MODEL BENCHMARKS FOR COASTAL LAGOONS

Catherine Aliaume, Antonio Bodini, Cristina Bondavalli, María Francisca Careño Fructuoso, Annie Chapelle, Pedro Duarte, Miguel Angel Esteve, Manuela Falcão, Annie Fiandrino, Gianmarco Giordani, Zacharo Kavakli,

Lionel Loubersac, Dimitar Marinov, Julia Martinez, Alain Norro, José Ozer, Antonio Pereira, Martin Plus, Francesca Somma, George Tsirtsis, Pierluigi

Viaroli and José- Manuel Zaldívar

Institute for Environment and Sustainability

2006

EUR 22216 EN

The mission of the Institute for Environment and Sustainability is to provide scientific and technical support to EU policies for the protection of the environment contributing to sustainable development in Europe. European Commission Directorate-General Joint Research Centre Institute for Environment and Sustainability Contact information Address:Via E. Fermi 1, TP 272 E-mail: jose.zaldivar-comenges@jrc.it Tel.:+39-0332789202 Fax: +39-0332785807 http://ies.jrc.cec.eu.int http://www.jrc.cec.eu.int Legal Notice Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use which might be made of this publication. A great deal of additional information on the European Union is available on the Internet. It can be accessed through the Europa server http://europa.eu.int EUR 22216 EN Luxembourg: Office for Official Publications of the European Communities © European Communities, 2006 Reproduction is authorised provided the source is acknowledged Printed in Italy

Cover: Logo of the DITTY Project

MODEL BENCHMARKS FOR COASTAL LAGOONS

Catherine Aliaume1, Antonio Bodini2, Cristina Bondavalli2, María Francisca Careño Fructuoso3, Annie Chapelle4, Pedro Duarte5, Miguel Angel Esteve3, Manuela Falcão6, Annie Fiandrino4, Gianmarco Giordani2, Zacharo Kavakli7, Lionel Loubersac4, Dimitar Marinov8, Julia Martinez3, Alain Norro9, José Ozer9, Antonio Pereira5, Martin Plus4, Francesca Somma8, George Tsirtsis7, Pierluigi Viaroli2 and José- Manuel Zaldívar8

1 ECOLAG, Montpellier University II, Montpellier, France 2 Department of Environmental Sciences, University of Parma, Italy 3 Department of Ecology and Hydrology, Murcia University, Spain 4 Ifremer, France 5 Centre for Modelling and Analysis of Environmental Systems, Fernando Pessoa

University, Portugal 6 IPIMAR, Olhão, Portugal 7 Dept. of Marine Sciences, School of Environmental Sciences, Aegean University,

Greece 8 Institute for Environment and Sustainability, Joint Research Centre, European

Commission, Italy 9 Management Unit of the North Sea, Royal Belgian Institute for Natural Sciences,

Belgium

DITTY PROJECT (Development of an information technology tool for the management of Southern European lagoons under the influence of river-basin runoff)

(European Commission FP5 EESD Project EVK3-CT-2002-00084)

D16. Model benchmarking and site comparisons

i

CONTENTS

1. INTRODUCTION 1

2. WATERSHED MODELLING BENCHMARKS 3

2.1. The Etang de Thau watershed 3

2.2. Comparison between SWAT and AGWFL 6

2.2.1. Data availability 6

2.2.2. Description of model runs 6

2.2.3. Result comparison 12

3. HYDRODYNAMIC MODELLING BENCHMARKS 14

3.1. The Etang de Thau lagoon 14

3.2. Comparison between COHERENS and MARS3D models 16

3.2.1. Description of simulation 16

3.2.2. Comparison of model results 18

3.3. The Gulf of Gera 23

3.4. Comparison between COHERENS and POM models 24

3.4.1. Introduction and forcing 24

3.4.2. Description and results of the test run 27

4. BIOGEOCHEMICAL MODELLING 30

4.1. LOICZ intercomparison 30

4.1.1. Introduction 30

4.1.2. LOICZ BM application to the DITTY sites 31

4.1.3. Results of the LOICZ BM applications and comparison 31

4.1.4. Conclusions 39

4.2. Object oriented approach to biogeochemical modelling 40

4.3. Phytoplankton modelling approaches 45

4.3.1. Phytoplankton growth models 46

5. CONCLUSIONS 49

REFERENCES 50

ANNEX 1. Results of the Comparison between COHERENS and

MARS3D models 54

ANNEX 2. Examples of interfacing FORTRAN, C AND C++ languages 66

ANNEX 3. Example of COHERENS using ECODYNAMO objects

with a Light class 71

ii

1

1. INTRODUCTION

To promote reliable, real-time management of coastal lagoons, increasingly

sophisticated numerical models have been developed within the DITTY project (WP4).

While models are diverse in design and scope, i.e. watershed, fluid-dynamics,

biogeochemical, all have the same fundamental goal, i.e. to account realistically for the

processes that drive the dynamic behaviour in coastal lagoons so that their status may

ultimately be predicted and the effects of mitigation actions be properly evaluated,

resulting on a series of good management practices that increase the sustainability of

these fragile ecosystems.

The intent of WP5 (Intercomparison analysis) is to establish a standard set of input

parameters and numerical “experiments” to be performed by various existing models so

that independent results could be meaningfully compared and evaluated, having in mind

the diversity of approach and systems (watershed, lagoon, adjacent coastal area).

Furthermore, though comparison, conceptual weakness could be identified and targeted

for further exploration by the DITTY partners as a whole. In this report (D16) , that

follows D15 in which we described summarised and analysed the employed models in

the DITTY community, we have carried out several intercomparison exercises taking

into account the diversity of the model employed.

Model intercomparison techniques have been developing over the years in a number of

environmental research communities (Røed et al., 1995; Hackett et al., 1995; Proctor,

1997 and 2002; Cramer and Field, 1999; Denning et al., 1999; Orr, 1999; Skogen and

Moll, 2000; Beckers et al., 2002; Davies et al., 2002; Caputo et al., 2003; Smith et al.,

2004; Delhez et al., 2004). For example, Smith et al. (2002) developed a distributed

model intercomparison project (DMIP) to compare simulation of distributed hydrologic

models to investigate several issues, such as: nature and impact of spatial variability of

basin physical characteristics and forcings, optimal level of basin disaggregation to

captures essential spatial variability, nature of error propagation through distributed

models. Concerning hydrodynamic models, Beckers et al. (2002) carried out an

intercomparison exercise on several water circulation models applied to the

Mediterranean Sea using the same forcing. The results show that no model performed

better than the others and that there was a similar correlation between model

characteristics and modeller’s skill in terms of results. Intercomparison analysis

concerning biogeochemical models are more scarce, for example a limited exercise was

carried out by Skogen and Moll (2000) concerning the primary production of the North

Sea using two ecological models. Both models gave similar results on the annual mean

primary production, its variability and the influence of the river inputs.

As already stated, the goals of an intercomparison exercise are always the same.

However, the processes we are interested in assess through the developed models are

quite different and the requests in terms of data input, forcing variables, validation data,

as well as calibration and sensitivity analysis are not completely similar. As we are

concerned with three fundamentally different realms of modelling (watershed,

hydrodynamic, biogeochemical modelling), we have decided to structure the document

in three separate sections.

The analysis will serve as a preliminary screening mechanism to select which

techniques/tools can (realistically) be implemented in the Decision Support System

2

(DSS). Furthermore, it has the objective of assessing the confidence in the model

outputs/predictions from the defined scenario analysis which in turn will support

decision making in coastal lagoons. The final outcome of the analysis will be a guide

for coastal lagoon modellers that may help the implementation of similar tools to other

coastal lagoons (D17 and D18). These two deliverables will be structured in the form of

a book that will also contain the software developed in the DITTY project and has as

the main objective to disseminate the information and knowledge generated during the

project.

3

2. WATERSHED MODELLING BENCHMARKS

2.1. The Etang de Thau watershed

The Etang de Thau watershed (Fig 1.) extends over about 280 km2 and is drained by

numerous small streams (3-13 km) with intermittent flows. It is delimited by the

Aumelas massif (altitude, about 300 m above sea level) to the north, the Gardiole massif

to its eastern rim and by the Hérault river basin to the west. Toward the south, a narrow

sandy strip separates the lagoon from the sea. This area, called Lido, accounts for only

5% of total watershed surface, with very low slopes and sandy sediments, which render

the inputs extremely diffuse. The rivers that flow on the northern border of the lagoon

(areas delimited by green lines on figure 1) drain a surface of 250 km2. Ten streams can

be clearly outlined from a hydrological modelling point of view as accounting for

almost all of the contributions to total inputs into the lagoon (table 1).

Figure 1. The Thau watershed, digital elevation model (50 m grid) and stream network (French

National Geographic Institute -IGN- database). Green lines outline the catchment area

and its sub-watersheds.

Table 1. Etang de Thau sub-watersheds characteristics.

From table 1, one can observe that two river basins (the Vène and the Pallas) cover

about the half of the whole catchment area, while other river basins are much smaller.

For the purpose of the benchmarking exercise, it was decided to focus only on the Vène

Surface (km2) % Cumulative %

Total watershed 250.3 100 Vène 66. 4 26.5 26.5 Pallas 52.2 20.9 47.4 Nègue-Vaques 33.0 13.2 60.6 Soupié 18.2 7.3 67.9 Aygues-Vaques 15.4 6.1 74.0 Lauze 9.3 3.7 77.7 Fontanilles 9.2 3.7 81.4 Mayroual 6.8 2.7 86.8 Joncas 5.5 2.2 89.0 Aiguilles 5.3 2.1 91.1

4

river, being it the largest of all streams in the watershed. Therefore, comparison

between the SWAT and AVGWLF models has been conducted on the Vène sub-

watershed.

The eastern part of the catchment area is composed of strongly karstic Jurassic

limestone, overlaid by Miocene marls in its central part. This area corresponds more or

less to the Vène river watershed, which is fed by two karstic resurgences, the Vène and

Issanka springs. By overlaying the geologic and the D.E.M. maps (figures 1 and 2), one

can also note that area of the Vène sub-watershed is characterised by steep slopes (mean

slope is about 2%, following Anonymous, 1997).

Figure 2. Schematic geomorphologic map of the Thau watershed.

Figure 3 presents a land cover map of the Thau watershed area. This map has been

drawn up on the basis of aerial photographs taken in 1996 at a 1/25000 scale (French

National Geographic Institute), and validation has been performed by field observations

(La Jeunesse, 2002).

Natural areas spread over a large part of the Thau catchment area, covering 103 km2.

They are principally located in the north east, and represent about 35% of the total

surface, with the natural Mediterranean sclerophyllous vegetation locally called

'garrigue', as the main component (93 km2). In some places the garrigue is mixed with

some pine woods as on the Gardiole massif or on the Sète city hill for example. The

garrigue (0.2-2 m high) consists of low scattered bushes (Quercus coccifera, Quercus

ilex, Cistus albidus, Juniperus oxycedrus, together with many common herbaceous

species Thymus vulgaris, Rosmarinus officinalis, etc.) with bare patches of rock or stony

ground between.

The typical crop landscape on the Thau watershed is composed of vineyards, covering

almost 40% (109 km2) of total watershed surface. Vineyards are mainly located in the

south west of the watershed, and represent a significant coverage in the Vène sub-

5

watershed. Other crops are mainly composed of durum wheat (7% of total watershed

surface, 20 km2), and some scarce parcels of fruit trees (in total 2 km

2).

Figure 3. Land cover map of the Thau watershed (1996).

Urban areas represent 10% of the total watershed surface, with 22 km2 covered. A

motorway crosses the watershed covering 1 km2 in total. Some bare soils (originating

from mineral extraction quarry and dump sites or simply being ravines and

escarpments), are also distinguishable and represent 2% of total watershed surface (6

km2). Urban areas are composed of small villages and little cities distributed all over the

catchment area (table 2 presents the respective importance of each urban areas). In

summer the population increases sensibly due to tourism, with an increase of population

ranging 7%-770% during June, July and August.

Table 2. List and importance of urban sites located on the Thau lagoon watershed. Identification

numbers (Id) are reported in figure 3. Id Name Surface (km

2) Permanent population

† Summer-increase

1 Sète 10.2 42 738 × 1.2 2 Balaruc 2.8 6 962 × 3.0 3 Bouzigues 0.7 1 014 × 1.8 4 Poussan 1.1 3 563 × 1.5 5 Gigean 0.7 2 847 × 1.1 6 Montbazin 0.7 2 490 × 1.1 7 Cournonsec 0.6 1 569 × 1.1 8 Villeveyrac 0.6 2 026 × 1.5 9 Loupian 0.4 1 399 × 1.8 10 Mèze 1.6 6 977 × 1.6 11 Pinet 0.4 944 × 1.1 12 Pomérols 0.7 1 837 × 1.5 13 Marseillan 1.6 5 432 × 8.7

† Number of inhabitants in 1998 (La Jeunesse, 2001)

The vine industry concerns directly the rivers located on the Thau watershed. In each

village, the winegrower cooperatives processing the grapes from the surroundings

6

vineyards discharge a large amount of organic wastes. In the ‘60s and ‘70s all

discharges were made directly in the rivers, but progressively, cooperatives have been

connected to the existing urban wastewater treatment plants, with or without pre-

treatment by flocculation-centrifugation or by evaporation tanks (table 3).

Table 3. Winegrowers cooperatives annual production (P, year 1996), date of connection to urban

wastewater treatment plant and type of treatment applied (data from La Jeunesse, 2001). Cooperative P (hectolitres) Connection Pre-treatment

Gigean† 138 000 1991 flocculation-centrifugation until 1993

evaporation tank since 1994 Marseillan 81 884 1985 " " " 1996 Mèze 40 900 1965 " " " 1993 Montbazin 35 284 1991 " " " 1991 Pinet 65 439 1988 " " " 1994 Pomérols 58 181 1988 " " " 1994 Villeveyrac 78 451 1980 " " " 1993

† The Gigean cooperative collects also Bouzigues, Cournonsec, Loupian and Poussan productions

2.2. Comparison between SWAT and AVGWLF

The SWAT and AVGWLF models were used to derive water flows and nitrogen,

phosphorus and sediment fluxes discharged by the Vène river into the Etang de Thau.

The POL was used to derive nitrogen and phosphorus fluxes only. No model details will

be offered in this section, as an accurate description of each model is presented in the

report “Comparison between different modelling approaches for coastal lagoons” (D15,

EUR 21817, this project). Model intercomparison is also presented in the same report.

All model runs have been performed under the same scenario set-up.

2.2.1. Data availability

A database (partially loaded in the project database - http://www.dittyproject.org) was

built from all available data. Datasets pertained to permanent installations present on the

Vène river and watershed, and from several measuring campaigns. A synoptic view of

data availability is presented in tables 4 and 5.

2.2.2. Description of model runs

SWAT. The ArcView environment was used to set up and run the SWAT model. After

preparation of the appropriate meteo and point source datasets, and loading of the

necessary coverages (DEM, soil, land use, river network, point sources, etc.) to the

ArcView project, the model was calibrated for the Vène catchment on the 1993-1994

period. All final values for the calibration parameters are listed in table 6. Simulations

were launched on a ten years period (1990-1999). However, in order to reach a

"biogeochemically" acceptable status of the model (all variables were set to 0 at the

beginning of the simulation), the results will be taken into account only after 3 years of

simulation, i.e. during 7 years (from 1993 to 1999). Figures 4 to 7 present the

comparison of simulated versus measured water flow, sediment production, nitrogen

and phosphorus at the Vène river outlet.

Table 4. Spatial coverages available for model building.

Theme Format Resolution Source

DEM Raster 50 m French National Geographic Institute Stream network Vector - French National Geographic Institute Geomorphology Vector - Othman (1997); Aonnymous (1997) Land cover Vector - Derived from aerial photos of the French National Geographic Institute Point source location Vector - Plus et al. (2003)

7

Table 5. Datasets for watershed model calibration and validation. Category Parameter Units Location Period Frequency Sampling

Rainfall Wind speed

mm m/s

Sète 1994-2004 Daily Continuous

Rainfall mm Marseillan 1994-2004 Daily Continuous Rainfall mm Mèze 1994-2004 Daily Continuous

Rainfall mm Montbazin 1994-2004 Daily Continuous

Rainfall mm Florensac 1994-2004 Daily Continuous

Mete

oro

log

ical

Air temperature Solar radiation Air humidity

°C MJ/m2/d %

Fréjorgue 1994-2004 Daily Continuous

Water flow m3/s Vène outlet 1994-2004 Daily Continuous

Sediments mg/l Vène outlet 1994-2004 Daily Sparse Nitrogen (in various forms) mg/l Vène outlet 1994-2004 Daily Sparse R

iver

ou

tlet

Phosphorus (in various forms) mg/l Vène outlet 1994-2004 Daily Sparse

Water flow m3/s Treatment plants* 1994-2004 Daily Sparse

Nitrogen (in various forms) mg/l Treatment plants* 1994-2004 Daily Sparse Phosphorus (in various forms) mg/l Treatment plants* 1994-2004 Daily Sparse

Water flow m3/s Springs** 1994-2004 Daily Sparse

Nitrogen (in various forms) mg/l Springs** 1994-2004 Daily Sparse

Po

int

so

urc

es

Phosphorus (in various forms) mg/l Springs** 1994-2004 Daily Sparse * see table 3 for a list of plants; **See figure 3 for spring location.

Table 6. Values of parameters deriving from the SWAT calibration process.

Parameter Description Value Unit

CN2 urban Runoff coef. for urban zones 85 - CN2 range Runoff coef. for fallow 77 - USLE_P vine USLE support practice coef. 0.15 - " garrigue USLE support practice coef. 0.1 - " range USLE support practice coef. 0.1 - " urban USLE support practice coef. 0.1 - OV_N range Manning's overland flow coef. 1.5 -

ALPHA_BF Base flow recession cst. 0.2 d GW_DELAY Delay time for recharge 20 d GW_REVAP Revap. coefficient 0.1 - REVAPMIN Threshold depth for revap. 0 mm

CH_K2 Hydraulic conduct. (streams) 150 mm/h

SURLAG Surface runoff lag time 1 d RCN Nitrogen in rain 0.669 mg/l BC1 Cst. rate for NH4→NO2 1 d

-1

BC2 Cst. rate for NO2→NO3 2 d-1

RK1 CBOD desoxygenation rate 0.02 d-1

RK2 Oxygen reaeration rate 1 d-1

RK4 Benthic oxygen demand 0 mg/m2/d

RS1 Algal settling rate 0.15 m/d AI1 Nitrogen fraction in algae 0.09 mg/mg AI2 Phosphorus fraction in algae 0.02 mg/mg AI5 O2 uptake/NH3 oxidation 3 mg/mg AI6 O2 uptake/NO2 oxidation 1 mg/mg TFACT Photosynth. active radiation 0.5 - P_N Algal preference for NH4 1 - MUMAX Algal maximum growth rate 3 d

-1

K_L light limitation coef. for algae 0.223 kJ/m2/mn

K_N N limitation coef. for algae 0.01 mg/l K_P P limitation coef. for algae 0.001 mg/l

Figure 4. Mean daily simulated (blue line) and measured (red rhombuses) Vène water flow. Empty

rhombuses are instantaneous measurements.

8

Figure 5. Daily simulated (blue line) and measured (red rhombuses) suspended matter (SM) loads

(A, log scale) and concentrations (B) in the Vène river. Empty rhombuses are calculated

values based on instantaneous measurements.

Figure 6. Simulated (blue line) and measured (red rhombuses) daily loads for ammonium (top),

nitrate (center), and nitrite (bottom) in the Vène river. Empty rhombuses are calculated

values based instantaneous measurements. Values are displayed in log scale, line breaks

are due to extremely low values.

9

Figure 7. Simulated (blue line) and measured (red rhombuses) daily loads for organic nitrogen

(top), dissolved inorganic phosphorus (center), and oraganic phosphorus (bottom) in the

Vène river. Empty rhombuses are calculated values based instantaneous measurements. Values are displayed in log scale, line breaks are due to extremely low values.

AVGWLF The ArcView environment was also used to set up and run the AVGWLF

model. After preparation of the appropriate meteo and point source datasets, and

loading of the necessary coverages (DEM, soil, land use, river network, point sources,

etc.) to the ArcView project, the model was calibrated for the Vène catchment on the

1993-1994 period. It must be noted here that input requirement for the two model is

conceptually similar. However, model set-up for AVGWLF is relatively simpler than for

SWAT: the former lacks a weather generator that “fills the blanks” in discontinuous

weather series, and requires only precipitation and air temperature data; no crop growth

is modelled, and as a consequence land use categories are much reduced in number

(only two classes for all arable and tree crops); no there is no river water routing and

water quality modelling. In other words, model parametrization is much simpler.

All final values for the calibration parameters are listed in table 7. Simulations were

launched on a ten years period (1990-1999). However, in order to reach a

10

"biogeochemically" acceptable status of the model (all variables were set to 0 at the

beginning of the simulation), the results will be taken into account only after 3 years of

simulation, i.e. during 7 years (from 1993 to 1999).

Table 7. Values of parameters deriving from the AVGWLF calibration process.

Parameter Description Category Value Unit

hay/pasture pasture 2 - cropland arable crop 43 - lo_int_dev low density urban 65 - C

N2

Runoff

coeffic

ient

hi_int_dev high density urban 82 - hay/pasture pasture 0.52 - cropland arable crop 0.45 - lo_int_dev low density urban 0.20 -

US

LE

_P

US

LE

suppor

t pra

ctic

e

coeffic

ient

hi_int_dev high density urban 0.20 - April 0.5 / 13 - May 0.6 / 15 - June 0.8 / 15 - July 0.9 / 15 - August 0.9 / 14 - September 0.7 / 12 - October 0.2 / 11 - November 0.1 / 9 - December 0.1 / 9 - January 0.2 / 9 - February 0.4 / 10 -

Ket /

Day h

ours

ET

cover

coeff

icie

nts

N

um

ber

of

daylig

ht hours

March 0.5 / 12 - Init_Unsat_Stor Initial unsaturated storage 10 cm Init_Sat_Stor Initial saturated storage 0 Recess_Coef Recession coefficient 0.12 d

-1

Sediment A Factor Empirically derived constant 1.092E-04

In figure 8 the comparison between simulated and measured flows at the Vène outlet are

presented. AVGWLF also simulates monthly discharge of sediments, nitrogen and

phosphorus. Simulation results are presented in figures 9 to 11. In the same figures

monthly estimates derived from scattered measurements are also displayed. The original

measurements are those presented in figures 5 to 7.

0

500,000

1,000,000

1,500,000

2,000,000

01-01-93 01-07-93 01-01-94 01-07-94 01-01-95 01-07-95 01-01-96 01-07-96 01-01-97 01-07-97 01-01-98 01-07-98 01-01-99

(m3/d)

Figure 8. Mean daily simulated (green line) and measured (blue rhombuses) Vène water flow.

Empty rhombuses are instantaneous measurements.

11

0

500

1000

1500

2000

Apr-90 Apr-91 Apr-92 Apr-93 Apr-94 Apr-95 Apr-96 Apr-97 Apr-98

To

tal

Se

dim

en

ts

(to

n/m

on

th)

Figure 9. Monthly AVGWLF simulated total sediments at the Vène river outlet (green line). The red

rhombuses represent monthly estimates derived from instantaneous daily measurements.

0

5000

10000

15000

20000

25000

30000

Apr-90 Apr-91 Apr-92 Apr-93 Apr-94 Apr-95 Apr-96 Apr-97 Apr-98

To

tal

Nit

rog

en

(k

g/m

on

th)

Figure 10. Monthly AVGWLF simulated total nitrogen at the Vène river outlet (green line). The

red rhombuses represent monthly estimates derived from instantaneous daily

measurements.

0

1000

2000

3000

4000

5000

Apr-90 Apr-91 Apr-92 Apr-93 Apr-94 Apr-95 Apr-96 Apr-97 Apr-98

To

tal

ph

os

ph

oru

s (

kg

/mo

nth

)

Figure 11. Monthly AVGWLF simulated total phosphorus at the Vène river outlet (green line). The

red rhombuses represent monthly estimates derived from instantaneous daily measurements.

12

2.2.3. Result comparison

The general flow regime of small Mediterranean streams is characterised by high values

during the autumn and winter months while extremely low or no flows are found during

summer. This general flow pattern is well simulated by both SWAT and AVGWLF, as well

as some sudden floods that usually occur under Mediterranean climates (see for

example the sudden flow increase, reaching 13 m3/s in the Vène river during two days

in May 1999). A comparison between model performances for water flow simulation is

presented in figure 12, along with the Nash and Sutcliffe efficiency coefficient. The

models simulates also fairly well dry spells, such as the one occurring in 1998, with a

very low flow all along the year (maximum flow at the Vène river was 1.8 m3/s and

mean annual flow: 0.14 m3/s). As no measurements were available for this year, these

simulated values still have to be validated.

Concerning sediments and nutrients, whatever the form, the discharges show a highly

variable pattern, with changes of several orders of magnitude that can occur from one

day to another. A general seasonal pattern can be noted: higher pikes are usually found

during the three months September, October and November, high values can also be

noticed during winter and spring while summer period is always characterised by

minimum loads. Since dissolved inorganic nitrogen concentrations (data not shown)

appeared to be much less variable than daily loads, it can be stated that the nitrogen

discharges at the river outlets are mainly driven by the river water flow.

The SWAT model is able to simulate sediment and nutrients flows satisfactorily (see

figures 5 to 7). In particular, SWAT allows to separate nitrogen and phosphorus yield at

the outlet in their various components (organic nitrogen, ammonia, nitrate, nitrite,

organic and mineral phosphorus). Such possibility is very useful when wanting to study

sediment and nutrient mobilization, which in ephemeral streams is a very important

phenomenon. On the contrary, AVGWLF provides only monthly values. Such

characteristic derives from the main finality of the model, which is the study of source

apportionment. In figures 9 to 11 such simulated values are displayed along with

monthly estimates derived from scattered measurements (the original data are those

appearing in figures 5 to 7). A better calibration for sediment and nutrient simulated

values could be performed if longer series of measured values were available. Such

longer series would also allow better estimates of monthly totals to compare with

simulated values. Here the comparison between monthly simulated values of total

sediments, total nitrogen and total phosphorus and estimated monthly totals is presented

simply as an exercise. Not judgment can be passed at this stage on model performance.

In terms of suitability to a Decision Support System, it must be emphasized that the

SWAT model is definitely not apt to fitting within such frame. Input requirements,

calibration and outputs are excessively cumbersome and more suited to the purposes of

applied research than to a technical purpose. The AVGWLF model instead has being

designed to be interfaced with a DSS, and in fact link to a DSS for analysis and

selection of Best Management Practices that allow to reduce point and diffuse source

pollution is already available (Evans et al., 2003).

13

ESWAT=0.61

0

500,000

1,000,000

1,500,000

2,000,000

01/09/1993 01/03/1994 01/09/1994 01/03/1995 01/09/1995 01/03/1996

(m3/d)

SWAT AVGWLF Observed Observed

EAVGWLF=0.69

Figure 12. Monthly AVGWLF simulated total phosphorus at the Vène river outlet (green line). The

red rhombuses represent monthly estimates derived from instantaneous daily

measurements.

14

3. HYDRODYNAMIC MODELLING BENCHMARKS

3.1. The Etang de Thau lagoon

The lagoon of Thau is located on the French Mediterranean coast, with an approximate

75 Km2 surface and an average depth of 4.5 m. It is under strong marine influence. The

lagoon is connected north to the sea by the canal of Sète (90% of exchanges) and south

by the Grau de Pisse. Saumes (10% of exchanges)

Figure 13. Location of the Etang de Thau.

The climate imposes a wide range of water temperatures and salinities with minima of

5°C in February and salinity near 27, and maxima of 29°C in August and a salinity of

40. Precipitation also shows large interannual variation (from 200 to 1000 mm per

year). Wind is often strong with a mean of 118.5 days per year above Beaufort force 5

(data from Météo-France), particularly when it is blowing from the Northwest (the so

called “Tramontane”).

The lagoon is home to an intensively developed shellfish farming activity (oysters and

mussels) that covers about 20% of the available water surface area and produces yearly

about 15000 tons of shellfish, more than 10% of French oyster production. The

economical annual revenue has been oscillating during the past few years around 40

MEuro providing work for approximately 2000 persons. This considerable production

depends to a large extent on nutrient inputs into the ecosystem, supplied mainly from

fresh water. Because of the weak tidal range, the residence time of water masses in the

Thau lagoon mainly depends on wind and barometric effects and it has been estimated

that the water renewal time is about 3 months. The catchment has already been

described in Ch. 2.1.

15

Due to the low water exchange, intensive shellfish farming activities and to urban and

agricultural pollution, the Thau lagoon has experienced, during several summers, acute

eutrophication problems with anoxic crises. During the last fifteen years, the Thau

lagoon has been extensively studied within the framework of PNOC and then PNEC

national research programmes, with investigations of the exchange between the water

column and sediments, the oysters farming activities, the impact of the watershed and

interactions with the Mediterranean sea. Various numerical models have been

developed, focusing on hydrodynamics, nitrogen and oxygen cycles, plankton

ecosystem, impact of shellfish farming, macrophytes.

A coupled biological-hydrodynamic three-dimensional integrating all above cited

lagoon compartments has been developed and allows relevant simulations. These

studies have contributed to a preliminary understanding of the Thau lagoon

biogeochemical cycles (Plus et al., 2006). It is also since 1998 under the influenced of

harmful algae blooms (Alexandrium) with direct impact on shellfish production and

commercialization.

Figure 14. Thau lagoon during a dystrophic episode (malaïgue). In August 1997, nearly one third of

the annual oyster annual production was lost.

Ifremer operates four monitoring networks for the evaluation of the environmental

quality of the Thau lagoon:

• RNO: chemical contaminants,

• REMI: microbiological quality of water and shellfish,

• REPHY: harmfull algae and phycotoxins,

• RSL: eutrophication

Figure 15. REMI control points in the Thau lagoons and typical graph displaying time series of

microbiological level quality in shellfish.

16

3.2. Comparison between COHERENS and MARS3D models

3.2.1. Description of the simulation

In this section, results obtained from an application of both COHERENS (Luyten et al.,

1999) and MARS3D (Lazure and Jégou, 1998; Lazure, 1992) models in the Thau

lagoon are discussed. A technical comparison between these models is presented in D15

(Chapelle et al., 2005). Models have been implemented on the same grid (see figure

16). The horizontal resolution is equal to 100 m in both directions. Ten sigma levels are

used along the vertical.

0 2 4 6 8 10 12 14 16 18

0

2

4

6

8

THAU

(100,60)

(139,46)

(146,34)

(149,49)

(69,25)

m

0.5

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

Figure 16. Grid used for the application of the models to the Thau lagoon. Distance along x and y

axis are given in km. Note that the x axis is on the tilt with respect to the East by

approximately 33° (counted positively in the anti-clockwise direction). The five nodes at

which model results will be compared are also indicated. For each node, its grid indexes

in the x and y directions are given.

The models are run, starting from the sea at rest, for a two months period (01.07.1994 at

00:00 – 01.09.1994 at 00:00). The comparison of the model results is done on the period

mid of July – mid of August. Main model forcing during this period are the tide, the

wind and the heat exchange at the sea surface. Rainfall and freshwater discharges are

negligible during the period of interest. As a consequence, density driven currents can

be considered as negligible and one may assume that the circulation in the lagoon is

mainly driven by the tide and the wind.

The lagoon is connected to the sea through the Sète channels that are very schematically

taken into account in this study (only one channel is used; see figure 16). At the open

sea boundary, the time evolution of the elevation of the free surface due to the tide is

prescribed. The amplitude and phase of 10 tidal constituents are taken into account.

These amplitudes and phases are in fact valid for the port of Marseille. The time series

of the elevation of the free surface due to the tide is presented on figure 17. The tidal

range is of the order of 0.2 m during spring tide and of the order of 0.1 m during neap

tide.

17

0.25

0.3

0.35

0.4

0.45

0.5

0.55

07/16 07/23 07/30 08/06 08/13

Ele

va

tio

n (

m)

Time

Marseille

Figure 17. Time series of the elevation of the free surface due to tide as imposed at the sea

boundary.

Wind data (three hourly values of wind speed and direction) are coming from

observations provided by Météo-France. Time series of these data are presented in

figures 18 and 19.

2

4

6

8

10

12

14

16

07/16 07/23 07/30 08/06 08/13

Win

d s

peed m

s-1

Time

Thau

Figure 18. Time series of wind speed (ms-1).

The wind speed is rarely smaller than 4 ms-1

. The strongest winds (around the 20th

of

July and around the 10th

of August) are observed when the wind is blowing from the

Northwest (the so-called “Tramontane”). The intercomparison of model results deal

with the time series of current speed computed by the two models at the five nodes

shown in figure 16.

18

0

50

100

150

200

250

300

350

400

07/16 07/23 07/30 08/06 08/13

Win

d d

irection

Time

Thau

Figure 19. Time series of the direction of the wind. The direction is that from which the wind is

blowing and it is counted positive clockwise from the North.

It must be noted that in the implementation of the COEHRENS model, no attempt has

been made to take into account, in a way or another, the presence of ‘oyster-beds. In

MASR3D, the horizontal diffusion of momentum is increased in the areas where those

‘oyster-beds are implemented.

3.2.2. Comparison of model results

In a first stage the focus is on the time series of the current speed near the bottom, bur

, at

mid-depth, mur

, and near the surface, sur

. At these levels, one computes, for each model,

a mean value, a minimum value, a maximum value and the standard deviation (σ ) for

each component of the velocity as well as for its modulus. A first inside the agreement

between the two models is gained by looking at the value of the correlation coefficient,

r, and at the value of the root mean squared error (rmse). All results are presented in the

tables below.

Node (69,25)

COHERENS MARS3D Mean Min Max σ Mean Min Max σ r rmse

bu 0.00 -0.05 0.04 0.02 0.00 -0.07 0.06 0.03 0.80 0.02

mu -0.01 -0.12 0.08 0.05 0.00 -0.11 0.09 0.05 0.79 0.03

su -0.01 -0.20 0.15 0.08 0.00 -0.15 0.13 0.07 0.79 0.05

bv 0.00 -0.04 0.05 0.01 0.00 -0.06 0.07 0.02 0.79 0.01

mv 0.00 -0.05 0.04 0.01 0.00 -0.06 0.06 0.02 0.58 0.02

sv 0.00 -0.12 0.07 0.03 0.00 -0.02 0.07 0.01 -0.20 0.03

bur

0.02 0.00 0.05 0.01 0.03 0.00 0.07 0.01 0.56 0.02

mur

0.04 0.00 0.12 0.02 0.05 0.00 0.12 0.02 0.46 0.03

sur

0.08 0.00 0.21 0.04 0.06 0.00 0.15 0.03 0.38 0.04

19

Node (100,60)

COHERENS MARS3D

Mean Min Max σ Mean Min Max σ r rmse

bu -0.01 -0.06 0.05 0.02 0.00 -0.05 0.05 0.02 0.39 0.02

mu -0.01 -0.09 0.07 0.03 0.01 -0.08 0.10 0.03 0.67 0.03

su -0.01 -0.15 0.12 0.05 0.02 -0.10 0.16 0.06 0.73 0.05

bv 0.00 -0.04 0.06 0.01 0.01 -0.03 0.06 0.02 0.44 0.02

mv 0.00 -0.04 0.04 0.01 0.00 -0.03 0.06 0.01 0.07 0.01

sv -0.01 -0.12 0.09 0.03 -0.01 -0.12 0.05 0.02 0.61 0.02

bur

0.02 0.00 0.06 0.01 0.02 0.00 0.06 0.01 0.26 0.01

mur

0.03 0.00 0.09 0.02 0.03 0.00 0.10 0.02 0.15 0.03

sur

0.05 0.00 0.16 0.03 0.06 0.00 0.18 0.04 0.07 0.05

Node (139,46)

COHERENS MARS3D

Mean Min Max σ Mean Min Max σ r rmse

bu 0.00 -0.04 0.06 0.02 -0.01 -0.08 0.05 0.02 0.48 0.03

mu 0.00 -0.05 0.05 0.01 0.01 -0.04 0.08 0.02 0.48 0.02

su 0.00 -0.08 0.06 0.03 0.00 -0.08 0.08 0.03 0.63 0.02

bv 0.01 -0.05 0.05 0.02 0.02 -0.04 0.08 0.02 0.21 0.03

mv 0.00 -0.05 0.07 0.01 -0.01 -0.07 0.04 0.02 0.28 0.02

sv -0.01 -0.12 0.09 0.03 -0.02 -0.16 0.07 0.04 0.63 0.03

bur

0.02 0.00 0.07 0.01 0.03 0.00 0.09 0.02 0.17 0.02

mur

0.02 0.00 0.07 0.01 0.02 0.00 0.08 0.01 0.00 0.02

sur

0.04 0.00 0.12 0.02 0.04 0.00 0.17 0.02 0.30 0.03

Node (146,34)

COHERENS MARS3D

Mean Min Max σ Mean Min Max σ r rmse

bu 0.00 -0.08 0.08 0.03 0.00 -0.07 0.07 0.02 0.69 0.02

mu -0.01 -0.14 0.13 0.05 0.02 -0.09 0.09 0.04 0.80 0.04

su -0.01 -0.22 0.15 0.07 0.02 -0.12 0.13 0.05 0.76 0.06

bv -0.01 -0.11 0.07 0.03 -0.01 -0.09 0.06 0.03 0.76 0.02

mv -0.02 -0.21 0.14 0.06 -0.02 -0.15 0.10 0.05 0.84 0.03

sv -0.03 -0.37 0.24 0.09 -0.03 -0.20 0.15 0.07 0.84 0.05

bur

0.04 0.00 0.13 0.02 0.03 0.00 0.11 0.02 0.64 0.02

mur

0.07 0.01 0.25 0.04 0.06 0.00 0.17 0.04 0.61 0.03

sur

0.10 0.00 0.40 0.06 0.08 0.00 0.22 0.05 0.62 0.05

20

Node (149,49)

COHERENS MARS3D

Mean Min Max σ Mean Min Max σ r rmse

bu 0.01 -0.05 0.05 0.02 0.02 -0.04 0.05 0.02 0.51 0.02

mu 0.01 -0.06 0.07 0.02 0.01 -0.03 0.05 0.01 0.41 0.02

su 0.01 -0.06 0.10 0.03 -0.01 -0.10 0.04 0.02 0.24 0.04

bv 0.00 -0.04 0.04 0.01 0.01 -0.02 0.05 0.01 0.41 0.02

mv 0.00 -0.05 0.04 0.02 0.00 -0.03 0.03 0.01 0.62 0.01

sv -0.01 -0.18 0.10 0.04 -0.01 -0.09 0.06 0.03 0.71 0.03

bur

0.02 0.00 0.06 0.01 0.02 0.00 0.06 0.01 0.27 0.01

mur

0.02 0.00 0.08 0.01 0.02 0.00 0.06 0.01 0.25 0.02

sur

0.04 0.00 0.20 0.03 0.03 0.00 0.11 0.02 0.29 0.03

Time averaged values of the components of the current are most often smaller than 0.01

ms-1

(absolute value). The largest absolute value, 0.03 ms-1

, is found in both model

results at the node (146,34) for the component of the current speed near surface in the

direction perpendicular to the main axis of the lagoon, sv .

At all nodes, the range of variation of the components of the current speed and the range

of variation of its modulus at all levels is consistent in the two sets of model results. The

correlation coefficient is positive almost everywhere. There is only one negative value

( sv node (69,25)). At the node (146,34), the correlation coefficient is above 0.7 for both

components of the currents at all the levels. The root mean squared error is varying

between 0.01 ms-1

and 0.06 ms-1

with a mean value close to 0.03 ms-1

.

In a second stage, we look at vector plots of wind and current near the surface. All

figures are given in Annex 1. The figure at node (146,34) is given in figure 20.

In both models, the current near the surface is highly influenced by the wind forcing. In

a last stage, we decided to look at the evolution in time of the profiles of the modulus of

the current speed. All figures are given in annex 2. Results at node (146,34) are

presented in figure 21. The influence of the wind is visible at all sigma levels.

21

Figure 20. Vector plots of wind speed and current speed near the surface at node (146,34). The top

panel is for the period 15/07/1994 – 25/07/1994. The middle panel is for the period

25/07/1994 – 04/08/1994. The bottom panel is for the period 04/08/1994 – 14/08/1994.

Each panel is divided in three parts: the upper part represents the wind speed, the

middle part represents the current near the surface as computed by COHERENS, the

lower part indicates the current near the surface as computed by MARS3D.

22

THAU (146,34)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

Figure 21. Wind speed and profiles of the modulus of the current at node (146,34). The top panel is

for the period 15/07/1994 – 25/07/1994. The middle panel is for the period

25/07/1994-04/08/1994. The bottom panel is for the period 04/08/1994-14/08/1994. Each

panel is divided in three parts: the upper part represents the wind speed, the middle

part represents the current near the surface as computed by COHERENS, the lower

part indicates the current near the surface as computed by MARS3D.

23

3.3. The Gulf of Gera

The gulf of Gera is a semi-enclosed water body located in the island of Lesvos, Greece

in the Aegean archipelago (figure 22). The surface area of the gulf is approximately 43

km2, and the mean depth of about 10 m. The gulf is connected to the open sea through a

channel, having a width of 200-800 m, length of 6.5 Km and depth ranging from 10 to

30 m. The surrounding area of approximately 200 km2, can be divided into two parts

with differences in geomorphology and land use.

The western part of 170 km2, is characterized by a rather smooth terrain cultivated

mainly with olive trees, the location of five villages with a total population of 7000

people and a rich hydrographic network of small rivers flowing mainly during winter.

The eastern part of the watershed of approximately 30 km2, is covered with olive trees

growing on rather steep terraced slopes.

Figure 22. The gulf of Gera on the Island of Lesvos, Greece.

The water circulation in the gulf is tidally driven and a cyclonic pattern is observed

most of the time of the year. The exchange of water between the gulf and the open sea

shows a fluctuation during the year due the morphological characteristics of the gulf.

During the warm months of the year (April to October), the physical conditions allow

the entrance of oligotrophic water masses from the Aegean Sea into the gulf, whereas

the hydrodynamic regime is reversed during the rest of the year and the renewal time of

the water can be up to 3 months.

The flux of nutrients from non-point sources (agricultural run-off) is considerable,

especially during the winter period, when the contribution to the total inorganic nitrogen

stock (the limiting nutrient in the area) varies between 40 to 60%. The most important

point discharges are untreated domestic wastewater and effluents from the local

industrial activities, especially olive oil processing by-products. The input of nutrients

and organic matter from the surrounding watershed and the low renewal rate result to

the development of eutrophication crises during the year.

Olive tree cultivation is a near monoculture in the area. Fruits and vegetables are also

produced and a small number of greenhouses exist. Fisheries and aquaculture are also

important for the local population. During the last decade, the development of tourism

24

became the top priority for the local population and many tourist resorts have been

constructed, especially on the western part of the coastal zone of the gulf. The

environmental impact of the development of tourism and the sustainability of the

ecosystem are of main concern for the local authorities (Municipalities of Gera,

Evergetoulas and Mytilene), responsible for the management of the area.

The dynamics of the coastal ecosystem and the land-sea interactions have been

extensively studied during the last decade using many quantitative techniques, including

statistical methodologies, simulation modelling and multicriteria analysis. The

application of these techniques aims at the development of an integrated procedure to

support planning and decision making in the coastal area.

3.4. Comparison between COHERENS and POM models 3.4.1. Introduction and forcing

In this subchapter we will describe what has been done in terms of intercomparison

between two different models applied to the Gera Bay. The description of the models as

well as the main features is found in the DITTY deliverable D15.

Simulations have been conducted in order to best compare the results of both models on

the Golf of Gera site.

Same bathymetry (figure 23) has been applied to both Princeton Ocean Model (POM)

that is applied to the Golf of Gera by the modelling team of the Aegean University and

the COHERENS model applied to the same domain by MUMM. The used grid is

151*56 cells of 100 m square. 10 sigma levels were considered for both models.

Figure 23. Bathymetry that has been used by POM Gera and COHERENS. Coordinates given in

grid point.

25

Tidal forcing is applied at the southern boundary of the domain using harmonic

components (see table).

M2 S2 N2 K1 M4 Amplitude (m) 0.1248 0.071 0.0417 0.0055 0.0082 Frequency (cph) 0.0805114 0.0833333 0.0789992 0.0417807 0.1610228 Phase MUMM 128.2 99.7 1.6 - - Phase Aegean (degrees) 132.28 167.94 121.64 189.69 336.72

We have to remark that the phases proposed here (Phase Aegean were not the one

computed and used for the first test at MUMM (figure 24) but were finally used in the

final test run (starting 26 March 1997) for COHERENS in order to have the same

forcing for both models.

Figure 24. Elevation (in m) computed by both models at the point close to the southern boundary,

note the difference in phase that was corrected for the final run. Time is accounted in days.

Both models were using same atmospheric forcing proposed by Aegean University.

For the test run, daily wind was considered. No heat nor salinity fluxes were considered

for this run. Monthly surface fields for temperature and salinity were provided to cover

the two months simulation period (26 March 1997- 22 May 1997). These surface fields

were computed from available data. In addition, southern boundary conditions were

provided with monthly profiles of temperature and salinity. Linear interpolation is used

to generate daily value feed in POM. Since COHERENS uses, as standard setup, heat

and salt fluxes, that part of the forcing was not included. At the time of writing we

received from Aegean three hourly data for atmospheric forcing including wind speed

and direction, air humidity and temperature and cloud cover and we will use these new

data to re-run the test case. Of course due to time delay, this cannot be proposed in this

report.

26

Figure 25. Components of the wind used as atmospheric forcing by COHERENS and POM.

As we did for the Thau lagoon we selected few points on the domain in order to achieve

intercomparison. Due to the specificity of Gera bay, these points (see figure 26) were

situated close to the southern boundary (2-48), inside the narrow channel located

between the open sea and the Bay (16,16) and finally inside the large and shallow bay

(141,24). Other points are located on 58,18 and 77,19.

Figure 26. Location of the intercomparison stations. Coordinates are in grid point, in this graph,

level means depth expressed in m.

27

3.4.2. Description and results of the test run

The test run starts on 26 March 1997 and ends on 22 May 1997. Due to the important

depth that can be found in the channel located between the open sea and the Gera Bay,

the 2D time step as been reduced to 2 sec.

It has been decided to make a test run starting from the simplest hydrodynamic

situation. By starting end of March, stratification is not yet established in the Bay and

initialisation can be from rest with constant value for T and S all along the domain.

Forcing is used as described in the last paragraph.

Because little difference still exist in the setup (exact use of drag formulation, turbulent

closure parameter, etc…) of both models and in the used forcing, we will not present

the results in the same way as we did before for the Thau lagoon test run but in a more

qualitative way.

0 10 20 30 40 50 60−0.3

−0.25

−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.151616−10−u−m

time in day

u c

om

po

ne

nt

of

the

ve

locity in

m/s

mumm

gera

Figure 27. Comparison between models results for the U component of the velocity at the surface.

Intercomparison station 16-16.

Figure 27 presents the intercomparison at station 16-16 located in the channel between

open sea and Gera Bay (figure 26). Same behaviour is observed even if COHERENS

(labelled here as MUMM) produce more intense current. Both models reproduce the

same physics. The V component of the velocity is presented at figure 28 where same

remarks can be made.

28

0 10 20 30 40 50 60−0.8

−0.6

−0.4

−0.2

0

0.2

0.4

0.6

0.81616−10−v−m

time in day

v c

om

ponent of th

e v

elo

city in m

/s

mumm

gera

Figure 28. Comparison between models results for the V component of the velocity at the surface.

Intercomparison station 16-16.

Moreover at time between 38 and 39 days the signal produced by POM change

direction and that is not observed in the COHERENS signal. Figure 25 showing at that

time a pick in the wind, that difference could be explained in the difference both models

have in the way they handle the wind stress.

By comparing the signal produced for the station 141-24 by POM and COHERENS we

can observe at Figure 29 and 30 that the behaviour given by both model is the same

meaning that the same physics is represented even if difference remains.

At the beginning of the simulation V component of the velocity vector shows opposite

sign between POM and COHERENS results but at the end of the simulation same sign

is produced.

In the other hand U component produced by COHERENS presents stronger current than

POM but with similar sign and behaviour.

After 35 days of simulation both U and V component produced by both models show

similar values for the intensity of the current.

To conclude it can be said that the differences observed in results of both models is

probably due the difference still existing in the setup and forcing used. But what is clear

is that same behaviour is observed and using the newly available set of data for the air-

sea interaction and heat and salt fluxes, the intercomparison will be better, just in the

same way as it has been shown for the Thau lagoon exercise at section 3.2.

29

Figure 29. Comparison between models results for the U component of the velocity at the surface.

Intercomparison station 141-24.

Figure 30. Comparison between models results for the V component of the velocity at the surface.

Intercomparison station 141-24.

30

4. BIOGEOCHEMICAL MODELLING

4.1. LOICZ intercomparison

4.1.1. Introduction

The LOICZ biogeochemical model (LOICZ BM) was developed in the framework of

the Land Ocean Interaction in Coastal Zone (LOICZ) project, a core project of the

International Geosphere Biosphere Programme (IGPB) that was funded from 1993 to

2002; actually the LOICZ II project started in 2003 and is funded by IGBP, the

International Human Dimension Programme (IHDP) and other partners for ten years

(http://www.loicz.org ).

The main characteristic of the LOICZ BM is its wide applicability with a minimal data

requirement to a large range of coastal systems. The results of the various applications

can be compared since they are based on a uniform methodology. Basically, the

modeling results are informative about CNP fluxes and related processes.

The LOICZ BM was applied by local researchers to about 200 coastal systems around

the globe and the results are available on the LOICZ Biogeochemical Modelling Node

web page (http://data.ecology.su.se/MNODE).

The LOICZ BM is based on the mass balance of materials: Materials with conservative

behaviour (such as water and salt) are used to estimate the mass movements of water.

Materials with non-conservative behaviour (nutrients) are used to estimate internal

transformations and important ecosystem processes such as net ecosystem metabolism

(NEM) (i.e. the difference between production and respiration) or the difference in the

rates of nitrogen fixation and denitrification (nfix-denitr).

The application of the model is based on the following steps:

1. Water budget: Establish a budget of freshwater inflows such as runoff (VR),

precipitation (VP), groundwater (VG), sewage or other input (VO) and evaporative

outflow (VE). There must be compensating outflow (or inflow) to balance the water

volume in the system: the residual flow (VR).

2. Salt budget: Salt must be conserved in the system. Therefore salt flux not accounted

for the salinities used to describe the freshwater flows in the previous step here above,

must be balanced by mixing flow (VX). VX account for the seawater that replaces a

water volume in the lagoon and can not be calculated with the water budget. If there is

no salinity difference between the system of interest and adjacent systems, or if the

pattern of water exchange is too complex to be described by the combined water and

salt budgets, some more complex form of circulation analysis such as hydrological

models are required. Steps 1 and 2 describe the exchange of water between the system

of interest and adjacent systems by the processes of advection and mixing.

3. Budgets of non-conservative materials: All dissolved materials and in particular

dissolved inorganic phosphorus (DIP) and nitrogen (DIN) exchange between the system

of interest and adjacent systems according to the criteria established in Steps 1 and 2,

above. Deviations of material concentrations from predictions based on these two

previous steps are indicated with ∆ and quantitatively attributed to net non-conservative

reactions or internal transformations of materials in the system. ∆DIP and ∆DIN are

considered as the net difference of the processes that result in a release of nutrients

(source) and the ones that contribute at their storage in the system (sink)

31

4. Stoichiometric relationships among non-conservative budgets: It can often be

assumed that the non-conservative flux of dissolved inorganic phosphorus is an

approximation of net metabolism at the scale of the ecosystem, because there is no gas

phase for phosphorus flux. Nitrogen and carbon both have other major flux pathways

(notably denitrification, nitrogen fixation, gas exchange across the air-sea interface, and

[in some systems] CaCO3 reactions).

The deviation of the fluxes of these materials from expectation based on C:N:P

composition ratios of reactive particles in the system can be assigned to other processes

in a quantitatively reproducible fashion. Non-conservative DIP flux (∆DIP) is assumed

proportional to NEM (primary production – respiration). Mismatch from “Redfield

expectations” for DIP and DIN flux is assumed proportional to (nfix-denitr).

Details on the LOICZ BM and the relative formulation are reported in Gordon et al.

(1996) and in Giordani et al. (2005).

4.1.2. LOICZ BM applications to the DITTY sites

The LOICZ BM was applied to all the 5 DITTY sites with differences based on the sites

characteristics and data availability. The details are indicated in the relative reports in

attachment. When real data were not available, results of the hydrological and

biogeochemical models or other available estimations were used. The model was

applied for the year 2002 for Ria Formosa (RF), for the 2003 for Mar Menor (MM)

(Carreño et al., 2005) and for an average year for Etang de Thau (ET) (Richard and

Aliaume, 2004). For the Sacca di Goro (GO), the model was applied for 1992 (GO92)

(Austoni et al., 2005) and 1997 (GO97) (Viaroli et al., 2001) on annual, seasonal and

monthly (only for GO92) basis. In between of these 2 years, hydraulic engineering

works were conducted to decrease the freshwater inputs and improve the water

exchanges with the sea. In GO92, dissolved organic nitrogen and phosphorus budgets

were investigated. For the Gulf of Gera (GE) (Kavakli et al., 2005), two different

periods were considered: from May to October (1996-97), when the system was

stratified (GE-S) and from November to April (1996-97) when the system was fully

mixed (GE-M). For all sites the one box – one layer model was applied.

4.1.3. Results of the LOICZ BM applications and comparison

Water budgets. The DITTY sites are quite different in size and mean depth (table 8).

GO is the lagoon that has the lower surface area (26 km2) and the lower average depth

(1.5m). The higher surface area was observed at MM (135 km2) and the maximum

mean depth at the GE (12.1 m). Thus, the larger water bodies are MM (6.1 x 108 m

3)

and GE (5.2 x 108 m

3), similar volumes were estimated for RF and ET (3.7 -3,4 x 10

8

m3) and the lower for GO (3.9 x 10

7 m

3).

The main freshwater inputs are from river runoff for GO and ET and from direct

precipitation for MM (Table 8). The runoff discharge is extremely high at GO, about

one order of magnitude higher than in the other systems; this is due to the location of

this lagoon, at the end of big drainage channels in the Po river delta. Direct precipitation

and river runoff have the same importance for RF and GE; for the latter also

groundwater is an important input. Water loads from waste water treatment plants are

important flow in RF, ET and GE. The freshwater inputs at GE are extremely low in the

stratification period (2.6 x 107 m

3 y

-1). Direct evaporation fluxes were estimated with

Hargreaves formulation (Shuttleworth, 1993) at RF and GO, following Linacre (1973)

at ET and measured directly at MM and GE sites. Only at GO direct evaporation is not

quantitatively important; at the other sites, this flux is estimated as a significant water

32

output in comparison to the other freshwater flows. In the MM site, evaporation is

higher than the sum of the freshwater inputs and a net inflow of seawater is request to

maintain the volume of the system. This water flux is called residual flow (VR) and it’s

estimated from evaporation minus the sum of all the freshwater inputs. It is negative

when water flows out from the system and positive in the opposite case. As expected

from the freshwater inputs data, the higher residual fluxes were observed for GO92 and

GO97.

Table 8. Data and results of the water and salt budgeting exercise for the 5 Ditty sites. Flows are

indicated with V: positive values are inputs and negative outputs.

System

Units Ria

Formosa (RF)

Mar Menor (MM)

Etang de Thau (ET)

Sacca di Goro

(GO92)

Sacca di Goro

(GO97)

Gera annual (GE)

year 2002 2003 mean 1992 1997 1996-97

area km2 105 135.24 75 26 26 43

depth m 3.5 4.5 4.5 1.5 1.5 12.1

VQ 106 m

3y

-1 63.8 27.4 110.1 721.2 365.9 12.5

VG 106 m

3y

-1 0.0 5.0 9.4 0.0 0.0 37.1

VP 106 m

3y

-1 55.4 50.6 43.1 14.7 15.5 28.1

VO 106 m

3y

-1 30.9 1.7 2.3 0.0 0.0 30.5

VE 106 m

3y

-1 -87.5 -219.5 -102.7 -10.2 -20.1 -13.6

VR 106 m

3y

-1 -62.6 134.8 -62.2 -725.7 -361.4 -94.6

Ssys PSU 36.32 43.90 35.80 25.00 22.64 38.86

Ssea PSU 36.25 38.35 36.85 28.30 28.25 38.95

VX 106 m

3y

-1 5412* 1034 2150 6466* 1640 24580*

τ d 24.5 190 55.7 2.0 7.1 8.0

n: considered negligible; *calculated from hydrodynamic models: RF2D Hydrodynamics – EcoDynamo for RF, COHERENS for the GO92 and POM for GE

Salt budget. The mean salinities of the DITTY sites are reported in table 8. Values

similar to the standard seawater are observed at RF, ET and GE indicating a strong

marine influence, higher values were observed at MM indicating low water exchanges

with the sea coupled to high evaporation rates. The low values observed at GO indicates

high river loads, in parallel, the low values observed in the adjacent sea are due to the

Po river plume influence.

The salt budgets calculated following the LOICZ biogeochemical model guidelines

were generally used to estimate the exchange flow between the system and the sea (VX).

This was done for MM, ET and for GO97. But for RF, GO92 and GE, 3D

hydrodynamic model estimations were used since they are developed in the Ditty

Project framework and, in these sites, salinity gradients were not wide enough to allow

a good estimation of VX. The models that have been used are: RF2D Hydrodynamics –

EcoDynamo for RF, COHERENS for the GO92 and POM for GE (Chapelle et al.,

2005). The results are reported in table 8. Low VX values were estimated for MM and

ET due to the narrow channels that connect these systems with the sea. High values

were observed at RF due to the tide influence; this is the only tidal system since the

other ones are in the Mediterranean Sea were tide excursion is lower than 1 meter. In

the VX estimation of RF, the water that move back and forth with tides but do not exit

from the system was keep into account. For GO, two different methods were used: the

COHERENS 3D hydrodynamic model for GO92 and the salt budget for GO97. The VX

estimated for these two periods are comparable and are within the range indicated for

GO for the 1989-1998 period reported in Zaldivar et al., 2001. In this report, the wide

changes of the sea opening width, observed in this period, was considered. The sea

33

opening was 1350 m wide in 1992 and 1536 m in 1997 (Simeoni et al., 2000). In

summer 1992-1993, a second sea mouth was opened by cutting the sand barrier among

the lagoon and the sea to improve the water circulation in the more confined lagoonal

area. In the following years, the connection among the sea and the lagoon become

wider. The higher VX observed in 1992 respect to 1997 is consistent with the different

opening width and with the larger salinity difference among the lagoon and the sea

observed in the 1997. The VX estimated for the GE was about 3 times higher than what

observed for RF and GO92 and this can be due to the large section of the gulf-sea

connection.

With the estimation of VR and VX, the theoretical water residence time can be calculated

since these fluxes account for the water renewal of the system. The relationship among

the water residence time and the surface area is shown in figure 31. MM which is the

bigger site has also the higher water residence time (more than 6 months), RF has a

residence time lower than one month due to the effective tidal flushing and GO only of

some days due to its high freshwater inputs and low water volume.

Figure 31. Estimated water residence time (ττττ) in relation to the surface area of the DITTY sites.

Budgets of non-conservative materials: DIN and DIP. The nutrient loads from the

watershed to the Ditty sites were estimated by multiplied the mean concentrations of

DIN and DIP for the various water loads for GO, ET and MM while, for RF and GE, the

activities that took place in the watershed were quantified and multiplied for standard

conversion factors as indicated in the LOICZ biogeochemical node web site:

http://data.ecology.su.se/MNODE/ (San Diego-McGlone et al., 2000). The nutrient

loads of the first 3 lagoons are reported in Table 9. For DIP, the main inputs are from

the river runoff even if for ET, waste water treatment plants and precipitations are also

important sources. For DIN, runoff is the main source except for ET where DIN input

from precipitation is about two times higher. The nutrient loads to RF are indicated in

figure 32: the main inputs are from urban waste for both DIN and DIP while agriculture

and livestock are important sources for DIN and DIP respectively. The nutrient loads

from the GE catchment are separated in point and non point sources (figure 33).

Approximately the non point sources are from agricultural activities and point sources

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150

area (km2)

Ta

u (

d)

MarMenor

RiaFormosa

EtangDuThau

GulfofGeraSaccaDiGoro

34

are from urban areas. For both DIN and DIP, the main inputs are from point sources (93

and 82% respectively).

Table 9. Nutrient loads from watershed and atmosphere estimated for Mar Menor (MM), Etang de

Thau (ET) and Sacca di Goro in 1992 (GO92) and 1997 (GO97). Unit 106 x mol y-1.

MM ET GO92 GO97

VQDIPQ 5.8 0.7 1.1 0.5

VGDIPG 0.0 0.0 0.0 0.0

VPDIPP 0.1 0.4 0.0 0.0

VODIPO 0.2 0.4 0.0 0.0

total DIP 6.1 1.4 1.1 0.5

VQDINQ 135.9 12.5 143.4 31.9

VGDING 8.5 0.3 0.0 0.0

VPDINP 6.1 26.8 1.4 1.5

VODINO 4.6 1.1 0.0 0.0

total DIN 155.1 40.7 144.9 33.4

The total nutrient loads to the Ditty sites were reported in table 9 and figures 32 and 33.

MM receives the higher inputs of both DIN and DIP but is the larger system. In figure

34, the total loads were shown per lagoonal surface area unit as these can be compared

among sites. The higher DIN loads were estimated for GO92 and GO97 (5.6 and 1.3

mol m-2

y-1

respectively); the lower values calculated for GO97 can be due to the

engineer works conducted in 1994 to limit the river discharges and to a general DIN

loads decrease observed in the last decades (Viaroli et al, 2006). Significant DIN inputs

were estimated also for the MM site (1.1 mol m-2

y-1

). The higher DIP loads were

estimated for the MM and GO92 sites (45 and 42 mmol m-2

y-1

respectively), lower

values were estimated for RF and GO97 (32 and 20 mmol m-2

y-1

). The lower DIN and

DIP loads were estimated for the GE with values lower than 50 and 1 mmol m-2

y-1

. The

N:P ratio in the loads can be used to estimate which nutrient is in excess considering as

balanced the Redfield value (16:1). The ratios estimated for the Ditty sites are reported

in figure 35. More or less balanced values were calculated for MM and ET while DIN

loads are dominant at GO and GE where agriculture, that is mainly a source of nitrogen

compounds, has a large diffusion in the catchment. The DIP loads are in excess at RF

probably due to the large urban areas located in the watershed where high DIP amounts

are produced.

35

66%

27%

4% 3%

NH4 p

NO3 p

NH4 np

NO3 np

82%

18%

DIP p

DIP np

DIP load:

1.1 x 104 mol y-1

DIN load:

1.8 x 106 mol y-1

Figure 32. Nutrient loads from the Ria Formosa catchment in the 2002.

Figure 33. Nutrient loads from the Gulf of Gera catchment in 1996-1997. p: point sources; np:

non point sources.

Figure 34. Dissolved inorganic nitrogen and phosphorus (DIN and DIP) loads from the watershed to the DITTY sites. The values for the DIN have to be multiplied for 10.

69%

28%

1%

2%

0%

Urban waste

Livestock

Aquaculture

Non-point agricultural

runoff

Manufacturing

59%

7%

0%

34%

0%

DIN load

16.2 x 106 mol y-1

DIP load

3.33 x 106 mol y-1

69%

28%

1%

2%

0%

Urban waste

Livestock

Aquaculture

Non-point agricultural

runoff

Manufacturing

59%

7%

0%

34%

0%

DIN load

16.2 x 106 mol y-1

DIP load

3.33 x 106 mol y-1

59%

7%

0%

34%

0%

DIN load

16.2 x 106 mol y-1

DIP load

3.33 x 106 mol y-1

0

100

200

300

400

500

600

RF MM ET GO92 GO97 GE

mm

ol m

-2 y

-1

Σ V*DIN (x10) Σ V*DIP

36

Figure 35. Molar DIN:DIP ratio in the DITTY sites loads. Green dotted line indicates the

Redfield value (16).

As consequence of the loads, water renewal and internal transformations, a mean

concentration of nutrient was set up in the water column of the DITTY sites (figure 36).

The highest concentrations of DIN were observed at GO97, higher than GO92 even if

the DIN loads were lower but this was probably compensated by a longer water

retention time. High DIN concentrations were measured also at the MM site while low

concentrations were observed at ET and GE. DIP concentrations lower than 1 mmol m-3

were observed at all sites and lower than 0.15 mmol m-3

in GE-S.

Figure 36. Mean DIN and DIP concentration in the water column of the DITTY sites. DIN values

have to be multiplied for 10.

Since nutrients do not have a conservative behaviour in the system, an estimation of

their internal transformations can be made considering all the significant inputs and

outputs and assuming steady state conditions. The DIP internal transformations,

considered as (sources – sinks) and indicated as ∆DIP in mmol m-2

y-1

, are shown in

figure 37. While RF and MM act as net DIP sinks, the other sites can be considered as

net sources. The highest sink and source are MM and GE with -42 and +45 mmol m-2

y-

0

25

50

75

100

125

150

RF MM ET GO92 GO97 GE

N:P

rati

o i

n t

he l

oad

s

0

1

2

3

4

5

6

7

RF MM ET GO92 GO97 GE

mm

ol m

-3

DINsys (x10)

DIPsys

37

1 respectively. The DIN internal transformations are reported in figure 4.1.8. For that

nutrient, MM and ET are net sinks and RF and GE are net sources. A big difference was

observed among GO92 and GO97: while GO92 is a net DIN sink, GO97 is a net source.

This wide changes are observed also by Zaldívar et al., 2001 and are typical of dynamic

systems affected by intensive blooms of nitrophilous macroalgae as the Ulva spp in GO

(Viaroli et al., 2006).

Figure 37. Internal transformation of DIP (∆DIP= sources – sinks) estimated from the mass

balance budgets in the DITTY sites.

Figure 38. Internal transformation of DIN (∆DIN= sources – sinks) estimated from the mass

balance budgets in the DITTY sites.

A selection of 79 LOICZ sites indicate that reasonable ranges for coastal system for

∆DIP and ∆DIN are -0.22 ÷ 0.16 and -14.8 ÷ 3.4 mol m-2

y-1

respectively (Buddemeier

et al., 2002) and include the values estimated for the DITTY sites.

In figure 39, the relationships among the DIP and DIN loads and the estimated ∆DIN

and ∆DIP are reported. Even if the low number of observations does not allow any

statistical investigations, a trend that indicate a decrement of both ∆DIN and ∆DIP as

the loads increase can be observed. This trend is also reported by Buddemeier et al.

(2002) from the global coastal zone LOICZ dataset.

∆DIP

-50

-40

-30

-20

-10

0

10

20

30

40

50

RF MM ET GO92 GO97 GE

mm

ol m

-2 y

-1

∆DIN

-8

-6

-4

-2

0

2

4

RF MM ET GO92 GO97 GE

mo

l m

-2 y

-1

38

Figure 39. Relationships among the DIP and DIN loads and the estimated ∆DIN and ∆DIP. Units:

mmol m-2 y-1 (left panel) and mol m-2 y-1 (right panel).

Stoichiometric relationships among nonconservative budgets. As described above, the

C:N:P of the reactive particles or the dominant primary producers is considered as the

stoichiometric link among the cycle of these elements in the system. The Redfield ratio

(106:16:1) was used for RF, ET and GE since phytoplankton is assumed to be the main

primary producer compartment. For MM and GO92, the C:N:P ratio was directly

measured in the dominant plant species: 393:13:1 for MM and from 113:13:1 to

644:40:1 for GO92 since monthly estimations were available. For GO97 the ratio of

335:35:1 reported by Atkinson and Smith., 1983 was used. The NEM values were than

calculated from these C:N:P ratios and ∆DIP estimations (NEM= -∆DIP x C:P). The

results are reported in figure 40. NEM represent the net difference between production

and respiration processes (p-r) averaged for the whole lagoon. The production results

largely dominant at MM with NEM values of 16.6 mol m-2

y-1

while respiration

dominate at GO97 with -8.2 mol m-2

y-1

. Production dominates also at RF and GO92

with 2.6 and 1.2 mol m-2

y-1

respectively while respiration dominates at GE (-4.8 mol m-

2 y

-1) and very low value was estimated for ET (-0.2 mol m

-2 y

-1). NEM and ∆DIP were

both positive for GO92 because different monthly C:N:P ratio were used and the annual

budgets are the weighted means of monthly budgets with both positive and negative

∆DIP (Austoni et al., 2005). Anyway, for GO, the NEM values can be affected by the

intense macroalgal blooms and dystrophic crisis that occurred in the lagoon in which

DIP recycling from the sediment can be relevant. Thus, in this lagoon, ∆DIP was

probably not only related to the balance between production and respiration processes

but also depended on water-sediment fluxes, as observed in real and simulated

dystrophic crises (Giordani et al., 1996; Viaroli et al., 1996).

y = -1.57x + 42.95

R2 = 0.70-50

-40

-30

-20

-10

0

10

20

30

40

50

0 10 20 30 40 50

DIP loads

∆∆ ∆∆D

IP

y = -1.38x + 0.98

R2 = 0.83

-8

-6

-4

-2

0

2

4

0 1 2 3 4 5 6

DIN loads

∆∆ ∆∆D

IN

39

Figure 40. Annual Net Ecosystem Metabolism (NEM) estimated for the DITTY sites.

Considering the N:P ratio of the main primary producers and ∆DIP, it is possible to

estimated the ∆DIN expected from production and respiration processes (∆DINexp). The

difference among the expected and observed ∆DIN can be considered as the net

difference among microbial processes as N fixation and denitrification (nfix-denitr).

This parameter is shown in figure 41. In RF and GO97, N fixation prevails over

denitrification while MM, ET, GO92 and GE result net denitrifiers. The concept of

(nfix-denitr) seems not working properly for GO or an important process affecting N

cycle is not considered because (nfix-denitr) values estimated for GO92 are one order of

magnitude higher than the typical values for coastal systems (+0.4 to –1 mol m-2

y-1

).

These values are in agreement with what estimated by Zaldívar et al. (2001). Even if no

measurement of N fixation were made in GO we can consider this process quite slow in

marine system (lower than 0.4 mol m-2

y-1

) and direct measures of denitrification found

values around 0.9 mmol m-2

y-1

with peaks of 12.8 mmol m-2

y-1

in some spots (Bartoli

et al., 2001). Thus the estimated (nfix-denitr) is quite far also from what expected from

direct measurements.

Figure 41. Annual N fixation minus denitrification fluxes [(nfix-denitr)] values estimated for the

DITTY sites.

4.1.4. Conclusions

The LOICZ biogeochemical model allowed the comparison among water and nutrient

fluxes among the DITTY sites. The results are in the range of the LOICZ database

NEM

-10

-5

0

5

10

15

20

RF MM ET GO92 GO97 GE

mo

l m

-2 y

-1

(nfix-denitr)

-8

-6

-4

-2

0

2

RF MM ET GO92 GO97 GE

mo

l m

-2 y

-1

40

values for the global coastal zone except for (nfix-denitr) values in the GO system.

Moreover some trends, emerging on global scale as the dominance of systems that

behave as nutrient sink at high nutrient loads, can be observed also for the DITTY sites.

4.2. Object oriented approach biogeochemical modelling

Over the last few decades, several modelling tools have been developed for the

simulation of hydrodynamic and biogeochemical processes in aquatic ecosystems. Until

late 70’s, coupling hydrodynamic models to biogeochemical models was not common

and today, problems linked to the different scales of interest remain. The time scale of

hydrodynamic phenomena in coastal zone (minutes to hours) is much lower than that of

biogeochemistry (few days). Over the last years, there has been an increasing tendency

to couple hydrodynamic and biogeochemical models in a clear recognition of the

importance of incorporating in one model the feedbacks between physical, chemical and

biological processes. However, different modelling teams tend to adopt different

modelling tools, with the result that benchmarking exercises are sometimes difficult to

achieve in projects involving several institutions. Therefore, the objectives of this work

are to analyse and compare some of the currently available modelling tools, to help

people choose among the diversity of available models, as a function of their particular

needs, and to propose a unified approach to allow modellers to share software code,

based on the object oriented programming potentiality. This approach is based on

having object dynamic link libraries that may be linked to different model shells. Each

object represents different processes and respective variables, e.g. hydrodynamic,

phytoplankton and zooplankton objects. Some simple rules are proposed to link

available objects to programs written in different source languages.

- The object oriented approach (EcoDynamo)

EcoDynamo (Ecological Dynamics Model) is a software application to simulate

physical, biogeochemical and anthropogenic processes in aquatic ecosystems. It is an

object oriented program application, built in C++, with a shell that manages the

graphical user interface, the communications between classes and the output devices,

where the simulation results are saved.

The simulated processes include:

• hydrodynamics of aquatic systems: water elevations, current speeds and

directions;

• thermodynamics: energy balances between water and atmosphere and water

temperature;

• biogeochemical: nutrient and biological species dynamics;

• anthropogenic: e.g. biomass harvesting.

The ecosystem characteristic properties are described in a model database with the

following files: Morphology file - geometric representation of the model and grid

dimensions; Classes file – list of available classes for a particular model, depending on

the processes and variables considered; Variables file – list of variable names and initial

values for each class; Parameters file – list of parameter names and their values; Loads

file – list and location of loads into the model domain. The class hierarchy in

EcoDynamo and files used by different classes are depicted in figure 42.

41

Other system specific...

Hydrodynamic objects Dissolved subtances Phytoplankton Zooplankton Other...

Ecodynclass

EcodynMorphology file

Classes file

•Variables file

•Parameters file

Tidal harmonics

Compiled as DLLs

Figure 42. EcoDynamo general class structure and files required by different classes.

The user can choose between file, chart or table to store the simulation results. These

output formats are compatible with some commercial software (like MatLab®)

products, enabling their later analysis. Different classes simulate different variables and

processes, with proper parameters and process equations. Classes can be selected or

deselected from shell dialogs determining its inclusion or exclusion in each model run.

This application has an interface module that enables communications with other

programs for external control. For example, the simulation runs can be controlled by

commands like start / stop / pause / restart / step simulation. Simulation activity can be

monitored with the help of log files, activated before the simulation run.

The idea of using object oriented programming in ecological modelling goes back at

least to Silvert (1993). Ferreira (1995) developed EcoWin and presented a detailed

analysis of OOP advantages in ecosystem models. Many aspects of EcoDynamo are

very similar to EcoWin. Perhaps, the most important difference is that EcoWin was

originally designed for box models, whereas EcoDynamo was designed for coupled

hydrodynamic-biogeochemical models. In fact, it is possible to use EcoWin (at least the

EcoWin98 version) in this last type of models (cf. Duarte et al., 2003). However,

several changes have to be carried out in the program shell code, mostly related to its

box structure. Typically, EcoWin handles a relatively small number of boxes of any size

and shape, connected as defined by the user. Therefore, one box may have many

connections to a number of other boxes. EcoDynamo domain is defined as a grid,

handling up to many thousands of cells of regular size and shape (at present, it handles

only Cartesian finite differences grids). Connections between cells are rigidly defined

by the matrix grid structure.

- Description of the structure of EcoDynamo objects

EcoDynClass is the base simulation class used by EcoDynamo. All the simulation

classes inherit from this one. It reads model morphology, initialises relevant fields and

implements the default behaviour of the public methods that can be inherited. This class

controls the model time step evolution and, as ancestor class of all the others present in

42

the simulation, it knows how many objects exist in the simulation and their

relationships.

The general outline of the running process is as follows. The shell invokes all active

classes using the Go method. Hydrodynamic class calculates velocity fields and water

elevations, other classes calculate local changes of their variables at each grid cell.

These local changes are partial derivatives that are integrated by the Integrate method.

After this first round of calculations is completed, the hydrodynamic class transports all

“transportable” variables across grid cells. Finally, cell geometry is updated by the

Reinitialise method.

The public methods that all classes inherit and must rewrite are:

Go – invoked in each model time step, responsible for all object calculations (as in

EcoWin, Ferreira (1995))

Integrate – responsible for time integrations within each grid cell calculations (as in

EcoWin, Ferreira (1995))

Update – update one internal variable value, requested by an external object

Inquiry – send to an external object the value of one internal variable

Reinitialize – update velocities, flows and system geometry to the next time step.

EcoDynamo performs the simulation as a cyclic loop. In each cycle:

“Go” is invoked for all the objects

Hydrodynamic object can adjust the model time step

“Integrate” is invoked for all the objects

“Reinitialize” is invoked for the hydrodynamic objects.

The constructors of each class inherit from the EcoDynClass (which knows the system

morphology) and are responsible by the initialisation of the particular variables and

parameters from the database files. Each class type is built as a Dynamic Link Library

(DLL) and can be integrated at run-time by the EcoDynamo application.

- Interfacing code from different modelling software

a/ Mixing code in C, C++ and Fortran with GNU ‘g++’, ‘gcc’ and ‘g77’ compilers: The

easiest way to mix code written in Fortran 77 and C / C++ languages is to use compilers

that are compatible in the object code generated. This could be done with GNU

compilers, namely those belonging to the MinGW project (Minimalist GNU for

Windows [http://www.mingw.org]). According to the project page, it supplies a

collection of freely available and freely distributable Windows specific header files and

import libraries, combined with GNU toolsets, allowing producing native Windows

programs that do not rely on any 3rd

-party C runtime DLLs.

b/ Call by Value / Call by Reference: In Fortran language, arguments are passed by

reference whereas in C and C++ languages, arguments are passed both by value and by

reference. This means that the normal way Fortran subroutines and functions are called

allows the modification of their argument variables inside the subroutine / function

code, while C and C++ use a slightly different syntax to allow the modification of

argument variables. To allow code mixing between Fortran and C / C++ languages all

the shared subroutines / functions / methods should pass arguments by reference.

c/ Splitting Code into Multiple Files:

Normally the program source code is separated into several files:

43

• One per function or subroutine (Fortran)

• One per function types (C)

• One per class (C++)

Each source code is compiled into an object file (a ‘.o’ file in Unix or a ‘.obj’ file in

Windows), and the various object files are linked together into a final single executable.

The advantages of splitting code up this way include:

• The possibility to use different languages for different portions of the program

• The possibility to have different programmers writing different functions

• More efficient compilation, since a change to one source file only requires its

object file to be recompiled and the object files to be re-linked, rather than

recompiling the entire body of code from scratch.

The use of the make utility to automate the build process of the program is a common

practice when the source code exceeds three or four files. Definitely is the best way,

with only one command, to rebuild the necessary object files when one or more changes

are made in the source files. In addition, it enables the name control of the object files,

independently of the operating system in use.

Nowadays, the Integrated Development Environments (IDEs) and the drag-and-drop

facility enable the inclusion or exclusion of source files in the program code very

intuitively, and hide the correspondent changes in the make files – most of the users

don’t even know about its existence and, the others, don’t bother about it, assuming that

it is well done. When the use of different languages in different parts of the program is

intended, it’s advisable to manipulate the make utility directly to control program

building with more detail.

d/ Internal Symbol Names: When the source code is compiled and turned to object files,

the compiler usually change the internal name of the variables, functions or subroutines

by appending / prepending underscores or other symbols. The GNU compilers used in

this study add a single underscore to all the names used by the code. The Fortran

language is case insensitive: the compiler converts all the symbols to lowercase letters.

Additionally, the Fortran compiler appends a single underscore to each symbolic name

or, if the name has yet an underscore, appends a double underscore. For interoperability

between C, C++ and Fortran the interface functions must have lowercase names and use

the C-style interface. This means that all the C++ interface functions must have an

‘extern “C”’ directive. Several examples are listed in Annex 2.

- Definition of a “linking protocol” between COHERENS and EcoDynamo

a/ Architectural choice

To allow Coherens developers to interact with EcoDynamo classes, an architectural

choice must be made:

1. Main program in Fortran - Fortran subroutines and functions, calling the

EcoDynamo classes using special interface functions, or

2. Main program in C++ - definition of an object-oriented wrapper code in C++

that handles the interaction with Coherens, invoking its functions and passing it

relevant data.

The first architecture option seems to be easier to use for Fortran-only programmers.

The idea is to define an interface to EcoDynamo with functions that can be invoked by

Fortran, and manipulate the C++ objects (create, use and destroy them, read their

44

properties from permanent storage and call their methods). This is even more

challengeable because there is no pointer system in Fortran.

A solution can be found using the concept of “logical units” of Fortran: the C++

interface will generate an integer reference number for all objects manipulated by

Fortran. The Fortran code will have to keep a map associating those reference numbers

to real objects. The solution proposed associates the address of the object in memory

with the reference number in Fortran (the 32-bit integer in Fortran has the same size as a

pointer in C).

The compilation phase is straightforward – each source file is compiled with its native

compiler. The linking process is a little bit more complex because to have objects

instantiated when needed and C++ special mechanisms activated, it is advisable to use

the C++ linker. As the main program is written in Fortran, the best way to do this is to

link all the files with the C++ linker, adding Fortran libraries as link options, as pointed

out in previous example 3.

It is important to notice that this can be very system dependant and caution must be

taken before rebuild the executable program in a new platform.

- Coherens using EcoDynamo objects

Singleton interface class For each class from EcoDynamo platform one singleton interface class must be defined.

This C-style interface provides a static method that returns the reference address of the

object instantiated by the constructor when the first call to that object is performed by

the Fortran code. Every time Fortran code wants to use methods (or read / write data)

from that object, the reference must be indicated by Fortran code or, in another way,

must be supplied by the singleton interface method. The singleton interface class must

follow the rules:

1. Definition of one public static method that returns the reference address of the

class.

2. Definition of one block ‘extern “C”’ (C-style interface directive) with all the

functions that can be called from Fortran:

a. The names of the functions must be lowercase and with underscores

appended;

b. All the parameters must be passed by reference.

3. Changes in the original source code (EcoDynamo C++ sources) must be

enclosed by the symbol _PORT_FORTRAN_ to enable compilation in both

projects (EcoDynamo application and EcoDynamo/Coherens program).

The makefile that builds EcoDynamo/Coherens program (compile and link) must

include the following rules:

1. The source directories of EcoDynamo classes must be added to compilation

flags as include directories.

2. The symbol _PORT_FORTRAN_ must be defined in the compilation flags.

3. The Fortran libraries must be added to link command.

The Fortran code must follow the rules:

1. Definition of one integer variable to save the reference address of the class.

2. The first interface function called must be the one that creates the class object.

45

3. Call the interface functions always with the reference variable.

An example for interfacing both programmes is given in Annex 3.

4.3. Phytoplankton modelling approaches Even though, there is a considerable quantity of different approaches in ecological

modelling, the fact that DITTY has concentrated its efforts on modelling Southern

European coastal lagoons has contributed to some extent to the fact that some modules

inside the developed ecological models are quite similar. For example, the sediment

module for Sacca di Goro, Ria Formosa and Mar Menor has been adapted from the

sediment module developed by Chapelle (1995) for Etang de Thau.

Therefore, in order to assess similarities differences between the proposed ecological

modelling approaches and due to the fact that each ecosystem have some differences,

e.g. macroalgal blooms in Sacca di Goro, rooted macrophytes in Thau lagoon, jellyfish

in Mar Menor, etc. we have decided to concentrate on phytoplankton which is a

common element in all ecological models developed.

As already explained in Chapelle et al. (2005) –D15-, there are universal equations that

can be used to determine how material is transferred between variables of an ecosystem

model. A general equation of population growth, which can accommodate most of the

limiting processes in a closed system, has been proposed by Wiegert (1979):

( ) ( ) ( )1 1

m md X jp f p fe X X Xj j kij j ij ij j j j jk jkkdt i k

µ ϕ ρτ τ= − + + −∑ ∑= =

The first sum represents the assimilated ingestion or uptake by species j from all other

modelled species or abiotic sources. The middle term represents losses due to

physiological causes, death or external factors (e.g. grazing) that are not explicitly

included in the model. The last summation represents the predation on species j by other

species. The coefficients are defined as follows: eij is the assimilation efficiency of

species j using resource i; τij is the maximum specific ingestion / uptake rate of species

j; pij is the preference of species j for resource i; fij is the limitation of ingestion / uptake

of resource i by species j; µj is the specific loss rate due to natural or externally imposed

mortality; φj is the specific loss rate due to excretion; ρj is the specific loss rate due to

respiration.

These coefficients may depend on a variety of physiological and behavioural

interactions making them non-linear functions of the species or abiotic sources. The

equations are as not well-defined as the physical equations of motion, because they are

not based on known quantitative laws, as those available in physics. It is common to

simplify most of these coefficients to either constant values, functions of time or space,

or functions of the physical forcing (Taylor, 1993).

Concerning phytoplankton normally there are several process that are modelled, these

are:

- Growth, which is generally expressed as a function of light and nutrient availability

and temperature.

- Mortality which is normally expressed as a linear function of biomass

- Grazing, which for the zooplankton is generally expressed as a function grazing rate,

temperature and zooplankton biomass, whereas for shellfish depends on filtration speed,

temperature, efficiency factor and shellfish biomass.

- Exudation, fraction that goes into DOC

46

- Sedimentation, normally used for diatoms.

In this work we will focus on the Growth expressions used in the various DITTY

ecological models.

The main differences between phytoplankton models resides in the fact that Goro model

has three differential equations to represent phytoplankton dynamics, whereas in all the

other models one differential equation represents the dynamics of one type of

phytoplankton population. In the Goro model, each phytoplankton community is

described by three state variables defined according to their metabolic function: the

monomers (S), the reserve products (R), and the functional and structural

macromolecules (F) (Lancelot et al., 2002; Tusseau et al., 1997). Concerning

phytoplankton communities, Thau considers two types: pico-nanophytoplankton and

microphytoplankton, Goro considers diatoms and flagellates, Gera and Mar Menor

considers only one type of phytoplankton.

4.3.1. Phytoplankton growth models

In phytoplankton growth models, the growth is normally expressed as a function of

maximum growth multiplied by light intensity, nutrients and temperature functions that

limit this maximum growth. The functions limiting this growth may have different

expressions. The first difference is on how these functions are combined. Whereas in

Thau and Mar Menor there is a multiplicative factor, i.e:

)()()(max TfNutfIfgrowthgrowth ⋅⋅=

in Gera and Goro, the factor is the minimum value between all of them:

)}()()(min{max TfNutfIfgrowthgrowth ⋅⋅⋅=

However, in Gera there is no temperature dependence.

Figure 43. Temperature growth limiting function for diatoms and flagellates in the Goro model.

- Temperature dependence:

In Goro model, temperature dependence is expressed as

−−=

2

exp)(width

opt

GT

TTTf

47

with optimal temperatures of 13.5° C and 20° C and width temperatures of 2.5° and 7.5°

C for diatoms and flagellates, respectively. These types of functions will produce a

Gaussian shape, see figure 43, with high values around the optimum temperature and

standard deviation according to Twidth. As can be seen diatoms would have a short

temperature period of high growth than flagellates according to these parameters.

- Nutrient dependence:

Similar expressions are employed in all the programs. For Thau, Gera, Ria Formosa and

Mar Menor, the Wrobleski (1977) formula is used:

NH

NH

NO KNH

NHe

KNO

NONutf

++

+=

+

+Ψ−

−

−+

][

][

][

][)(

4

4][

3

3 4

This function normally gives no limitations until it approaches KNO and KNH, see figure

44.

Figure 44. Nutrient growth limiting function for phytoplankton in the Gera, Mar Menor and

Thau models.

Whereas for Goro also phosphorous and Silica (for diatoms) are considered as limiting

nutrients and nitrates and ammonium are considered as a single term:

++=

POK

PO

NK

NNutf

POtotN

tot ,min)(

+++=

POK

PO

NK

N

SiOK

SiONutf

POtotN

tot

Si

,,min)(

- Light dependence:

Different relationships that have been employed to simulate productivity-irradiance

curves for phytoplankton have been recently summarised by Macedo and Duarte

48

(2006). In this case, Goro, Mar Menor and Thau use the Steele’s equation (Steele, 1962)

for I<Iopt:

−=

optopt I

I

I

IIf 1exp)(

where Iopt is the optimal light intensity. Normally this function is corrected to take into

account the depth as well as shading effects of phytoplankton biomass. For example in

Goro at a given depth, z, I is calculated as:

])][(exp[0 zkChlkII wc +⋅−=

where I0 is the photosynthetically active irradiance at the water surface and [Chl] is the

chlorophyll a concentration.

In Gera this function is calculated following the model proposed by Taylor and Joint

(1990):

zk

KzkI

KI

If

I

I

⋅

⋅−+

+

=

/)exp(1

1

ln

)(

0

0

where KI is the half saturation light intensity for phytoplanktonic growth and k is the

light extinction coefficient, k=kw+kc[Chl], see figure 45.

Figure 45. Light intensity growth limiting function for phytoplankton in the Gera model.

As can be seen from the above mentioned intercomparison, the parameterization even

though the models are quite similar, is different from model to model. This is due to the

fact that the equations that describe the dynamic behaviour of organisms are as not well-

defined as the physical equations of motion, because they are not based on known

quantitative laws, as those available in physics. Therefore, the models normally tend to

be based on empirical observations and the fitting of selected equations. Furthermore,

ecological models have been adapted to the specific characteristics of the system to be

49

modelled which make difficult to assess which of the coexistent models would produce

the better results.

5. CONCLUSIONS

The main objective of this work was to provide a intercomparison between several of

the modelling tools developed, implemented and/or applied within the DITTY project.

Furthermore the LOICZ budgeting methodology has been applied to carry out an

intercomparison between the different sites to complement the one carried out using the

IFREMER classification scheme (Austoni et al., 2004).

From this analysis, it is possible to extract relevant information concerning the

modelling of coastal lagoons. The main conclusions are, which also follow from D15:

- Hydrodynamic models of the watershed are essential for the management of coastal

lagoons. This is even more important for Southern European lagoons where extreme

events may account for a high percentage of nutrient inputs.

- Standard hydrodynamic models have to be tailored for these shallow environments by

adding a sediment module and in some cases a dry/wet scheme.

- Ecological models may be developed by coupling several modules according to the

main characteristics of each lagoon. Afterwards, a calibration of the main parameters is

always necessary. An object-oriented approach with a library of modules and a common

coupling mechanism would be a useful tool to develop a new model in a less resources

consuming way.

In general, a 3D coupled biogeochemical model highly sophisticated and require a

trained operator to run them. Therefore, their introduction in a DSS for management use

is not the adequate option. The models are useful tools to understand the behaviour of

the coastal lagoons for the Scenario analysis. Afterwards, a summary of the main

features can be extracted and incorporated in the DSS in form of 0D model with

different forcing outputs from watershed or in form of extracted information obtained

after running the scenarios.

50

REFERENCES

Atkinson, M.J., Smith, S.V., 1983. C:N:P ratios of benthic marine plants. Limnol.

Oceanogr. 28, 568-574.

Austoni, M., Viaroli, P., Giordani, G. and Zaldivar, J. M., 2004. Intercomparison among

the test sites of the DITTY project using the IFREMER classification scheme for

coastal lagoons. EUR Report n° 212876 EN. EC, JRC.

Austoni M., Giordani G., Castaldelli G., Zaldívar J.M., Marinov D., Viaroli P., 2005.

Sacca di Goro Lagoon. In Giordani G., Viaroli P., Swaney D.P., Murray C.N.,

Zaldívar J.M. and Marshall Crossland J.I.. Nutrient fluxes in transitional zones of

the Italian coast. LOICZ Reports & Studies No. 28, LOICZ, Texel, the Netherlands.

pag. 29-40.

Bacher C, Héral M, Deslous-Paoli and Razet D (1991) Modèle énergétique uniboite de

la croissance des huîtres (Crassostrea gigas) dans le bassin de Marennes-Oléron.

Can. J. Fish. aquat. Sci. 48, 391 – 404.

Bacher, C. 1989. Capacité trophique du bassin de Marennes-Oléron : couplage d’un

modèle de transport particulaire et d’un modèle de croi ssance de l’huître

Crassostrea gigas. Aquat. Living Resour. 2, 199-214.

Baretta J and Ruardij P (eds) (1988) Tidal flat estuaries. Simulation and analysis of the

Ems Estuary. Springer-Verlag, Berlin.

Bartoli M., Castaldelli G., Nizzoli D., Gatti L.G., Viaroli P., 2001. Benthic fluxes of

oxygen, ammonium and nitrate and coupled-uncoupled denitrification rates within

communities of three different primary producer growth forms. In F.M. Faranda, L.

Guglielmo, G. Spezie (eds), Structure and processes in the Mediterranean

ecosystems, Chapter 29: 227-235. Springer Verlag Italia, Milano.

Bax, N. & J.-E Eliassen, 1993. Multispecies analysis in Balsfjord, northern Norway:

Solution and sensitivity analysis of a simple ecosystem model. J. Cons. Int. Explor.

Mer. 47, 175-204.

Bird, S. and R. Hall. 1988. Coupling hydrodynamics to a multiple box water quality

model. Tech. Rep. EL-88-7, Waterways Experiment Station, Corps of Engineers,

Vicksburg, MS.

Blumberg A.F ,Mellor G., 1987 A description of the three dimensional coastal

circulation model. In: Three dimensional coastal model. Editor Heaps,N. AGU.

Brock, T.D., 1981. Calculating solar radiation for ecological studies. Ecological

Modelling 14, 1-9.

Buddemeier, R.W., Smith S.V., Swaney D.P. and Crossland C.J. 2002.The role of the

coastal ocean in the disturbed and undisturbed nutrient and carbon cycles. LOICZ

Reports & Studies No. 24, ii + 83 pages, LOICZ IPO, Texel, The Netherlands.

Carreño M. F., Martínez J., Pardo M.T., Esteve M.A., 2005. Nutrients salt and volume

fluxes in the Mar Menor coastal lagoon in the South East Spain. LOICZ budget.

Report on task 4.2. DITTY Project (EVK3-CT-2002-00084). Development of an

information technology tool for the management of Southern European lagoons

under the influence of river-basin runoff .

Chapelle, A., 1995. A preliminary model of nutrient cycling in sediments of a

Mediterranean lagoon. Ecol. Model. 80, 131-147.

Chapelle A., Duarte P., Esteve M.A., Fiandrino A., Galbiati L., Marinov D., Martinez

J., Norro A., Plus M., Somma F., Tsirtsis G., Zaldívar J.M. 2005. Comparison

between different modelling approaches for coastal lagoons. European Communities

report n. EUR 21817 EN. 80 pp.

51

Duarte, P., Meneses, R., Hawkins, A.J.S., Zhu, M., Fang, J. and J. Grant, 2003.

Mathematical modelling to assess the carrying capacity for multi-species culture

within coastal water. Ecological Modelling, 168,109-143.

Fasham, M. J. R., Ducklow, H. W. & McKelvie, S. M. (1990). A nitrogen-based model

of plankton dynamics in the oceanic mixed layer. Journal of Marine Research 591-

639.

Ferreira, J.G., 1995. EcoWin – an object-oriented ecological model for aquatic

ecosystems. Ecological Modelling 79, 21-34.

Ferreira, J.G., Duarte, P. and B. Ball. 1998. Trophic capacity of Carlingford Lough for

oyster culture – analysis by ecological modelling. Aquatic Ecology 31, 361 – 378.

Fulton, E.A., Smith, A.D.M. and C.R. Johnson, 2004. Effects of spatial resolution on

the performance and interpretation of marine ecosystem models. Ecological

Modelling 176, 27-42.

Gertsev, V.I. and V.V, Gertseva, 2004. Classification of mathematical models in

ecology. Ecological Modelling 178, 329-334.

Giordani G., Bartoli M., Cattadori M., Viaroli P., 1996. Sulphide release from anoxic

sediments in relation to iron availability and organic matter recalcitrance and its

effects on inorganic phosphorus recycling. Hydrobiologia 329, 211-222.

Giordani G., Viaroli P., Austoni M., Zaldivar J.M., 2005. LOICZ Biogeochemical

Model Guidelines. Report on Tasks 4.2 and 5.3 (WP4-WP5). DITTY Project

(EVK3-CT-2002-00084). Development of an information technology tool for the

management of Southern European lagoons under the influence of river-basin

runoff.

Gordon, Jr., D. C., P. R. Boudreau, K. H. Mann, J.-E. Ong, W. L. Silvert, S. V. Smith,

G. Wattayakorn, F. Wulff and T. Yanagi, 1996. LOICZ Biogeochemical Modelling

Guidelines. LOICZ Reports & Studies No 5, 1-96.

Herman, P.M.J., 11193. A set of models to investigate the role of benthic suspension

feeders in estuarine ecosystems. In: Dame, R.F. (ed.) Bivalve filter feeders in

estuarine and coastal ecosystem processes, NATO ASI Series, Series G: Ecological

Sciences, Vol. 33, Springer-Verlag, p. 421-454.

Kavakli Z., Papapetrou P., Spatharis S., Tsirtsis G., 2005. An integrated coastal zone

management approach based on scenario development and LOICZ budget analysis:

application to a coastal ecosystem in the Aegean, Eastern Mediterranean. Report on

task 4.2. DITTY Project (EVK3-CT-2002-00084). Development of an information

technology tool for the management of Southern European lagoons under the

influence of river-basin runoff.

Linacre A. T., 1973. A simple empirical expression for actual evapotranspiration rate.

Agricultural Meteorology, 11(3), 451-452.

Luyten P.J, Jones J.E, Proctor R.,Tabor A.,Tett P., Wild-Allen K. 1999. COHERENS-

A coupled hydrodynamical-Ecological model for regional and shelf seas. Users

documentation. Mumm Report, Management Unit of the Mathematical Models of

the North Sea, 914 pp.

Macedo, M. F. and Duarte, P. 2006. Phytoplankton production modelling in three

marine ecosystems-static versus dynamic approach. Ecol. Model. 190, 299-316.

Phillips N.A, 1957. A coordinate system having some special advantages for numerical

forecasting. Journal of Meteorology 14, 184-185.

Portela, L.I. and R. Neves, 1994. Modelling temperature distribution in the shallow Tejo

estuary. In: Tsakiris and Santos (Editors), Advances in Water Resources

Technology and Management, Balkema, Rotterdam, pp. 457 - 463.

52

Raillard, O. and A. Ménesguen. 1994. An ecosystem model for the estimating the

carrying capacity of a macrotidal shellfish system. Mar. Ecol. Prog. Ser. 115, 117–

130.

Richard A., Aliaume C., 2004. Modèle biogéochimique LOICZ appliqué à la Lagune de

Thau. Rapport de stage de 3ième année. Polytech’ Montpellier 15 février - 15 juin

2004

San Diego-McGlone M.L., Smith S.V. and Nicolas V. 2000. Stoichiometric

interpretations of C:N:P ratios in organic waste materials. Marine Pollution Bulletin.

40, 325-330.

Shin, P.K.S. & R.S.S. Wu, 2003. Estimating the environmental carrying capacity for

sustainable fish culture. In: Yu, H. and Bermas, N. Determining environmental

Carrying capacity of coastal and marine areas: Progress, constraints, and future

options. PEMSEA Workshop Proceedings No. 11, Global Environmental

Facility/United Nations Development Programme/ International Maritime

Organisation Regional Programme on Building Partnerships in Environmental

Management for the Seas of East Asia (PEMSEA), Quezon City, Philippines, p. 85-

97.

Shuttleworth, W.J. 1993. Evaporation, Ch.4, In D.R. Maidment (ed.), Handbook of

Hydrology, Mcgraw-Hill, various pagings.

Silvert, W., 1993. Object-oriented ecosystem modelling. Ecological Modelling 68, 91-

118.

Simeoni, U., Fontolan, G. and Ciavola, P., 2000, Morfodinamica delle bocche lagunari

della Sacca di Goro, Studi Costieri 2, 123-138.

Steele, J.H., 1962. Environmental control of photosynthesis in the sea. Limmol.

Oceanogr. 7, 137-150.

Taylor, A.H., 1993. Modelling climatic interactions of the marine biota. In: Willebrand,

J., Anderson, D.L.T. (eds.) Modelling oceanic climate interactions, NATO ASI

Series, Series I: Global Environmental Change, Vol. I, Springer-Verlag, p. 373-413.

Taylor, A.H. and Joint, I.,1990. A steady-state analysis of the microbial loop in

stratified systems. Mar. Ecol. Prog. Ser. 59, 1-17.

van der Tol, M.W.M. & H. Scholten, 1998. A model analysis on the effects of

decreasing nutrient loads on the biomass of benthic suspension feeders in the

Oosterschelde ecosystem (SW Netherlands). Aquatic Ecology, 31 (4), 395-408.

Viaroli P., Bartoli M., Bondavalli C., Christian R.R., Giordani G., Naldi M., 1996 (1).

Macrophyte communities and their impact on benthic fluxes of oxygen, sulphide

and nutrients in shallow eutrophic environments. Hydrobiologia 329, 105-119.

Viaroli P., Giordani G. , Bartoli M., Naldi M., Azioni R., Zizzoli D., Ferrari I., Zaldívar

J. M., Bencivelli, S., Castaldelli G., Fano E. A., 2006. The Sacca di Goro and an

arm of the Po river. The handbook of Environmental Chemistry (Ed. in Chief: O.

Hutzinger) Volume 5. Water pollution: estuaries. Volume editor: P.J. Wangersky

Springer-Verlag, Berlin. 197-232.

Viaroli P., Giordani G., Cattaneo E., Zaldìvar J.M., Murray C.N., 2001. Sacca di Goro

Lagoon. In: Coastal and estuarine systems of the Mediterranean and Black Sea

regions: carbon, nitrogen and phosphorus fluxes. Edited by: V. Dupra, S.V. Smith,

J.I. Marshall Crossland and C.J. Crossland. LOICZ (Land-Ocean Interaction in

Coastal Zone) Reports & Studies No. 19. pag.36-45

Wiegert, R.G., 1979. Population models: experimental tools for the analysis of

ecosystems. In: Horn D.J., Mitchell, R., Stairs, G.R. (eds.) Proceeding of

colloquium on analysis of ecosystems, Ohio State University Press, p. 239-275.

53

Wrobelski, J., 1977. A model of phytoplankton plume formation during Oregon

upwelling. J. Mar. Res. 35, 357-394.

Yanagi, T., 2003. Environmental carrying capacity of an oyster culture in a bay: A case

study from Japan. In: Yu, H. and Bermas, N. Determining environmental Carrying

capacity of coastal and marine areas: Progress, constraints, and future pptions.

PEMSEA Workshop Proceedings No. 11, Global Environmental Facility/United

Nations Development Programme/ International Maritime Organisation Regional

Programme on Building Partnerships in Environmental Management for the Seas of

East Asia (PEMSEA), Quezon City, Philippines, p. 78-84.

Zaldivar J.M., Cattaneo E., Viaroli P., Giordani G., 2001. A Biogeochemical Model of a

Mediterranean Lagoon: (Sacca di Goro (I) 1989-1998). European Communities,

report n. EUR 19925 EN. 64pp.

54

APPENDIX 1: RESULTS OF THE COMPARISON BETWEEN

COHERENS AND MARS3D MODELS

Vector plots of wind speed and current speed near the surface

55

56

57

58

59

60

Time evolution of the profiles of the modulus of the current.

61

THAU (69,25)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

62

THAU (100,60)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

63

THAU (139,46)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

64

THAU (146,34)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

65

THAU (149,49)

0.20.40.60.8

0.20.40.60.8

-10.

10.

15/07 17/07 19/07 21/07 23/07 25/07

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

25/07 27/07 29/07 31/07 02/08 04/08

if

mu

wind

0.20.40.60.8

0.20.40.60.8

-10.

10.

04/08 06/08 08/08 10/08 12/08 14/08

if

mu

wind

m/s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

66

APPENDIX 2. EXAMPLES OF INTERFACING FORTRAN, C AND C++ LANGUAGES The next examples will help to understand what is necessary to do in the source files to

interface conveniently the three languages.

All the examples use three subroutines: one defined in C++ language (cppfunc),

another defined in C language (cfunc) and the last one defined in Fortran language

(ffunc). Each subroutine accepts one variable as argument and changes its value inside

the subroutine.

The main program defines one variable (x) and calls previous subroutines to change that

variable’s value. Each example has the main program written in a different language.

Example 1: Main program in C with subroutines in C, C++ and Fortran C requires the subroutines “call by reference” syntax to make the changes in the

variable persistent.

The invocation of the Fortran function by the main program is made with ‘ffunc_’.

The file with the “cppfunc” function must define its prototype enclosed by the

‘extern “C”’ directive indicating that it is called from a C-style interface. The source

files are:

File: cprog.c (main program)

#include <stdio.h>

int main()

{

float x;

x = 1.0;

printf(“Before running C function: x=%f \n”, x);

cfunc(&x);

printf(“AFTER running C function: x=%f \n\n”, x);

cppfunc(&x);

printf(“AFTER running C++ function: x=%f \n\n”, x);

ffunc_(&x);

printf(“AFTER running FORTRAN function: x=%f \n\n”, x);

return 0;

}

File: cfunc.c

void cfunc(float *a)

{

*a++;

}

File: cppfunc.cpp

extern “C” {

void cppfunc(float *a);

}

void cppfunc(float *a)

{

*a += 2;

}

67

File: ffunc.f

SUBROUTINE FFUNC (A)

A = A + 3

END

To compile and link the program the easiest way is to write the file for the make utility: File: makefile

cprog: cprog.o cfunc.o cppfunc.o ffunc.o

gcc –o cprog.exe cprog.o cfunc.o cppfunc.o ffunc.o

cprog.o: cprog.c

gcc –c cprog.c –o cprog.o

cfunc.o: cfunc.c

gcc –c cfunc.c –o cfunc.o

cppfunc.o: cppfunc.cpp

g++ –c cppfunc.cpp –o cppfunc.o

ffunc.o: ffunc.f

g77 –c ffunc.f –o ffunc.f

Each program file is compiled into an object file using the appropriate compiler and the

–c flag. The linker invoked is the default C linker. It’s assumed that all the files are in

the same directory.

To compile the program simply invoke make cprog. After the generation of the object

and executable files run the program with cprog. The results are:

Before running C function: x=1.000000

AFTER running C function: x=2.000000

AFTER running C++ function: x=4.000000

AFTER running FORTRAN function: x=7.000000

Example 2: Main program in C++ with subroutines in C, C++ and Fortran When the main program is written in C++ it must specify to the C++ compiler that the

C and Fortran subroutines will be called with a C-style interface, and also that the

Fortran compiler will append one underscore to the Fortran function name.

On the other hand, the “cppfunc” function could be called in a native C++ way.

The new source files are:

File: cppprog.cpp (main program)

#include <iostream.h>

extern “C” {

void ffunc_(float *a);

void cfunc(float *a);

}

void cppfunc(float *a);

68

int main(void)

{

float x;

x = 1.0;

cout << “Before running C function: x=” << x << endl;

cfunc(&x);

cout << “AFTER running C function: x=” << x << endl;

cppfunc(&x);

cout << “AFTER running C++ function: x=” << x << endl << endl;

ffunc_(&x);

cout << “AFTER running FORTRAN function: x=” << x << endl << endl;

return 0;

}

File: cppfunc2.cpp

void cppfunc(float *a)

{

*a += 2;

}

To compile and link the new program the file for the make utility is changed, adding

the new files to compile and link: File: makefile

cppprog: cppprog.o cfunc.o cppfunc2.o ffunc.o

g++ –o cppprog.exe cppprog.o cfunc.o cppfunc2.o ffunc.o

cppprog.o: cppprog.c

g++ –c cppprog.c –o cppprog.o

cppfunc2.o: cppfunc2.cpp

g++ –c cppfunc2.cpp –o cppfunc2.o

The linker now used is the default C++ linker.

To compile the program simply invoke make cppprog. After the generation of the

object and executable files run the program with cppprog. The results are:

Before running C function: x=1

AFTER running C function: x=2

AFTER running C++ function: x=4

AFTER running FORTRAN function: x=7

Example 3: Main program in Fortran with subroutines in C, C++ and Fortran In the main Fortran program the C and C++ subroutines are called without any

underscore but as the compiler will append one to the function name internally, in both

C and C++ files the prototype must have an underscore after the function name.

Also the ‘extern “C”’ directive must be used in the C++ file, indicating that the C++

function is called with a C-style interface.

The new source files are:

69

File: fprog.f (main program)

PROGRAM FPROG

REAL X

X = 1.0

WRITE (*,*) 'Before running C function: x=',X

CALL CFUNC (X)

WRITE (*,*) 'AFTER running C function: x=',X

WRITE(*,*) ' '

CALL CPPFUNC (X)

WRITE (*,*) 'AFTER running C++ function: x=',X

WRITE(*,*) ' '

CALL FFUNC (X)

WRITE(*,*)'AFTER running FORTRAN function: x=',X

STOP

END

File: cfunc3.c

void cfunc_(float *a)

{

*a += 1;

}

File: cppfunc3.cpp

extern "C" {

void cppfunc_(float *a);

}

void cppfunc_(float *a)

{

*a += 2;

}

To compile and link the new program the file for the make utility is changed, adding

the new files to compile and link:

File: makefile

fprog: fprog.o cfunc3.o cppfunc3.o ffunc.o

g++ -o fprog.exe fprog.o cfunc3.o cppfunc3.o ffunc.o -lfrtbegin -lg2c

fprog.o: fprog.f

g77 -c fprog.f -o fprog.o

cfunc3.o: cfunc3.c

gcc -c cfunc3.c -o cfunc3.o

cppfunc3.o: cppfunc3.cpp

g++ -c cppfunc3.cpp -o cppfunc3.o

70

The best way to link this program is with the default C++ linker with 2 special libraries

to treat the Fortran symbols (-lfrtbegin and –lg2c). 1

To compile the program simply invoke make fprog. After the generation of the object

and executable files run the program with fprog. The results are:

Before running C function: x= 1.

AFTER running C function: x= 2.

AFTER running C++ function: x= 4.

AFTER running FORTRAN function: x= 7.

1 The program could be generated with the default Fortran compiler (g77) but, in this case, more libraries

must be included and the result command will be more complicated.

71

APPENDIX 3: EXAMPLE OF COHERENS USING ECODYNAMO OBJECTS WITH A LIGHT CLASS In EcoDynamo there is a class to compute light intensity at sea level, as a function of

cloud cover, latitude, date and time, using standard formulations described in Brock

(1981) and Portela and Neves (1994). Submarine light intensity is computed from the

Lambert-Beer law as a function of depth and a water light extinction coefficient. Results

from this object are used by other classes, to calculate the water heat budget and

photosynthetic rates. The Light Class was chosen to exemplify the usage of EcoDynamo

classes from the Coherens.

Header file in C++ code To provide an interface to the Light class (TLight symbol), the header file of the class

must include the following major changes:

1. Inside the class definition, add one public static method that returns the

reference address of the class:

public:

#ifdef _PORT_FORTRAN_

static TLight* getLight(TLight* plight);

#endif

2. Outside the class definition, add one ‘extern “C”’ with all the functions that

can be called from Fortran. These functions must reflect all possible interactions

between Coherens code and EcoDynamo class. As an example:

/* Functions that can be called from Fortran */

#ifdef _PORT_FORTRAN_

extern "C" {

void light_(int* plight, int* nc, int* nr, int* nz,

float* latitude, float* kvalue, float* depth,

float* sigma, float* cloudcover, float* airtemperat);

void light_go__(int* plight, float* curtime, float* julianday);

void light_getvalues__(int* plight, int* ic, int* ir, int* iz,

float* totallight, float* parlight,

float* parhorizontallight, float* hoursofsun,

float* horizontallight, float* noonpar,

float* photicdepth, float* subsurfacelight,

float* parsubsurfacelight, float* atmosphericir);

}

#endif

Source file in C++ code The source files of Light class must implement the methods referred in the header file.

As an example:

#ifdef _PORT_FORTRAN_

/*

* Singleton provider - TLight class method

*/

TLight* TLight::getLight(TLight* plight)

{

TLight* Plight = plight;

if (plight == 0)

PLight = new TLight();

return PLight;

}

void light_(int* plight, int* nc, int* nr, int* nz,

72

float* latitude, float* kvalue, float* depth,

float* sigma, float* cloudcover, float* airtemperat)

{

TLight* ptr;

ptr = TLight::getLight((TLight*) *plight);

*plight = (int)ptr;

ptr->SetNumberOfColumns(*nc);

ptr->SetNumberOfRows(*nr);

ptr->SetNumberOfLayers(*nz);

ptr->SetLatitude(*latitude);

ptr->SetKValue(*kvalue);

ptr->SetDepth(*depth);

ptr->SetLayers(*sigma);

ptr->SetCloudCover(*cloudcover);

ptr->SetAirTemperature(*airtemperat);

}

void light_go__(int* plight, float* curtime, float* julianday)

{

TLight* ptr = (TLight*) *plight;

int jd = *julianday;

if (*plight == 0)

return;

ptr->SetCurrentTime(*curtime);

ptr->SetJulianDay(jd);

ptr->Go();

}

void light_getvalues__(int* plight, int* ic, int* ir, int* iz,

float* totallight, float* parlight,

float* parhorizontallight, float* hoursofsun,

float* horizontallight, float* noonpar,

float* photicdepth, float* subsurfacelight,

float* parsubsurfacelight, float* atmosphericir)

{

TLight* ptr = (TLight*) *plight;

char* classname;

int boxNumber;

double Value;

char MyParameter[65];

if (*plight == 0)

return;

classname = ptr->GetEcoDynClassName();

/*

* the Fortran arrays are indexed by layer, line and column

*/

boxNumber = (*iz - 1)

+ ptr->GetNumberOfLayers() * (*ir – 1)

+ (ptr->GetNumberOfLines() * ptr->GetNumberOfLayers())

* (*ic – 1);

strcpy(MyParameter, "Total surface irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*totallight = Value;

strcpy(MyParameter, "PAR surface irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*parlight = Value;

strcpy(MyParameter, "Mean horizontal water PAR irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

73

*parhorizontallight = Value;

strcpy(MyParameter, "Daylight hours");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*hoursofsun = Value;

strcpy(MyParameter, "Mean horizontal water irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*horizontallight = Value;

strcpy(MyParameter, "Noon surface PAR");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*noonpar = Value;

strcpy(MyParameter, "Photic depth");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*photicdepth = Value;

strcpy(MyParameter, "Sub-surface irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*subsurfacelight = Value;

strcpy(MyParameter, "Sub-surface PAR irradiance");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*parsubsurfacelight = Value;

strcpy(MyParameter, "Atmospheric IR");

ptr->Inquiry(classname, Value, boxNumber, MyParameter, 0);

*atmosphericir = Value;

}

#endif

Invoking C++ code from Fortran To use the Light class from Coherens code it is necessary:

1. Declare one integer to store the light class reference in memory and invoke the

function that build the Light object2:

INTEGER LIGHTOBJ

C build Light object - called first time only

CALL LIGHT(LIGHTOBJ, NC, NR, NZ, DLAT, KVALUE,

H2ATC, GZ0, CLOUD2, SAT2)

2. At each time step the Light object must update their internal values (“Go”

method) to be used by Coherens:

C run the Light object

CALL LIGHT_GO(LIGHTOBJ, HOUR, IDATE)

C get values from Light object

DO 100 I=1,NC

DO 100 J=1,NR

IF (NWD(J,I).EQ.1) THEN

DO 101 K=NZ,1,-1

2 The LIGHTOBJ variable must have the value 0 (zero) when the program starts.

74

LIGHT_GETVALUES(LIGHTOBJ, I, J, K, TLIGHT, PLIGHT, PHLIGHT,

HSUN, HLIGHT, NOONPAR, PDEPTH, SSLIGHT,

PSSLIGHT, ATMIR)

C do what you want with the values

C code must be included here

101 CONTINUE

ENDIF

100 CONTINUE

The reference of the Light object (in this example) can be used by Fortran to pass it to a

new C++ object (for instance, the WaterTemperature) and put the C++ objects related

one to the other:

INTEGER WATEROBJ

C build WaterTemperature object with Light reference

CALL WATERT(WATEROBJ, LIGHTOBJ, NC, NR, NZ, H2ATC, GZ0,

SAT2, S, RO, WINDU2, WINDV2, T, HUM2)

More than that, several objects can be instantiated only in one call from Fortran (all

objects needed to the simulation). One interface function can deal with that, accepting

all the mandatory parameters and references, and mixing the relationships with the

classes internally.

Makefile In the makefile the new C++ sources must be appended to the main program with

compilation flags and link options changed accordingly. For instance:

## makefile adapted to generate Coherens program

# with TLight C++ object class manipulation

#

SRC_HDR = C:/DITTY/EcoDyn_V6

SRC_EDC = C:/DITTY/EcoDyn_V6/EcoClass

SRC_LIT = C:/DITTY/EcoDyn_V6/Liteobjt

SRC_ECODYN = EcoDyn_sources

CFLAGS = -D_PORT_FORTRAN_ -I$(SRC_HDR) -I$(SRC_EDC) -I$(SRC_LIT)

FC = g77 -c

CPPC = g++ -c $(CFLAGS)

LINK32 = g++ -v -o "$@"

LIBS = -lfrtbegin -lg2c

(…)

OFILES3 = testlight.o LiteObjt.o EcoClass.o

OFILESMAIN = $(OFILES1) $(OFILES2) $(OFILES3)

MAINCOM = coherens.exe

(…)

## creating executable code

# main program

75

$(MAINCOM): mainprog.o $(OFILESMAIN)

$(LINK32) mainprog.o $(OFILESMAIN) $(LIBS)

mainprog.o: $(IFILES) mainprog.f

$(FC) mainprog.f -o $@

(…)

testlight.o: $(SRC_ECODYN)/testlight.inc $(SRC_ECODYN)/testlight.f

$(FC) $(SRC_ECODYN)/testlight.f -o $@

LiteObjt.o: $(SRC_LIT)/LiteObjt.cpp $(SRC_LIT)/LiteObjt.h \

$(SRC_EDC)/EcoDynClass.h $(SRC_HDR)/ecodyn.rh

$(CPPC) $(SRC_LIT)/LiteObjt.cpp -o $@

EcoClass.o: $(SRC_EDC)/EcoDynClass.cpp $(SRC_EDC)/EcoDynClass.h

$(CPPC) $(SRC_EDC)/EcoDynClass.cpp -o $@

European Commission

EUR 22216 EN – DG Joint Research Centre, Institute for Environment and Sustainability Luxembourg: Office for Official Publications of the European Communities 2006– 82 pp. – 21 x 29.7 cm Scientific and Technical Research series

Abstract

To promote reliable, real-time management of coastal lagoons, increasingly sophisticated numerical models have been developed within the DITTY project. While models are diverse in design and scope, i.e. watershed, fluid-dynamics, biogeochemical, all have the same fundamental goal, i.e. to account realistically for the processes that drive the dynamic behaviour in coastal lagoons so that their status may ultimately be predicted and the effects of mitigation actions be properly evaluated, resulting on a series of good management practices that increase the sustainability of these fragile ecosystems. The intent of the Intercomparison analysis work package is to establish a standard set of input parameters and numerical “experiments” to be performed by various existing models so that independent results could be meaningfully compared and evaluated, having in mind the diversity of approach and systems (watershed, lagoon, adjacent coastal area). Furthermore, though comparison, conceptual weakness could be identified and targeted for further exploration by the DITTY partners as a whole. In this report (D16), that follows D15 in which we described summarised and analysed the employed models in the DITTY community, we have carried out several intercomparison exercises taking into account the diversity of the model employed.

The mission of the JRC is to provide customer-driven scientific and technical support for the conception, development, implementation and monitoring of EU policies. As a service of the European Commission, the JRC functions as a reference centre of science and technology for the Union. Close to the policy-making process, it serves the common interest of the Member States, while being independent of special interests, whether private or national.