data for hindcasting areassimilated into a model to compile sets ofhistoric fields and distributions (typicallymonthly or annually) of variables such as seasurface elevation, water temperature, salinity,nutrients, radio-nuclides, metals, fish stockassessments, etc.

Model generations

Numerical modelling has been used in marinescience for almost 50 years. A convenientdistinction is as follows:Generation 1: models where algorithms,numerical grids and schemes are beingdeveloped often utilising specificmeasurements focused on process studies.Generation 2: pre-operational models with(effectively) fully-developed codesundergoing appraisal and development,generally against temporary observationalmeasurements or test-bed data sets.Generation 3: operational models in routineuse and generally supported by a permanentmonitoring network.

A cascade time of approximately 10 years is typically required to migrate between eachGeneration.

Operational oceanography is defined as the activity of routinely making,

disseminating, interpreting measurements ofseas, oceans, and the atmosphere to provideforecasts, nowcasts, and hindcasts.

Forecasting

Forecasting includes real-time numericalprediction of processes such as storm surges,wave spectra, sea ice occurrence, climaticstatistical forecasts, and seasonal and inter-annual variability. Forecasts on a climatic orstatistical basis may extend forward for hours,days, months, years, or even decades.Accumulation of errors, both from modelinaccuracies and from uncertainties in forcing,limit realistic future extrapolations.

Nowcasting

In nowcasting, observations are assimilated innumerical models and the results are used tocreate the best estimates of fields at the presenttime, without forecasting. These observationsmay involve daily or monthly descriptions ofsea ice, sea surface temperature, toxic algalblooms, state of stratification, depth of the mixed layer, or wind-wave data.

Figure 8: Operational oceanography: an example of the MERCATOR system

26

Appendix 5: Terms of Reference and Membership

Terms of Reference for Marine BoardWorking Group on InnovativeModelling of Coastal and Shelf Seas

Goals of the Marine Board – ESF includepromoting the quality of, and access to, the science needed for effective managementcoastal and marine resources – underpinninggovernance to ensure sustainable exploitationof this invaluable resource. Parameters ofinterest include: surges, waves, currents,temperature, salinity, ice, sediment transportthrough to an ever expanding range ofbiological and chemical tracers The diversityin nature, usage and hence challenges inEuropean coasts requires fostering of localisedscientific expertise with access to s-o-afacilities to maintain excellence, andeffectiveness.

Models synthesise theory into algorithms anduse observations to set-up, initialise, force,assimilate and evaluate simulations inhindcast, nowcast and forecast modes. Thusthe use of models range from: gaining insightand understanding, hypotheses testing,quantifying the stage of scientific development,forecasting (flood warning to scenario testing),to sensitivity testing of dependence onalgorithms, computational resolution,accuracy and extent of observational data. The scope of the models involves linkagesacross ocean-atmosphere-seas-coasts andbetween physics-chemistry-biology-geology-hydrology, this connectivity spans:meteorological agencies, satellite missions,international scientific and survey programmessuch as IGBP, CLIVAR, GOOS etc.

Extensive European collaboration has beenfostered via initiatives such as: the ESF GrandChallenges, EC Framework projects,EuroGOOS Regional Task Teams and Panelsetc. Whilst these activities have been highlysuccessful, longer-lasting initiatives arenecessary to maintain European leadership inthe range of technologies involved. This ESF

Marine Board Working Group will aim toidentify such initiatives for innovativemodelling of Coastal and Shelf Seas.

Issues to be examined by the Working Groupwill include:• fostering of a European Marine Coastal

Modelling Community• coupling aspects of meteorology-physics-

ecology-hydrology• requirements for test-bed experiments and

long-term monitoring (networks)• future opportunities and requirements from:

in situ, satellite and other remote sensinginstruments

• development of assimilation techniques inthe coastal zone

• the range and success of community-modelgroups

• future plans and requirements of engineeringconsultancies

• the needs for infrastructure investment forintegrated modelling

• training and career planning.

Membership

Dr David Prandle (Chair)Proudman Oceanographic LaboratoryJoseph Proudman Building6 Brownlow StreetLiverpool L3 5DAUKTel.: +44 151 795 4861Fax: +44 151 795 4801Email: dp@pol.ac.uk

Professor Wolfgang FennelInstitut für Ostseeforschung Warnemünde (IOW)Postfach 30 11 6118119 RostockGermanyTel.: +49 381 5197110Fax: +49 381 5197114Email: wolfgang.fennel@io-warnemuende.de

27

Professor João Gomes FerreiraIMAR – Centre for Ecological ModellingDepartamento de Ciências e Engenharia do AmbienteFaculdade de Ciências e TecnologiaUniversidade Nova de Lisboa2825-114 Monte da CaparicaPortugalTel.: +351 21 2948300 Ext 10117Fax: +351 21 2948554Email: joao@hoomi.com

Dr Michael HartnettMarine Modelling CentreDepartment of Civil EngineeringMartin Ryan Institute & National University of IrelandGalwayIrelandTel: +353 91 524411 Ext 2502Email: Michael.Hartnett@nuigalway.ie

Dr Peter HermanHead of Department Spatial EcologyNetherlands Institute of Ecology (NIOO-KNAW)Centre for Estuarine and Marine EcologyP.O. Box 1404400 AC YersekeThe NetherlandsTel.: +31 113 577475Fax: +31 113 573616Email: p.herman@nioo.knaw.nl

Dr Michiel KnaapenDepartment of Civil EngineeringUniversity of TwenteP.O. Box 2177500 AE EnschedeThe NetherlandsTel.: +31 534 892 831Fax: +31 534 895 377Email: m.a.f.knaapen@ctw.utwente.nl

Dr Hans LosDelft HydraulicsP.O. Box 1772600 MH DelftThe NetherlandsTel.: +31 15 2858549Email: hans.los@wldelft.nl

Professor Morten PejrupInstitute of GeographyUniversity of CopenhagenOester Voldgade 101350 Copenhagen KDenmarkTel.: +45 3532 2505Fax: +45 3532 2501Email: mp@server1.geogr.ku.dk

Dr Thomas PohlmannInstitut für MeereskundeTroplowitzstr. 722529 HamburgGermanyTel.: +49 40 4123 3547Fax: +49 40 4123 4644Email: Pohlmann@ifm.uni-hamburg.de

Dr Roger ProctorProudman Oceanographic LaboratoryJoseph Proudman Building6 Brownlow StreetLiverpool L3 5DAUKTel.: +44 151 795 4856Fax: +44 151 795 4801Email: rp@pol.ac.uk

Dr Yann-Hervé De RoeckHead of DepartmentDynamics of Coastal EnvironmentIFREMERBP 7029280 PlouzanéFranceTel.: +33 2 98 22 44 95Fax: +33 2 98 22 47 60Email: yhdr@ifremer.fr

Dr Karline SoetaertNetherlands Institute of Ecology (NIOO-KNAW)Centre for Estuarine and Marine EcologyP.O. Box 1404400 AC YersekeThe NetherlandsTel.: +31 113 577300Fax: +31 113 234234Email: k.soetaert@nioo.knaw.nl

Dr Takvor SoukissianNational Centre for Marine ResearchInstitute of Oceanography POSEIDON Project47 km Athens-Sounion RoadP.O. Box 712 AnavyssosAttica 190 13GreeceTel: +30 22910 76399Fax: +30 22910 76323Email: tsouki@ncmr.gr

28

Dr Tapani Stipa Finnish Institute of Marine ResearchP.O. Box 3300931 HelsinkiFinlandTel.: +358 9 613941Fax: +358 40 643 9786Email: Tapani.Stipa@fimr.fi

Dr Georg UmgiesserOceanographyISMAR – CNR1364 S. Polo30125 VeneziaItalyTel.: +39 041 5216 875Fax: +39 041 2602 340Email: georg.umgiesser@ismar.cnr.it

Dr Waldemar WalczowskiMarine Dynamics DepartmentInstitute of OceanologyPolish Academy of Sciencesul. Powstancow Warszawy 5581-712 SopotPolandTel.: +48 58 551 72 81Fax: +48 58 551 21 30Email: walczows@iopan.gda.pl

Marine Board SecretariatDr Niamh ConnollyExecutive Scientific SecretaryMarine Board European Science Foundation 1, Quai Lezay-MarnésiaBP 9001567080 Strasbourg CedexFranceTel.: +33 (0)3 88 76 71 66

Email: nconnolly@esf.orgFax: +33 (0)3 88 25 19 54

European Science Foundation1 quai Lezay-Marnésia, B.P. 9001567080 Strasbourg cedex, FranceTel: +33 (0)3 88 76 71 00Fax: +33 (0)3 88 37 05 32 www.esf.org

European Science Foundation1 quai Lezay-Marnésia, B.P. 9001567080 Strasbourg cedex, FranceTel: +33 (0)3 88 76 71 00Fax: +33 (0)3 88 37 05 32 www.esf.org

EUROPEAN SCIENCE FOUNDATION

June

200

5 –

Prin

t run

: 200

0