Observing system and forecasting

Observing Systems and Forecasting

A Site for the Observation of a Highly-EnergeticCoastal Marine System: the Straits of Messina

A. Bergamasco1, E. Crisafi1, F. Decembrini1, E. Messina1, E. Tamiro1, G.Giunta21, Institute for Coastal Marine Environment, CNR, Messina, Italy2, Horcynus Orca Foundation, Messina, [email protected]

Abstract

The use of a platform of opportunity has been experimented both as a scien-tific observatory in a peculiar coastal system and to support the development of newtechnologies for energy production from marine currents. During 2008, CNR-IAMCMessina, in cooperation with the Horcynus Orca Foundation, has installed onboardthe ENERMAR-Kobold platform, deployed in the Straits of Messina, an environ-mental and engineering monitoring kit that includes an automatic system for themeasurement of marine currents, water temperature and meteorological parametersand an underwater camera. Engineering parameters are simultaneously measured tocharacterize the functioning of the turbine and the pitch/roll of the platform. All thedata are collected by the acquisition system and stored in a relational database. Awireless link through a web server allows the real-time access and data visualization.The marine current intensity distribution proved to be fitted through the Weibull dis-tribution so allowing the comparison of the energy production potential of differentsites. CNR-IAMC is promoting the inclusion of the ENERMAR-Kobold structurein the research national network of installations currently in operation. The per-spective of an increasing energy demand from renewable sources makes it possiblethe forthcoming presence of a global network of platforms hosting state-of-the-artinstruments to observe the marine environment.

1 Introduction

Coastal environments are very complexmarine ecosystems. The Straits of Messinais a peculiar case, characterized by highbiological productivity and diversity anda relevant spatial and temporal variabilityof the natural physical and biogeochemi-cal processes. The long-term systematicobservation of such environments and theirinternal equilibria through innovative tech-nologies such as automatic measurementsystems is today the first step towards a

balanced development of the managementpolicies of the area. In early 2000s, the re-search activity of the Istituto SperimentaleTalassografico of the National ResearchCouncil (today CNR-IAMC Messina) wassupported by the National project Real-izzazione ed attivazione di una rete inte-grata di piattaforme costiere e mezzo mo-bile attrezzati per sistemi avanzati di mon-itoraggio delle acque (Cluster 10-SAM),funded by the Italian Research Ministry.In this context a platform was deployed inthe Straits of Messina to measure meteo-


rological parameters (air temperature, at-mospheric pressure, solar radiation, windspeed and direction) and physico-chemicaland trophic characteristics of the water col-umn at different depths up to 25m (temper-ature, salinity, oxygen, fluorescence, tur-bidity, ammonium, nitrites, nitrates, or-thophosphates). The data were collectedautomatically and forwarded in quasi real-time through SMS messages via GSMnetwork to the institute where they weremerged with the data coming from peri-odic coastal surveys and stored in SAM-BA the common database of the projectSAM accessible via internet. At the endof the decade, the process of transferringthe developed technology and the exper-tise acquired during the research activitieswas initiated by considering the use on-board platforms of opportunity, in or-der to combine the continuation of the de-velopment and an immediate application.The presence of a platform deployed inthe Straits of Messina by a private com-pany to study the potential of energy pro-duction from marine currents triggered thisevolution. The Straits of Messina sep-arates Sicily from the Italian peninsulaand has the characteristic inverted-funnelshape, with a total length of about 40 kmin the north-south direction and a variablewest-to-east width ranging from 2.8 kmnear the Tyrrhenian edge to 25 km at the Io-nian open boundary. The narrowest cross-section (0.3 Km2) has a depth of 80 mand coincides with the presence of a sillwhich divides the area into a northern NE-to-SW oriented sector and a southern N-to-S one, reaching about 200 m and 1200m of depth respectively. From an oceano-graphic point of view the Straits of Messinaexhibits very strong tidal currents (max.5 knots). It is occupied by different wa-ter masses, the Atlantic Water (AW) at the

surface and the deeper Levantine Interme-diate Water (LIW), whose mixing by up-welling currents produces a new body oftypical water [1]. This enrichment condi-tion results in a stimulating effect on theliving component enhancing the biologi-cal richness. The Straits of Messina con-stitutes a direct route of communicationbetween the eastern and western Mediter-ranean basins, as indicated by the cross-ing of swordfishes, tuna fishes, cetaceansand turtles. The climate in the region istemperate-humid with an average tempera-ture above 22 C and rainfall concentratedin the cold period (fall-winter). The aver-age annual temperature is around 18.5 C(1916 to 1999, CV= 0.03, [2]), and the an-nual rain precipitation is about 1000 mm.Objectives of the present paper is to de-scribe the laboratory for the observation ofthe Straits of Messina that has been imple-mented onboard the ENERMAR-Koboldplatform. The main aspects of the researchwork performed in the last years are illus-trated and an overview of the most relevantfindings is presented.

2 The hosting platformand the measurementsystem

The ENERMAR-Kobold platform is a pro-totype built to extract energy from the ma-rine current flow by utilizing a vertical-axisturbine [3]. It mounts a 3-bladed rotor be-low a round-shaped hull of 10 m diame-ter (see Figure 1). The platform is mooredin the Straits of Messina, 200 m offshorein front of the village of Ganzirri (15 kmnorth of the city) where the water depth isabout 25 m. The plant has been perma-nently connected to the Italian power grid

2188

Marine research at CNR

Figure 1: The ENERMAR-Kobold platform in the Straits of Messina: a prototype builtto extract energy from the marine current flow (patented in 1998 and owned by Ponte diArchimede International).

in 2008. The energy is produced by the sys-tem at any angular velocity, depending onlyon the current speed. The prototype hasbeen deployed to study the performancesof the turbine in relation to the environmen-tal conditions and evaluate the potential ofthe production site. To these scopes, during2008, CNR-IAMC Messina, in cooperationwith the Horcynus Orca Foundation, hasinstalled onboard the ENERMAR-Koboldplatform, an environmental and engineer-ing monitoring system to provide the inputdata to the turbine controller (e.g. the ac-tual values of current velocity, pitch androll of the hull, etc.) and assess the envi-ronmental performances of the platform. Awireless link ensures the remote control ofthe turbine operations and the real-time ac-cess and visualization of the environmentaldata through a web server. The systems ar-

chitecture includes the following main sixblocks:Turbine controller. This control sys-tem measures the most important electro-mechanical parameters during the tur-bine functioning: turbine speed, genera-tor torque and generator power. The con-troller is an industrial PC-type acting onthe frequency converter and the brake sys-tem, while the control software was devel-oped in Matlab. A VNC (Virtual NetworkComputer) connection allows a completeremote control of the turbine.Environmental data acquisition system.The installed system is currently employ-ing a development of SAM-BA, a hard-ware/software architecture and a relationaldata structure specifically designed for themanagement and elaboration of meteo-oceanographic and environmental data col-

2189


Figure 2: SAM-BA: The scheme of the relational structure developed to manage theenvironmental data collected onboard the ENERMAR-Kobold platform in the Straits ofMessina.

lected by automatic platforms and fieldcampaigns [4]. The hardware includes aPC and a serial multiport connected to themeasurement devices, that can be directlysensors (e.g. a temperature probe) or com-plex instruments (e.g. a current meter).Hardware expandability is straightforwardand transparent to the end user through thesharing of the tasks among several dedi-cated PCs that can be geographically dis-tributed, each with a specific task (collec-tion from sensors - onboard; database stor-age/query, presentation via web). The soft-ware has been developed by CNR-IAMCMessina on a open-source LINUX plat-form. The management of sensors/devicesis ensured by a fully expandable plug-in structure. The software communicateswith a PostgreSQL database for data stor-age. The frequency of acquisition and stor-age onboard is programmable according to

the different characteristics of the sensorsand a backup of local data is performed ona remote database using a secured ssh con-nection.The measurement system includes a LinuxPC and a serial multiport to connect:1. a DAVIS Vantage Pro 2 meteo station:

wind speed and direction, air temper-ature, atmospheric pressure, humidity,precipitation;

2. a NORTEK Aquadopp Profiler (ADCP)deployed at the seabed (20 layers, each1 m thick) for 3D profiling. The currentmeter mounts a high sensitivity pressuresensor able to provide wave measure-ments (under development);

3. sea temperature at the seabed (built-inin the ADCP) and at 1 m of depth (by aSBE39);

4. a fluorometer and a turbidimeter, placedat -1.5 m;

2190


5. a CROSSBOW CXTILT 2D inclinome-ter for evaluating the pitch and roll ofthe platform during strong currents pe-riods and/or wave solicitations.

The oceanographic data (3 measurementsevery 15 min for currents and sea temper-ature at seabed, 1 measure per second forSBE39) are stored onboard in the Post-greSQL database, while meteo data areavailable in quasi real-time and then stored.Camera control. Below the platform (atabout 2 m depth), inside of a waterproofcontainer, a dome type video color cam-era with PTZ (Pan-Tilt-Zoom) capabilityhas been installed that is able to rotate onboth horizontal and vertical axis in addi-tion to variable zoom. The main purposeof this camera is to monitor the fish activ-ities near the turbine blades and the inter-action with them. The analog video signalfrom the camera is acquired by a MPEG4codec/video server and converted in a dig-ital format suitable to be transported overthe LAN. The access of the video stream-ing and the camera PTZ controls from thelocal network or Internet is ensured by aweb server by using a web browser.Wireless link. The system includes twowireless bridges with directional antennas,operating in IEEE 802.11g standard at 2.4GHz. In the actual point-to-point configu-ration, and given the distance of about 8 km(required to connect the Kobold platform tothe Internet Service Provider), the wirelesslink ensures a bit-rate of 36 Mbps.The SAM-BA database. The managementof the data collected by the ENERMAR-Kobold platform is performed by the SAM-BA database, which has been tailored andfurther developed from its first version dat-ing on 2004. In the frame of Clus-ter 10 Programme, the SAM project [5]set the problem of regional coastal mon-itoring for Sicily. Among the other is-

sues, the management of the collected data,coming from very different sources wasapproached through the development ofSAM-BA a relational structure tailored tostore:

i) meteorological data and water qualitydata collected and transmitted via SMStechnology by a network of automaticcoastal platforms

ii) data from lab analyses on water, sedi-ment and biota samples collected duringinterdisciplinary campaigns at sea

iii) the data collected by an underway con-tinuous system

iv) CTD profiles, etc...Furthermore on line queries and graphicvisualization of selected data were pro-vided through standard web browser ap-proach. The core of SAM-BA was de-veloped on the ANSISQL standard, sothat the database can be transported toa different ANSISQL-compliant RDBMS.This ensures to the developers the pos-sibility of easily migrating SAM-BA onhigher-performing hardware/software ar-chitectures without changing the interfacecodes to the database. The SAM-BA struc-ture is modular (see Figure 2). There isa fundamental information (the measuredvalue), around which a set of ancillary in-fos specify further important aspects neces-sary to track the life cycle of the measuredvalue and to link it with other values in thedatabase. Though maintaining a commonarea where the values are stored and cor-related (the Values Table), in this wayit is possible to associate to each recordancillary infos that can vary in relation tothe nature of the record itself (e.g. arecord coming from an automatic platformor a measure produced in the lab). Thequerying was developed as a web inter-face on the HTTPS protocol; the used lan-guage was PHP, coupled with the APACHE

2191


web server. Beyond the known reliabilityof both these softwares, a further note ison their availability for a wide number ofhardware/software platforms. The use ofthe HTTPS protocol, with a bilateral ex-change of certificates, makes it possible theencoding of the exchanged infos and thetracking of the accesses.Web portal. A website for the real-time publication of the collected data hasbeen developed and is currently available athttp://kobold.horcynusorca.it/. The pagesare conceived to publish all the currentlyacquired data (actually meteo and currentvelocity) onboard the platform with auto-matic periodic updates (typically few min-utes) which can be programmed accordingto the research and monitoring needs. Thesite makes available both raw data tablesand their graphical presentation. The web-site shows also the archive of previous data(24 hours for marine currents, full datasetfor meteo data).

3 The different aspects ofthe research

3.1 Meteo-oceanographic as-pects: the observed pro-cesses

Hydrography. From the hydrograph-ical point of view the Straits exhibitsvery relevant tidal currents driven by bothbarotropic and baroclinic processes whichdepend on strong bathymetric constraintsexerted by the sill and coastal morphology[6]. Important features are also the pres-ence of an amphidromic M2 point close tothe sill and the recurrent presence of twowater masses, the Atlantic Water (AW) atthe surface and the Levantine Intermedi-

ate Water (LIW) underneath [7]. In bothTyrrhenian and Ionian basins the interfacebetween these two water layers is generallyat a depth of 150 m whereas near the sillit uplifts to a 30 m of depth. The waterfluxes through the Straits average 233x103m3 s1 and depend on tidal rhythms [8].Tidal currents velocities often exceed thethreshold of 2.0 ms1 near the sill at theNE entrance and are induced by the oppo-site phase in tidal amplitudes which existsbetween the Tyrrhenian (12 cm) and the Io-nian (6 cm) seas. Due to the strong mix-ing phenomena that involve the whole wa-ter column the Straits generates a peculiarwater mass (called Messina Mixing Wa-ters) that flows southward and can be iden-tified up to 150 km along the ionian coastof Sicily [9]. This phenomenology hasbeen accurately described since the worksof Vercelli and Picotti of early decades ofthe previous century [10], and continueduntil now based on oceanographic cruiseswith classical sampling methods. There-fore a systematic observational approachproducing a long-term eulerian time seriesis lacking. The continuous observation ofthe environmental system made possible bythe floating laboratory provides the oppor-tunity to describe in a greater detail phys-ical and oceanographic processes, alreadyobserved through classical methods. Someof the findings are reviewed below.Currents. The current measurementswere collected by the 3D ADCP (NortekAquadopp) every 15 min and stored on-board after a preliminary averaging pro-cess. The considered period covered 98days for a total of 9395 valid measures.Current intensity showed a maximum of2.24 ms1 (in S-to-N or montante con-ditions) and a mean value of 0.47 0.34ms1 (Figure 3). In terms of frequencydistribution (Figure 4), after the subdivi-

2192


Figure 3: Straits of Messina (Ganzirri): Current intensity (measured every 15-min andaveraged in the 5-to-10m layer) observed at the ENERMAR-Kobold platform from Oc-tober 10th, 2008 to January 9th, 2009.

sion into 45 equal-amplitude classes ofspeed, the percentage of time characterizedby a current speed greater than 1 ms1was found to be 7%, which becomes 2%if we consider currents greater than 1.5ms1, and only 0.1% for currents greaterthan 2 ms1. The cumulative frequencyhas the median of 0.4 ms1, i.e. for 50%of the time, currents did not exceed suchvalue. Following the approach used in lit-erature [11, 12] for the ocean surface cur-rents, the distribution that best fits the ex-perimental data is the Weibull distribution,that is characterized by only two param-eters a (the scale parameter) and b (theshape parameter). In our case, the fittingprocedures calculates the following values(with 95% confidence interval): a=0.52and b=1.44 (Chi-square Goodness-of-Fittest, P


Figure 4: Frequency distribution of the current intensity data collected on the Straits ofMessina (Ganzirri) and their fit with Weibull distribution.

tion of a generic tidal site through onlytwo parameters, the scale and shape ofthe Weibull distribution, explicitly linkedto the average and the standard deviationof the current intensities respectively. Fi-nally, it is possible to describe some impor-tant features of the currents in the measure-ment site. Figure 5 shows the tidal ellipseof the current speed (averaged in the layer5-10 m of depth). An evident dissymme-try in the order of 3 to 1 between the tidalphases (montante and scendente) canbe noted: due to bathy-morphological con-straints the site is far more exposed to tidalcurrents from S to N.Water temperature. The thermal features

of the waters in the study site give insighton the heating/cooling processes and theperiodic alternance of the different watermasses. Figure 6 shows the temporal trendof the water temperature at the seabed (dur-ing a period of about 3 months). In Octobervalues close to 23 C indicate a still devel-oped late summer condition while the slowseasonal process of cooling begins at theend of November and ends in January. Thehigh dynamics of the area is highlighted bythe wide thermal range (about 7 C) dueto the presence of water masses of differ-ent origin in relation to the tide (Tyrrhe-nian Surface Waters, 23 C; IntermadiateIonian Waters, 14.5 C). A further process

2194


Figure 5: Tidal ellipse of the current speed (averaged in the layer 5-10 m of depth) col-lected near the ENERMAR-Kobold platform in the Straits of Messina.

is the periodic alternance of periods char-acterized by a low or high thermal dailyvariation depending on the moon phases:neap tides associated with low currents re-duce the presence of Ionian colder water.This process disappears towards the wintermonths when the Tyrrhenian surface waterscool down.Meteorology. The possibility to describein detail particular meteorological eventsis a further value of an automatic observa-tion system. The records of a 1-day winterstorm are shown in Figure 7. In the first 36hours, in normal weather conditions with

a stable atmospheric pressure and a mod-erate wind, small periodic asymmetric os-cillations of the platform induced by tidescan be noted. The abrupt decrease of thepressure (from 1013 to 1002 mbar) and thestrong increase of the wind speed (from 20to 75 kmh1) that follow induce an evi-dent increase of the platform pitch ampli-tudes ( 8 deg). After the storm, the resid-ual oscillation due to the persistence of arough sea can be observed for at least 12 h.

2195


Figure 6: Temporal trend of the water temperature at the seabed (during a period of about3 months) collected near the ENERMAR-Kobold platform in the Straits of Messina.

3.2 Technological and environ-mental aspects: the Energyproduction

An important engineering parameter in theoperation of a floating vertical-axis tur-bine is the actual angle between the axisitself and the water flow. Due to thedynamic response of the platform system(hull+moorings) to the surface movements(tidal variations, waves, etc.), the angle candecrease with respect to its optimal value(90 deg): this reduces the performancesof the turbine and provokes frictions andstresses to the mechanical parts. To addressthe study of this aspect we have used thedata collected by the inclinometer in springtide and calm sea conditions (duration of

the observation 3 days; sampling frequency1 Hz) to obtain the harmonic response ofthe platform system (Figure 8). The contin-uous value (at 0 Hz) indicated the need ofa fine levelling of the platform by adjustingthe ballast. In the tidal band (up to 2x104

Hz) the moon semidiurnal M2 componentinduces inclinations of about 1 deg mainlyalong the roll (S-to-N) axis. In the waveband (greater than 10-2 Hz) a small sig-nal with a period of about 3.7 sec is theplatform swing. The evaluation of the pro-totype efficiency and the assessment of itsenvironmental and social impact have beenachieved during the pre-operational phasein 2009. The local analysis on the marineenvironment showed that the ENERMAR-Kobold prototype has a very low overall

2196


Figure 7: Weather conditions (wind speed and atmospheric pressure), and its effects onplatform oscillation (pitch component) measured during the storm.

impact. The joint measurement of currentsand produced mechanical power made pos-sible to evaluate in the order of 10% the ef-ficiency of the prototype in extracting en-ergy from the current flow. The proto-type produces energy (about 10 kW) assoon as the currents exceed 0.8 ms1 es-pecially during montantephase, which isstronger than scendente phase in the site.Losses in the electrical system due to anon-optimal control of the turbine are stillrelevant at low currents (Figure 9). If weassume that the Weibull distribution forcurrents holds, the total theoretical poweravailable in a generic site is not a directfunction of speed, but only depends on theparameters a e b of the Weibull distribution,and can be calculated as:

PT0 = +0

P0()f()d

=12Sa3(1 +

3b

). (3)

On the other hand, the total theoreticalpower available (per unit area) derivedfrom direct measurements of current inten-sity can be expressed (using the ergodicityof the process) as follows:

PSpS

=12

+TfTi

w3(t)dt

= 12

ki=1

w3i piT, (4)

where k is the number of classes of subdivi-sion of the distribution (in our case 45), wiis the velocity of the ith class, pi is the ab-solute frequency (number of times per unittime) for which there was the ith speed, andT is the unit of time (15 minutes = 900sec). If we compare the two values (re-

2197


Figure 8: Harmonic response of the ENERMAR-Kobold platform in calm sea conditions.The inclinations induced by the K1 (diurnal, 23.2 h) and M2 (semidiurnal, 12.6 h) tidalcomponents are indicated.

spectively 158 Wm2 and 169 Wm2 )obtained through the two approaches wecan note that they differ by only 6.5%.Hence, considering a specific energy avail-ability (theoretical PT0) of 160 Wm2and a measured efficiency () of about 10%it is possible to estimate the total energy Epproduced by the ENERMAR-Kobold pro-totype in a 25-years life cycle (LC) ofcontinuous operation in the actual deploy-ment site in the Straits of Messina. Sincethe efficient area S is about 30 m2 we ob-

tain:

Ep = PT0 S |DeltaLC = 3 105MJ.(5)

From the total inventory calculated by us-ing the real metric computation (materi-als, metalworks, manufacturing, etc.) andthe Life Cycle Assessment approach, theenergy consumed by the prototype dur-ing its entire life cycle can be estimatedin less than 2.5x105 MJ. Although theENERMAR-Kobold is the first release ofthis new technology for energy production

2198


Figure 9: ENERMAR-Kobold energy production in the Straits of Messina between Jan-uary, 7th and January 12th , 2009. Time series of the observed current intensity and themeasured mechanical power.

from renewable sources, the energy bal-ance is therefore positive.

4 Future developmentsAll sources of renewable energy are be-coming an important option for a grow-ing number of communities especially indeveloping countries. The rapid growthof population, the increasing energy de-mand, the need for a better quality oflife in remote and isolated places, attractmore attention and interest on renewableenergy also from marine sources. Drivenby the developments and assessments onthe ENERMAR-Kobold prototype in theStraits of Messina, UNIDO (the United Na-tion Industrial Development Office) has re-cently launched an international partner-ship with South-Eastern Asian countries(Indonesia, Philippines, China) aiming atexploring the potential of exportability ofthis kind of technology in specific opera-

tional scenarios (small islands). The possi-bility of modelling the tidal currents inten-sities through the Weibull distribution al-lows to characterize a generic site by usingonly two parameters (the Weibull parame-ters a and b) directly related to the meanand standard deviation of the current in-tensities, that can be then estimated froma time series of current measurements witha duration of at least half lunar cycle. Fur-thermore, the theoretical density of powerof the current flow available in each sitecan be calculated from the same two pa-rameters, so that the comparison of the en-ergy production potential of different sitesis straightforward. As an example, Figure10 shows the available density of powerversus the mean velocity of the tidal currentflow for several values of standard devia-tion. Some selected geographical sites aremapped on these curves according to theirmeasured signature (in terms of mean andstandard deviation of their tidal currents)

2199


Figure 10: Power density available depending on the average speed of current flow forvarious values of standard deviation (dotted curves). Straits of Messina: data of thisstudy; Chesapeake bay: data of the buoy BrownShoals-Delaware USA, January-February2009; Venice Lagoon: source Magistrato alle Acque di Venezia (extrapolated data takenat the inlets). Lombok Strait and Parola Strait: data from field surveys in 2007 funded byUNIDO. Islay Channel: available on the web.

so that their energy production potentialcan be then easily compared. The labo-ratory for the observation of the Straits ofMessina system currently hosted onboardthe ENERMAR-Kobold platform can pro-vide the following output to the researchand management communities:

continuous collection of Meteo-oceanographical parameters available inreal-time via internet;

information needed to the optimizationof the ENERMAR-Kobold prototype andsimilar technologies;

possibility to host instrumentation and/or

sensors from different research teams toperform joint programmes (e.g. on bio-diversity and climate changes).

The CNR-IAMC is promoting the inclu-sion of the ENERMAR-Kobold platformin the research national network of instal-lations currently in operation. The per-spective of an increasing request of energyfrom renewable sources makes it possiblethe forthcoming presence of a global net-work of platforms hosting state-of-the-artinstrumentation to observe the marine en-vironment.

2200


5 Acknowledgements

The authors are grateful to Mr. MicheleFurnari and Mr. Francesco Soraci (CNR-IAMC) for their field collaboration andto Dr. Santino Smedile (Horcynus Orca

Foundation) for his technical support.Thanks go to technical staff of Ponte diArchimede SpA for their support duringthe study. A special thanks to Dr. ElioMatacena to have inspired and funded theoverall ENERMAR-Kobold initiative.

References[1] E. Bohm, G. Magazzu`, L. Wald, and M.L. Zoccolotti. Coastal currents on the

Sicilian shelf south of Messina. Oceanologica Acta, 10(2):137142, 1987.

[2] C. Agnese, V. Bagarello, and G. Nicastro. Alterazione di alcuni caratteri del regimepluvio-termometrico siciliano nel periodo 1916-1999. Proceedings of the NationalConference of the Italian Association of Agrometeorology (AIAM) Catania 6-7 June2002 (in Italian), 2002.

[3] G. Calcagno, F. Salvatore, L. Greco, A. Moroso, and H. Eriksson. Experimentaland Numerical Investigation of an Innovative Technology for Marine Current Ex-ploitation: the Kobold Turbine. Proceedings of the ISOPE-2006. Conference, SanFrancisco, USA., 2006.

[4] A. Bergamasco. Relazione tecnico-scientifica finale Workpackage 1-A5 Acqui-sizione, gestione, elaborazione e restituzione dei risultati, Piano Ambiente Marino,Programma Cluster 10, Progetto n.12, C2 SAM (Sistemi Automatici di Monitorag-gio). CNR-IAMC UOS Messina internal Report (in Italian), 2004.

[5] AA.VV. Final Report of the Project SAM: Realizzazione ed attivazione di unarete integrata di piattaforme costiere e mezzo mobile attrezzati per sistemi avan-zati di monitoraggio delle acque (Piano Ambiente Marino, Programma Cluster 10Potenziamento del. Report edited by the CNR-IAMC UOS Messina, in Italian.,2004.

[6] T.S. Hopkins, E. Salusti, and D. Settimi. Tidal forcing of the water mass interfacein the Straits of Messina. Journal of Geophysical Research, 89:20132024, 1984.

[7] M. Battaglia, A. Caserta, and E. Salusti. Transient phenomena in the Straits ofMessina. In: The Straits of Messina ecosystem, Edited by Guglielmo, L. and others.University of Messina, Department of Marine Biology and Ecology, pages 3147,1995.

[8] F. Mosetti. Tidal and other currents in the Straits of Messina. In: Guglielmo, L.and Manganaro, A. and De Domenico, E. (Eds), The straits of Messina ecosystem:present knowledge for an eco-hydrodynamical approach. Proceedings of sympo-sium held in Messina, April 4-6 1. University of Messina, Dipartimento di BiologiaAnimale ed Ecologia Marina, pages 1330, 1995.

2201


[9] F. Decembrini, T. Hopkins, and F. Azzaro. Variability and sustenance of the deep-chlorophyll maximum over a narrow shelf, Augusta Gulf (Sicily). Chemistry andEcology, 20(1):S231S247, 2004.

[10] F. Vercelli and M. Picotti. Il regime fisico chimico delle acque nello Stretto diMessina - Parte II, Commissione Internazionale del Mediterraneo - DelegazioneItaliana, Crociere per lo studio dei fenomeni nello Stretto di Messina, Campagnedella R. Nave Marsigli negli anni 1. 1926.

[11] P.C. Chu. Probability distribution function of the upper equatorial Pacific currentspeeds. Geophysical Research Letters, (35), 2008.

[12] P.C. Chu. Weibull Distribution for Global Surface Current Speeds Obtained fromSatellite Altimetry. Proceedings of 16th Conference on Satellite Meteorology andOceanography, Phoenix, 11-15 January 2009, 2009.

[13] A.H. Monahan. The probability distribution of sea surface wind speeds. Part-1:theory and SeaWinds observations. Journal of Climate, 19:497519, 2006.

2202

The CNR operational Sea Surface TemperatureProducts in the Framework of MyOcean Project

B. Buongiorno Nardelli, C.Tronconi, V. Forneris, E. Bohm, R.SantoleriInstitute of Atmospheric Sciences and Climate, CNR, Roma, [email protected]

Abstract

Different remotely-sensed Sea Surface Temperature (SST) products have beendeveloped by the Gruppo di Oceanografia da Satellite (GOS) of the Istituto di ScienzedellAtmosfera e del Clima (ISAC) and are operationally produced and distributed innear-real time in the framework of GMES (Global Monitoring for Environment andSecurity) MyOcean project. The products are based on the infrared images collectedby the sensors mounted on several satellite platforms, and they cover the Mediter-ranean Sea, including the western Atlantic Ocean, and the Black Sea. The SST pro-cessing chain includes several steps, from the data extraction and preliminary qualitycontrol, to the images compositing and merging. A two-step algorithm finally allowsto interpolate SST data at high (1/16) and ultra-high (1/100) spatial resolution, ap-plying optimal techniques. The basic design of the MyOcean processing chain andthe main algorithms are briefly described hereafter.

1 Introduction

The ISAC-GOS (Istituto di ScienzedellAtmosfera e del Clima-Gruppo diOceanografia da Satellite) is involved inboth operational and R&D (Research andDevelopment) activities related to the re-trieval of the Sea Surface Temperature(SST) from satellite data. These activ-ities, started within Medspiration, MF-STEP and MERSEA projects, are nowprimarily carried out in the frameworkof two international initiatives: the Euro-pean MyOcean project, and the GODAE-GHRSST (Global Ocean Data Assimila-tion Experiment-Group for High Resolu-tion SST, [1]). The first one is the Euro-pean project devoted to the implementa-tion of the GMES (Global Monitoring forEnvironment and Security) Marine CoreService for the 2009-2012 period. My-

Ocean project is co-funded by the Euro-pean Union through the 7th FrameworkProgramme for Research and Develop-ment, and aims to develop the first con-certed and integrated pan-European capac-ity for ocean monitoring and forecasting.On the other hand, the GHRSST is an inter-national scientific and technical frameworkspecifically set up to address the need foraccurate high resolution SST products, asrequired by several different kinds of sci-entific and institutional users, as well asby private companies and enterprises. Infact, the SST is a fundamental parameterfor the scientific investigation of the oceanand atmosphere dynamics and climate, andit is clearly needed also by the meteoro-logical and marine operational forecastingsystems to constrain their numerical pre-diction models (mainly through direct as-similation). Moreover, an increasing num-


Figure 1: Scheme of the Mediterranean and Black Sea MyOcean SST processing chain.

ber of private companies (working, for ex-ample, on fisheries, tourism, marine trans-portation, marine environment and securitymanaging, offshore exploration and extrac-tion, etc.) is now requesting operationalaccess to SST data. Actually, the mainresults and recommendations of GHRSSThave been reflected in the design of theMyOcean SST-TAC (Thematic AssemblyCentre). Within the SST-TAC, ISAC-GOShas the responsibility of the developmentand implementation of the operationalchains for the production, validation anddissemination of the SST Level 4 (L4) datacovering the Mediterranean and Black Sea.These L4 data correspond to daily griddedoptimally interpolated satellite estimatesof the SST at high (1/16) and ultra-high(1/100) spatial resolution, built from allavailable infrared measurements. In thispaper, the present (V0) GOS-ISAC My-Ocean SST processing chain is describedtogether with the evolution to the V1 pro-duction chain, that will be fully operationalonly at the end of 2010. The data used andthe main modules of the chain are intro-duced, together with a description of thealgorithm for the merging/collating of thesatellite passes and of the optimal interpo-lation techniques adopted.

2 Input data

The SST L4 processing chain is based onthe merging and interpolation of the night-time images acquired by the infrared sen-sors installed on both geostationary and po-lar orbit satellites. The sensors (and plat-forms) presently ingested in ISAC-GOSsystem include: the AATSR (flying onENVISAT), MODIS (on both Aqua andTerra satellites), AVHRR (on METOP andNOAA satellites), SEVIRI (installed onMSG). The raw data are collected and pro-cessed up to the Level 2 (L2, i.e. asSST values retrieved on the native sensorswath) by several different institutions, andare successively made available in a sin-gle common format through the GHRSSTGlobal and Regional Data Assembly Cen-tres (GDAC, RDAC; see www.ghrsst.orgfor more details). The GHRSST L2 dataare written as netCDF files and containSST observations, geo-location data, errorestimates (bias error and standard devia-tion), land and ice flags, as well as addi-tional auxiliary fields for each pixel, re-ferred to as dynamic flags (therefore, thesefiles are called L2P data).In particular, L2Pdynamic flags include estimates of the sur-face wind field (that could be used to

2204


identify areas subjected to intense diurnalvariations), surface solar irradiance (SSI),aerosol optical depth (useful to flag ar-eas contaminated by atmospheric dust) andsea ice concentration. Through the MAs-ter MEtadata Repository (MMR) system,GHRSST provides also metadata recordsfor each satellite. These metadata im-age are based on international standards(FGDC, ISO19115, INSPIRE) and containinformation on the single L2P file content(coverage, acquisition time, etc.), whichare used in our system for a pre-selectionof the files that are needed by the process-ing chain.

3 Processin chain

The ISAC-GOS SST processing chain isdesigned as a two-step process: the highresolution (HR) processing and interpola-tion is performed first, while the ultra-highresolution (UHR) is run in sequence (seethe scheme in Figure 2). This architecturereflects the specific algorithm used for theUHR interpolation (see section 3.4). Thewhole chain is organized in five main mod-ules/packages (M1-5), which are managedthrough a specific System Controller. TheSystem Controller governs the sequence ofoperations and includes all error handlingand communication procedures, as well asthe internal and external interfaces moni-toring. The five modules are described hereas logical steps, that apply to both theHR and the UHR L4 processing, while thespecific software packages have necessar-ily been adapted or configured to run dif-ferently when activated by one processingor by the other. All modules have beendesigned in order to allow an easy con-figuration of the main parameters requiredby each operation (through configuration

files). A brief description of their mainfunctionalities is presented in the follow-ing:

3.1 External data collection(M1)

This module manages the external inter-face to the input data providers and the in-ternal input data archive. It includes con-figurable connection protocols and editablelists of the sensors to be considered. It col-lects metadata from GHRSST GDAC andRDAC and downloads only the L2P datacovering at least a fraction of the interpola-tion areas. This module is unique for bothHR and UHR, and it is activated at regulartime intervals (every three hours).

3.2 L2P data extraction, prelim-inary data quality controland remap to L3 (M2)

The second module deals with the datapre-processing. SST data are first extractedbasing on the geographical coverage andlocal time of the observations, on pixel ba-sis. To avoid diurnal warming contamina-tion (see [2]), the ISAC-GOS system onlyselects the observations collected between11 p.m. and 6 a.m. (local pixel time).L2P data quality control is then performed,starting from the quality flags and confi-dence values associated to each pixels. Allthe parameters used at this stage are con-figurable. Presently, only highest qualityflag data are retained. Selected valid dataare finally remapped over the final interpo-lation grids (Level 3, L3). The MyOceanMediterranean HR grid corresponds to thefirst level of the MyOcean MediterraneanModelling and Forcasting Centre (MFC)1/16 grid. A similar grid has been de-

2205


Figure 2: Example of MyOcean Mediterranean HR SST L4

veloped for the Black sea (at 1/16). AllUHR grids have been generated from theglobal land/sea tag file NAVOCEANO(http://www.ghrsst-pp.org/GHRSST-PP-NAVO-Land-and-sea-Mask.html).

3.3 Sensor bias adjustment, L3super-collating and addi-tional cloud screening (M3)

Due to computational limitations (see sec-tion 3.4), the space-time interpolation al-gorithm implemented at ISAC-GOS is ableto ingest a single value of SST per pixeland per day. Consequently, the differ-ent images acquired by all sensors withinthe pre-defined time interval need to bepreliminarly merged into a single super-collated image (L3 super-collated). How-ever, the SST estimated from one sen-sor might significantly differ from that re-trieved by another, mainly as a conse-quence of the differences among the sen-sors (number of bands, spectral resolution,scanning/viewing geometry, etc.) and/orof the different algorithms applied, which,in particular, might correct very differ-ently the atmospheric contribution to themeasured brightness temperature. Rapidchanges in the atmospheric conditionsmight additionally lead to spurious biases

at large spatial scales also between imagesacquired by the same sensor at differenttimes of the day. Consequently, in order toavoid artefacts in the super-collated data, abias adjustment procedure needs to be ap-plied to L3 data before merging. This mod-ule thus represents one of the most difficultand crucial steps in SST data processing,which has a very strong impact on the qual-ity of the final product. Two different algo-rithms are used in the HR and UHR pro-cessing. The HR scheme (originally devel-oped within MERSEA Integrated Project,but significantly optimized for MyOcean)builds a single merged image per day byselecting the best measure available foreach pixel and correcting the biases amongthe images by adjusting them to some ref-erence sensor measurement. Here, bestis defined through a pre-determined sen-sor validation statistics, and a simple hi-erarchy of sensors is identified coherently.In practice, a reference sensor list is cho-sen (sensors whose SST values do not needto be corrected, e.g. AATSR). Similarly, alist of sensors to be adjusted is defined andordered for quality. For each pixel, onlythe observation from the best sensor avail-able is retained. However, before addingthe new data to the merged map, the largescale bias between each new image and

2206


the pixels that have already been mergedis estimated and removed. In this phase,an additional check on cloud contamina-tion is performed by flagging the pixelsthat result to be colder (by a fixed thresh-old) than the previous day value, as mea-sured in the corresponding super-collatedL3. In this way, the reference map is up-dated every time a new valid observationcomes in. To apply a smooth bias cor-rection (i.e. a correction that doesnt cre-ate artefacts or spurious gradients at im-age edges), a specific new algorithm hasbeen developed for MyOcean. An ini-tial bias correction map is estimated as theSST difference, i.e. by subtracting the ref-erence values to the uncorrected observa-tions. This map will display valid data onlywhere the images overlap, while it will befilled by NaN (Not a Number) values else-where. The bias is then set to zero in thepixels where an estimate of the SST dif-ference is not present, provided they areat a distance of at least 1500 km from theoverlap zones. The final step is the itera-tion (100 cycles) of a 10-grid point mov-ing window average (which ignores NaNvalues), which thus extrapolates the biascorrection from the available SST differ-ence towards the areas set to zero. At eachiteration, the bias correction is set backto the first iteration values if these origi-nally contained valid numbers (namely, ifthey were zero or valid SST difference val-ues). A validation of this algorithm hasbeen performed by analyzing the super-collated L3 SST andSST (where indi-cates time difference) gradients. Actually,the merging applied at 1/16 degree reso-lution might anyway lead to spurious SSTgradients at UHR resolution, in case of un-even distribution of data and/or scatteredclouds that are present in the highest accu-racy sensor images. As a consequence, the

UHR scheme is based on a different defi-nition of best measure, which keeps intoaccount the continuity of the data presentin the single image. In practice, the biasis not estimated with respect to the higheraccuracy sensor data but between each im-age and the first guess field, which is builddirectly from the HR L4 SST (see section3.4). This bias is estimated and removedlocally (50 km). The best data are thenselected basing on a measure of each im-age data sparseness (spotty/scattered dataare qualified as worse). Data sparsenessis quantified by computing an SST gradientmap after assigning an unrealistic negativevalue to non-valid data (e.g. -99). In thisway, the gradients are much higher nearcloud borders, and the images that displaythe lower gradients easily identify the lesssparse data source for each pixel.

3.4 Interpolate L3 data usingspace-time optimal interpo-lation (M4)

The classical optimal interpolation (OI)method [3, 4] has been adapted to theMediterranean and Black Sea MyOceanSST HR and UHR processing, startingfrom the previous experiences of ISAC-GOS. OI gives an estimate of an anomalyfield with respect to a first guess, assum-ing the statistical characteristics of the vari-ability are known (background error co-variance and observation error covariance).The SST L4 analysis is then obtained asa linear combination of the observations(namely, of the SST anomalies with respectto the first guess), weighted directly withtheir correlation to the interpolation pointand inversely with their cross-correlation.The MyOcean HR OI module has beenoriginally developed in the framework of

2207


the MERSEA Integrated Project, while theUHR one has been specifically designedfor MyOcean. The MERSEA HR statisti-cal interpolation scheme uses a daily pen-tad climatology as first guess (built from21 years of AVHRR Pathfinder data), andall its characteristics (first guess computa-tion, covariance model, data sub-samplingstrategy, inversion technique, etc.) are fullydocumented in Marullo et al. [5]. Marulloet al. focused on the re-analysis of thePathfinder data [6] that was performed dur-ing MFSTEP project (see also [7]), but theyused exactly the same OI scheme chosenfor the MERSEA HR processing. Conse-quently, the common HR and UHR fea-tures will be only briefly introduced inthe following, while more details will begiven on the strategy chosen for UHR.More specifically, the approach proposedto solve the problem of interpolating theSST at an effective resolution of 1 km(UHR) is based on a simple concept con-cerning scales separation/decomposition.In practice, if we recognize that in thelower resolution maps the small scale vari-ability has already been filtered out, boththrough binning on a smaller resolutiongrid and through statistical filtering by OI,it may be assumed that the lower resolu-tion map, properly re-binned on the UHRgrid (through a simple bilinear algorithm),may be used as first guess for a secondinterpolation step, where the SST anoma-lies, estimated now at 1 km, only containthe small scale signals. A similar approachhas been adopted by the Japan Meteo-rological Agency (JMA) to develop theirMerged Global Development SST prod-uct (http://goos.kishou.go.jp/rrtdb-cgi/jma-analysis/jmaanalysis.cgi). Actually, a sub-optimal statistical interpolation scheme isused both at HR and UHR. The scheme issub-optimal in the sense that for each in-

terpolation point the input data are selectedonly within a limited sub-domain (within aspace-time influential radius), while an op-timal scheme would, in theory, require thatall available observations are used. Giventhe near-real-time requirements of our sys-tem, this step is needed to reduce the num-ber of operations required by the statisti-cal interpolation (mainly by the inversionof the covariance matrix). In practice, thesuper-collated L3 data at HR are collectedwithin a temporal window of ten days,while at UHR only the measurements rel-ative to the same day are used. The spa-tial influential radius ranges between 300and 900 km at HR and is limited to 20km at UHR. Moreover, in order to avoiddata propagation across land from one sub-basin to the other and to speed up theinput data search, both the HR and theUHR schemes are built to drive multipleanalyses. In practice, the interpolation isrun several times, applying different datasearch and interpolation grid masks, whichare built cropping the original grid. Sixsub-basin grids have been defined for theHR processing (eastern Atlantic, WesternMediterranean basin, Tyrrenian sea, Adri-atic sea, Levantine basin, Aegean sea), and175 masks have been constructed for theUHR. In order to avoid artifacts at the bor-der of the different masks, two differentgrids are used for each sub-basin/mask, oneidentifying the interpolation points and onefor the selection of the observations. Thislast includes buffer zones at the borders,whose dimension are defined by the spa-tial input data search radius. Before en-tering the influential data selection withineach analysis, super-collated L3 imagesare checked for residual cloudy pixels atboth resolutions. In practice, cloud mar-gins are first eroded, flagging all valueswithin a distance of m pixels to a pixel al-

2208


ready flagged as cloudy. A second checkis then performed through the comparisonto the closer (in time) L4 analysis avail-able, that is used as a reference only ifthe analysis error is lower than a fixedvalue. Data that differ from the refer-ence field for more than a defined thresh-old (usually 2, where represents theaverage standard deviation between con-secutive night images =.7 C) are not in-cluded in the analysis. Even within thelimit imposed by the influential data se-lection, only n observations can be effec-tively used because of computational timelimitations. As a consequence, a strategyto remove the most cross-correlated/leastcorrelated (to the interpolation point) datais needed. The method chosen here sortsthe data as a function of their correlationto the interpolation point. The most cor-related observation is selected first, whileall successive data are selected only ifthey are found along a new direction inthe space-time (until n observations arefound). This allows a more balanced cov-erage within the influential bubble, evenselecting a small number of observations.As already pointed out, the background co-variance functions used at HR are basedon the correlations used within MERSEA,originally estimated during the EuropeanSpace Agency Medspiration project. Onthe opposite, different background covari-ance scales have been tested at UHR. TheUHR covariance model has thus been de-fined as a result of empirical tests, eventhough looking at some of the quantita-tive metrics suggested by the GHRSST,as the gradient of the SST and the gradi-ent of the SST. The spatial decorrelationlengths have thus been varied between 5and 15 km, but very little differences havebeen found on the interpolated maps in thewhole range considered. The smaller value

is currently used in the operational chain.On the opposite, the time scale (and cor-responding input search radius) had to bereduced to 1 day, in order to avoid the sud-den growth of fake spatial SST gradients inareas differently covered by satellite obser-vations during successive days. However,more detailed validations, including com-parisons to in situ data, are still in progress.

3.5 Compute SST anomalies,re-format output, man-age SFTP server andTHREDDS catalogue (M5)

The last module deals with the prepara-tion of L4 data for the dissemination to theusers. It manages the external interfacesto the users and the internal output dataarchive. Both L4 SST and L4 SST anoma-lies (with respect to the climatology used asfirst guess, see section 3.4) are delivered asMyOcean core products. All data are writ-ten in NetCDF format, following the spec-ifications from GHRSST. Two external in-terfaces are used for the MyOcean V0 SSTproducts at ISAC-GOS: an FTP site and aTHREDDS server. This module thus up-dates the FTP site and the THREDDS cat-alogue at the end of each processing cycle(i.e. daily before 12 UTC).

4 ConclusionsThe basic design and the most innovativemodules of the SST operational process-ing chain, as implemented at ISAC-GOS inthe framework of the European MyOceanproject, have been described in this paper.In fact, ISAC-GOS is responsible of theproduction of SST L4 analyses coveringthe Mediterranean and Black Seas at highand ultra-high resolution. These L4 data

2209


are included in the MyOcean core prod-uct catalogue and are delivered daily to sci-

entific, institutional and private users allaround Europe.

References[1] C. Donlon and the GHRSST-PP Science Team. The Recommended GHRSST-PP

Data Processing Specification GDS. 2005.

[2] B. Buongiorno Nardelli, S. Marullo, and R. Santoleri. Diurnal variations in AVHRRSST fields: a strategy for removing warm layer effects from daily images. Rem. Sens.Env., 95(1):4756, 2005.

[3] L.S. Gandin. Objective analysis of meteorological fields. Leningrad. Hydromet.Press. Translated from Russian by Israel Program for Scientific Translations.Jerusalem, 1965.

[4] F. P. Bretherton, R. E. Davis, and C. B. Fandry. A Technique for Objective Analysisand Design of Oceanographic experiments applied to MODE-73. Deep-Sea Res.,23:599582, 1976.

[5] S. Marullo, B. Buongiorno Nardelli, M. Guarracino, and R. Santoleri. ObservingThe Mediterranean Sea from Space: 21 years of Pathfinder-AVHRR Sea SurfaceTemperatures (1985 to 2005). Re-analysis and validation. Ocean Sci., 3:299310,2007.

[6] K.A. Kilpatrick, G.P. Podesta, and R. Evans. Overview of the NOAA/NASA Ad-vanced Very High Resolution Radiometer Pathfinder algorithm for sea surface tem-perature and associated matchup database. J. Geophys. Res., 106:91799197, 2001.

[7] B. Buongiorno Nardelli, G. Larnicol, E. DAcunzo, R. Santoleri, and S. Marulloe P.Y. Le Traon. Near Real Time SLA and SST products during 2-years of MFS pilotproject: processing, analysis of the variability and of the coupled patterns. AnnalesGeophysicae, 20:119, 2002.

2210

The Ocean Colour Satellite Observing System

S. Colella, G. Volpe, E. Bohm, R. Santoleri, C. Tronconi, V. FornerisInstitute of Atmospheric Sciences and Climate, CNR, Roma, [email protected]

Abstract

The synoptic view and regular data coverage provided by satellite data makethem essential to monitor marine ecosystems. Ocean colour (OC) Satellite Observ-ing System is an essential component of operational ocean observing and forecast-ing systems currently developed for the global ocean and the European Seas. Inthe framework of this European effort, the Satellite Oceanography Group (GOS) ofISAC Rome has developed a system that produces satellite OC images and data forthe Mediterranean and the Black Seas meeting the growing demand for near real-time OC products for applications in operational oceanography. The GRID basedsystem has been developed to produce: 1) fast delivery images for environmentalmonitoring applications and operational support to oceanographic cruises; 2) accu-rate OC products for data assimilation in ecosystem models; 3) temporally consis-tent reanalysis products for climate change studies. The OC data processing usinga specific regional algorithm developed by GOS for the Mediterranean representsan improvement with respect to the global algorithms that significantly overestimatechlorophyll concentration. Since 1999 Near Real Time and Delayed Time data areprovided daily through an ad hoc automatic procedure that processes satellite dataand makes higher level products available to the users within an hour of raw data ac-quisition from space agencies ground segments. GOS is now extending this regionalalgorithm to optically complex case 2 waters such as the Adriatic Seas.

1 Introduction

A significant proportion of the world eco-nomic and social activities depend on thesea. These activities are subject to uncer-tainty, loss of efficiency and direct costsand damages caused by the varying impactand hostility of the marine environment. Toensure a sustainable use of the marine re-sources, an accurate description and a reli-able prediction of the ocean state and vari-ability is crucial. In the last twenty yearsobservations of the ocean by sensors onEarth orbiting satellites have become an es-sential element of 21st century oceanogra-

phy. In fact, it is now widely recognizedthat to monitor the ocean with the neces-sary sampling frequency in both space andtime, it is essential to supplement conven-tional in situ analysis methods with dataderived using remote sensing technology,primarily from Earth observing satellites.It is also recognized that it is essential tointegrate the satellite and in situ measure-ments through the use of numerical oceanmodels, in order to provide timely informa-tion about the state of global ocean and Eu-ropes seas.Since the 90s the Unesco/IOC action on theGlobal Ocean Observing System (GOOS)


and its Coastal Ocean Observing Panel es-tablished as main objective the develop-ment of a world-wide network for the realtime exchange and use of ocean data in pre-dictive models of the marine environment,from physical fields to marine ecosystemvariables. The European component ofGOOS strongly sustained the developmentof a European operational oceanography(OO) system based on the integration ofexisting regional systems. Since the early80s, oceanographic remote sensing tech-nology has shown a cost-effective means toaddress this issue. In this context, duringthe last 15 years several research projectswere directly targeted the development ofOO capabilities, and have been funded bythe EC Framework Programmes 4, 5 and6. Similarly, the European Space Agency(ESA) funded several projects to developsatellite missions dedicated to OO. This ef-fort led to the establishment of a number ofresearch centres in Europe with advancedknowledge for the exploitation of satelliteoperational products. However, it is onlyrecently that an ocean forecasting system,analogous to the meteorological commu-nitys, has become the marine componentof the Global Monitoring for Environmentand Security (GMES) program: namely theMarine Core Service (MCS). In this con-text, physical properties of the ocean suchas surface temperature and slope, waveheight and surface winds are currentlymeasured globally at high resolution pro-viding reliable inputs to ocean circulationmodels. Similarly, OC measurements ofphytoplankton pigment concentration (i.e.,chlorophyll, CHL) are now used to validatemarine ecosystem models and as input tobio-geochemical models. This is the basisthe new operational ocean observing andforecasting systems that are currently beingdeveloped for global ocean and European

seas in the framework of GMES. In the last15 years, Satellite Oceanography Group(GOS) of ISAC Rome has developed a sys-tem that provides satellite ocean colour im-ages and data covering the Mediterranean(MED) and the Black (BLS) seas. Thissystem constitutes the Mediterranean com-ponent of the European Ocean Colour Ob-serving System (OCOS) and was built tomeet the growing demand for near real-time OC products for applications in OOand climate studies. The system was de-signed to produce: 1) fast delivery imagesfor environmental monitoring and opera-tional support to oceanographic cruises; 2)accurate OC products for data assimilationin ecosystem models; 3) consistent reanal-ysis product for climate change studies.This paper describes the GOS OC observ-ing system and reviews the major scien-tific and technological steps made to de-velop and maintain this system. Section 2presents the OCOS architecture, whereasthe quality improvements of CHL esti-mates in both open and coastal waters isdescribed in Section 3. Finally, we discussthe future development of this system in theframework of Ocean Colour Thematic As-sembling Centre (OCTAC) of MCS.

2 The Ocean Colour Ob-serving System Archi-tecture

The architecture of the GOS OC systemis based on four main modules: the DataCapture and acquisition Facility, the pro-cessing system, the data output reformat-ting and data archive and dissemination.These modules have correspondence withthe four main functions described in thefollowing sections and summarized in Fig-

2212


ure 1. The system is based on GRID envi-ronment and it has a modular design com-posed by three separate processing chains(SeaWiFS, MODIS and MERIS), to facil-itated the maintenance taking into accountnew sensor/satellite and software upgrades.The processing module (Figure 1, mid-dle panel) is the interface between inputdata from space agencies ground segments(NASA and ESA, Figure 1, left panel) andthe GOS data archive and disseminationsystem (Figure 1, right panel). This pro-cessing module consists of a set of shellscripts, IDL and SeaDAS procedures de-veloped by GOS. The system operates intwo modes:1) operational mode used to produce NearReal Time (NRT) and Delayed Time (DT)products, and2) on demand mode used to produce re-analysis or end-user defined products. NRTdata are produced once a day, in less thanan hour from input data availability usingclimatological ancillary data. In generaldata are available to the end-users withinone day from the satellite overpass. DTproducts are generated when ancillary dataare obtained from NASA (in general 2-3day delay). DT products are higher qualitythan NRT and thus are used for data assim-ilation and validation of ecosystem modelsand to produce value-added products (e.g.,phytoplankton primary production).

2.1 The input data and acquisi-tion facility

The satellite data input to the GOS OS arethe full resolution (1.1 km) LAC (LocalArea Coverage) SeaWiFS, MODIS-Aquaand MERIS passes covering the MEDand BLS domain. Historically, SeaWiFSLevel-0 (L0) data were acquired locally by

GOS receiving station (HROM). The ac-quisition function is performed by the Quo-rum HRPT Data Capture Engine installedat GOS. This station is operational sincethe SeaWiFS launch in 1997, and was theonly SeaWiFS real-time receiving stationwith the complete coverage of the MEDarea, among the 9 other NASA authorizedstations worldwide. More than 6000 satel-lite passes have been acquired and pro-vided to the NASA DAAC from 1997 to2004. GOS is still acquiring SeaWiFSfor scientific use. The input for the ac-quisition function is the encrypted HRPTdata continuously broadcasted by the satel-lite when it is within the acquisition circleof the GOS antenna. LAC acquisition isperformed two-three times daily accordingto the Orbview-2 direct broadcast sched-ule, and is processed to L0 after decryp-tion. MODIS L0 data are acquired auto-matically from the Goddard Space FlightCenter, NASA, where data covering thewhole globe are distributed in 5-minutegranules. The system automatically selectsand downloads the granules covering theMED and BLS area of interest. SimilarlyMERIS L2 data covering the MED are ac-quired from ESA rolling archive. Ancil-lary data, e.g., satellite telemetry and at-mospheric fields are also acquired fromNASA. Ancillary data include wind, at-mospheric pressure, precipitable water, andozone data from National Center for Envi-ronmental Prediction (NCEP). The sourceof the ozone data is EP/TOMS.

2.2 OC processing systemThe SeaWiFS and MODIS processingchains are designed to process data fromL0 (raw radiance counts) to L3 (geophys-ical products) and L4 (multi-day, multi-sensor products) and consist of five main

2213


Figure 1: The GOS OC system: 1) Acquisition (Data Capture and acquisition Facilitymodule): Data from several sensor are collected by CNR-GOS Data Capture and acquisi-tion Facility module and sent to the processing system module. 2) Processing (processingsystem and data output reformatting modules): Raw input data get into the processingchain running into the Grid environment. Output data are then reformatted and storedinto the ISAC database. 3) Dissemination (data archive and dissemination modules): in-put (raw) and output data are stored into the GOS archive. Output data are published onthe GOS website and THREDDS catalogue.

processors, whereas MERIS processingchain only deals with L2 to L4 data (Fig-ure 1). Below is a detailed description ofeach step.L0 to L1 processor. In this step L0 dataare transformed into standard L1A data(HDF files containing raw data, telemetryand navigation information). During thisstep, MODIS contiguous granules are firstmerged into a single L0 and then processedto L1A.L1 to L2 processor. Here, L1A rawdata are processed to obtain geophysicalparameters. The main issue related tothis step is the application of the atmo-spheric correction procedure and of thebio-optical algorithm to retrieve ocean pa-

rameters. The processing is currentlycarried out using SeaDAS v5.1.5 soft-ware package available from NASA web-site (seadas.gsfc.nasa.gov), which now al-lows for other than standard algorithms(i.e., MED-specific algorithms) to be im-plemented. L1A data are processed upto L2 applying the dark pixel atmosphericcorrection scheme [1]. The result of thisstep is the Rrs at all wavelenghts which arethen used as input for the bio-optical al-gorithm for oceanic products retrival. L2(HDF) files contain: Remote Sensing Re-flectance (Rrs) at all wavelengths, whichcan be used to produce additional marineOC parameters (e.g., Coloured DissolvedOrganic Matter, CDOM, Total Suspended

2214


Matter, TSM) diffuse attenuation coeffi-cient (Kd), aerosol optical thickness at allwavelengths (for atmospheric application),CHL concentrations using both MED andstandard algorithms, photosynthetically ac-tive radiation (PAR), quality flags [2], andinformation about the viewing geometrysuch as the satellite zenith angle. Withinthis step Quasi True Colour (QTC) im-ages of each satellite pass are also created.QTC is generated by combining the threeOC bands that most closely represent red,green and blue (RGB) in the visible spec-trum, creating an image that is fairly closeto what the human eye and brain would per-ceive. These data can be useful for environ-mental monitoring. For example, SeaWiFSQTC were recently used in the frameworkof the EU-funded project ADIOS to mon-itor the occurrence of Saharan dust eventsin the Mediterranean Sea [3].L2 to L3 processor. This step is com-mon to MODIS, SeaWiFS and MERISprocessing. Here, relevant parametersfor each application/scientific project areextracted and remapped into single-bandproducts over a common equirectangulargeographical projection covering the en-tire MED and BLS domain (27.6-48.4N,9.5W, -43.5E). This processor containscustomized and standard procedures. Thestandard procedure remaps the L2 productsat high resolution (1.1 km at nadir).Specific maps, in terms of domain, for-mat and resolution, are produced for spe-cific national and international scientificprojects. In the framework of the Ma-rine EnviRonment and Security for theEuropean Area (MERSEA) and MyOcean(MyO) EU-funded projects, daily prod-ucts are realized applying standard mask-ing criteria for detecting clouds or othercontamination factors, i.e., land, sun glint,atmospheric correction failure, high total

radiance, large solar zenith angle (70),large spacecraft zenith angle (56), coc-colithophores, negative water leaving ra-diance, and normalized water leaving ra-diance at 555 nm below 0.15Wm2sr1

[2]. In the context of the Mediter-ranean Forcasting System Project, a spe-cific chlorophyll product is realized by bin-ning daily data over the 1/16 of degree(ca. 7 km) ecological model resolution gridwith reduced spatial gaps. In the frame-work of the ADRICOSM Project, dailyCHL data are generated for assimilationinto numerical ecosystem models. Thisproduct has a nominal spatial resolution of2 km matching the model grid and cov-ers the entire Adriatic Sea. ADRICOSMCHL is produced within the DT process-ing mode. Other OC parameters not avail-able in the L2 files, such as CDOM andTSM, are required by this project and areproduced using ad hoc bio-optical algo-rithms from Rrs data. In the context ofboth ADRICOSM and ECOOP projects, aCase1-Case2 waters merged CHL is pro-duced and more fully discussed in Section3.2. Since MERIS L2 data are produced byESA using standard chlorophyll algorithma specific code was developed and imple-mented in the processing system to derivea MED-suitable CHL product from Rrs us-ing the MedOC4ME algorithm (see Sec-tion 3.1). Static data (e.g., jpeg, png andgif) images are produced daily and postedon the GOS website.

2.3 Data format, archive anddelivery system

Daily and reanalysis data files have beenproduced in HDF and lately in NetCDFformat. The OC data format was up-dated from HDF to NetCDF 3.5, following

2215


the Climate and Forecast convention, IN-SPIRE, EN-ISO 19115, 19119 and 19139.NetCDF is a machine-independent formatfor representing scientific data. It is self-describing, portable, appendable (e.g., datamay be appended to a properly structuredNetCDF file without copying the datasetor redefining its structure), sharable (onewriter and multiple readers may simul-taneously access the same NetCDF file),archivable (access to all earlier forms ofNetCDF data is supported by current andfuture version of the software). Moreoverit allows direct access: a small subset ofa large dataset may be accessed efficiently,without first reading through all the preced-ing data. The Data Archive (DA) is basedon a file system and stored on a worksta-tion cluster within the CNR IT infrastruc-ture. DA is based on both an internal andan external archive. The internal database,accessible only by GOS, maintains all theinput and output data of the processing sys-tem. The aim of the external archive is toprovide end-users with access to the finalproduct through the dissemination system.This system was developed in order to bea delivery, discovering and viewing systemas requested by the INSPIRE directive, andconsists of different services: 1) data ac-cess either via ftp from GOS website forstatic images, or through specific e-mail re-quest for HDF or NetCDF data. 2) A The-matic Real-time Environmental DistributedData Services (THREDDS) catalogue hasbeen set up on GOS website for NetCDFdata access. THREDDS provides meta-data and data access and is built on existingtechnologies and protocols. THREDDScatalogues are logical directories of on-line data resources, encoded as XML doc-uments, which provide a place for anno-tations and other metadata about the dataresources to reside. THREDDS enables

end-users to find out what data are avail-able from data providers. 3) A Live Ac-cess Server (LAS) has been set up onGOS website for NetCDF data downloadand dynamic images creation. LAS is ahighly configurable web server designed toprovide flexible access to geo-referencedscientific data. LAS manages NetCDF,ASCII or binary data formats. Variablesand specific data subsets (both in spaceand time) can be visualized on-the-fly andeventially saved into a user-selected fileformat. LAS allows access to backgroundreference material about the data (meta-data) and to compare (difference) variablesfrom distributed locations. Moreover it canpresent distributed datasets as a unified vir-tual database through the use of Opendapnetworking.

2.4 Reanalysis and reprocessingsystems

The entire SeaWiFS archive, combiningHRPT acquired at GOS with the MergedLocal Area Coverage (MLAC) available onNASA website (totalling more than 8000passes) was reprocessed by GOS takinginto account new algorithm developmentsand software updates. In general, reanal-ysis products are generated at least once ayear or everytime a new algorithm or soft-ware update become available, by repro-cessing everytime the entire archive (forconsistency). The reprocessing exploits theGRID environment in order to reduce therunning time. For example, the reprocess-ing of the entire SeaWiFS archive takes6 days using the CNR-ARTOV GRID in-frastructures. Apart from the unique soft-ware or algorithm version, the archive re-processing does not differ from the DTprocessing. In the reprocessing required

2216


by the MERSEA Project, each L2 (1 kmspatial resolution) was spatially binned ona sinusoidal grid projection (2 km spatialresolution) and quality controlled by ap-plying the standard NASA processing L2flags. These L2 spatially binned data werethen merged to obtain daily images data.The obtained L3 timely binned daily im-ages were transferred (via ftp) to the dataarchive connected to Live Access Serverto be available to the scientific commu-nity. The reprocessing of the entire archiveallows the construction of a scientificallyhomogeneous dataset, which can then beused for long-term analysis. For example,weekly, monthly, seasonally to yearly aver-ages are routinely computed from daily im-ages for such a purpose. Moreover, averagefields at different space and time resolutionare used to validate ecosystem models.

2.5 Operational system moni-toring

The operational processing chain takescare to monitor itself: for each step, eachprocedure automatically alerts the opera-tor by sending e-mails reporting on its sta-tus. When the system fails in any of itssteps, it automatically tries to manage theerror. For example, in case of missing L0files from source sites, the procedure itera-tively tries to download data until success-ful. When the error is unrecoverable (or fa-tal), the system reports on the steps affectedby failure. There are three kinds of alerts:1) info, which informs about the process-ing chain status (start/stop, produced filesand so on); 2) warning, which informsthat something (explained in the alert itself)went wrong. In this case, the system is stillable to produce some results (typical error:one of the L0 file is missing from external

source sites, but others are present); 3) er-ror, which informs about errors preventingthe completion of the chain (i.e., no L0 dataavailable from the external source sites).

3 Mediterranean regionalchlorophyll algorithms

3.1 Basin scaleHistorically, an extensive calibration andvalidation activity was performed overSeaWiFS OC data by the SeaWiFS andSIMBIOS Projects. The result of thisCAL/VAL activity was the development ofempirical bio-optical algorithms (OC2v4and OC4v4 [4]) for the CHL opera-tional retrieval in the open ocean. Thiseffort laid the basis for the analogousbio-optical algorithms development forMODIS (OC3,[4]) and MERIS OC sen-sors (Algal1, [5]). Although these algo-rithms were demonstrated to perform ade-quately at the global scale [4, 6], they wereshown to perform generally worse at theregional scale. In the Mediterranean Sea,the standard NASA algorithms (OC2v4and OC4v4) lead to a significant over-estimation of the SeaWiFS-derived CHL(above 70% for chlorophyll below 0.2mg m-3) when compared to in situ data[7, 8, 9, 6]. Using the most spatiallyand temporally representative in situ bio-optical dataset for the MED, GOS quanti-fied the uncertainties of existing regionaland global OC algorithms using SeaW-iFS, MODIS and MERIS sensors. Thisanalysis led GOS to identify and developoptimal algorithms for the production ofhigh quality OC datasets for this basin:namely the MedOC4 for SeaWiFS [10], theMedOC3 for MODIS and the MedOC4MEfor MERIS [11]. This work was part of

2217


Figure 2: Mediterranean bio-optical dataset (crosses) adapted for three OC sensors(MODIS (a), MERIS (b) and SeaWiFS (c)). Continuous lines represent the regional al-gorithms while dashed lines represent standard algorithms. Snapshots of SeaWiFS passover central Mediterranean Sea on 2 July 2004 re-processed with both the NASA stan-dard processing algorithm (d, OC4v4) and the Mediterranean adapted one (e, MedOC4)their difference is also shown (f).

the MERSEA Project, which aimed, be-sides other objectives, to provide high qual-ity satellite products for data assimilationand validation of global and regional mod-els. The identification of the best-suitedMED CHL algorithm and its associated un-certainty was thus an essential step to pro-ceed to the reprocessing of the entire Sea-WiFS mission (Section 2.4). More specif-ically, this work led to the production ofa new chlorophyll dataset suitable for theassessment of and assimilation into thecoupled biochemical and ecosystem mod-els, as required by the modelling com-munity of the MERSEA Project. Figure2 shows the three Mediterranean-adaptedalgorithms (continuous lines) along withtheir respective standard versions (dashed

lines) for SeaWiFS, MODIS and MERIS.Under the reasonable assumption that thein situ dataset (crosses in Figure 2(a-c))well represents the actual field conditions,it is clear that the standard algorithms,for all of the three sensors overestimatein situ CHL for values below approxi-mately 1 mg m-3. It is worth mention-ing that this oligotrophic condition repre-sents more than 70% of the basin space-time variability. In addition to overestimat-ing low values, standard algorithms sig-nificantly underestimate high CHL. There-fore, in the context of a new and reliableoperational ocean observing and forecast-ing system, such as the one foreseen inthe framework of GMES, the use of re-gional adapted algorithms is strongly rec-

2218


ommended. In this respect, Figure 2 (d-f)shows a snapshot of a SeaWiFS pass overthe central Mediterranean processed withboth the standard (OC4v4, Figure 2(d)) andthe regional (MedOC4, Figure 2(e)) algo-rithms. It is clear from Figure 2(f) thatthere might be an order of magnitude dif-ference at sub-basin scale (see for exam-ple the Tyrrhenian Sea) due to the imple-mentation of the standard algorithm in lieuof the more reliable regional one. Apartfrom the qualitative picture offered by Fig-ure 2(d-f), GOS focused on quantitativelyassessing the uncertainties associated withstandard and other than GOS-developed re-gional algorithms [10]. This validation ex-ercise shows that, when these algorithmsare applied to satellite-derived Rrs, GOS-developed algorithms perform better thanany other algorithm built for application inthe MED (with a relative percentage differ-ence of 35% for SeaWiFS, [10]) uncertain-ties exceed 100% when the standard NASAalgorithm, OC4v4, is used instead.

3.2 Coastal System and prod-ucts: the Adriatic case

Coastal waters represent the areas of theworld ocean where the human impact, in-tuitively due to their close and constant in-fluence, is most intensive. Given the well-known difficulties of retrieving OC data incoastal waters along with both the ecolog-ical and the socio-economical importanceof these areas, the quality-controlled dataproduction for this area remains a verychallenging issue. Under the umbrella ofseveral national and international projectsthe Adriatic Sea has been investigated withboth in situ and remotely-sensed data. Thisbasin thus represents an excellent test bedfor the evaluation of existing algorithms,

for the development of new and more re-liable algorithms, and for their quality as-sessment. In this context, GOS main ob-jectives were first to assess and develop areliable OC product over optically complexwaters (i.e., Case2 or coastal waters), andsecond to develop a method for mergingCase1 and Case2 products into a single val-idated product. These are not trivial issues,as it is not uncommon to observe artifactsdue to the implementation of the wrong al-gorithm for example when using a specificCase1 waters algorithm over Case2 waters.Another source of misleading results cancome from the erroneous combination ormerging of different algorithms within thesame image, which in turn can give riseto fronts which actually do not exist. Oneway to overcome this problem is to processthe image twice, once with a Case1 wateralgorithm (for example the MedOC4) andonce with a Case2 water algorithm (for ex-ample the AD4,[12]). At this point, oneof the most challenging tasks when merg-ing products retrieved from different algo-rithms pertaining to different water types isthe exact identification of such water types.Currently the SeaWiFS and MODIS turbidwater flag is set when the Rrs(670) exceedsby 25% the value expected for pure waterin this band [13]. The principle underlin-ing this approach is that it is known thatin Case1 water the contribution of the wa-ter leaving radiance at this band is fairlynegligible, whereas in Case2 waters it isnot, due to different constituents that donot exhibit covariance with CHL. Here, amethod is described that takes into accountthe whole light spectrum from blue to NIRbands for both water types. The rationalefor this is that computing an average watertype spectral signature from in situ mea-surements for both Case1 and Case2 wa-ters should, in theory, give more confidence

2219


when screening the two water types. More-over, and even more important, this lastapproach gives the opportunity of smooth-ing out spurious gradients that might re-sult from the application of different algo-rithms. For the computation of these twoaverage spectra two distinct in situ datasetswere used for Case1 and Case2 watersthe MedOC4 [10] and CoASTS [14, 15]datasets, respectively. Both datasets wereused either to compute the spectral signa-ture of the respective water type and tobuild the bio-optical algorithm (AD4 forthe CoASTS,[12]). In practice, the methodworks as follows: a pixel-by-pixel spectralcomparison is performed between satelliteand reference spectra. The distance fromeach of the two average spectra is thencomputed and used as weight for merging(i.e., averaging) the two chlorophyll prod-ucts into a single pixel value.

4 Conclusions and FutureProspective

The GOS OCOS architecture has been pre-sented focusing on its evolution from a sys-tem for acquisition, storage and deliveryto specific tasks to an OCTAC informa-tion system that is an integrating part ofthe GMES MCS. The main elements ofthis evolution include the setup of a dataacquisition methodology, a fast process-ing and delivery system, quality assuranceof the products, and finally the special-ization of regional OC to coastal regionsthat are important to the MED. An im-portant aspect of the whole system is con-cerned with its errors traceability and per-

formance monitoring. Scientifically, GOShas focused on algorithm development toimprove MED regional products with re-spect to the Global algorithm data qual-ity. GOS, in its MyOcean OCTAC leadingrole, is now a European centre that deliv-ers locally-produced value-added regionaldatasets for the MED, global and Europeanseas, as well as regional products generatedat the OCTAC partner facilities. It servesas central repository and contact centre forthe MyOcean Information Service (MIS)that routes the end and intermediate-userrequests to GOS. With the objectives of en-suring a reliable supply of data to both in-ternal MyO users (e.g, Modelling and Fore-casting Centre) and external intermediateusers (environmental agencies), OCTACwill ensure operational delivery with guar-anteed success percentages above agreed-upon thresholds. The product portfolio en-visaged for OCTAC systems stems fromthe MERSEA legacy and includes baselineL3 products of CHL and Inherent OpticalProperties (IOPs) from different data pro-ducers. Early, in the MyO service evolu-tion, formats will be homogenized to en-sure uniform approach to the data productsregardless of the provider. Subsequently,L3 products will be intercalibrated and col-lated among satellite/sensors suites to en-sure both optimal geographic coverage andreliable error bars for all products in viewof improved assimilation into models. Aconsiderable R&D effort (for which GOSwill seek support external to MyO) willbe spent to improve existing algorithmsperformance and multi-sensor data merg-ing techniques. This will in turn benefitthe OCTAC processing chain and thus theoverall MyO service.

2220


References[1] D.A. Siegel, M. Wang, S. Maritorena, and W. Robinson. Atmospheric correc-

tion of satellite ocean color imagery: the black pixel assumption. Applied optics,39(21):35823591, 2000.

[2] C.R. McClain, K.R. Arrigo, W. Esaias, M. Darzi, F.S. Patt, R.H. Evans, C.W.Brown, J.W. Brown, R.A. Barnes, and L. Kumar. SeaWiFS Quality Control Masksand Flags: Initial Algorithms and Implementation Strategy. pages 37, 1995.

[3] G. Volpe, V. F. Banzon, R. Evans, R. Santoleri, A.J. Mariano, and R. Scia-rra. Satellite observations of the impact of dust in a low nutrient low chloro-phyll region: fertilization or artefact? Global Biogeochemical Cycles - GB3007,doi:10.1029/2008GB003216., 23, 2009.

[4] J.E. OReilly, S. Maritorena, D. Siegel, M.C. OBrien, D. Toole, B.G. Mitchell,M. Kahru, F.P. Chavez, P. Strutton, G. Cota, S.B. Hooker, C.R. McClain, K.L.Carder, F. Muller-Karger, L. Harding, A. Magnuson, D. Phinney, G.F. Moore, andJ. Aiken. Ocean color chlorophyll a algorithms for SeaWiFS, OC2, and OC4: ver-sion 4. SeaWiFS Postlaunch Technical Report Series, SeaWiFS postlaunch calibra-tion and validation analyses: part 3, 11:923, 2000.

[5] A. Morel and D. Antoine. Pigment index retrieval in case 1 waters, MERISAlgorithm Theoretical Basis Document. ATBD2, 9, 25, available from:http://envisat.esa.int/instruments/meris/pdf/atbd 22 09.pdf, 2000.

[6] W.W. Gregg and N.W. Casey. Global and regional evaluation of the SeaWiFSchlorophyll data set. Remote Sensing Of Environment, 93(4):463479, 2004.

[7] A. Bricaud, E. Bosc, and D. Antoine. Algal biomass and sea surface temperaturein the Mediterranean Basin - Intercomparison of data from various satellite sensors,and implications for primary production estimates. Remote Sensing of Environment,81(2-3):163178, 2002.

[8] H. Claustre, A. Morel, S.B. Hooker, M. Babin, D. Antoine, K. Oubelkheir,A. Bricaud, K. Leblanc, B. Queguiner, and S. Maritorena. Is desert dust mak-ing oligotrophic waters greener? Geophysical Research Letters, 29(10):art.n.1469,2002.

[9] F. DOrtenzio, S. Marullo, M. Ragni, M. R. dAlcala, and R. Santoleri. Validation ofempirical SeaWiFS algorithms for chlorophyll- alpha retrieval in the MediterraneanSea - A case study for oligotrophic seas. Remote Sensing of Environment, 82(1):7994, 2002.

[10] G. Volpe, R. Santoleri, V. Vellucci, M. Ribera dAlcala, S. Marullo, andF. DOrtenzio. The colour of the Mediterranean Sea: Global versus regional bio-optical algorithms evaluation and implication for satellite chlorophyll estimates.Remote Sensing of Environment, 107:625638, 2007.

2221


[11] R. Santoleri, G. Volpe, S. Marullo, , and B. Buongiorno Nardelli. Open Waters Op-tical Remote Sensing of the Mediterranean Sea, in Remote Sensing of the EuropeanSeas. Remote Sensing of the European Seas, pages 103116, 2008.

[12] J.F. Berthon and G. Zibordi. Bio-optical relationships for the northern Adriatic Sea.International Journal of Remote Sensing, 25(7-8):15271532, 2004.

[13] F.S. Patt, R.A. Barnes, R.E. Eplee, B.A. Franz, W.D. Robinson, G.C. Feldman, S.W.Bailey, J. Gales, P.J. Werdell, M. Wang, R. Frouin, R.P. Stumpf, R.A. Arnone, R.W.Gould, P.M. Martinolich, J.E. OReilly, and J.A. Yoder. Algorithm Updates for theFourth SeaWiFS Data Reprocessing. 22:3440, 2003.

[14] J.F. Berthon, G. Zibordi, J.P. Doyle, S. Grossi, D. van der Linde, and C. Targa.Coastal Atmosphere and Sea Time Series (CoASTS). Part2: Data Analysis,page 25, 2002.

[15] G. Zibordi, D. DAlimonte, D. van der Linde, J.F. Berthon, S.B. Hooker, J.L.Mueller, G. Lazin, and S. McLean. The eighth SeaWiFS Intercalibration Round-Robin Experiment (SIRREX-8). page 39, 2002.

2222

Observation Networks and NRT Data Transmis-sion: Case Studies and Integration

G. Stanghellini1, M. Bastianini2, M. Ravaioli1, R. Colucci3, E. Paschini4,S. Carluccio1, G. Bortoluzzi1, T. Minuzzo2, C. Fonda5, P. Focaccia11, Institute of Marine Sciences, CNR, Bologna, Italy2, Institute of Mar

Documents

Observing system and forecasting