K. Tsukamoto, T. Kawamura, T. Takeuchi, T. D. Beard, Jr. and M. J. Kaiser, eds.Fisheries for Global Welfare and Environment, 5th World Fisheries Congress 2008, pp. 317–334.© by TERRAPUB 2008.

Harmful Algal Blooms and Ocean ObservingSystems: Needs, Present Status and

Future Potential

Donald M. Anderson

Biology Department, MS #32Woods Hole Oceanographic Institution

Woods Hole, MA 02543, USA

E-mail: danderson@whoi.edu

Harmful algal blooms (HABs), commonly called “red tides” are increasinglycommon worldwide. These are caused by the growth and aggregation of mi-croscopic algae, leading to negative impacts of many types, including illnessand death in human consumers of contaminated fish and shellfish, as well asmortalities of wild and farmed fish, marine mammals, and other animals. Thediversity of HAB species and their impacts, as well as the oceanographic com-plexity of these phenomena all present significant challenges to those respon-sible for the management of coastal resources and the protection of publichealth. A promising development in this regard is the advent of ocean observ-ing systems (OOSs)—arrays of moored and mobile instruments that can col-lect and transmit data continuously from remote locations to shore-based sci-entists and managers. The potential benefits from ocean observing systemsare many, and improved monitoring and management of HABs are frequentlycited as example benefits to justify the investment of resources in these sys-tems. During this era of accelerated instrument development and deploymentthrough national and international OOS programs, it is instructive to examinethe needs, present status, and realistic potential of ocean observatories astools for HAB monitoring and management. HABs represent a biological com-ponent of coastal waters that challenges present technologies, in part be-cause of the need for species- or toxin-specific detection capabilities. Thispaper discusses the observing system capabilities or assets that are neededfor effective HAB monitoring and management, highlighting and evaluatingnew technologies and future capabilities. Examples are given of HABs world-wide, with special emphasis on paralytic shellfish poisoning (PSP) outbreaksin the northeastern United States as a HAB system with representative chal-lenges in observatory design and capabilities.

KEYWORDS HABs; ocean observing systems; monitoring; management

318 D. M. ANDERSON

1. Introduction

The Integrated Earth Observation System(GEO 2007) is becoming a reality, as satel-lites, ocean buoys, weather stations andin-situ earth observing instruments are be-ing deployed worldwide and their data as-similated into advanced numerical modelsto provide the analysis and understandingneeded to address critical societal issues suchas climate change, drought, fisheries man-agement, or pollution, to name but a few.This reflects the growing recognition thatdecision making benefits from access to awide variety of environmental, biological,economic, statistical, and other data, andmore importantly, the ability to integrate,analyze and evaluate these types of datawithin a common framework shared by manycountries and end users.

Within this large and visionary context,ocean observing systems (OOSs) are beingplanned and deployed on a major scale.Oceanography has entered a new era—onein which the traditional ways of obtainingdata from the ocean using research vesselsare being replaced by arrays of moored andmobile instruments that can collect and trans-mit data continuously from remote locationsto shore-based scientists and managers. Justas networks of meteorological stations andnumerical models of atmospheric dynamicsrevolutionized our understanding of theweather and greatly improved our ability toprovide accurate forecasts of weather events,OOSs and their associated numerical mod-els of ocean dynamics have the potential todocument long-term patterns and changes inthe sea, to detect infrequent events that pre-viously went unobserved, and to make pre-dictions or forecasts about these and otherphenomena that directly affect humanpopulations and marine ecosystems. Ad-vances in communications, robotics, com-puting, platform design, power systems, andsensor technology now make it possible to

get broader and deeper views of the oceansover longer periods and to share that infor-mation in real time among scientists,policymakers, resource managers, educators,and students.

The potential benefits from OOSs aremany, and include the detection and predic-tion of climate variability, facilitation of safeand efficient marine operations, ensuring na-tional security, preserving and restoringhealthy marine ecosystems, mitigating natu-ral hazards, managing living resources, andensuring public health (NORLC 1999). Un-der the last three topics, harmful algal blooms(HABs) are frequently cited as phenomenathat can be better understood and managedusing ocean observatories (e.g., ORION Ex-ecutive Steering Committee 2005). HABsare highly visible phenomena that affect thegeneral public and coastal resources in manyways, and clear economic and managerialbenefits would accrue from advance warn-ing and forecasting capabilities. This poten-tial is based on a need to understand the bio-logical, physical, and chemical factors con-trolling the dynamics of individual HAB spe-cies at appropriate scales, but it is also a rec-ognition that management of HABs can ben-efit from improved cell and toxin detectioncapabilities, coupled with modeling and fore-casting of bloom transport and landfall(Ramsdell et al. 2005). However, HABs rep-resent a biological component of coastalwaters that challenges present technologies,in part because of the need for species- ortoxin-specific detection capabilities.

In recent years, progress has been madein molecular and immunological cell andtoxin detection technologies, opening thedoor to an era where remote, subsurface, nearreal-time detection of specific HAB taxa andtoxins can be envisioned. Given this poten-tial, a logical conclusion would be that in-struments and observatory systems currentlybeing planned or deployed would includesome that will detect HAB cells or toxins,and that these systems are being deployed

Harmful algal blooms and ocean observing systems 319

in areas where HABs are a serious and re-current problem. This, unfortunately, is gen-erally not the case, as other scientific priori-ties have been used in observatory siting de-cisions, and available funds are being usedto build and deploy instruments that are wellproven for oceanographic measurements, butwhich do not provide the species-specificdata needed for HABs. In effect, HAB-spe-cific observatory instrumentation is not yetready for operational deployment (Sellner etal. 2003; Paul et al. 2007). Here the objec-tive is to examine the capabilities that areneeded specifically for HAB detection andforecasting in regional monitoring and man-agement programs, and to assess where weare with new technologies, and where weneed to go if we are to meet the expectationsof those who have funded the OOSs. It ishoped that this review will help to guide re-search and funding programs, and that it willinject a level of realism that moderates thelofty expectations and claims about futurecapabilities.

2. Harmful Algal Blooms

Over the last several decades, countriesthroughout the world have experienced anescalating trend in the incidence of “harm-ful algal blooms” (HABs; Anderson 1989;Hallegraeff 1993). HAB events are charac-terized by the proliferation and occasionaldominance of particular species of toxic orharmful algae. When toxic algae are filteredfrom the water as food by shellfish, their tox-ins accumulate in those shellfish to levels thatcan be lethal to humans or other consumers.Another type of HAB impact occurs whenmarine fauna are killed by algal species thatrelease toxins and other compounds into thewater. HABs also cause mortalities of wildfish, seabirds, whales, dolphins, and othermarine animals. Non-toxic blooms of algaecan cause harm, often due to the highbiomass that some blooms achieve, and thedeposition and decay of that biomass, lead-

ing to anoxia.A poorly defined but potentially signifi-

cant concern relates to sublethal, chronic im-pacts from toxic HABs that can affect thestructure and function of ecosystems. Adultfish can be killed by the millions in a singleoutbreak, with long- and short-term ecosys-tem impacts (Okaichi et al. 1989; Kim et al.1999). Likewise, larval or juvenile stages offish or other commercially important speciescan experience mortalities from algal toxins(White et al. 1989). Chronic toxin exposuremay have long-term consequences that arecritical with respect to the sustainability orrecovery of natural populations at highertrophic levels (Ramsdell et al. 2005).

2.1. Paralytic shellfish poisoning in theGulf of Maine

It is useful to view the issues involving HABsand observing systems in the context of areal-world problem—i.e., recurrent HABoutbreaks in an important site that has beenwell studied and characterized, and that of-fers representative challenges and constraintsto the successful deployment of a HAB ob-serving and forecasting capability. The ex-ample selected is the Gulf of Maine in thenortheastern United States, the site of wide-spread outbreaks of PSP caused by thedinoflagellate Alexandrium fundyense.

A dominant feature underlying A.fundyense regional bloom dynamics is theMaine Coastal Current or MCC (Fig. 1;Lynch et al. 1997)—a composite of multi-ple segments and branch points. The twomajor transport features in this system arethe eastern and western segments of theMCC, hereafter termed the EMCC andWMCC. Conceptual models of A. fundyensebloom dynamics have been provided byAnderson et al. (2005b) and McGillicuddyet al. (2005). Key features in the models aretwo large cyst “seedbeds”—one in the Bayof Fundy and the other offshore of mid-coastMaine (Fig. 2; Anderson et al. 2005b). Cystsgerminate from the BOF seedbed, causing

320 D. M. ANDERSON

recurrent coastal blooms that are self-seed-ing with respect to future outbreaks in thatarea. The blooms also contribute to popula-tions in the EMCC, as some cells escape theBay of Fundy and enter the EMCC wherethey bloom. Some cells travel south and westwith the EMCC, while others deposit cystsin the mid-coast Maine seedbed. In subse-quent years, these latter cysts (combined withcells from the EMCC) inoculate WMCCblooms that cause toxicity in western por-

tions of the Gulf and possibly offshore wa-ters as well.

Numerical modeling of A. fundyense andPSP dynamics in the Gulf of Maine utilizesa hierarchy of physical–biological models.The current Alexandrium sub-model formu-lation follows Stock et al. (2005) andMcGillicuddy et al. (2005) and includes ger-mination, growth, mortality, and nutrientlimitation. Each year, germination from cystseedbeds provides the inoculum for subsequent

Fig. 1. Gulf of Maine surface circulation, with eastern and western segments of the MaineCoastal Current system identified (EMCC and WMCC). Other abbreviations: GOMCP—Gulf ofMaine coastal plume (from Keafer et al. 2005); MA—Massachusetts; ME—Maine; NH—New Hamp-shire; NS—Nova Scotia (Modified from Pettigrew et al. 2005).

Harmful algal blooms and ocean observing systems 321

growth in the overlying waters, while windevents influence delivery onto the coast.Currently, realistic nowcasts of bloom de-velopment are possible using observed cystdistributions, cyst germination and vegeta-tive cell growth rates, and continuous realtime river flow and hydrographic data (Heet al. in press). Forecasts are also being gen-erated during the bloom season, but only onan experimental basis at present.

PSP outbreaks in the Gulf of Maine aresufficiently well characterized and modeled(McGillicuddy et al. 2005; Stock et al. 2005;He et al. in press) that they could benefit fromautomated observations through an ocean ob-servatory system. In the sections that follow,the technologies and instrumentation that canbe used to achieve this goal will be high-lighted, and the challenges to realization ofthis potential discussed.

Fig. 2. Conceptual model of A. fundyense bloom dynamics and PSP toxicity. Solid black linesdenote the eastern and western segments of the Maine Coastal Current system (EMCC and WMCC,respectively). Solid black lines also depict the circulation around Georges Bank. Short, dashedblack lines delimit the cyst seedbeds in the BOF and mid-coast Maine. The shaded areas withinthe Gulf represent portions of the EMCC, WMCC, and BOF where A. fundyense blooms tend tooccur with the highest color intensity denoting areas with higher cell concentrations. Dashedgrey lines show the transport pathways of these water masses and their associated Alexandriumcells. Modified from Anderson et al. (2005b).

322 D. M. ANDERSON

3. Observational and Analytical Needsfor HAB Monitoringand Management

Anderson et al. (2001) highlight the dif-ferent approaches adopted by countries andcommercial enterprises worldwide to moni-tor and manage HABs in coastal waters. Thisis typically accomplished through the estab-lishment of programs for toxin and cell de-tection (and quantitation) in water, aerosols,shellfish, fish, etc., development of bloomforecasting and early warning capabilities aswell as medical intervention and therapeu-tic strategies, and to a growing extent, bloomprevention and mitigation strategies. Thereare, however, many challenges associatedwith these activities, due to the complexityand diversity of HAB phenomena. Resourcemanagers and regulatory officials must dealwith multiple toxins and multiple toxic al-gal species, multiple toxic fisheries re-sources, and large- and small-scale HABevents that occur intermittently. Many newtechnologies are emerging that can addressthese management challenges, however, asdiscussed herein. A comprehensive reviewon this topic is provided by Sellner et al.(2003), and a more general review by Paulet al. (2007) in the context of biological sen-sors that are not HAB-specific.

3.1. Sampling platforms

There are many possible ocean observatoryconfigurations, varying in geographic scaleas well as in the manner in which instrumentsare deployed, powered, and utilized. Instru-ment packages can be located at specific sitesand depths using surface moorings, they canbe sequestered on the ocean bottom, risingthrough the water column to obtain profiledata, and they can be mounted on underwa-ter vehicles of various types that travelthrough the water in a given area on pre-programmed missions. They can also be

mounted on surface vessels or simple plat-forms that house instruments that collectwater at depth for automated analysis at thesurface. Each of these design features hasadvantages and disadvantages with respectto HAB monitoring. For example, samplingposition in the water column is a critical de-sign constraint, as there can be considerablevariability in the vertical distribution of HABspecies. Some are often found in subsurface,thin layers that are difficult to detect and sam-ple (e.g., Xie et al. 2007), but others are pre-dominantly confined to the surface mixedlayer, and thus would be amenable to detec-tion using an instrument package mooredwithin that layer. In the Gulf of Maine, forexample, considerable information has beenobtained from observations of the surfacedistribution of Alexandrium fundyense (e.g.,Townsend et al. 2001; Anderson et al. 2005a).Thin, subsurface layers of Alexandrium havealso been detected in some areas of the Gulf(Townsend et al. 2001) but the importanceof these accumulations remains unknown inthe context of shellfish toxicity and generalbloom dynamics. A surface-deployed instru-ment package would therefore be highly in-formative in that region, but one that couldobtain profiles of cell distributions would beeven more so. In this latter instance, the valueof the vertical profile data must be balancedagainst the additional costs incurred. Thehigher resolution of the vertical profile wouldrequire more reagents, filters, sensors, orarrays, and the instrument package wouldtherefore need to be serviced more frequentlythan an instrument that samples at a singledepth. This is an important considerationgiven the cost of supplies as well as person-nel and vessel time for the servicing opera-tions.

Horizontal coverage is an equally im-portant factor in HAB monitoring, andagain, the Gulf of Maine provides a goodexample of the issues that need to be con-sidered in instrument siting. As describedabove, A. fundyense is transported to the

Harmful algal blooms and ocean observing systems 323

south and west in two coastal currents—theEMCC and the WMCC (Fig. 1). Not onlyare the Alexandrium distributions non-uni-form within these currents (Townsend et al.2001), but the location of these water massesvaries through time as well (Pettigrew et al.2005). With appropriate wind and hydrody-namic forcings, the EMCC and its associ-ated A. fundyense populations can be carriedoffshore into the central Gulf of Maine, and/or delivered toward shore (Luerssen et al.2005; Keafer et al. 2005). This variability inalongshore transport has direct implicationsto the patterns of PSP toxicity in nearshoreshellfish (Luerssen et al. 2005). A single ob-servatory mooring placed in these coastalwaters would therefore not provide accuratecell abundance information. In this instance,multiple instruments arrayed along a cross-shore transect would be an expensive, butmuch more informative configuration.

These concerns about vertical and hori-zontal resolution can be addressed with mo-bile autonomous underwater vehicles(AUVs), but for most HABs, considerabledevelopment is needed if these vehicles areto provide species-specific or toxin-specificdata. With the exception of Karenia brevisand the BreveBuster described by Robbinset al. (2006), most HAB species are not ame-nable to optical detection due to their lackof distinctive pigments or other features thatcan be distinguished optically. As a result,water samples must be collected, manipu-lated and complex chemistries performed,and this in turn requires power, space, androbotic capabilities that are not possible inAUVs at present. For most HABs, and forthe near future, it therefore appears thatAUVs will be used to supply contextual data(e.g., salinity, temperature, turbidity, chlo-rophyll) that will help in the understandingthe patterns of HAB abundance and toxicitythat are observed with other instruments.This limitation in HAB detection capabili-ties will change when miniaturized massspectrometers are configured to detect HAB

toxins dissolved in seawater (see below), orwhen other features of these cells or toxinsare identified that can be readily measuredwith the relatively simple type of roboticsand detectors that can be deployed withinpower- and space-limited AUVs.

The foregoing discussion highlights oneof the major obstacles that has slowedprogress in the detection of HAB cells andtoxins at ocean observatories—the need toprocess water samples through filters or otherconcentrating devices, and to manipulatethose samples for extraction and analysis oftoxins or the cellular targets needed for spe-cies identification and enumeration. Tech-nologies are available for many of theseanalyses, but they need to be incorporatedinto an instrument that can be deployed un-derwater and that can perform the series ofrobotic functions needed for each analysis.

One instrument that provides these ca-pabilities and that can be configured for usefor HAB cell and toxin detection in oceanobserving systems is the EnvironmentalSampling Processor (ESP; Goffredi et al.2006; Scholin et al. 1998, in press). The ESP(Fig. 3) autonomously collects discrete wa-ter samples from the ocean subsurface, con-centrates microorganisms (particulates), andautomates application of molecular probesto identify specific microorganisms and theirgene products (Scholin et al. 2006). The cur-rent prototype is the second generation ESPor 2G ESP. It consists of three major com-ponents: the core sample processor, analyti-cal modules, and sampling modules. Thecore ESP is designed to collect and processsmall- to moderate-sized samples (mLs toseveral liters) at depths to 50 m. Analyticalmodules are stand-alone detection systemsthat can be added to the core ESP to impartdifferent analytical functions downstream ofcommon sample processing operations (e.g.,PCR, capillary electrophoresis, competitiveimmunoassays, etc.).

The ESP currently utilizes DNA probeand protein arrays to detect target molecules

324 D. M. ANDERSON

indicative of species and the substances theyproduce. As described below, within the coreESP, DNA probe arrays specific for targetHAB species capture ribosomal RNA(rRNA) from a crude sample homogenateusing a quantitative sandwich hybridizationassay (SHA; Fig. 4A). A quantitative PCRmodule is under development as an alterna-tive approach to cell enumeration. Proteinarrays utilize a competitive ELISA (enzyme-linked immunosorbent assay) technique fordetecting target substances such as domoicacid (e.g., Fig. 4b). The ESP can also archivesamples for laboratory analyses after the in-strument is recovered, including fluorescentin situ hybridization (FISH), various nucleic

acid analyses (cloning, sequencing) and al-gal toxin measurement (e.g., Greenfield etal. 2006).

Sample manipulations are carried out inreaction chambers called “pucks” that areloaded into and removed from various sta-tions by robotic mechanisms. The automated

Fig. 3. Second generation ESP instrument(16”d × 30”h). bottom left: puck magnified ~2xrelative to those loaded in the carousel. Seealso: http://www.mbari.org/ESP/esp2G_compare.htm

Fig. 4. In situ detection of Pseudo-nitzschia spp. and domoic acid (DA) inMonterey Bay, March 2006 using the ESP.A) DNA array with probes for Pseudo-nitzschiaaustralis, P. multiseries/pseudodelicatisima,Alexandrium catenella, and Heterosigmaakashiwo (HA array). Only control spots areoutlined as targeted species are below de-tection level. B) Protein array for DA, corre-sponds to (A). Circled spots are IgG controlsconfirming consistency of detection chem-istry. Spots in red rectangles are DA conju-gate. Maximum intensity in DA conjugatespots indicates DA levels are below detec-tion limit, whereas decreasing spot intensityindicates presence of DA (competitiveELISA). C) HAB array from a later time indi-cates the presence of P. australis (spotsbounded by dashed lines) and P. multiseries/pseudodelicatissima (spots bounded by dot-ted lines), both of which can produce DA.D) DA array that corresponds to (C); circledspots are IgG controls; spots in rectanglesare DA conjugate; the DA conjugate spotsare considerably dimmed, indicating the pres-ence of this toxin (from Doucette et al. andGreenfield et al., in prep.).

Harmful algal blooms and ocean observing systems 325

process from collection of a live sample tobroadcast of an imaged DNA or proteinprobe array takes ~2 hours and can occursubsurface. The instrument can perform~30–40 of those operations before servicingis required. The current limitation is thenumber of pucks stored in the carousel. Thereare, however, design options for increasingthe number of sampling/analytical events anddecreasing power consumption.

The instrument can be bundled with con-textual sensors such as a CTD, fluorometer,transmissometer, and nutrient analyzer. Datafrom the external sensors along with resultsof the probe assays are uploaded periodicallyfrom the deployed instrument to a shore sta-tion for analysis and interpretation. Two-waycommunication allows for rescheduling ofmission sampling profiles if desired. Furtherdetails on the instrument’s design and op-eration are described elsewhere (Greenfieldet al. 2006; Scholin et al. 2006; Roman etal. 2007; Paul et al. 2007; see also http://www.mbari.org/esp).

As promising as this instrument is, it isnot yet commercially available, and thus test-ing has been predominantly through researchgrants to the developers and collaborators.To date, the ESP has been deployed multi-ple times in surface waters for periods of sev-eral weeks to a month, during which time ithas successfully automated application ofthree classes of DNA probe arrays (HABs,bacteria/archaea, invertebrate larvae) and thedomoic acid assay (Goffredi et al. 2006;Greenfield et al. 2006; Paul et al. 2007; Joneset al. in press; C. Scholin, G. Doucette,unpub. data). Further details on the instru-ment’s design and operation are describedin these citations and at http://www.mbari.org/esp.

At the present time, the only other ad-vanced robotic instrument capable of in situwater collection, processing, and sophisti-cated molecular and biochemical analysis isthe Autonomous Microbial Genosensor(AMG), designed to detect specific micro-

bial targets in coastal or oceanic waters (Paulet al. 2007). The current prototype of theAMG collects water samples, filters cells andextracts RNA, and performs amplificationautonomously. A second generation AMGwill incorporate microfluidic liquid process-ing, array technology, and intensified lightdetection.

3.2. Toxin detection

Of paramount importance to many HABmonitoring programs are methods to detectand quantify the toxins produced by HABspecies, which include a broad spectrum ofcompounds ranging in size, potency, andsolubility. In all cases, the marine HAB tox-ins that cause the human poisoning syn-dromes consist of families or groups of struc-turally related compounds, with individualderivatives exhibiting potencies that can sig-nificantly differ from other congeners (VanDolah 2000). During food web transfer, HABtoxins can also be metabolized or biotrans-formed into structurally different com-pounds. The broad chemical and structuraldiversity of algal toxins and their derivativesand metabolites, coupled with differences intheir potency account for many of the chal-lenges associated with their detection inocean observatory programs.

Traditionally, biotoxin monitoring pro-grams have relied on measurements of tox-ins in shellfish samples collected weekly orbi-weekly from key locations in areas af-fected by HABs (e.g., Shumway et al. 1988).This procedure works well and provides ap-propriate public health protection if the sta-tions are well sited, and the toxin assays arerun at relatively frequent intervals. Toxinmeasurement methods can be grouped intothree main types: chemical, in vitro, and invivo assays (Hallegraeff et al. 2003). Thelatter (bioassays) have had a long history inHAB toxin detection, but are obviously notamenable to automation and high-through-put analysis in ocean observatories, so the

326 D. M. ANDERSON

only options in that context are measure-ments of toxin in seawater using eitherchemical analyses or in vitro assays. This im-mediately introduces some concerns, asconsiderable work will be needed to relatemeasurements of toxins dissolved inseawater, or in particulate form in that wa-ter, to the risk to human consumers of shell-fish or fish. In the Gulf of Maine example,where measurements of toxicity in shellfishtissues are used for regulatory decisions,measurements of Alexandrium toxins in thewater column at ocean observatories willlikely be used as supplementary informationin assessing risk, both current and future. Itwill be many years before sufficient data areaccumulated to allow such measurements tobe used for regulatory purposes for nearshoreshellfish.

Chemical methods for toxin analysis in-clude high performance liquid chromatog-raphy (HPLC), and mass spectrometry cou-pled to liquid chromatographic separation(Quilliam 1996). Of these two alternatives,only mass spectrometry shows the potentialfor use in ocean observatories, and there thechallenges remain significant due to the di-versity, size, and solubility of the toxins, aswell as the matrices in which they occur (e.g.,particulate versus dissolved). Another con-straint is the need to perform spectrometryin a vacuum and underwater, which posessignificant engineering challenges. Progresshas been good, however. For example, asmall, modular mass spectrometer has beendeveloped and mounted in an AUV (Wenneret al. 2004). That system consists of an in-situmembrane-introduction linear-quadrupolemass spectrometer capable of detecting dis-solved gases and volatile organic compoundsat sub parts-per-billion concentrations. Thisinstrument is still under development and hasnot been configured for HAB toxins, but fu-ture designs may permit the analysis of HABtoxins that occur dissolved in seawater (e.g.,brevetoxins, domoic acid, okadaic acid).Analysis of toxins in particulate form will

require a different approach, such as LaserDesorption Mass Spectrometry (LDMS),which is widely employed in analytical labo-ratories due to its simplicity of operation andrapid analysis times. One benefit of LDMSis that many different types of materials canbe vaporized and ionized by a tightly focusedlaser beam (Cotter 1997). This can avoidsample purification or preparative tech-niques, which is critical to deployment ofsuch technologies in a moored or mobile con-figuration in an OOS, as it will greatly re-duce sampling and handling requirements,and thus power drain, space needs, and rea-gent needs as well. LDMS has been used forthe detection of bacterial spores, vegetativecells, viruses, and toxins in aerosol environ-ments (Fenselau and Demirev 2001), and ef-forts are underway to apply this method toHAB cells and dissolved toxins in seawater(A. Place, pers. comm.).

Another important and rapidly develop-ing group of HAB toxin detection methodscomprises the in vitro assays. One sub-group—the functional assays—relies on de-tection of a toxin’s biochemical activitywhile the other—structural assays—dependson recognition of chemical structure at themolecular level (reviewed in Cembella et al.2003; Van Dolah and Ramsdell 2001). A va-riety of functional assays have been devel-oped for the detection of HAB toxins, includ-ing cytotoxicity assays (e.g., Manger et al.1995), enzyme inhibition assays (e.g., DellaLoggia et al. 1999), and receptor bindingassays (e.g., Van Dolah et al. 1994). Never-theless, retention of the biological activityof a cell line or a receptor preparation out-side the laboratory remains a significant, andthus far, insurmountable obstacle to in situuse of these assays (Sellner et al. 2003).

In contrast, structural assays show con-siderable promise for automated deploymentin an observatory system. These assays relyon the structural or conformational interac-tion of a toxin with a recognition factor suchas an antibody. Antibody-based assays have

Harmful algal blooms and ocean observing systems 327

been developed for a variety of HAB toxinsand many of these tests are now commerciallyavailable (Laycock et al. 2001; Cembella etal. 2003). In many ways, the procedures usedfor toxin immunoassays are similar to thoseused for HAB cell detection using oligonu-cleotide probes (described below), and thusthese technologies have the potential to becombined in a single instrument that can de-tect HAB cells and toxins simultaneously.Doucette and co-workers (unpub. data) aredeveloping an immunoassay-based methodfor detection of domoic acid in robotic fash-ion on board the ESP, described above. Thisanalysis utilizes a known quantity of domoicacid–antibody conjugate immobilized onmembranes as replicate spots. These mem-branes are then exposed to a simple extractof filtered plankton, and a competitiveELISA performed. The resulting arrays canbe imaged (Fig. 4B) and the data sent to shoreelectronically. This is an example of a toxin-detection technology that can be automatedand performed in situ in a moored instrument.

Other investigators are developing alter-native immunosensors that also have the po-tential for in situ deployment, though mosthave only been configured for laboratory-based, bench-top formats at present. Onenovel immunoassay utilizes surface plasmonresonance (SPR) in a portable system devel-oped for rapid field quantification of toxinlevels in both shellfish and seawater.(Stevens et al. 2007). The SPR assay had alimit of detection of 3 ppb domoic acid anda quantifiable range from 4 to 60 ppb. Com-parison of analyses with standard HPLCprotocols gave an excellent correlation. Thissame technology should also function for de-tection of domoic acid (and other algal tox-ins for which antibodies are available) in con-centrated algal extracts or high dissolved lev-els in seawater. With refinement of the ex-traction protocols and generation of higheraffinity monoclonal antibodies, detection ofmuch lower levels of toxin should be possi-ble, leading to eventual application of auto-mated SPR biosensors on moorings.

Another novel and potentially useful ap-proach for in situ observations is a competi-tive immunoassay using screen-printed elec-trodes (SPEs; Kreuzer et al. 2002; Micheliet al. 2004). Excellent sensitivity and accu-racy has been achieved with HAB toxinssuch as okadaic acid, brevetoxin, and domoicacid. For all toxins investigated, results com-pared favorably with other toxin analysistechniques. The advantages of speed ofanalysis, simplicity of design, in situmeasurement capability, stability (storage upto four weeks prior to use), and disposabil-ity make SPE immunosensors good candi-dates for observatory instrumentation. Ad-aptation of this and other immunoassay tech-nologies to robotic systems and deploymentin remote locations is thus possible, but willrequire further development effort.

3.3. Cell detection

Two approaches have been followed to im-prove on traditional light microscope countsof HAB species in field programs. One uti-lizes optical characters that are unique to thetarget organism. The only success in this re-gard is for Karenia brevis, the Florida redtide organism, which produces a pigmentcalled gyroxanthin-diester. This carotenoidis found in other fucoxanthin-containingdinoflagellate species as well, but in someareas, such as the Gulf of Mexico, it is suffi-ciently unique to be a useful biomarker forK. brevis and other toxic or potentially toxicKarenia species (Kirkpatrick et al. 2000;Richardson and Pinckney 2004). Instrumentshave been developed that can quantify thispigment in water samples, and these havebeen mounted on board research vessels(Kirkpatrick et al. 2003) and inside an AUVcalled the BreveBuster (Robbins et al. 2006).This approach thus has great potential formonitoring of those HAB species that havethis unique pigment, but for the vast majorityof other species, alternative approaches to celldetection are needed.

328 D. M. ANDERSON

The second approach involves the de-velopment of species- or strain-specific“probes” which can be used to label HABcells of interest so they can then be detectedvisually, electronically, or chemically.Progress has been rapid and probes and as-says of multiple types are already availablefor many of the HAB species. The mostpromising of these approaches in the con-text of ocean observing systems are shortpieces of synthetic DNA (probes or primers)that bind to complementary portions of thosemolecules in the target HAB species. Thesetargets can be visualized and/or quantifiedusing a variety of techniques such as whole-cell fluorescent in situ hybridization (FISH;Anderson et al. 2005b; Hosoi-Tanabe andSako 2005), sandwich hybridization assays(SHA; Scholin et al. 1996; Diercks et al.2008), and a variety of PCR-based assays(e.g., Penna and Magnani, 1999; Guillou etal. 2002). Of these, the FISH technique isnot amenable to in situ use in observatorysystems, so future applications will likelyutilize either the SHA or quantitative PCR(qPCR).

The SHA involves chemical lysis of thealgal cells to release ribosomal RNA targetmolecules that are then “captured” by a probeimmobilized on a surface, and visualized us-ing a colorimetric, fluorometric, or chemi-luminescent reporting system linked to a sec-ond (“signal”) probe in solution. The SHAallows for rapid, high throughput sampleanalysis and has been effectively automatedin a variety of formats, including in the ESP(Fig. 4A; Scholin et al. 2006; Greenfield etal. 2006). One advantage of this approach isthat it utilizes a crude plankton lysate foranalysis—i.e., no RNA purification isneeded. Another is its sensitivity—with de-tection limits of a few thousand cells/L ofPseudo-nitzschia species, and 100 cells/Lwith Alexandrium species, given the presentpumping and filtering specifications (C.Scholin, pers. comm.). With higher volumesample concentration, currently under devel-

opment, detection limits can drop to levelsof a few cells/mL, sufficient to identify theearliest stages of blooms. For example,STMicroelectronics offers the In-Check®

platform, a microfluidic chip that combinesPCR amplification and probe array detectionfunctions. Integrated devices like this couldfind application in an ocean observatory set-ting.

Another rapidly emerging approach tothe detection of HAB species is qPCR (e.g.,Bowers et al. 2000, 2006; Galluzzi et al.2004; Coyne et al. 2005). With respect toin situ applications in robotic or moored sys-tems, it is of note that quantitative PCR pro-cedures require extraction and purificationof nucleic acids from samples, an enzymaticreaction mixture, and application of one ormore thermocycling protocols. Sample han-dling and processing are thus important con-siderations for in situ measurements. Nev-ertheless, portable instruments suitable forfield applications are being developed andshow promise for inclusion in ocean observa-tory-based HAB monitoring and researchprograms in the near future.

One example of the manner these con-straints are being addressed is the portablesensor technology developed for detectionand enumeration of the HAB species K. brevisusing nucleic acid sequence-based amplifi-cation (NASBA; Casper et al. 2004). NASBAis an isothermal method for the amplifica-tion of RNA, so this simplifies the powerand manipulation requirements of the assaywhich otherwise would require multipleheating and cooling steps. To address theproblems with extraction of RNA from wa-ter samples, a simple procedure was devel-oped that requires no special equipment ortraining, and that performs as well as expen-sive, commercial kits (Casper et al. 2007).Further progress was made with the devel-opment of a handheld sensor that providesreal-time fluorescence plotting of theamplification. Results using the handheldNASBA analyzer compare favorably to

Harmful algal blooms and ocean observing systems 329

laboratory-based technologies. This extrac-tion protocol and detection sensor are nowbeing incorporated into an autonomous plat-form called the AMG, described above.

3.4. Modeling and forecasting

The value of data from ocean observatoriesis greatly enhanced by numerical modelingtechniques that can lead to forecasts of HABtransport and dynamics. A region in whichHAB-specific instruments are deployedshould therefore take steps to develop andvalidate numerical models that incorporatelocal HAB species into hydrodynamic mod-els of the region (physical-biological mod-els). The ultimate goal is to obtain data onHAB cells and toxins through instrumentsin an observatory system, assimilate thesedata and contextual meteorological and ocea-nographic observations into the models, andprovide continually updated forecasts ofbloom behavior. The first step in this processis the formulation of conceptual models thatexplain in words and simple diagrams howHABs occur in a given area. Examples ofconceptual models developed forAlexandrium blooms in the Gulf of Maineare given in Anderson et al. (2005b) andMcGillicuddy et al. (2005). If a verbal de-scription of a model can be formulated thatis consistent with observations and data overan extended interval of time, it is much easierto formulate a numerical model that capturesthe same dynamics (McGillicuddy et al.2005). For many areas of the world, thereare significant challenges to achieving thisgoal, as neither conceptual models nor nu-merical models exist for regional HAB prob-lems.

A significant constraint to numericalmodel development is the need to identifyinitial conditions for the biological fields(i.e., the HAB species’ distribution). In theGulf of Maine example, cyst maps in bottomsediments are used as the initial condition,with germination of those cysts producing

the vegetative cells that ultimately grow andform the bloom (McGillicuddy et al. 2005;He et al. in press). In other HAB systems,and in particular those without cystpopulations, cell concentrations measured byinstruments in an observing system may wellprovide the initial conditions for subsequentmodel runs and forecasts of bloom dynam-ics.

In addition to forecasting, numericalmodels can help in the identification of keylocations at which instruments capable of de-tecting HAB cells and toxins can be de-ployed. Figure 5 demonstrates one of thechallenges associated with siting decisionsfor observatory instruments. Given the concep-tual model described earlier for A. fundyensein the Gulf of Maine (Fig. 2) it is possible toidentify key locations in the different trans-port pathways that would facilitate HABdetection and forecasting in this large region.With a hypothetical set of 9 moored instru-ment packages, scientists would place instru-ments at key locations and branch points,with horizontal variability in the coastal cur-rents and Alexandrium distributions ad-dressed through cross-shore arrays of instru-ments at each of the key locations (Fig. 5A).However, shellfish managers would chooseto locate the same number of instrumentsalong a line just offshore or upstream of im-portant shellfish growing areas (Fig. 5B),providing alongshore resolution, but none inthe cross-shore direction. In this instance, themanagers opt for early warning through di-rect cell detection rather than through nu-merical model forecasting. Resolution of thedifference in viewpoints depicted in Fig. 5would require a demonstration that a scien-tifically or hydrographically based mooringconfiguration, coupled with a numericalmodel, can provide early warning informa-tion suitable for management purposes.

Another justification for model devel-opment in an area where an ocean observa-tory is planned is that the model can be usedto identify the locations where data on HAB

330 D. M. ANDERSON

species abundance or toxicity would be themost useful. Termed observing system simu-lation experiments (OSSEs), these numeri-cal analyses have been used in dynamic me-teorology (e.g., Charney et al. 1969) and arebecoming an important tool in the planningof oceanographic sampling systems (e.g.,Robinson et al. 1998; McGillicuddy et al.2001). OSSEs can help to optimize thenumber and location of instruments neededto provide a necessary level of coverage. Theultimate goal is to have an array of instru-ments located in strategic spots to capturethe information that can then be assimilatedinto numerical models through time, greatlyincreasing their accuracy and utility. Onecannot over-emphasize the importance ofproper instrument siting or validated numeri-cal models if the true potential of ocean ob-serving systems is to be realized in HAB re-search and monitoring programs.

4. Summary

Improvements in HAB monitoring and fore-casting are frequently cited as justificationsfor the deployment of ocean observing sys-tems, yet few of the observatory systems be-ing deployed worldwide have any HAB com-ponents. This is largely because the technolo-gies needed to achieve HAB cell or toxin de-tection are still under development. Only asmall number of HAB species can be de-tected using optical measurements, eitherin situ or remotely from space, and there-fore instruments that can detect the vast ma-jority of HAB species need to have capabili-ties for sample collection, concentration, andmanipulation. The chemistries and proce-dures for cell identification and enumerationusing molecular probe assays of varioustypes are well established, so the challengenow is to incorporate these assays into au-tonomous instruments that are capable of

Fig. 5. Map showing possible locations of ESPs or other instruments capable of detectingAlexandrium fundyense cells and/or toxins in the Gulf of Maine. A) Configuration suggested byscientists to capture horizontal variability in cell distributions and water masses along key trans-port pathways. B) Configuration suggested by managers, who simply want early warning at majorshellfish growing areas.

Harmful algal blooms and ocean observing systems 331

References

Anderson DM. Toxic algal blooms and red tides: a global perspective. In: Okaichi T, Anderson DM, NemotoT (eds). Red Tides: Biology, Environmental Science and Toxicology. Elsevier. 1989; 11–16.

Anderson DM, Andersen P, Bricelj VM, Cullen JJ, Rensel JE. Monitoring and Management Strategies forHarmful Algal Blooms in Coastal Waters. Asia Pacific Economic Program, Singapore, and Intergovern-mental Oceanographic Commission, Paris. 2001; 268 pp.

Anderson DM, Kulis DM, Keafer BA, Gribble KE, Marin R, Scholin CA. Identification and enumeration ofAlexandrium spp. from the Gulf of Maine using molecular probes. Deep-Sea Res. II 2005a; 52(19–21):2467–2490.

Anderson DM, Stock CA, Keafer BA, Bronzino Nelson A, Thompson B, McGillicuddy DJ, Keller M, MatraiPA, Martin J. Alexandrium fundyense cyst dynamics in the Gulf of Maine. Deep-Sea Res. II 2005b;52(19–21): 2522–2542.

Bowers HA, Tengs T, Glasgow HB, et al. Development of real-time PCR assays for rapid detection of Pfiesteriapiscicida and related dinoflagellates. Appl. & Environ. Microbiol. 2000; 66(11): 4641–4648.

Bowers HA, Tomas C, Tengs T, Kempton JW, Lewitus AJ, Oldach DW. Raphidophyceae [Chadefaud ex Silva]systematics and rapid identification: Sequence analyses and real-time PCR assays. J. Phycol. 2006; 2:1333–1348.

Casper ET, Paul JH, Smith MC, Gray, M. Detection and quantification of the red tide dinoflagellate Kareniabrevis by realtime nucleic acid sequence-based amplification. Appl. Environ. Microbiol. 2004; 70(8),4727–4732.

Casper ET, Patterson SS, Bhanushali P, Farmer A, Smith M, Fries DP, Paul JH. A handheld NASBA analyzer forthe field detection and quantification of Karenia brevis. Harmful Algae 2007; 6: 112–118.

Cembella AD, Doucette GJ, Garthwaite I. In vitro assays for phycotoxins. In: Hallegraeff GM, Anderson DM,Cembella AD (eds). Manual on Harmful Marine Microalgae. IOC-UNESCO, Paris. 2003; 297–345.

Charney JG, Halem M, Jastrow R. Use of incomplete historical data to infer the present state of the atmos-phere. J. Atmos. Sci. 1969; 26: 1160–1163.

Cotter RJ. Time-of-Flight Mass Spectrometry. ACS Professional Reference Books, Washington DC. 1997.

sample collection and processing whilemoored or deployed within an AUV. Severalinstruments have been developed that havethis capability in moored configurations (i.e.the ESP and AMG) and others will surely bedeveloped in the future given the need forthis type of capability in many other typesof environmental monitoring.

In a similar manner, methods for toxindetection have been developed that can beincorporated into moored instrument pack-ages, typically based on structural assayssuch as immunoassays. Here again, incor-poration of these methods into instrumentscapable of the appropriate processing hasonly just begun. Development is rapid on allof these fronts, but will be greatly acceler-ated if research and development funding istargeted to technologies for species-specificdetection of HAB cells and metabolites.

All too often these days, ocean observa-

tory resources are dedicated to existing,proven technologies that are ready for de-ployment. If the development of HAB-spe-cific instrumentation lags too far behind thepace of observatory infrastructure construc-tion, observatory assets will be located inareas or at sites that are not optimal for HABdetection or will not have the appropriatepower or mooring capabilities, and thus thegoals of improved HAB monitoring andmanagement through ocean observatorieswill not be realized.

Acknowledgements

This effort was supported in part by the followinggrants to D.M. Anderson: NOAA CooperativeAgreement NA17RJ1223; NOAA ECOHAB grantNA06NOS4780245; NOAA grant (through Universityof new Hampshire) NA05NOS4191149; NIEHS Grant1 P50 ES012742; and NSF Grants OCE-0430724 andOCE-0402707.

332 D. M. ANDERSON

Coyne KJ, Handy SM, Demir E, Whereat EB, Hutchins DA, Portune KJ, Doblin MA, Cary SC. Improvedquantitative real-time PCR assays for enumeration of harmful algal species in field samples using anexogenous DNA reference standard. Limnol. Oceanogr. Methods 2005; 381–391.

Della Loggia R, Sosa S, Tubaro. A Methodological improvement of the protein phosphatase inhibition assayfor the detection of okadaic acid in mussels. Nat. Tox. 1999; 7: 387–392.

Diercks S, Medlin LK, Metfies K. Colorimetric detection of the toxic dinoflagellate Alexandrium minutumusing sandwich hybridization in a microtiter plate assay. Harmful Algae 2008; 7: 137–145.

Fenselau C, Demirev P. Characterization of intact microorganisms by MALDI mass spectrometry. MassSpectrom. Rev. 2001; 20: 157.

Galluzzi L, Penna A, Bertozzini E, Vila M, Garces E, Magnani M. Development of a real-time PCR assay forrapid detection and quantification of Alexandrium minutum (a dinoflagellate). Appl. Environ. Microbiol.2004; 70: 1199–1206.

Goffredi SK, Jones W, Scholin C, Marin R, Hallam S, Vrijenhoek RC. Molecular detection of marine larvae.Marine Biotechnol. 2006; 8: 149–160.

Greenfield DI, Marin R III, Jensen S, Massion E, Roman B, Feldman J, Scholin C. Application of the Environ-mental Sample Processor (ESP) methodology for quantifying Pseudo-nitzschia australis using ribos-omal RNA-targeted probes in sandwich and fluorescent in situ hybridization. Limnol. Oceanogr. Meth-ods 2006; 4: 426–435.

Group on Earth Observations (GEO). The Full Picture. GEO, Geneva, Switzerland. 2007; 278 pp.Guillou L, Nézan E, Cueff V, Erard-Le Denn E, Cambon-Bonavita M-A, Gentien P, Barbier G. Genetic diver-

sity and molecular detection of three toxic dinoflagellate genera (Alexandrium, Dinophysis, and Karenia)from French coasts. Protist 2002; 153: 223–238.

Hallegraeff GM. A review of harmful algal blooms and their apparent global increase. Phycologia. 1993; 32:79–99.

Hallegraeff GM, Anderson DM, Cembella AD (eds). Manual on Harmful Marine Microalgae. Monographson Oceanographic Methodology, 11, UNESCO. 2003; 794 pp.

He R, McGillicuddy DJ, Keafer BA, Anderson DM. Gulf of Maine harmful algal bloom in summer 2005—Part2: Coupled bio-physical numerical modeling. J. Geophys. Res. (in press).

Hosoi-Tanabe S, Sako Y. Rapid detection of natural cells of Alexandrium tamarense and A. catenella(Dinophyceae) by fluorescence in situ hybridization. Harmful Algae 2005; 4: 319–328.

Jones WJ, Preston C, Marin R III, Scholin C, Vrijenhoek R. A robotic molecular method for in situ detectionof marine invertebrate larvae. Molecular Ecology Notes (in press).

Keafer BA, Churchill JH, McGillicuddy DJ, Anderson DM. Bloom development and transport of toxicAlexandrium fundyense populations within a coastal plume in the Gulf of Maine. Deep-Sea Res. II2005; 52(19–21): 2674–2697.

Kim CS, Lee SG, Lee CK, Kim HK, Jung J. Reactive oxygen species as causative agents in the icthyotoxicityof the red tide dinoflagellate Cochlodinium polykrikoides. J. Plankt. Res. 1999; 21: 2105–2115.

Kirkpatrick GJ, Millie DF, Moline MA, Schofield O. Optical discrimination of a phytoplankton species innatural mixed populations. Limnol. Oceanogr. 2000; 45(2): 467–471.

Kirkpatrick GJ, Orrico C, Moline MA, Oliver M, Schofield OM. Continuous hyperspectral absorptionmeasurements of colored dissolved organic material in aquatic systems. Appl. Opt. 2003; 42(33):6564–6568.

Kreuzer MP, Pravda M, O’Sullivan CK, Guilbault GG. Novel electrochemical immunosensors for seafoodtoxin analysis. Toxicon. 2002; 40: 1267–1274.

Laycock MV, Jellett JF, Belland ER, Bishop PC, Theriault BL, Russell-Tattrie AL, Quilliam MA, Cembella AD,Richards RC. Mist Alert: A rapid assay for paralytic shellfish poisoning toxins. In: Hallegraeff GM,Blackburn SI, Bolch CJ, Lewis RJ (eds). Harmful Algal Blooms 2000. IOC-UNESCO, Paris, 2001; 254–256.

Luerssen RM, Thomas AC, Hurst J. Relationships between satellite-measured thermal features and Alexandrium-imposed toxicity in the Gulf of Maine. Deep-Sea Res. II 2005; 52(19–21): 2656–2673.

Lynch DR, Holboke MJ, Naimie CE. The Maine coastal current: spring climatological circulation. Cont. ShelfRes. 1997; 17: 605–634.

Manger RL, Leja LS, Lee SY, Hungerford JM, Hokama Y, Dickey RW, Granade HR, Lewis R, Yasumoto T,Wekell MM. Detection of sodium channel toxins: directed cytotoxicity assays of purified ciguatoxins,brevetoxins, saxitoxins, and seafood extracts. J. Assoc. Anal. Chem. Int. 1995; 78: 521–527.

Harmful algal blooms and ocean observing systems 333

McGillicuddy D Jr, Lynch DR, Wiebe P, Runge J, Durbin EG, Gentleman WC, Davis CS. Evaluating thesynopticity of the U.S. Globec Georges Bank broad-scale sampling pattern with observational systemsimulation experiments. Deep-Sea Res. II 2001; 48: 483–499.

McGillicuddy D Jr, Anderson DM, Lynch DR, Townsend DW. Mechanisms regulating large-scale seasonalfluctuations in Alexandrium fundyense populations in the Gulf of Maine: Results from a physical-biological model. Deep-Sea Res. II 2005; 52(19–21): 2698–2714.

Micheli L, Radoi A, Guarrina R, Massaud R. Disposable immunosensor for the determination of domoicacid in shellfish. Biosensors and Bioelectronics 2004; 20: 190–196.

National Ocean Research Leadership Council (NORLC). An Integrated Ocean Observing System: A Strat-egy for Implementing the First Steps of a U.S. Plan. National Oceanographic Partnership ProgramOffice, Consortium for Oceanographic Research and Education, Washington DC. 1999.

Okaichi T, Ochi T, Nishio S, Takano H, Matsuno K, Morimoto T, Murakami T, Shimida S. The cause of fishkills associated with red tides of Chattonella antiqua (HADA) ONO. In: Miyachi S, Karube I, Ishida Y(eds). Current Topics in Marine Biotechnology. Japanese Society for Marine Biotechnology. 1989;185–188.

ORION Executive Steering Committee. Ocean Observatories Initiative Science Plan. Washington DC. 2005;102 pp.

Paul J, Scholin CA, van den Engh G, Perry MJ. In situ Instrumentation. Oceanography 2007; 20: 58–66.Penna A, Magnani M. Identification of Alexandrium (Dinophyceae) species using PCR and rDNA-targeted

probes. J. Phycol. 1999; 35: 615–621.Pettigrew NR, Churchill JH, Janzen CD, Mangum LJ, Signell RP, Thomas AC, Townsend, DW, Wallinga JP,

Xue H. The kinetic and hydrographic structure of the Gulf of Maine Coastal Current. Deep-Sea Res. II2005; 52(19–21): 2369–2391.

Quilliam MA. Liquid chromatography-mass spectrometry of seafood toxins. In: Barcelo D (ed). Applicationsof LC-MS in environmental science. Elsevier, Amsterdam. 1996; 415–444.

Ramsdell JS, Anderson DM, Glibert PM (eds). HARRNESS, Harmful Algae Research and Response: A Na-tional Environmental Science Strategy 2005–2015. Ecological Society of America, Washington DC.2005; 96 pp.

Richardson TL, Pickney JL. Monitoring of the toxic dinoflagellate Karenia brevis using gyroxanthin-baseddetection methods. J. Appl. Phycol. 2004; 16: 315–328.

Robbins IC, Kirkpatrick GJ, Blackwell SM, Hillier J, Knight CA, Moline MA. Improved monitoring of HABsusing autonomous underwater vehicles (AUV). Harmful Algae 2006; 5(6): 749–761.

Robinson AR, Lermusiaux PFJ, Sloan NQS III. Data assimilation. The Sea 1998; 10: 541–594.Roman B, Scholin C, Jensen S, Massion E, Marin R III, Preston C, Greenfield D, Jones W, Wheeler K. Control-

ling a robotic marine water sampler with the ruby scripting language. J. Amer. Lab. Automation 2007;12: 56–61.

Scholin C, Massion E, Mellinger E, Brown M, Wright D, Cline D. The development and application ofmolecular probes and novel instrumentation for detection of harmful algae. Ocean Community Con-ference, Marine Technology Society. 1998.

Scholin C, Jensen S, Roman B, Massion E, Marin R, Preston C, Greenfield D, Jones W, Wheeler K. TheEnvironmental Sample Processor (ESP)—An Autonomous Robotic Device for Detecting MicroorganismsRemotely Using Molecular Probe Technology. Proceedings, OCEANS 2006 MTS/IEEE Conference.Boston, MA. Marine Technology Society, Columbia, MD. 2006.

Scholin CA, Buck KR, Britschgi T, Cagelosi G, Chavez FP. Identification of Pseudo-nitzschia australis (Bacillario-phyceae) using rRNA-targeted probes in whole cell and sandwich hybridization formats. Phycologia1996; 35: 190–197.

Scholin CA, Doucette GJ, Cembella AD. Prospects for developing automated systems for in situ detectionof harmful algae and their toxins. In: Babin M, Roesler C, Cullen J (eds). Monographs on oceano-graphic methodology. UNESCO (in press).

Sellner K, Sellner G, Doucette GJ, Kirkpatrick GJ. Harmful algal blooms: causes, impacts and detection.J. Ind. Microbiol. Biotechnol. 2003; 30: 383–406.

Shumway SE, Sherman-Caswell S, Hurst JW. Paralytic shellfish poisoning in Maine: Monitoring a monster.J. Shellfish Res. 1988; 7: 643–652.

Stevens RC, Soelberg DD, Eberhart BT, Sylvester Spencer L, Wekell JC, Chinowsky TM. Detection of thetoxin domoic acid from clam extracts using a portable surface plasmon resonance biosensor. HarmfulAlgae 2007; 6: 166–174.

334 D. M. ANDERSON

Stock CA, McGillicuddy, DJ, Solow AR, Anderson DM. Evaluating hypotheses for the initiation and develop-ment of Alexandrium fundyense blooms in the western Gulf of Maine using a coupled physical-biologi-cal model. Deep-Sea Res. II 2005; 52(19–21): 2715–2744.

Townsend DW, Pettigrew NR Thomas AC. Offshore blooms of the red tide dinoflagellate, Alexandrium sp.,in the Gulf of Maine. Cont. Shelf Res. 2001; 21: 347–369.

Wenner PG, Bell RJ, van Amerom FHW, Toler SK, Edkins JE, Hall ML, Koehn K, Short RT, Byrne RH. Environ-mental chemical mapping using an underwater mass spectrometer. Trends in Anal. Chem., Specialissue on Deploying mass spectrometers in harsh environments 2004; 23: 288–295.

White AW, Fukuhara O, Anraku M. Mortality of fish larvae from eating toxic dinoflagellates or zooplanktoncontaining dinoflagellate toxins. In: Okaichi T, Anderson DM, Nemoto T (ed). Red Tides: BiologyEnvironmental Science and Toxicology. Elsevier, NY. 1989; 395–398.

Van Dolah FM. Marine algal toxins: origins, health effects, and their increased occurrence. Environ. HealthPerspect. 2000; 108S: 133–141.

Van Dolah FM, Ramsdell JS. Review and assessment of in vitro detection methods for algal toxins. J. Assoc.Anal. Chem. Int. 2001; 84: 1617–1625.

Van Dolah FM, BL Finley, GJ Doucette, PD Moeller, JS Ramsdell. Development of rapid and sensitive highthroughput pharmacologic assays for marine phycotoxins. Natural Toxins 1994; 2: 189–196.

Xie H, Lazure P, Gentien P. Small-scale retentive structures and Dinophysis. J. Mar. Sys. 2007; 64: 173–188.