A new transportable floating mesocosm platform with autonomous sensors for real-time data acquisition and transmission for studying the pelagic food web functioning

394

Marine ecosystems are subjected to a variety of natural andanthropogenic perturbations that have been occurringincreasingly frequently over the past few decades. Mediumand long-term field observations using various observationalmethods have revealed relationships between changes inphysical and chemical parameters and biological shifts (abun-dance, biomass, and production), showing marine systemresponses to environmental change (Behrenfeld et al. 2006;

Boyce et al. 2010). However, field observations do not provideinsight into the mechanisms and processes involved in theresponse of marine organisms to these perturbations. Forstudying biological responses, an experimental design simu-lating a single perturbation, or combination of perturbationsapplied to an adequate isolated volume of water is required.Only a few milliliters or liters of water and only a few hours ordays of incubation are required for physiological studies at thecell level, such as growth rate of a phytoplankton species, orfor studying the interactions between two or more popula-tions at one or two trophic levels, such as microzooplanktongrazing rates.

Large enclosed systems that can be maintained for severalweeks are generally used for studying interactions betweenseveral trophic levels and the community responses to one ormore forcing factors (Grice and Reeve 1982; cited by Takeda etal. 1999). For studying at the communities and food webslevel, the required isolated water volume is more than 1 m3 upto 1000 m3. Such enclosed systems are commonly called“mesocosm” (Bloesch 1988).

A new transportable floating mesocosm platform withautonomous sensors for real-time data acquisition andtransmission for studying the pelagic food web functioningBehzad Mostajir1,2*, Emilie Le Floc’h1,2, Sébastien Mas1,2, Romain Pete1, David Parin1,2, Jean Nouguier1, Eric Fouilland1, and Francesca Vidussi11Ecologie des Systèmes Marins côtiers UMR 5119 ECOSYM (Université Montpellier 2, CNRS, IRD, IFREMER, UniversitéMontpellier 1), Montpellier, France2Centre d’écologie marine expérimentale MEDIMEER (Mediterranean Center for Marine Ecosystem Experimental Research)UMS 3301 (Université Montpellier 2, CNRS), Montpellier, France

AbstractWe describe a new transportable floating mesocosm platform with autonomous sensors. The platform has 9

separate units that can be transported by medium-sized research vessels and positioned in coastal waters. Thein situ mesocosms are equipped with a set of sensors for measuring water temperature, conductivity, chloro-phyll a fluorescence (Chl a), and dissolved oxygen concentration. It can take measurements every 2 min, storethese measurements, and transmit them in real time. Each mesocosm has a pump with regulated flow to mixthe water column. One of the floating units is used as an in situ observatory to monitor the water temperatureand Chl a in the water around the mesocosms as well as weather data and the incident light. The main datalogger on the platform sends all the data collected to a remote PC computer. This floating mesocosm platformwas successfully deployed in 2010 and 2011 in Mediterranean coastal waters (Thau lagoon and Cretan Sea,respectively). Simultaneous, automatic, high temporal resolution monitoring of physical, chemical, and bio-logical parameters in the mesocosms proved to be a powerful, noninvasive, and effective approach for i) mon-itoring the variations in physical and chemical parameters in real time and ii) assessing the short-term varia-tions in Chl a and the pelagic food web metabolism (e.g., the community respiration, gross primary production,and net community production) in the mesocosms without any manipulation of water samples.

*Corresponding author: E-mail: [email protected]

AcknowledgmentsWe should like to thank Dinet T., Geoffroy T., Roques C., Pitta P.,

Tsagaraki T., Simonelli P., Hennequin P., Captain M. Cantou of the SèteMarine Station, Captain M. Kokos and the crew of the F/R Philia, HCMRresearch vessel, for their assistance during the operations in the Thaulagoon (2010) and in the Cretan Sea (2011), and Pringault O. for fruit-ful discussions about diel oxygen technique. This research leading tothese results was funded under the European Union Seventh FrameworkProgram (FP7/2007-2013), grant agreement N° 228224, MESOAQUA.DOI 10.4319/lom.2013.11.394

Limnol. Oceanogr.: Methods 11, 2013, 394–409© 2013, by the American Society of Limnology and Oceanography, Inc.

LIMNOLOGYand

OCEANOGRAPHY: METHODS

Mesocosms are generally set up in a fixed location (e.g., TheNorwegian National Mesocosm Centre, University of Bergen,Williams and Egge (1998) and the Multiscale ExperimentalEcosystem Research Center, University of Maryland, Oviatt etal. 1984) and are mostly used to study local issues. A large vol-ume of water can be taken from nearby areas and transportedusing a boat or truck and used to fill the mesocosms (e.g.,Belzile et al. 1998). As there are few permanent marine meso-cosm facilities around the world, most of marine systems can-not be studied at food web level using existing mesocosm facil-ities. It was clear that there was a need for a mobile mesocosmplatform that could be deployed at various sites of interestwhere there is no nearby permanent mesocosm facility. Such amobile mesocosm platform can be deployed in differentmarine systems to carry out identical, comparable mesocosmexperiments with standardized protocols to show the responsesof a variety of marine systems to a given global perturbation.

In recent years, several mobile mesocosms have beendesigned and deployed in various marine ecosystems. For exam-ple, an enclosure bag (2.5 m diameter and 16 m deep) wasdeployed to assess the phytoplankton succession after nutrientsfertilization of an oligotrophic water in coastal North Pacific(Patricia Bay, Saanich Inlet, Canada, Suzuki et al. 1997). Two ofthese enclosures were also deployed in the same area to studymicronutrient dynamics in the plankton ecosystem (Nishiokaet al. 2001) and the effect of microzooplankton on a diatombloom (Suzuki et al. 2002). An Open-Ocean Enclosure (OOE, 2m diameter and 42 m deep, 132 m3) was used to study the roleof iron in the biological CO2 pump in the subarctic NorthPacific Ocean (Takeda et al. 1999). Eight mesocosm units sus-pended from a floating platform (21 m deep with an effectivevolume of 25 m3) were deployed to study the response of pico-phytoplankton to changes in light and nutrient manipulationin Johnson Cove, near the Spanish Antarctic Base Livingston(Agawin et al. 2002). Six clean mesocosms, 2.3 m diameter and14.4 m deep (52 m3) were deployed in the Western Mediter-ranean Sea (Corsica), to study the effect of the deposition ofatmospheric dust from Africa (Guieu et al. 2010). During thepast few years, the Kiel Off-Shore Mesocosms for Future OceanSimulations (KOSMOS, 2 m diameter and 15 to 25 m deep, witha volume of 50-75 m3) were developed and deployed to studyocean acidification at sites such as the Kongsfjord on the westcoast of Spitsbergen (Riebesell et al. 2012). Each KOSMOS can betransported by medium-sized research vessels and is designedfor operation moored and free-floating.

For mesocosm studies, it is useful to be able to monitorchanges over different temporal scales to track the physiologi-cal and ecological mechanisms involved in these changes. How-ever, it is logistically difficult to take samples from in situ meso-cosms, especially floating mesocosms, several times a day totrack daily variations over a medium to long-term experiment(several days or weeks). Automated sensors have been devel-oped to monitor physical, chemical, and biological parametersof interest with a high temporal resolution. These sensors have

been used for fixed stations as well as, recently, for gliders (John-son et al. 2009). The latest generation of automated sensorsmeasure the main parameters, such as water temperature, con-ductivity, and chlorophyll and oxygen concentrations, to pro-vide essential information about physical, chemical, and bio-logical variations (Mignot et al. 2011). Using a pack ofautomated sensors that monitor the physical, chemical, andbiological parameters during an experiment is an original, non-invasive way of obtaining ecological information. Until now,this approach has not been used for mesocosm experiments.

We present here an innovative, easily transportable floatingmesocosm platform that differs from the existing mesocosmsystems that have been described in the literature as it has auto-mated sensors recording physical, chemical, and biologicalparameters every 2 min in real time. We demonstrate the use ofautomated sensors in the mesocosm experiments providingimportant information about metabolism parameters andabout the control of the water column mixing over the courseof an experiment. The parameters measured by the sensors werealso measured in the water around the mesocosms togetherwith weather data and incident light, using the same real-timestorage and transmission protocol. This mobile mesocosm plat-form was successfully deployed in the shallow coastal waters ofthe Thau Lagoon (northwestern Mediterranean Sea) and in theCretan Sea (eastern Mediterranean Sea). As an example, chloro-phyll fluorescence and oxygen concentration, monitored every2 min over a 17-d experiment carried out in 2010, are presentedand discussed. The use of automated sensors in mesocosm stud-ies provided new insight into the short-term variation in a phy-toplankton biomass proxy (chlorophyll fluorescence) as well asinto the community respiration, gross primary production, andnet community production (derived from oxygen sensor mea-surements) in the mesocosm experiments.

Materials and proceduresFloating units and cover domes

The mobile mesocosm platform is composed of 9 floatingunits that can be deployed separately or connected in theform of a raft depending on the experimental design (Fig. 1).This platform is called Lite aquatic Automated MesocosmsPlatform (LAMP). Each floating unit has 4 side sections madeof two tubes of high-density polyethylene (HDPE) heldtogether at one end by short lengths of tube welded at right-angles (Technosea SRL). The sections are 2.971 m long, 0.944m wide, and weigh 90 kg. The diameter of the tubes is 0.25 m,and the two tubes are 0.25 m apart. Each section has a slidingupright bracket (0.125 m diameter, 0.593 m high) for mount-ing the handrail. At each end, the HDPE tubes have circularflanges (0.36 m diameter), which are used to connect the sec-tions together into a square with 12 stainless steel bolts perflange (total of 96 bolts for one unit). The unit has outsidedimensions of 3.915 m square and a total weight of about 400kg when fully operational. The handrail is a pipe (diameter0.11 m) mounted on the 4 upright brackets, at a height of 1

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

395

m. The handrail is used by scientists when taking samples andis also used for attaching the mesocosm bag. Each unit occu-pies 15.3 m2, with a central opening of 2.03 × 2.03 m and aheight of 1.5 m (including the buoyancy components andhandrail). The mesocosm bag is lowered through the centralopening. All 9 units can be used for in situ mesocosms, allow-ing the replication of the experimental conditions dependingon the experimental design. The structures can be easily tiedtogether with straps, separated by plastic fenders. Connectedtogether, the 9 units form a complete mesocosm platformmeasuring about 40 m long (9 units plus fenders) and 4 mwide (Fig. 1 showing 7 units with 6 in situ mesocosms).

The units can be easily assembled and dismantled for stor-age in a standard 40 feet intermodal-shipping container. Theycan be transported easily on medium-sized research ships anddeployed at sea or a small boat can tow and deploy a set of upto 3 units near the coast.

Transparent (crystal clear PVC film) domed covers wereused to prevent any atmospheric contamination and ensurethat no undesirable material found its way into the mesocosmbags during the experiment. The frame for the cover was madeof 4 stainless steel arms joined together in the middle by astainless steel disc and fixed to the handrail in 4 places usinga clamp. Each arm had a ring for suspending the mixing sys-tem and sensors. The bottom of the cover overhung thehandrail providing extra protection against water spray anddriving rain. There was also a ventilation system to preventcondensation due to the greenhouse effect on sunny days.One segment of the cover had a zip to give access to the meso-cosms for sampling and for checking the equipment inside.Setting up, filling, and closing the bags

LAMP was successfully deployed in 2010 and 2011, and thefinal, improved configuration (2011 deployment) is describedbelow, and the changes from the 2010 version are summarizedin Table 1. The mesocosm bags were made of two transparent,UV stabilized, 200 μm thick vinylacetate mixed polyethylenefilms separated by a reinforcing nylon mesh (Insinööri-

toimisto Haikonen Ky). The bags were 2 m diameter and 6.5m high (Fig. 2).

The walls had four vertical channels holding 10 mm ropesused to attach the bag to the handrail at the upper end andattach at the bottom end to a hatch. Four horizontal channelsheld stainless steel rings to maintain the cylindrical shape.The rings were made in 4 sections assembled into a 2 m diam-eter ring by stainless steel plates and bolts. The ring sectionswere inserted into the four horizontal channels before beingassembled. The bottom of the bags was conical, tapering from2 m diameter down to 1 m diameter.

The closing system consisted of a polyethylene hatch (0.8m diameter) attached to the conical end of the bag with apolyethylene fastening belt that pinched the plastic bag inbetween an inner and outer section. The hatch door, contain-ing a stainless steel plate, acted as a weight and caused the bot-tom of the bag to sink (Fig. 2).

The hatch door was left open and bags were lowered downto 6.5 m. Once the bags had reached the depth required, theupper aperture was pushed below the water surface to allowwater circulation through the bags for at least 48 h. This cir-culation was increased by the mixing pump placed at middledepth in the bag. When the water column had been thor-oughly homogenized, the upper part of the bag was raised andattached to the handrail and the hatch door was closed bydivers. In this way, all the mesocosms were closed within lessthan 30 min to prevent discrepancies between mesocosms. Itwas 4 times faster compared with the 2010 procedure wheremesocosms had flat, closed bottoms and were filled by lower-ing the bags to a depth of 5 m, letting the water enter themesocosms and pulling the bags up (Table 1).Continuous automatic measurement, storage, and trans-mission of physical, chemical, and biological parameters

Automated sensors in the mesocosms and in the sur-rounding waters

During the two deployment tests, a pack of optical andphysical, chemical, and biological sensors was installed in

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

396

Fig. 1. Lite aquatic Automated Mesocosm Platform (LAMP) deployment in the Bay of Dia Island in the Cretan Sea (eastern Mediterranean Sea), Sep-tember 2011.

three of the mesocosms at 2.5 m depth. Each pack comprised:two Combo fluorometer-scattering meters (2 scattering and 1fluorescence, BB2FL ECO-Puck, Wetlabs), an oxygen sensor(Aanderaa, Optode 3835 and 4175, Data Instrument AS) andan electromagnetic induction conductivity sensor (Aanderaa

3919, Data Instrument AS). In addition, three temperaturesensors (Thermistor probe 107, Campbell Scientific Ltd.) wereset up at 0.5, 2.5, and 4.5 m. The set of sensors was installedin a carrier to make it easier to immerse them in the mesocosmand reduce the risk of perturbations due to the sensors touch-

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

397

Table 1. Different characteristics of LAMP 2010 and 2011 deployments.

2010 2011 Sites Coastal lagoon (Thau, south of France),

NW Mediterranean Sea Marine Bay (Dia Island, north of Crete), SE Mediterranean Sea, Cretan Sea

Trophic status Mesotrophic Oligotrophic Site depth 8 m 9-31 m Number of units deployed 9 7 Boat used for deploying the units and associated equipment and discrete sampling

Small coastal boat Length: 7.5 m Beam: 2.8 m GT: 2 tons 115 bhp outboard motor, with center console

Coastal oceanographic boat (Philia) Length: 26.10 m Beam: 7.25 m Freeboard: 3.20 m GT: 143 tons, Deck Area: 35 m2, 1 Hiab Sea Crane 8.30 m boom

Method of transporting units to the sites

Up to 3 floating units towed at a time 3 to 4 units on deck deployed using the crane

Platform anchorage 6 x 25 kg anchors (2 at each end and 2 evenly spaced each side) and 30 m cable comprising 10 m steel chain and 20 m rope

4 x 400 kg concrete blocks and 4 x 30 m cables, longest distance between mooring 110 m, 4 yellow marker buoys to signal the platform

Bag characteristics 6 cylindrical bags with flat bottom Height: 6.5 m Diameter: 2 m

6 cylindrical bags with conical bottom Height: 6.5 m Diameter at the top: 2 m Diameter at the bottom: 1 m

Mesocosm filling procedure

Bags with closed flat bottom lowered and then pulled up

Bags lowered with bottom open, closed by divers after 2 days

Sampling system Niskin bottle (10 L): Samples were taken at three different depths (0.5, 2.5 and 4.5 m) and mixed

Manual suction: Three silicone tubes placed in each mesocosm (0.5, 2.5 and 4.5 m), a tube was connected to a 20 L carboy which was in turn connected to a manual air/water pump. The sampling method did not require any power and prevented contamination during sampling

Maximum of diel difference between the surface and bottom temperature of the 3 instrumented mesocosms: homogeneity of water column

1.52°C: less stabilized homogeneity 0.70°C: high stabilized homogeneity by regulation of the pump voltage

ing the wall of the bag, in particular the Wetlabs optical sen-sors. Three of the other mesocosms had three temperature sen-sors only which were at the same depths as in the other meso-cosms. The water temperature measurements at three depthsgive an indication of the temperature uniformity of the watercolumn in the mesocosm, and therefore, the effectiveness ofthe mixing pump.

The Wetlabs ECO-Puck sensors combine single-angle(117°) backscatter meters at two wavelengths and a singleexcitation (Ex) and emission (Em) fluorescence response. Bysetting two of these sensors in each mesocosm, fourbackscatter measurements (880, 650, 532, and 470 nm) andtwo fluorescence responses (colored dissolved organic mat-ter, CDOM fluorescence: 370/460 nm for Ex/Em; and Chl afluorescence: 470/695 nm Ex/Em) are provided. Thebackscatter meters and fluorometers allowed to have the fac-tory calibrated output values for the volume scattering coef-ficient β (117°, λ) (sensitivity: 10–5 m–1 sr–1), the Chl a con-centration (sensitivity: 0.01 μg L–1) and the CDOMconcentration (sensitivity: 0.09 relative units). In the presentpaper, only results for the continuous Chl a fluorescencemeasurements are presented and discussed. The instrumentprovides an output directly in concentration units (μg L–1)using a factory calibration curve.

One of the units was used as an “in situ observatory.” It wasequipped with a set of sensors immersed in natural water out-side the mesocosms. The sensors included three temperatureprobes (Campbell Scientific Ltd.) at the same depths as thetemperature probes in the mesocosms and a fluorometer(ECO-FLSNTU, Wetlabs) at 2.5 m depth measuring the Chl afluorescence (470/695 nm Ex/Em) and turbidity. The “in situobservatory” also had a weather station (Weather TransmitterWX520) recording the air temperature, relative humidity,atmospheric pressure, wind direction, and speed and precipi-tation (accumulated, duration, and intensity) and 3 light sen-sors (Skye, England) measuring incident photosyntheticallyactive radiation (PAR: 400-700 nm, Quantum SKP 215) andultraviolet A and B radiation (315-380 nm, SKU 420 and 280-315 nm, SKU 430, respectively). A solar panel (SX320, BPSolar) supplied the energy for the “in situ observatory.”

Before deployment, the three Wetlabs ECO-Puck sensorsand the “in situ observatory” fluorometer (Wetlabs Fluorome-ter, ECO-FLSNTU) were cross-calibrated with Chl a measuredby a spectrofluorometer (LS 45, PerkinElmer).Regulated mixing system

Each mesocosm was equipped with a pump to ensure a uni-form water column obviating the need for several sensors atdifferent depths. The 12V 11W pump (Model 5001) had a flowof 1250 L h–1, a blade wheel flow meter (BamoFlu 100) andPVC and crystal PVC pipe (22 mm diameter). The pump was50 cm below the surface, and the water was pumped from thesurface to the bottom of the mesocosm at 4.5 m depth.Control module, data acquisition, storage, and transmission

The LAMP sensors in each mesocosm were connected bywatertight cables and connectors to a secondary data logger(CR1000 Campbell), which was included in a watertight cylin-der installed at the base of the handrail upright.

The secondary data loggers stored the data from the sensorsand transmitted it (radio transceiver ARM-U8 868 MHz, 25mW range 500 m, ATIM) to the main data logger (CR 3000Campbell) on the “in situ observatory” (Fig. 3). The main dat-alogger stored copies of the data from the secondary datalog-gers on a 2Gb Compact Flash card for data security and to cen-tralize the data. The main data logger also controlled thegeneral operation of the sensors and could be used for relay-ing messages for adjusting the sensors. On request from aremote PC on the research vessel or on land up to 5 km away,the main datalogger transmitted all the data collected (radiomodem SATEL 869 MHz, 0.5 W). In addition, some adjust-ments, such as changes in pump flow rate to adjust the meso-cosm turnover time, could be transmitted from the remote PCto the main datalogger to be relayed to the appropriate sec-ondary dataloggers. The Campbell PakBus network protocolwas used for all data transmission between the CR1000 sec-ondary dataloggers, the CR3000 main datalogger, and theremote PC.

The main data logger also controlled the sensors in the “insitu observatory,” the navigation sensors (GPS, tilt-compen-

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

398

Fig. 2. A single unit with mesocosm bag and associated equipment.

sated 3-axis compass) and the radio communications qualityby measuring the RF signal level (RSSI).

The sensors were controlled directly by software in the dat-aloggers. Each datalogger recorded measurements in tables infiles at specified intervals. Each row of the tables had a timestamp followed by all the parameter values separated by com-mas. The measurement and storage frequencies were not nec-essarily the same. Each value of the table could, therefore, bea single measurement or the average of several measurements.In this case, the minimum value, maximum value, and stan-dard deviation were also stored. Each variable should berecorded at a frequency relevant for the speed of fluctuation.

Each secondary datalogger (CR1000) had 6 serial interfaces,16 analog inputs, and 1 switch input. For all the mesocosmsensors, including the flow meters, the measurements werestored every 2 min.

The main datalogger (CR3000) managed the RS232 naviga-tion and communication units (OEM GPS, Garmin; TCM2.5tilt-compensated 3-axis compass module PNI; radiomodemSATEL; and OEM micropower radio modem ATIM). In addi-tion, the weather data were stored at 10 min intervals. Thewind direction was corrected in real time using the compassreading. Analog sensors were used to measure water tempera-ture and PAR. The incident radiation was measured using aplatform with gimbals and software to keep the sensors as hor-izontal as possible using the angular measurements from the

3-axis compass and to correct for bias in the irradiance mea-surement. Additional serial and analog sensors can be addedwith Campbell SDM-SI01 interface modules (RS232, RS485,RS422 protocols) and AM16/32 Analog Multiplexers.Multi-sensor approach: applying the diel oxygen tech-nique to mesocosms

Combining salinity, temperature, and oxygen continuousmeasurements at the same time in one mesocosm enable theestimation of community respiration (CR), gross primary pro-duction (GPP), and net community production (NCP = GPP –CR) using the ‘free-water’ diel oxygen technique. The mixingby the pump ensures a homogenous water column, and there-fore, the physicochemical characteristics at 2.5 m depth arerepresentative of the entire mesocosm.

The dissolved oxygen concentration was measured every 2min. The concentrations were post-corrected for salinity andtemperature as measured by the conductivity and temperaturesensors.

The diel oxygen technique for the calculation of commu-nity metabolism is adapted from a detailed procedure byStaehr et al. (2010). The air-water gas flux (F) was calculatedfrom the concentration difference between the water and theatmosphere (as in Garcia and Gordon 1992) multiplied by theexchange coefficient k (also called the piston velocity) and bythe inverse of the mesocosm water column height (to obtaina volumetric gas flux). The k is usually modeled using the

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

399

Fig. 3. Data acquisition and transmission schema for the mesocosm units and in the in situ observatory.

wind speed but as the mesocosms were covered by a roof,there was no significant wind effect. However, wave propaga-tion through the walls and pump mixing induced turbulenceand so k could not be considered to be negligible. The k wastherefore taken from the literature where oxygen transfervelocities between air and water were experimentally mea-sured as a function of turbulence in mesocosms. Two extremevalues for k, 0.002 m h–1 (from Alcaraz et al. 2001) and 0.029m h–1 (from Petersen et al. 1998) were used in the presentstudy to give a realistic range of community metabolismparameters.

The sum of all others processes that could induce changesin oxygen concentrations (organic matter photo-oxidation,diffusion through mesocosm plastic wall) are taken to be neg-ligible.

Dissolved oxygen concentrations over a 24-h period fol-lowed two main patterns: an increase in the first part of thelight period (in this study over about 8 h from 1-2 h afterdawn to 1-2 h after zenith, Fig. 4) and a decrease over the sec-ond part (16 h during this study, Fig. 4). The first derivative offunctions fitted to the increasing and decreasing periods of thedaily variations (data were first smoothed by moving averageof nine successive values) was used to calculate for each timestep the evolution in oxygen concentration (Δ[O2]w/Δt). In thiscase, the best fit was using sigmoid functions (SigmaPlot soft-ware version 10.0), but depending on the plankton commu-nity, other functions can be applied to the oxygen time course(Pomeroy et al. 1994; Briand et al. 2004; Pringault et al. 2007).

The instantaneous net community production (NCPh) was thesum of the fitted Δ[O2]w/Δt curve and smoothed F values (mov-ing average of nine successive values) at each time step. TheCR rate during nighttime (CRdark) was assumed to be equiva-lent to the average of instantaneous community respirationrate (CRh) calculated from sunset to sunrise (from 21:30 to06:00 local time).

Departing from Staehr et al. (2010) and following Pringaultet al. (2007), the CR rate during daytime (CRlight) was not con-sidered as equal to nighttime respiration. It was assumed to beequivalent to the CRh rate averaged over the 1 h period justafter dark when microbial respiration still relies on the photo-synthesis products released during daylight (Markager et al.1992).

Considering the daily CR rate over 24h (CRd) as a sum ofCRdark and CR light, daily GPP (GPPd) and daily NCP (NCPd =GPPd – CRd) rates were then calculated.

Fig. 4 shows the results of fitting the data from the contin-uous oxygen measurements.Deployment tests and experimental design

Two tests were carried out successfully in western and east-ern Mediterranean coastal waters in May-June 2010 and inSeptember 2011, respectively (Table 1). During the twodeployments, the platform was anchored as it was deployednear the coast. Thau is a productive coastal Mediterraneanlagoon on the south coast of France, whereas the Cretan Sea isan oligotrophic area of the Mediterranean on the south ofGreece. During both deployment tests, phosphorus was added

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

400

Fig. 4. Diel variations of continuous oxygen measurements (dark blue dots), theoretical oxygen concentration at saturation (red line), modeled in situconcentration (light blue line), and NCPh estimates for low (light green line) and high (dark green line) air-sea gas exchange coefficients. The gapbetween the two periods of the NCPh estimates is a result fitting the increasing and decreasing parts of the oxygen concentration curve separately. Datafrom day 7 of the experiment in mesocosm 1 (see below). Shaded areas represent the dark period.

to trigger changes in the chemistry and biology of theenclosed water column and create differentials between thesensors in the different mesocosms. Phosphorous was chosenas it is recognized that the western Mediterranean basin ismoderately phosphorous limited and that the easternMediterranean basin is extremely phosphorous limited (Kromet al. 1991; Thingstad et al. 2005).

As this article only presents and discusses results from the2010 deployment, only the 2010 experimental design will bedescribed. The 3 mesocosms with sensors received differenttreatments: the first mesocosm did not receive any addednutrients, the second mesocosm received organic phospho-rus (Glycerol-3-Phosphate, final concentration in the meso-cosm 1.5 μM), and the third mesocosm received inorganicphosphorus (KHPO4, final concentration in the mesocosm1.5 μM). The three mesocosms were sampled on days 1, 2, 3,4, 5, 6, 7, 10, 13, and 17. Samples were not taken for oxygenconcentration and incubation on day 17. Samples weretaken from 09:30 to 10:30 at three different depths (0.5, 2.5,and 4.5 m) with a 10-L Niskin bottle. For each mesocosm,the water samples were mixed and stored for a maximum of2 h in darkened 10-L and 20-L acid-washed carboys beforesubsampling.Chl a analysis and measurement using HPLC

For Chl a analysis, 1 L water was subsampled from the 20 Lcarboy from each mesocosm and filtered on glass fiber filters(25 mm, GF/F, Whatman), frozen in liquid nitrogen andstored at –80°C until analysis by high performance liquidchromatography (HPLC). Pigments (chlorophylls includingChl a and carotenoids) were extracted in 2 mL of 95% MeOH(1 h at –20°C) then sonicated for few seconds, extracted for anadditional hour at 4°C, and then filtered onto a glass fiber fil-ter (25 mm, GF/F, Whatman). The pigments were analyzedusing the method described by Zapata et al. (2000) with fewmodifications to the HPLC system as described by Vidussi etal. (2011). The HPLC system comprised a pump (600 E,Waters) equipped with a 200 mL loop, an automatic injector(717plus Autosampler, Waters) and a photodiode detectorarray (2996 PDA, Waters). Chlorophylls were identified usingretention time and absorbance spectra, and the concentra-tions were calculated using absorption at 667 nm. The HPLCsystem was calibrated using commercial standards (DHI andSigma).Oxygen production and respiration measurements usingthe Winkler method

For oxygen measurements and related incubations, nine120 mL borosilicate bottles were carefully filled from each car-boy using a silicone tube. Three bottles were immediatelyfixed (time zero) using reagents prepared following the rec-ommendations of Caritt and Carpenter (1966). Three otherbottles were incubated in the surface waters of Thau Lagoon(0.5 m depth) under natural light from 09.00 to 18.00. Threedarkened borosilicate bottles were incubated from 09.00 to18.00 in a temperature-controlled bath set to the same tem-

perature as the lagoon. After incubation, all six bottles werethen fixed as described above. Dissolved oxygen concentra-tions were determined using an automated Winkler titratorbased on potentiometric determination (Crisson). Then, theNCP, CR, and GPP rates were calculated and expressed in g O2

m–3 d–1. It should be noted that to express the rates in dayunits, GPP was multiplied by the length of the day (15.5 h)and CR was multiplied by 24 h.

AssessmentAutonomous continuous recording and transmission ofphysical, chemical, and biological parameters

Physical and chemical parametersMonitoring the mesocosm water temperatures every 2 min

during the experiment showed a daily average increase from16.5°C on day 2 to 20°C on day 8, then a decrease to 17.8°Con day 14, increasing again to 22.3°C on day 17 (Fig. 5A, 5B,and 5C). Diel variations of water column temperature in themesocosms followed that of the natural water (Fig. 5D) beingaffected by the air temperature.

The conductivity in the mesocosms 1, 2, and 3 showed anincrease in salinity of 0.07, 0.39, and 0.02, respectively (datanot shown). It should be noted that the water temperatureand salinity values are required for post-correction of the oxy-gen measurements.

The measurements of oxygen concentration showed aslight decrease during the first 2 days and then remained con-stant until day 9 (Fig. 6). The oxygen concentrations thendecreased until day 12 and either increased (mesocosm 2) orremained more or less stable (mesocosms 1 and 3) until theend of the experiment. Diel oxygen variations always showedthe same pattern with an increase of oxygen from dawn to themiddle of the afternoon and then decreasing (Fig. 4).Biological parameters: example of phytoplankton Chl a

The Chl a fluorescence measured by the sensors was com-pared with Chl a concentrations in samples measured byHPLC. The measurements from the sensors were averaged overthe period when the samples were taken (09:30 to 10:30 localtime). For all 3 mesocosms, the relative standard deviation ofthe Chl a fluorescence measurements over the samplingperiod ranged from 1.8% to 6.1% showing that the fluores-cence was stable during this period. The measurements fromthe sensors were well correlated with the HPLC measurementsfor all the mesocosms (n = 29, r = 0.91, P < 0.00001, data notshown) showing that the Chl a fluorescence sensor mea-surements were consistent.

The Chl a fluorescence measurements during the experi-ment in the three mesocosms are given in Fig. 7A and showthat Chl a fluorescence decreased during the first 2 d, asalready showed for oxygen concentration (Fig. 6), which cancorrespond to the requested time for organism acclimation tomesocosm conditions. Then, Chl a fluorescence increaseduntil day 9 or 10, indicating the start of phytoplanktonbloom. Thereafter, Chl a fluorescence decreased until day 17

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

401

suggesting the post-bloom phase. The different patternsbetween Chl a fluorescence (Fig. 7) and oxygen concentration(Fig. 6) can be explained by the fact that oxygen concentra-tion is not only affected by phytoplankton variations, but alsoby other biological parameters (i.e., community respiration)and physicochemical ones (salinity, temperature, pH). At thesame time, the Chl a fluorescence is influenced not only byphytoplankton variations in abundance, but also by cellularphysiological acclimation of phytoplankton to the environ-ment as explained hereafter.

There was a clear diel pattern (Fig. 7B shows day 7 for meso-cosm 1) with the Chl a fluorescence doubling in the 14 h start-ing 3 h after dawn and then falling back to the original valueover the next 10 h. The increase of Chl a from dawn to duskcan be attributed to Chl production and, possibly, phyto-plankton biomass accumulation whereas its decrease fromdusk to dawn suggests a loss in Chl related to a combinationin biomass loss due to grazing and to photophysiological cellacclimation and/or cell cycle (Neveux et al. 2010).

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

402

Fig. 5. Continuous monitoring of water temperatures every 2 min in the mesocosms (A, B, and C) during the experiment. Diel variations of the natu-ral water column temperature outside the mesocosms (D). Surface, mid-depth, and bottom are at depths of 0.5, 2.5, and 4.5 m, respectively.

Fig. 6. Oxygen concentration during the experiment in the three mesocosms.

The continuous Chl a measurement also provides an inter-esting insight into physiological phytoplankton processes. Forinstance, the slight dip in fluorescence detected at noon(around 13:45, local time) may be due to the nonphotochem-ical quenching as a photoprotection mechanism under stronglight (Xing et al. 2012, Sackmann et al. 2008; Cullen and Lewis1995). Photoprotection mechanisms only occur over veryshort periods when the natural incident light is at its greatestand can only be detected by continuous measurement.

To improve the interpretation of the data and elucidate thedetails of diel variations of the key parameters in the systemfunctioning, more information is required about some param-eters. For instance, one sample was taken a day in the morn-ing for HPLC analysis but, because photoacclimation canoccur during the daytime, several discrete Chl a measurementsby HPLC will be necessary to show the potential role of pho-toacclimation so that the fluorescence sensor data over a 24-hperiod can be fully interpreted.Community metabolism

Simultaneous measurements of temperature, conductivity,and dissolved oxygen in the water enclosed in the mesocosmswere used to derive several metabolic parameters, such as com-munity respiration (CR), net community production (NCP),and gross primary production (GPP).

The highest NCPh during daylight was often around zenith(13:45 local time) or 1 to 2 h before depending on the effectof cloud cover on PAR with cloud cover (data not shown). Thelowest values of NCPh during the night (coinciding with thehighest CRh) were measured from around sunset (21:30) to 3-4 h after sunset.

Based on the values from sensors or by Winkler method, nosignificant difference between the mesocosms was found forCRd, GPPd, NCPd.

Comparison between community metabolisms based oncontinuous oxygen measurements with those of Winkler incu-bations showed no correlation between them (Fig. 8). The val-ues of CRd and GPPd from continuous oxygen measurementswere almost always 2-3 times higher than those measuredusing Winkler incubations, whereas NCPd from continuousoxygen measurements were lower except in 3 cases (Fig. 8).The use of the lower k value (0.002 m h–1, Alcaraz et al. 2001)or higher k value (0.029 m h–1, Petersen et al. 1998) did notsubstantially affect the daily CRd and GPPd rates as shown inFig. 9 whereas it did affect NCPd.

The dynamics of CRd, GPPd and NCPd based on the contin-uous oxygen measurements using the lower value of the air-sea exchange coefficient k (0.002 m h–1) for the duration of theexperiment is presented as an example for mesocosm 2

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

403

Fig. 7. Chlorophyll a fluorescence during the experiment in the three mesocosms (A) and the diel variation of chl a fluorescence over a full 24 h period(day 7 for mesocosm 1) (B).

Fig. 8. Comparison of CRd, GPPd, and NCPd calculated using Winklerincubations and continuous oxygen measurements.

(Fig. 10) as all the mesocosms had the same trend (notshown). Both CRd and GPPd increased after day 4, reachingtheir highest values on day 14. The tracking between CRd andGPPd with NCPd values fluctuating around zero, indicates abalanced ecosystem metabolism up to day 10 when a peak ofChl a was observed. After day 11, NCPd was permanently pos-itive showing an autotrophic-dominated community eventhough the Chl a concentration was already decreasing.

Discrepancies between metabolism estimates from contin-uous measurements and incubations are not surprising andhave often been reported in the literature (e.g., Garcia-Martinet al. 2011). On the one hand, the so-called bottle effect

(divergence of an isolated community, absence of mixing),using a two-point calculation for oxygen concentration (dif-ference between an end value and a starting value overlookingthe variations in between), the depth of incubation and esti-mating the respiration from dark incubations only are factorsthat will affect estimates from incubation. On the other hand,assuming a constant respiration rate in the light and in thedark, assuming a constant coefficient for oxygen flux betweenair and sea and inferring estimates for the whole mesocosmfrom a measurement at mid-depth (even if the water columnis well mixed) can affect the estimates from continuous oxy-gen measurements.

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

404

Fig. 9. CRd (A), GPPd (B), and NCPd (C) calculated using Winkler incubations (left-hand side boxplots), continuous oxygen measurements using thelower value of the air-sea exchange coefficient k (0.002 m h–1, middle boxplots), and higher k (0.029 m h–1, right-hand side boxplots) (Sigmaplot).

Assuming CRlight to be constant and equal to sunset respira-tion values will underestimate GPPd and CRd. However, theeffect on NCPd would be smaller as the underestimation ofGPPd is partly compensated by the underestimation of CRd.These results from continuous dissolved oxygen mea-surements confirmed the conclusions of previous studies (e.g.,Pringault et al. 2009) that there is a need for an alternativemethod of measuring community respiration rates in the fieldas has already been developed for in vitro measurements(using the production of 18O2 in the light, Robinson andWilliams 2005 and references therein).

The rate of air-sea oxygen flux could be better determinedusing a tracer gas (Czerny et al. 2013a) combined with correctestimation of the mesocosm volume (Czerny et al. 2013b).Water column homogeneity and mixing regime

The water column in the mesocosms was mixed by animmersed electrical pump. The mixing efficiency waschecked by comparing the water temperature at three depthsin the three mesocosms (Fig. 5). A temperature gradientinside the mesocosms appeared when there was stratificationwith a maximum difference of 1.52°C because of a lack ofpump regulation. The pump voltage was, therefore, con-trolled for the 2011 deployment which reduced the maxi-mum difference between the surface water and bottom waterto 0.70°C.

The control of turnover in the mesocosm is an improve-ment in the mixing regime rarely tried in in situ meso-cosms, although the mixing of the water column affectsmost of the biological components (Weithoff et al. 2000).For example, a change in mixing regime can shift the phy-toplankton community composition, and therefore, ecosys-tem functioning by changing the availability of nutrientsand altering UVBR photoadaptation capabilities (Fouillandet al. 2003).

Electric power requirement and managementThe autonomy of the mesocosm sensors and other plat-

form equipment depends on the electrical power requirementand how it is supplied. The power can be supplied to the maindatalogger and its sensors for a 3-week experiment using a 40W solar panel with a controller SS-10L-12 (Morningstar CorpUSA) and a 36 Ah 12 V battery. This power supply arrange-ment is also suitable for one mesocosm secondary dataloggerwith its set of sensors for a sampling interval of 10 min.

The mesocosms in this platform have a pump to ensurethat the water column is uniform as part of the experimentaldesign in this study. The power requirement for a mesocosmpump providing a turnover of 30 m3 d–1 is about 11 W at 12VDC and a standard 12 V 70 Ah car battery was used to powerthe water column mixing but it was exchanged regularly(every 2-3 d) for a fully charged battery. More power isrequired for the mixing pumps than the electronics in theLAMP. Off-the-shelf solar panels cannot easily supply enoughpower given the limited surface area available to avoid shad-ing the mesocosms. In addition, an alert system is required toallow for the mesocosm experiment to continue on cloudydays. In this case, the batteries must be changed every fewdays as mentioned above. An autonomous power supply usinghigh capacity Zn/air batteries should be considered to signifi-cantly reduce the weight and volume of the batteries (29 kg /20 dm3) required for 21 days continuous operation withoutbattery replacement.

DiscussionToward highly autonomous mesocosm experiments

There is no doubt that experiments carried out in meso-cosms can provide original results in aquatic ecology, espe-cially for the effect of global climate on the marine food web,such as an increase in ultraviolet-B radiation (e.g., Mostajir etal. 1999), an increase in CO2 and acidification (e.g., Riebesellet al. 2007), an increase in water temperature (e.g.,Lewandowska and Sommer 2010), and simultaneous increasesof water temperature and ultraviolet-B radiation (Vidussi et al.2011). New significant fundamental ecological concepts. suchas food web functioning. have emerged from results of themesocosm experiments (i.e., Thingstad et al. 2007). However,to perform the same mesocosm experiment in differentaquatic systems where there is no suitable infrastructure,requires a transportable mesocosm platform.

LAMP, a new floating transportable mesocosm platformand associated automated sensors, was designed, constructed,and then deployed successfully in two different coastal sites:an enclosed, shallow, mesotrophic site (Thau lagoon) and amore open, deeper, oligotrophic site (a bay in the Cretan Sea).As LAMP is easily transportable and equipped with automatedsensors that can measure a given set of parameters in an iden-tical and reproducible manner, it can be used to compare theresponses of the aquatic food web to forcing factors, for exam-ple those resulting from global climate change, by carrying

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

405

Fig. 10. The dynamics of CRd (red vertical bars), GPPd (green verticalbars), and NCPd (blue line) for the whole of the experiment based on thecontinuous oxygen measurements using the lower value of the air-seaexchange coefficient k (0.002 m h–1) for mesocosm 2.

out the same reproducible experiment and standardized pro-tocols in different aquatic ecosystems. Results of identicalmesocosm experiments undertaken in tropical, temperate,polar marine and freshwater systems will help to find localrules as well as generic rules for the response of aquatic sys-tems to the forcing factors studied.

In addition, the non-invasive, autonomous set of sensorsdeployed in the mesocosms provided a new insight into meas-uring key metabolism parameters such as CR, GPP, and NCP.

For future studies, once instantaneous GPP estimates canbe calculated using realistic CRlight values, it will be possibleto estimate instantaneous primary production rates nor-malized to biomass using continuous Chl a measurementsusing fluorescence sensors at the same time as particulatecarbon measurements using backscatter sensors (Alkire etal. 2012). Simultaneous PAR measurements make it possibleto estimate the real daily primary production rate in eachmesocosm.

In future mesocosm experiments, additional sensors bothin the mesocosms and in the “in situ observatory” will be ableto provide very useful details about the response of other envi-ronmental parameters and food web dynamics to a given forc-ing factor. Combining the data from a nutrient sensor (e.g.,Cycle Phosphate Sensor, Wetlabs; SUNA UV nitrate sensor, Sat-lantic) with the data from the physical, chemical, and biolog-ical sensors may help to provide a better understanding of thediel variations of various parameters (e.g., Chl a fluorescence,oxygen concentration, and calculated metabolic parameters).However, there are many other sensors for studying microor-ganisms and particle dynamics, such as submersible flowcytometers (Olson and Sosik 2007; Thyssen et al. 2008), in situimagery systems (Gorsky et al. 2000; Lunven et al. 2012), andsubmersible laser particle size analyzers (Agrawal and Pott-smith 2000), as well as a new generation of sensors providinginformation about aquatic microorganism diversity (e.g.,Environmental Sampling Processor (ESP), Scholin et al. 2006),which could help to increase and deepen our knowledge ofmarine ecology. However, most of these sensors cannot be fit-ted into mesocosms, because of their size, the need to cleanthem frequently, especially in eutrophic waters, or their highpower consumptions. For future mesocosm applications, notonly do the sensors need to be miniaturized but also the elec-tronic circuits.

The “in situ observatory” has several sensors that providethe physical, chemical, and biological characteristics of thenatural water surrounding the mesocosms. Comparing the“in situ observatory” with the mesocosm data, especially datafrom the control mesocosm (natural water in the mesocosmwithout applying the treatment), can indicate any discrep-ancy between changes to the natural water and the waterconfined in the mesocosms. This information is not usuallyavailable and is very useful in interpreting the mesocosm dataand providing the limits to the extrapolation of the meso-cosm results.

Beyond several advantages of the transportable floatingmesocosm platform with autonomous sensors presented inthis article, there are also some limitations and unknownpoints. We made the choice of a light-floating structure withthe buoyant capacity on the waterline. The raft can thereforebe used as platform to facilitate the access for scientists and fortechnical maintenance on sensors and data loggers and sam-pling. This kind of design is not optimal to tolerate big wavesin open waters and the main risk is the bags being torn due toopposite forces. Therefore, LAMP deployment could be limitedto sheltered areas or periods of the year with moderate waveaction. The other constraint related to the deployment of thepumps, sensors, and dataloggers is the power supply auton-omy, because manpower is frequently necessary to change thebattery pack. The main energy demand is due to the pumpsinsuring water mixing (12 W h–1 per mesocosm unit). Forexperimental designs, where no water column mixing is nec-essary, and only the sensors are operated (1 W h–1 per meso-cosm unit), power supply with solar panels will suffice. Itshould be noted that it is unknown in which extent theLAMP’s power supply could be affected by the temperaturearound or below zero. The filling procedure has proven to beimproved in efficiency and rapidity by using open bottombags fitted with closing hatch. The drawback of this procedureis that the closing step still relies on divers and can be delayeddue to rough weather. Finally, the cost of each unity fullyequipped with sensors (floating infrastructure, mesocosm bag,sensors, datalogger and associated materials) is about 37 k€(exchange rate 1 € = 1.31 US $) and the cost of consumable (ie,5 m length mesocosm bags and covers) for each unity is about1.5-2 k€. If the cost of consumable is relatively moderate for aminimal experience (2-3 replicate treatments and 2-3 controlsfor a total cost of 8-12 k€), the cost of a fully equipped plat-form for the same minimal experience will be relativelyexpensive (148-222 k€). In addition, the supplementary unitsmean also extra moorings and involve extra costs due to boattime needed for deployment and due to freight volume. Thecost of each extra unit could limit the experimental design(number of treatments and the number of replicates per treat-ment) and duration of the experiment and implies that thecost of this kind of large infrastructures should be shared toperform the multidisciplinary researches.Advices on experiment duration

In the mesocosm experiments, one challenge is to choosethe adequate duration of the experiment. The duration of theexperiments in the mesocosms depends on the scientificobjective of the project before all other considerations. One ofthe characteristics of the mesocosm experiments is to enclosevarious organisms of the food webs. We call this biologicalscale under experimentation and observation “Web Scale.”Organism acclimation to the new mesocosm environmentgenerally requires one or some days. Then, establishment ofdifferent interactions between presented organisms and con-sequent retroaction between them depends of the generation

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

406

time of the organisms in the experiment, applied treatments,as well as local environmental conditions (temperature, nutri-ent concentration, day length, water transparency, etc.). Inthe pelagic experimentations, the “Web Scale” includes mostlyorganisms from virus to zooplankton, but sometimes com-prises also the organisms of the higher trophic levels such asfishes. There is a positive relationship between the length ofthe “Web Scale” and the duration of the experiment. If amesocosm experiment deals with microorganisms, length ofthe “Web Scale” is from virus to microzooplankton with thegeneration time of hours to days, so the duration of suchexperiment can be from one to some weeks. In the contrary, ifthe “Web Scale” in a mesocosm experiment end up to the fish,as their generation time is in the order of months or year, theduration of the experiment can be in the same order of time.In the case of LAMP, as for all other in situ mesocosm struc-tures developed actually for marine waters, the adequate dura-tion of an experiment is of the order of weeks, principally dueto high environmental constrains.

Comments and recommendationsIn the future, some improvements of the infrastructure can

be realized such as a link using an Argos satellite transceivercould be incorporated to relay the position of the LAMP, itsstatus, and selected data while providing the ability to sendthe platform a limited range of emergency commands bydownlink.

For handling additional sensors, the software controllingthe Campbell dataloggers should be improved to provide bet-ter adaptability to different experimental configurations(modification of the sensor organization, data acquisition,etc.). Furthermore, the hardware could be made more compactby designing interface modules specifically for attaching sen-sors on a printed circuit board directly connected to the data-logger rather than being mounted on a connector and thesurge suppressor PCB.

ReferencesAgawin, N. S. R., S. Agusti, and C. M. Duarte. 2002. Abun-

dance of Antarctic picophytoplankton and their responseto light and nutrient manipulation. Aquat. Microb. Ecol.29:161-172 [doi:10.3354/ame029161].

Agrawal, Y. C., and H. C. Pottsmith. 2000. Instruments for par-ticle size and settling velocity observations in sedimenttransport. Mar. Geol. 168:89-114 [doi:10.1016/S0025-3227(00)00044-X].

Alcaraz, M., C. Marrasé, F. Peters, L. Arin, and A. Malits. 2001.Sea-water-atmosphere O2 exchange rates in open-top labo-ratory microcosms: application for continuous estimates ofplanktonic primary production and respiration. J. Exp. Mar.Biol. Ecol. 257:1-12 [doi:10.1016/S0022-0981(00)00328-2].

Alkire, M., and others. 2012. Estimates of net community pro-duction and export using high-resolution, Lagrangian mea-surements of O2, NO3

–, and POC through the evolution of

a spring diatom bloom in the North Atlantic. Deep-Sea Res.I 64:157-174 [doi:10.1016/j.dsr.2012.01.012].

Behrenfeld, M. J., and others. 2006. Climate-driven trends incontemporary ocean productivity. Nature 444:752-755[doi:10.1038/nature05317].

Belzile, C., and others. 1998. An experimental tool to studythe effects of ultraviolet radiation on planktonic communi-ties: A mesocosm approch. Environ. Technol. 19:667-682[doi:10.1080/09593331908616723].

Bloesch, J. 1988. Mesocosm studies. Hydrobiologia159(3):221-222 [doi:10.1007/BF00008235].

Boyce, D. G., M. R. Lewis, and B. Worm. 2010. Global phyto-plankton decline over the past century. Nature 466:591-596[doi:10.1038/nature09268].

Briand, E., O. Pringault, S. Jacquet, and J. P. Torreton. 2004.The use of oxygen microprobes to measure bacterial respi-ration for determining bacterioplankton growth efficiency.Limnol. Oceanogr.: Methods 2: 406-416.

Carritt, D. E., and J. H. Carpenter. 1966. Comparison and eval-uation of currently employed modifications of the Winklermethod for deter-mining dissolved oxygen in sea water. ANASCO report. J. Mar. Res. 24:286-318.

Cullen, J. J., and M. R. Lewis. 1995. Biological processes andoptical measurements near the sea-surface: some issues rel-evant to remote sensing. J. Geophys. Res. 100(C7):13255-13266 [doi:10.1029/95JC00454].

Czerny, J., K. G. Schulz, A. Ludwig, and U. Riebesell. 2013a.Technical Note: A simple method for air–sea gas exchangemeasurements in mesocosms and its application in carbonbudgeting. Biogeosci. 10:1379-1390 [doi:10.5194/bg-10-1379-2013].

———, K. G. Schulz, S. A. Krug, A. Ludwig, and U. Riebesell.2013b. Technical Note: The determination of enclosedwater volume in large flexible-wall mesocosms "KOSMOS",Biogeosci. 10:1937-1941 [doi:10.5194/bg-10-1937-2013].

Fouilland, É., and others. 2003. Effect of enhanced ultraviolet-B radiation and mixing regime on nitrogen metabolism ofa natural plankton community shifting from a nitrate to asilicate deficiency. Limnol. Oceanogr. 41:18-30[doi:10.4319/lo.2003.48.1.0018].

Garcia, H. E., and L. I. Gordon. 1992. Oxygen solubility in sea-water: Better fitting equations. Limnol. Oceanogr. 37:1307-1312 [doi:10.4319/lo.1992.37.6.1307].

Garcia-Martin, E. E., P. Serret, and M. Pérez-Lorenzo. 2011.Testing potential bias in marine plankton respiration ratesby dark bottle incubations in the NW Iberian shelf: incuba-tion time and bottle volume. Cont. Shelf Res. 31(5):496-506 [doi:10.1016/j.csr.2010.07.006].

Gorsky, G., M. Picheral, and L. Stemmann. 2000. Use of theunderwater video profiler for the study of aggregatedynamics in the North Mediterranean. Estuar. Coast. ShelfSci. 50(1):121-128 [doi:10.1006/ecss.1999.0539].

Grice, G. D., and M. R. Reeve. 1982. Introduction and descrip-tion of experimental ecosystems, p. 12-19. In G. D. Grice

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

407

and M. R. [eds.], Marine mesocosms. Springer.Guieu, C., and others. 2010. Large clean mesocosms and sim-

ulated dust deposition: a new methodology to investigateresponses of marine oligotrophic ecosystems to atmos-pheric inputs. Biogeosci. Discuss. 7:2681-2738[doi:10.5194/bgd-7-2681-2010].

Johnson, K., and others. 2009. Observing biogeochemicalcycles at global scales with profiling floats and glidersprospects for a global array. Oceanography 22:216-225[doi:10.5670/oceanog.2009.81].

Krom, M. D., and others. 1991. Phosphorus limitation of pri-mary productivity in the eastern Mediterranean-Sea. Lim-nol. Oceanogr. 36:424-432 [doi:10.4319/lo.1991.36.3.0424].

Lewandowska, A., and U. Sommer. 2010. Climate change andthe spring bloom: a mesocosm study on the influence oflight and temperature on phytoplankton and mesozoo-plankton. Mar. Eco. Prog. Ser. 405:101-111 [doi:10.3354/meps08520].

Lunven, M., and others. 2012. In situ video and fluorescenceanalysis (VFA) of marine particles: applications to phyto-plankton ecological studies. Limnol. Oceanogr. Methods10:807-823 [doi:10.4319/lom.2012.10.807].

Markager, S. A., and others. 1992. Diel changes in dark respi-ration in a plankton community. Hydrobiologia 238: 119-130 [doi:10.1007/BF00048780].

Mignot, A., and others. 2011. From the shape of the verticalprofile of in vivo fluorescence to Chlorophyll-a concentra-tion. Biogeosciences 8:2391-2406 [doi:10.5194/bg-8-2391-2011].

Mostajir, B., and others. 1999. Experimental test of the effectof Ultraviolet-B radiation in a planktonic community. Lim-nol. Oceanogr. 44:586-596 [doi:10.4319/lo.1999.44.3.0586].

Neveux, J., and others. 2010. Phytoplankton dynamics in thesouthern New Caledonian lagoon during a southeast tradewinds event. J. Mar. Syst. 82:230-244 [doi:10.1016/j.jmarsys.2010.05.010].

Nishioka, J., S. Takeda, and C. S. Wong. 2001. Change in theconcentration of iron in different size fraction during aphytoplankton bloom in controlled ecosystem enclosures.J. Exp. Mar. Biol. Ecol. 258:237-255 [doi:10.1016/S0022-0981(01)00219-2].

Olson, R. J., and H. M. Sosik. 2007. A submersible imaging-in-flow instrument to analyze nano- and microplankton:Imaging FlowCytobot. Limnol. Oceanogr. Methods 5:195-203 [doi:10.4319/lom.2007.5.195].

Oviatt, C. A., and others. 1984. Recovery of a polluted estuar-ine system: a mesocosm experiment. Mar. Ecol. Prog. Ser.16:203-217 [doi:10.3354/meps016203].

Petersen, J. E., L. P. Sanford, and W. M. Kemp. 1998. Coastalplankton responses to turbulent mixing in experimentalecosystems. Mar. Ecol. Prog. Ser. 171:23-41 [doi:10.3354/meps171023].

Pomeroy L. R., J. E. Sheldon, and W. M. Sheldon. 1994.Changes in bacterial numbers and leucine assimilation dur-ing estimations of microbial respiratory rates in seawater bythe precision Winkler method. Applied and Environ.Microb. 60: 328-332.

Pringault, O., V. Tassas, and E. Rochelle-Newall. 2007. Conse-quences of light respiration on the determination of pro-duction in pelagic systems. Biogeosciences 4:105-114[doi:10. 5194/bg-4-105-2007].

———, S. Tesson, and E. Rochelle-Newall. 2009. Respiration inthe light and bacterio-phytoplankton coupling in a coastalenvironment. Microb. Ecol. 57: 321-334.

Riebesell, U., and others. 2007. Enhanced biological carbonconsumption in a high CO2 ocean. Nature 450:545-548[doi:10.1038/nature06267].

———, and others. 2012. Technical note: A mobile sea-goingmesocosm system—new opportunities for ocean changeresearch. Biogeosci. Discuss. 9:12985-13017 [doi:10.5194/bgd-9-12985-2012].

Robinson, C., and P. J. le B. Williams. 2005. Respiration and itsmeasurement in surface marine waters. In del Giorgio, P. A.and P. J. le B. Williams [eds.], Respiration in aquatic ecosys-tems. Oxford Univ. Press.

Sackmann, B. S., M. J. Perry, and C. C. Eriksen. 2008. Seagliderobservations of variability in daytime fluorescence quench-ing of chlorophyll-a in Northeastern Pacific coastal waters.Biogeosci. Discuss. 5(4):2839-2865 [doi:10.5194/bgd-5-2839-2008].

Scholin, C., and others. 2006. The Environmental SampleProcessor (ESP): An autonomous robotic device for detect-ing microorganisms remotely using molecular probe tech-nology, p. 1-4. In Proceedings, OCEANS 2006 MTS/IEEEConference, Boston, MA. Marine Technology Society.

Staehr, P. A., and others. 2010. Lake metabolism and the dieloxygen technique: State of the science. Limnol. Oceanogr.Methods 8:628-644 [doi:10.4319/lom.2010.8.0628].

Suzuki, K., N. Handa, T. Nishida, and C. S. Wong. 1997. Esti-mation of phytoplankton succession in a fertilized meso-cosm during summer using high-performance liquid chro-matographic analysis of pigments. J. Exp. Mar. Biol. Ecol.214:1-17 [doi:10.1016/S0022-0981(97)00003-8].

———, and others. 2002. Grazing impact of microzooplank-ton on a diatom bloom in a mesocosm as estimated by pig-ment-specific dilution technique. J. Exp. Mar. Biol. Ecol.271(1):99-120 [doi:10.1016/S0022-0981(02)00038-2].

Takeda, S., J. Nishioka, C. S. Wong, F. A. Whitney, W. K. John-son, and T. J. Soutar. 1999. Application of open-oceanenclosures to study the control of biological CO2 pump inthe subarctic North Pacific Ocean, p. 583-586. In Proceed-ings, 2nd International Symposium CO2 in the Oceans,Tsukuba, Japan, January 1999, CGER-1037-’99, CGER/NIES.

Thingstad, F., and others. 2005. Nature of phosphorus limita-tion in the ultraoligotrophic Eastern Mediterranean. Sci-

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

408

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

409

ence 1068-1071 [doi:10.1126/science.1112632].Thingstad, T. F., and others. 2007. Ability of a “minimum”

microbial food web model to reproduce response patternsobserved in mesocosms manipulated with N and P, glucose,and Si. J. Mar. Sys. 64:15-34 [doi:10.1016/j.jmarsys.2006.02.009].

Thyssen, M., G. A. Tarran, M. V. Zubkov, R. J. Holland, G. Gre-gori, P. H. Burkill, and M. Denis. 2008. The emergence ofautomated high-frequency flow cytometry: revealing tem-poral and spatial phytoplankton variability. J. Plankton Res.30:333-343 [doi:10.1093/plankt/fbn005].

Vidussi, F., and others. 2011. Effects of experimental warmingand increased ultraviolet B radiation on the Mediterraneanplankton food web. Limnol. Oceanogr. 56:206-218[doi:10.4319/lo.2011.56.1.0206].

Weithoff, G., A. Lorke, and N. Walz. 2000. Effects of water-col-umn mixing on bacteria, phytoplankton, and rotifers

under different levels of herbivory in a shallow eutrophiclake. Oecologia 125:91-100 [doi:10.1007/PL00008896].

Williams, P. J. L., and J. K. Egge. 1998. The management andbehaviour of the mesocosms. Estuar. Coast. Shelf Sci. 46:3-14 [doi:10.1006/ecss.1998.0330].

Xing, X., H. Claustre, S. Blain, F. D’Ortenzio, D. Antoine, J.Ras, and C. Guinet. 2012. Quenching correction for in vivochlorophyll fluorescence measured by instrumented ele-phant seals in the Ker- guelen region (Southern Ocean).Limnol. Oceanogr. Methods 10:483-495.

Zapata, M., F. Rodriguez, and J. Garrido. 2000. Separation ofchlorophylls and carotenoids from marine phytoplankton:A new HPLC method using a reversed phase C-8 columnand pyridine-containing mobile phases. Mar. Ecol. Prog.Ser. 195: 29-45 [doi:10.3354/meps195029].

Submitted 20 December 2012Revised 25 June 2013

Accepted 28 June 2013

Mostajir et al. Lite aquatic Automated Mesocosm Platform (LAMP)

410