Marine Ecology Progress Series 487:176

THEME SECTION

Biophysical coupling of marine hotspots

Editors: E. L. Hazen, R. M. Suryan, S. J. Bograd, Y. Watanuki, R. P. Wilson

Marine Ecology Progress Series Vol. 487, pages 176304

Hazen EL, Suryan RM, Santora JA, Bograd SJ,Watanuki Y, Wilson RPINTRODUCTION: Scales and mechanisms ofmarine hotspot formation...................................177183

Prairie JC, Ziervogel K, Arnosti C,Camassa R, Falcon C, Khatri S,McLaughlin RM, White BL, Yu SDelayed settling of marine snow at sharpdensity transitions driven by fluid entrain-ment and diffusion-limited retention................185199

Boucher JM, Chen C, Sun Y, Beardsley RCEffects of interannual environmental variability on the transport-retentiondynamics in haddock Melanogrammusaeglefinus larvae on Georges Bank..................201215

Nishikawa H, Ichiro Y, Komatsu K, Sasaki H,Sasai Y, Setou T, Shimuzu MWinter mixed layer depth and spring bloomalong the Kuroshio front: implications for theJapanese sardine stock ......................................217229

Smith CE, Hurley BJ, Toms CN, Mackey AD,Solangi M, Kuczaj SA IIHurricane impacts on the foraging patterns ofbottlenose dolphins Tursiops truncatus inMississippi Sound...............................................231244

Pardo MA, Silverberg N, Gendron D,Beier E, Palacios DMRole of environmental seasonality in the turn-over of a cetacean community in the south-western Gulf of California .................................245260

Thorne LH, Read AJFine-scale biophysical interactions drive preyavailability at a migratory stopover site for Phalaropus spp. in the Bay of Fundy, Canada ...261273

Drew GS, Piatt JF, Hill DFEffects of currents and tides on fine-scaleuse of marine bird habitats in a SoutheastAlaska hotspot ....................................................275286

Santora JA, Veit RRSpatio-temporal persistence of top predatorhotspots near the Antarctic Peninsula ..............287304

FREEREE ACCESSCCESS

CONTENTS

sponsored by

North Pacific Marine Science Organization(www.pices.int)

MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 487: 177183, 2013doi: 10.3354/meps10477

Published July 30

What is a biological hotspot?

The term hotspot is used with increased fre-quency in marine biology and conservation litera-ture. The concept of a hotspot of biodiversity hasa longer history in the terrestrial community, withMyers (1988) defining hotspots as areas featuringboth high endemism and risk to habitat (Myers et al.2000). These concepts translate well to more staticmarine habitats such as coral reefs and kelp forests,but are less easily applied to pelagic systems, whereboth boundaries and features are dynamic. Here, webuild upon previous studies that have defined pelagichotspots based on bathymetric variation (Dower &Brodeur 2004) and ocean features in the NorthPacific (Sydeman et al. 2006), to identify the biophys-ical mechanisms that result in hotspot formation. This

requires defining the concept of a marine hotspot,particularly when it consists of mobile features.

We have taken a biophysical approach to definingmarine hotspots, focusing on their ecological ratherthan their conservation importance. Understandingmechanisms that result in hotspot formation is criticalto identify areas of high ecological importance andultimately conservation concern. Hotspots in marinesystems can be defined by (1) important life historyareas for a particular species, (2) areas of high biodi-versity and abundance of individuals, and (3) areasof important productivity, trophic transfer, and bio-physical coupling (Dower & Brodeur 2004, Sydemanet al. 2006, Santora & Veit 2013, this Theme Section).Areas of high trophic transfer are of particular inter-est, because predictable and recurrent productivityhotspots often serve as the foundation of pelagic food

Inter-Research 2013 www.int-res.com*Email: [email protected]

INTRODUCTION

Scales and mechanisms of marine hotspot formation

Elliott L. Hazen1,2,*, Robert M. Suryan3, Jarrod A. Santora4, 5, Steven J. Bograd1,Yutaka Watanuki6, Rory P. Wilson7

1National Oceanic and Atmospheric Administration, Southwest Fisheries Science Center, Pacific Grove, California 93950, USA2Institute of Marine Sciences, University of California, Santa Cruz, 100 Shaffer Road, Santa Cruz, California 95060, USA

3Oregon State University, Hatfield Marine Science Center, 2030 SE Marine Science Dr., Newport, Oregon 97365, USA4Farallon Institute for Advanced Ecosystem Research, 101 H Street, Suite Q, Petaluma, California 94952, USA

5Center for Stock Assessment and Research, University of California Santa Cruz, 110 Shaffer Road, Santa Cruz, California 95060, USA

6Graduate School of Fisheries, Hokkaido University, Hakodate, Hokkaido 041-8611, Japan7Biosciences, College of Science, Swansea University, Singleton Park, Swansea SA2 8PP, UK

ABSTRACT: Identifying areas of high species diversity and abundance is important for under-standing ecological processes and conservation planning. These areas serve as foraging habitator important breeding or settlement areas for multiple species, and are often termed hotspots.Marine hotspots have distinct biophysical features that lead to their formation, persistence, andrecurrence, and that make them important oases in oceanic seascapes. Building upon a session atthe North Pacific Marine Science Organization (PICES), this Theme Section explores the scalesand mechanisms underlying hotspot formation. Fundamentally, understanding the mechanisms ofhotspot formation is important for determining how hotspots may shift relative to ocean featuresand climate change, which is a prerequisite for determining management priorities.

KEY WORDS: Hotspot Ocean features Aggregations Bottom-up processes Biodiversity Marine conservation

Resale or republication not permitted without written consent of the publisher

FREEREE ACCESSCCESS

Contribution to the Theme Section Biophysical coupling of marine hotspots

Mar Ecol Prog Ser 487: 177183, 2013

webs. Fundamentally, hotspot formation operatesacross a suite of spatial and temporal scales (dis-cussed in the next section; see Fig. 1).

Life-history hotspots are critical to a large propor-tion of a species or population, particularly at sen -sitive life history stages. Examples of life-historyhotspots include spawning aggregations, juvenilesettling habitat, and areas providing unique foragingresources. For example, bluefin tuna migrate acrossthe Atlantic and regularly use the warm waters inthe Gulf of Mexico to spawn (Teo et al. 2007), andgrouper species form spawning aggregations in pre-dictable regions (Beets & Friedlander 1999, Sala et al.2001, De Mitcheson et al. 2008). Current speed, eddyactivity, or shelf break habitat within these regionsmay be important for larval dispersal or retention tomaximize survival (Teo et al. 2007, Heyman & Kjer-fve 2008). Benthic habitats such as seagrass beds canserve as settlement areas for pelagic fish (Ford et al.2010) and as foraging habitat for juvenile turtles(McClellan & Read 2007, Casale et al. 2012). Forag-ing hotspots close to a breeding colony can support alarge portion of each species population while alsoserving as important areas of trophic energy transferfrom the physical environment to phytoplankton toseabirds (Santora et al. 2012a).

At the broadest scales, biodiversity hotspots mostfrequently occur in upwelling systems, coral reef eco-systems, and along some continental shelves (Titten-sor et al. 2010). Where tropical and temperate habi-tats meet, there are consistent peaks in oceanicpredator biodiversity (Worm et al. 2003). The Califor-nia Current and North Pacific transition zone standout as particular high biodiversity and high use hot -spots (Bograd et al. 2010, Block et al. 2011). Coralreefs often contain high biodiversity, as they provideimportant structure and habitat near coastlines sur-rounding tropical and sub-tropical waters (Robertset al. 2002, Bellwood et al. 2004). High biodiversityallows multiple paths of trophic transfer bufferingagainst wasp-waist dominance of marine food webs(Field et al. 2006, Cury et al. 2008).

Trophic transfer hotspots often translate biophysi-cal processes across multiple trophic levels by sup-porting a suite of mid-trophic organisms and, inturn, their predators. These areas often have alarge ecosystem effect even though they may onlysupport a subset of a predators population or maynot be areas of highest biodiversity. Aggregationsof mid-trophic species can be important hotspotsfor top predators that migrate large distances tooptimize foraging opportunities (Cotte & Simard2005, Bailey et al. 2010). The mechanisms underly-

ing trophic hotspots can include island/ seamountwake effects (Johnston & Read 2007, Morato et al.2010), up welling shadows (Nur et al. 2011, Wing-field et al. 2011, Pardo et al. 2013, this Theme Sec-tion), wind or eddy-driven upwelling (Croll et al.2005, Atwood et al. 2010, Thorne & Read 2013, thisTheme Section), or bathymetric features (Croll etal. 2005, Gende & Sigler 2006). Fundamentally,changes in these hotspots may have indirect conse-quences on ecosystem functioning that cascadethrough to top predators.

Scales of hotspots

Inherent to all studies of marine hotspots is the con-cept of scale ecological phenomena interact at dis-crete and often multiple spatial and temporal scales.A Stommel diagram of hotspot mechanisms showshow processes vary across space and time, and forsimplicity assumes that the scaling between time andspace is linear (Fig. 1). However, meso- and fine-scale studies of aggregative responses among oceanphysics, predators and prey have revealed complexnon-linear interactions (Hunt & Schneider 1987, Piatt1990, Fauchald et al. 2000).

Research on the scale of physical and biologicalhotspots is often dictated by the sampling methodolo-gies and technology employed. For example: (1)Satellite-based observations of ocean conditions offerthe greatest flexibility by sampling broadly throughspace and time at fine to basin scales, but are limitedto surface conditions and only sample up to a proxyfor primary production in chlorophyll a (Palacios etal. 2006). (2) Satellite-tracking of vertebrate predatorsis dependent on the resolution of tracking devices(e.g. GPS, pop-up archival), the predators move-ment, and initial tagging location, but offer excep-tional insight into top predator behavior, distribution,and their use of multiple marine ecosystems (e.g.hotspot connectivity, residency: Block et al. 2011,Bailey et al. 2012, Hazen et al. 2012; tagging throughthe stages: Montevecchi et al. 2012). (3) Shipboardsurveys employ a variety of discrete and continuoussampling devices (e.g. nets, acoustics, visual obser-vations) to quantify vertical and horizontal dimen-sions of seascapes and preyscapes, simultaneouslyoffering insight into hotspot mechanisms and metricsof biodiversity (Santora et al. 2010, 2012a, Sigler et al.2012), but are expensive and highly influenced byweather conditions. Ultimately, the integration acrossmultiple types of observations should help resolvespatio-temporal mismatches.

178

Hazen et al.: Introduction to Theme Section on biophysical coupling of marine hotspots

Studies of global hotspots, especially diversity/richness hotspots, generally focus on relatively largegrid size resolutions (e.g. 10 km; months to decades)and may cover entire marine ecosystems (Tittensor etal. 2010, Block et al. 2011). This approach is useful forcomparing biodiversity and risk across ecosystemsand identifying important areas warranting fine-scale study (Halpern et al. 2008). Fine-scale researchis necessary to elucidate mechanisms of biophysicalhotspot formation and persistence and to under-stand critical species interactions within hotspots. Atthe global scale, most metrics for physical variabilityidentified as underlying hotspots are proxies for keyspecies interactions (Dawson et al. 2011, Hazen et al.2013, Mokany et al. 2013).

Mesoscale structure (10 to 1000 km; days tomonths) underlying physical and biological compo-nents of marine ecosystems often determine thestrength and recurrence of marine hotspots, and canprovide criteria for defining areas of high trophictransfer. Studies at fine spatial scales (1 to 10 km)examining predator preyoceanography relationshipsare critical to describe the mechanisms that deter-mine which meso scale hotspots are formed and per-sist. For example, tidal flow and internal waves pass-ing over topographic features in the Gulf of Maineresult in dense aggregations of both krill and sandlance, which support seasonal foraging for hump-

back and fin whales (Stevick et al. 2008,Hazen et al. 2009). Through the integratedassessment of physics, primary produc -tivity, and secondary production, the cou-pling of fine and mesoscale samplingoffers promising directions for studies ofmarine hotspots (Cury et al. 2008).

Mechanisms of hotspot formation andpersistence

Physical processes leading to hotspotformation are varied, ranging from nutri-ent input to retention or aggregation ofsubsequent biological production. Mecha-nisms of nutrient input into the euphoticzone include freshwater run-off (Chase etal. 2007, Planquette et al. 2011), aeoliansources (Fan et al. 2006), and up welling ofdeep, nutrient rich water (Mes khidze etal. 2007). Up welling (or water column mixing) can be seasonal or episodic whenwind driven; however, upwelling-enhancedproductivity can also be highly persistent,

especially when resulting from relatively static orcyclical processes, such as the interaction of theEquatorial Undercurrent meeting the western Gala-pagos Islands (Palacios et al. 2006) or small scale,tidally driven current interactions with bathymetricfeatures within a bay (Drew et al. 2013, Thorne &Read 2013, both in this Theme Section). Likewise,anticyclonic eddies that form in coastal regions andspin-off into or form in the oceanic domain canentrain or upwell macronutrients, leading to travel-ing open ocean hotspots of productivity (Crawfordet al. 2007) that are often utilized by upper trophiclevel predators (Ream et al. 2005).

Hotspots can be a function of biophysical aggrega-tion where physical features such as shelf breaks orocean fronts lead to increased densities of phyto-plankton or zooplankton, or bottom-up processes,such as where increased nutrient levels lead togreater primary productivity, greater densities ofgrazers, and so on up to mid-trophic and top-levelpredators (Mann & Lazier 1996, Genin 2004). Due totransport mechanisms and temporal lags, a foraginghotspot may not coincide with a region of enhancedprimary productivity. In these cases, spatial dis -cordance can result from downstream transport ofprey, such as zooplankton for foraging birds (Hunt etal. 1998, Thorne & Read 2013, this Theme Section).There are good examples of bathymetric features

179

Sp

ace

Time

Tidalwake

eddies

Shelf break &canyon effects

Eddies,jets &fronts

Island &seamount

effects

Gyreboundaries

Westernboundarycurrents

Thinlayers

Upwellingdynamics

Hours

109 m

Basin

Meso

Fine

106 m

103 m

1 m

Days DecadesMonths Years

Fig. 1. Stommel diagram. Spatial scale (y-axis) plotted against temporalscale, focusing on persistence (x-axis). Blue: bathymetric-driven hotspots;green: dynamic features that can move throughout the ocean. For exam-ple, upwelling dynamics operate on a temporal scale of days to monthswhile tidally driven mixing lasts for hours. Black outline: features that arepersistent throughout time rather than recurrent; dashed outline: features

that are ephemeral and neither persistent nor recurrent


creating important aggregative habitat for speciessuch as krill where they can reduce their exposure tocurrents while being close to foraging needs (Fiedleret al. 1998, Cotte & Simard 2005, Santora et al. 2010,Santora et al. 2011). These aggregations then be -come important foraging features for large predatorssuch as baleen whales that require high densities ofprey to maximize their foraging efficiency and ener-getic demands (Cotte & Simard 2005, Sigler et al.2012). Top-down hotspots are rare, but facilitated for-aging, where pelagic predators such as tuna forceforage fish towards the surface and make them moreavailable to seabirds, can result in higher biodiver-sity at pelagic hotspots (Maxwell & Morgan 2013).

The spatial and temporal dynamics of marinehotspot occurrence and persistence is quite impor-tant, yet infrequently examined (but see Gende &Sigler 2006, Sigler et al. 2012, Suryan et al. 2012,Santora & Veit 2013, this Theme Section). Manyinvestigations have focused on long-term averages(i.e. spatial climatologies) of physical and biologicalconditions to map hotspots, but the greatest chal-lenges for research on pelagic hotspots require stud-ies melding space and time to quantify persistence ofhotspots. Quantifying their spatio-temporal persist-ence will require highly replicated observations toestablish a baseline scale of variability and measurethe frequency of anomalies (Suryan et al. 2012). Dueto their ability to sample local to global spatial scalesover days to years, satellite-based observations ofocean conditions offer the greatest opportunity toexamine the spatio-temporal persistence of manymarine hotspots (Palacios et al. 2006). Moored obser-vatories (e.g. Neptune, Diemos, Mars), repeated glidertransects, or regular surveys can allow enhancedtemporal observations at marine hotspot locations (Biet al. 2007, Moustahfid et al. 2012).

Contributions to the Theme Section

This Theme Section arose out of a session at the2011 Annual Meeting of the North Pacific MarineScience Organization (PICES). The session builtupon previous efforts to identify hotspots in the NorthPacific and examine the biophysical mechanisms thatresult in their formation (Dower & Brodeur 2004,Sydeman et al. 2006). The session consisted of 19(total) presentations and posters. This Theme Sectioncontains 8 studies with topics ranging from hotspotsof marine snow to migratory top predators.

Prairie et al. (2013) in a laboratory experimentdemonstrated mechanisms by which particles can be

temporarily retained when encountering density gra-dients. The density of the aggregate relative to thedensity of the bottom layer in the gradient is the pri-mary determinant of the extent that marine snow willaggregate, thereby enhancing food web develop-ment at the boundary layer.

Boucher et al. (2013) used oceanographic and indi-vidual-based movement models to examine larvalhaddock dynamics in the Gulf of Maine. In goodyears, increased retention lead to hotspots of larvalhaddock on the bank but additional factors played arole in the magnitude of haddock recruitment in agiven year.

Nishikawa et al. (2013) examined how water col-umn characteristics lead to phytoplankton blooms inthe western North Pacific that develop into importantspawning habitat for Japanese sardine Sardinopsmelanostictus. A deeper mixed layer and higherphytoplankton density resulted in increased spawn-ing habitat north of the Kuroshio.

Smith et al. (2013) examined how a hurricane pass-ing through the Gulf of Mexico may have temporarilyincreased foraging hotspots for bottlenose dolphins.The decline in foraging habitat months after Hurri-cane Katrina suggests some hotspots, such as sea-grass beds, may have been lost or disrupted by thehurricane.

Pardo et al. (2013) studied the role of environmen-tal seasonality in a cetacean community within asmall bay in the Gulf of California. Different speciesused the bay as the season progressed; periods ofmixing and pycnocline shoaling resulted in increasedhabitat for blue whales and 2 dolphin species, whileother whales were more common during periods ofstratification.

Thorne et al. (2013) evaluated the biophysical pro-cesses that structure foraging habitat for phalaropes,a surface-feeding, planktivorous seabird in the Bayof Fundy. Their model indicates that copepods arephysically upwelled and advected down current,highlighting the potential for spatial mismatch oflower trophic level food web processes and predatorforaging.

Drew et al. (2013) used visual transect surveys toexamine how foraging strategies of seabirds resultedin differential habitat use in Glacier Bay, Alaska.Bathymetric variability results in differential currentspeeds and consequently a high diversity of hotspottypes in the bay; modeling approaches may help tounderstand the development of fine scale foragingbehavior that has so far been difficult to quantify.

Santora & Veit (2013) examined species richnessand abundance in top predators to identify persistent

180


hotspots near the Antarctic Peninsula. They identi-fied 15 richness hotspots associated with either theAntarctic Circumpolar Current, major breedingcolonies, or submarine canyons.

Future directions

To cross spatio-temporal boundaries, more syn-thetic approaches to hotspot research are necessary,such as combining Eulerian and Lagrangian meas-urements of hotspot dynamics. Combining tag-basedmovement data with shipboard surveys can provideinformation on behavioral ecology and biodiversityto address the suite of physical and ecological pro-cesses that result in formation and prolonged useof marine hotspots at multiple trophic levels (seeBenoit-Bird et al. 2013). Future studies of biophysicalhotspots should explicitly address the scale andscope of their defined hotspot so that syntheses andcomparisons can be made across studies of disparatemarine ecosystems (Santora et al. 2012b).

One category of marine hotspots that remainsunder-researched is the deep scattering layers (DSLs)of the open ocean, which are made up of a complexof species including fish, shrimp, jellies, and squid.DSLs have been observed at various depths aroundthe world, yet little is known on their extent or vari-ability (Barham 1963). They serve as a critical preyresource in otherwise oligotrophic ocean basins, sounderstanding the spatial and temporal distributionof DSL hotspots is critical (Benoit-Bird & Au 2004).Recent studies in Monterey Bay have shown a highdegree of temporal variability in distribution, bothvertical and horizontal, and abundance seasonally, ofDSL organisms (Urmy et al. 2012). Broad scale spatialpatterns of deepwater fishes in the southern Califor-nia Bight were recently described, with low oxygenlevels proposed as a primary mechanism determin-ing vertical distributions (Koslow et al. 2011).

An increasing number of studies focus on areas inthe ocean that are important for conservation. TheGlobal Ocean Biodiversity Initiative (GOBI) is aninternational organization working to define Eco -logical and Biologically Significant Areas (EBSAs) inthe worlds oceans (Williams et al. 2010). The clearidentification of hotspots and the establishment of ahotspot repository would ensure the effective studyof hotspot mechanisms and persistence, and subse-quently inform management and conservation efforts.Dynamic management has been proposed and evenimplemented in a few systems, but with ocean useprojected to increase in the future, new tools are

required to optimize ecological services and ecosys-tem functioning (Hobday et al. 2010, Dunn et al.2011, Grantham et al. 2011, Ronconi et al. 2012).Examination of the overlap between human-use hot -spots and the temporal persistence of ecological hot -spots will enable real-time management approachesto allow human uses when hotspots are absent andprotect habitats when hotspots are persistent.

LITERATURE CITED

Atwood E, Duffy-Anderson JT, Horne JK, Ladd C (2010)Influence of mesoscale eddies on ichthyoplankton assem -blages in the Gulf of Alaska. Fish Oceanogr 19: 493507

Bailey H, Mate BR, Palacios DM, Irvine L, Bograd SJ, CostaDP (2010) Behavioural estimation of blue whale move-ments in the Northeast Pacific from state-space modelanalysis of satellite tracks. Endang Species Res 10: 93106

Bailey H, Benson SR, Shillinger GL, Bograd SJ and others(2012) Identification of distinct movement patterns inPacific leatherback turtle populations influenced by oceanconditions. Ecol Appl 22: 735747

Barham EG (1963) Siphonophores and the deep scatteringlayer. Science 140: 826828

Beets J, Friedlander A (1999) Evaluation of a conservationstrategy: a spawning aggregation closure for red hind,Epinephelus guttatus, in the U.S. Virgin Islands. EnvironBiol Fishes 55: 9198

Bellwood DR, Hughes TP, Folke C, Nystrom M (2004) Con-fronting the coral reef crisis. Nature 429: 827833

Benoit-Bird KJ, Au WWL (2004) Diel migration dynamics ofan island-associated sound-scattering layer. Deep-SeaRes I 51: 707719

Benoit-Bird KJ, Battaile BC, Heppell SA, Hoover B and oth-ers (2013) Prey patch patterns predict habitat use by topmarine predators with diverse foraging strategies. PLoSONE 8: e53348

Bi HS, Ruppel RE, Peterson WT (2007) Modeling the pelagichabitat of salmon off the Pacific Northwest (USA) coastusing logistic regression. Mar Ecol Prog Ser 336: 249265

Block BA, Jonsen ID, Jorgensen SJ, Winship AJ and others(2011) Tracking apex marine predator movements in adynamic ocean. Nature 475: 8690

Bograd SJ, Block BA, Costa DP, Godley BJ (2010) Biologgingtechnologies: new tools for conservation. Introduction.Endang Species Res 10: 17

Boucher JM, Chen C, Sun Y, Beardsley RC (2013) Effectsof interannual environmental variability on the trans-port-retention dynamics in haddock Me lanogrammusaeglefinus larvae on Georges Bank. Mar Ecol Prog Ser487: 201215

Casale P, Broderick AC, Freggi D, Mencacci R, Fuller WJ,Godley BJ, Luschi P (2012) Long-term residence of juve-nile loggerhead turtles to foraging grounds: a potentialconservation hotspot in the Mediterranean. Aquat Con-serv 22: 144154

Chase Z, Strutton PG, Hales B (2007) Iron links river runoffand shelf width to phytoplankton biomass along the U.S.West Coast. Geophys Res Lett 34LO4607, doi:10.1029/2006GL028069

Cotte C, Simard Y (2005) Formation of dense krill patchesunder tidal forcing at whale feeding hot spots in the St.Lawrence Estuary. Mar Ecol Prog Ser 288: 199210

181

http://dx.doi.org/10.3354/meps288199http://dx.doi.org/10.1002/aqc.2222http://dx.doi.org/10.3354/meps10462http://dx.doi.org/10.3354/esr00269http://dx.doi.org/10.1038/nature10082http://dx.doi.org/10.3354/meps336249http://dx.doi.org/10.1371/journal.pone.0053348http://dx.doi.org/10.1016/j.dsr.2004.01.004http://dx.doi.org/10.1038/nature02691http://dx.doi.org/10.1023/A%3A1007404421518http://dx.doi.org/10.1126/science.140.3568.826http://dx.doi.org/10.1890/11-0633http://dx.doi.org/10.3354/esr00239http://dx.doi.org/10.1111/j.1365-2419.2010.00559.x


Crawford WR, Brickley PJ, Thomas AC (2007) Mesoscaleeddies dominate surface phytoplankton in northern Gulfof Alaska. Prog Oceanogr 75: 287303

Croll DA, Marinovic B, Benson S, Chavez FP, Black N, Ter-nullo R, Tershy BR (2005) From wind to whales: trophiclinks in a coastal upwelling system. Mar Ecol Prog Ser289: 117130

Cury PM, Shin YJ, Planque B, Durant JM and others (2008)Ecosystem oceanography for global change in fisheries.Trends Ecol Evol 23: 338346

Dawson TP, Jackson ST, House JI, Prentice IC, Mace GM(2011) Beyond predictions: biodiversity conservation in achanging climate. Science 332: 5358

De Mitcheson YS, Cornish A, Domeier M, Colin PL, RussellM, Lindeman KC (2008) A global baseline for spawningaggregations of reef fishes. Conserv Biol 22: 12331244

Dower JF, Brodeur RD (2004) The role of biophysical cou-pling in concentrating marine organisms around shallowtopographies. J Mar Syst 50: 12

Drew GS, Piatt JF, Hill DF (2013) Effects of currents and tideson a fine-scale use of marine bird habitats in a SoutheastAlaska hotspot. Mar Ecol Prog Ser 487: 275286

Dunn DC, Boustany AM, Halpin PN (2011) Spatio-temporalmanagement of fisheries to reduce by-catch and increasefishing selectivity. Fish Fish 12: 110119

Fan SM, Moxim WJ, Levy H II (2006) Aeolian input ofbioavailable iron to the ocean. Geophys Res Lett 33: L07602, doi: 10.1029/2005GL024852

Fauchald P, Erikstad KE, Skarsfjord H (2000) Scale-depen-dent predator-prey interactions: the hierarchical spatialdistribution of seabirds and prey. Ecology 81: 773783

Fiedler PC, Reilly SB, Hewitt RP, Demer D and others (1998)Blue whale habitat and prey in the California ChannelIslands. Deep-Sea Res II 45: 17811801

Field JC, Francis RC, Aydin K (2006) Top-down modelingand bottom-up dynamics: linking a fisheries-based eco-system model with climate. Prog Oceanogr 68: 238270

Ford JR, Williams RJ, Fowler AM, Cox DR, Suthers IM(2010) Identifying critical estuarine seagrass habitat forsettlement of coastally spawned fish. Mar Ecol Prog Ser408: 181193

Gende SM, Sigler MF (2006) Persistence of forage fish hotspots and its association with foraging Steller sea lions(Eumetopias jubatus) in southeast Alaska. Deep-Sea ResII 53: 432441

Genin A (2004) Bio-physical coupling in the formation ofzooplankton and fish aggregations over abrupt topogra-phies. J Mar Syst 50: 320

Grantham HS, Game ET, Lombard AT, Hobday AJ and others (2011) Accommodating dynamic oceanographicprocesses and pelagic biodiversity in marine conserva-tion planning. PLoS ONE 6: e16552

Halpern BS, Walbridge S, Selkoe KA, Kappel CV and others(2008) A global map of human impact on marine ecosys-tems. Science 319: 948952

Hazen EL, Friedlaender AS, Thompson MA, Ware CR,Weinrich MT, Halpin PN, Wiley DN (2009) Fine-scaleprey aggregations and foraging ecology of humpbackwhales Megaptera novaeangliae. Mar Ecol Prog Ser 395: 7589

Hazen EL, Maxwell SM, Bailey H, Bograd SJ and others(2012) Ontogeny in marine tagging and tracking science: technologies and data gaps. Mar Ecol Prog Ser 457: 221240

Hazen EL, Jorgensen S, Rykaczewski RR, Bograd SJ and

others (2012) Predicted habitat shifts of Pacific top pred-ators in a changing climate. Nat Clim Change 3: 234238

Heyman WD, Kjerfve B (2008) Characterization of transientmulti-species reef fish spawning aggregations at Glad-den Spit, Belize. Bull Mar Sci 83: 531551

Hobday AJ, Hartog JR, Timmiss T, Fielding J (2010)Dynamic spatial zoning to manage southern bluefin tuna(Thunnus maccoyii) capture in a multi-species longlinefishery. Fish Oceanogr 19: 243253

Hunt GL Jr, Schneider DC (1987) Scale-dependent pro-cesses in the physical and biological environment of mar-ine birds. In: Croxall JP (ed) Seabirds: feeding ecologyand role in marine ecosystems. Cambridge UniversityPress, Cambridge

Hunt GL Jr, Russell RW, Coyle KO, Weingartner T (1998)Comparative foraging ecology of planktivorous aukletsin relation to ocean physics and prey availability. MarEcol Prog Ser 167: 241259

Johnston DW, Read AJ (2007) Flow-field observations of atidally driven island wake used by marine mammals inthe Bay of Fundy, Canada. Fish Oceanogr 16: 422435

Kobayashi DR, Cheng IJ, Parker DM, Polovina JJ, KamezakiN, Balazs GH (2011) Loggerhead turtle (Caretta caretta)movement off the coast of Taiwan: characterization of ahotspot in the East China Sea and investigation ofmesoscale eddies. ICES J Mar Sci 68: 707718

Koslow JA, Goericke R, Lara-Lopez A, Watson W (2011)Impact of declining intermediate-water oxygen on deep-water fishes in the California Current. Mar Ecol Prog Ser436: 207218

Mann KH, Lazier J (1996) Dynamics of marine ecosystems,Vol 389. Blackwell Science, Cambridge, MA

Maxwell SM, Morgan LE (2013) Foraging of seabirds onpelagic fishes: implications for management of pelagicmarine protected areas. Mar Ecol Prog Ser 481: 289303

McClellan CM, Read AJ (2007) Complexity and variation inloggerhead sea turtle life history. Biol Lett 3: 592594

Meskhidze N, Nenes A, Chameides WL, Luo C, MahowaldN (2007) Atlantic Southern Ocean productivity: fertiliza-tion from above or below? Global Biogeochem Cycles 21: GB2006, doi: 10.1029/2006GB002711

Mokany K, Harwood TD, Ferrier S (2013) Comparing habi-tat configuration strategies for retaining biodiversityunder climate change. J Appl Ecol 50: 519527

Montevecchi WA, Hedd A, Tranquilla LM, Fifield DA andothers (2012) Tracking seabirds to identify ecologicallyimportant and high risk marine areas in the westernNorth Atlantic. Biol Conserv 156: 6271

Morato T, Hoyle SD, Allain V, Nicol SJ (2010) Seamounts arehotspots of pelagic biodiversity in the open ocean. ProcNatl Acad Sci USA 107: 97079711

Moustahfid H, Jech MM, Weise MJ, Horne JK, ODor R,Alexander C (2012) Advancing bio sensor integration withOcean Observing Systems to support ecosystem basedapproaches. Proc Oceans 2012. IEEE, New York, NY

Myers N (1988) Threatened biotas: hotspots in tropicalforests. Environmentalist 8: 187208

Myers N, Mittermeier RA, Mittermeier CG, da FonsecaGAB, Kent J (2000) Biodiversity hotspots for conservationpriorities. Nature 403: 853858

Nishikawa H, Ichiro Y, Komatsu K, Sasaki H, Sasai Y, SetouT, Shimuzu M (2013) Winter mixed layer depth andspring bloom along the Kuroshio front: implications forthe Japanese sardine stock. Mar Ecol Prog Ser 487:217229

182

http://dx.doi.org/10.3354/meps10201http://dx.doi.org/10.1038/35002501http://dx.doi.org/10.1038/nature01610http://dx.doi.org/10.1007/BF02240252http://dx.doi.org/10.1073/pnas.0910290107http://dx.doi.org/10.1016/j.biocon.2011.12.001http://dx.doi.org/10.1111/1365-2664.12038http://dx.doi.org/10.1029/2006GB002711http://dx.doi.org/10.1098/rsbl.2007.0355http://dx.doi.org/10.3354/meps10255http://dx.doi.org/10.3354/meps09270http://dx.doi.org/10.1093/icesjms/fsq185http://dx.doi.org/10.1111/j.1365-2419.2007.00444.xhttp://dx.doi.org/10.3354/meps167241http://dx.doi.org/10.1111/j.1365-2419.2010.00540.xhttp://dx.doi.org/10.1038/nclimate1686http://dx.doi.org/10.3354/meps08108http://dx.doi.org/10.1126/science.1149345http://dx.doi.org/10.1371/journal.pone.0016552http://dx.doi.org/10.1016/j.jmarsys.2003.10.008http://dx.doi.org/10.1016/j.dsr2.2006.01.005http://dx.doi.org/10.3354/meps08582http://dx.doi.org/10.1016/j.pocean.2006.02.010http://dx.doi.org/10.1016/S0967-0645(98)80017-9http://dx.doi.org/10.1029/2005GL024852http://dx.doi.org/10.1111/j.1467-2979.2010.00388.xhttp://dx.doi.org/10.3354/meps10304http://dx.doi.org/10.1016/j.jmarsys.2004.04.002http://dx.doi.org/10.1111/j.1523-1739.2008.01020.xhttp://dx.doi.org/10.1126/science.1200303http://dx.doi.org/10.1016/j.tree.2008.02.005http://dx.doi.org/10.3354/meps289117http://dx.doi.org/10.1016/j.pocean.2007.08.016http://dx.doi.org/10.3354/meps09857


Nur N, Jahncke J, Herzog MP, Howar J and others (2011)Where the wild things are: predicting hotspots of seabirdaggregations in the California Current System. EcolAppl 21: 22412257

Palacios DM, Bograd SJ, Foley DG, Schwing FB (2006)Oceanographic characteristics of biological hot spots inthe North Pacific: a remote sensing perspective. Deep-Sea Res II 53: 250269

Pardo MA, Silverberg N, Gendron D, Beier E, Palacios DM(2013) Role of environmental seasonality in the turnoverof a cetacean community in the southwestern Gulf ofCalifornia. Mar Ecol Prog Ser 487: 245260

Piatt JF (1990) The aggregative response of common murresand Atlantic puffins to schools of capelin. Stud Avian Biol14: 3651

Planquette H, Sanders RR, Statham PJ, Morris PJ, Fones GR(2011) Fluxes of particulate iron from the upper oceanaround the Crozet Islands: A naturally iron-fertilizedenvironment in the Southern Ocean. Global BiogeochemCycles 25: GB2011, doi: 10.1029/2010GB003789

Prairie JC, Ziervogel K, Arnosti C, Camassa R, Falcon C,Khatri S, McLaughlin RM, White BL, Yu S (2013)Delayed settling of marine snow at sharp density transi-tions driven by fluid entrainment and diffusion-limitedretention. Mar Ecol Prog Ser 487: 185199

Ream RR, Sterling JT, Loughlin TR (2005) Oceanographicfeatures related to northern fur seal migratory move-ments. Deep-Sea Res II 52: 823843

Roberts CM, McClean CJ, Veron JEN, Hawkins JP and oth-ers (2002) Marine biodiversity hotspots and conservationpriorities for tropical reefs. Science 295: 12801284

Ronconi RA, Lascelles BG, Langham GM, Reid JB, Oro D(2012) The role of seabirds in Marine Protected Areaidentification, delineation, and monitoring: introductionand synthesis. Biol Conserv 156: 14

Sala E, Ballesteros E, Starr RM (2001) Rapid decline of Nas-sau grouper spawning aggregations in Belize: fisherymanagement and conservation needs. Fisheries 26: 2330

Santora JA, Veit RR (2013) Spatio-temporal persistence oftop predator hotspots near the Antarctic Peninsula. MarEcol Prog Ser 487: 287304

Santora JA, Reiss CS, Loeb VJ, Veit RR (2010) Spatial asso-ciation between hotspots of baleen whales and demo-graphic patterns of Antarctic krill Euphausia superbasuggests size-dependent predation. Mar Ecol Prog Ser405: 255269

Santora JA, Sydeman WJ, Schroeder ID, Wells BK, Field JC(2011) Mesoscale structure and oceanographic deter -minants of krill hotspots in the California Current: im -plications for trophic transfer and conservation. ProgOceanogr 91: 397409

Santora JA, Field JC, Schroeder ID, Sakuma KM, Wells BK,Sydeman WJ (2012a) Spatial ecology of krill, micronek-ton and top predators in the central California Current: Implications for defining ecologically important areas.Prog Oceanogr 106: 154174

Santora JA, Sydeman WJ, Schroeder ID, Reiss CS and others(2012b) Krill space: a comparative assessment ofmesoscale structuring in polar and temperate marineecosystems. ICES J Mar Sci 69: 13171327

Sigler MF, Kuletz KJ, Ressler PH, Friday NA, Wilson CD,Zerbini AN (2012) Marine predators and persistent preyin the southeast Bering Sea. Deep-Sea Res II 6570: 292303

Smith CE, Hurley BJ, Toms CN, Mackey AD, Solangi M,Kuczaj SA II (2013) Hurricane impacts on the foragingpatterns of bottlenose dolphins Tursiops truncatus inMississippi Sound. Mar Ecol Prog Ser 487: 231244

Stevick PT, Incze LS, Kraus SD, Rosen S, Wolff N, Baukus A(2008) Trophic relationships and oceanography on andaround a small offshore bank. Mar Ecol Prog Ser 363: 1528

Suryan RM, Santora JA, Sydeman WJ (2012) New approachfor using remotely sensed chlorophyll a to identify sea-bird hotspots. Mar Ecol Prog Ser 451: 213225

Sydeman WJ, Brodeur RD, Grimes CB, Bychkov AS, McKin-nell S (2006) Marine habitat hotspots and their use bymigratory species and top predators in the North PacificOcean: Introduction. Deep-Sea Res II 53: 247249

Teo SH, Boustany A, Block B (2007) Oceanographic prefer-ences of Atlantic bluefin tuna, Thunnus thynnus, on theirGulf of Mexico breeding grounds. Mar Biol 152: 11051119

Thorne LH, Read AJ (2013) Fine-scale biophysical interac-tions drive prey availability at a migratory stopover sitefor Phalaropus spp. in the Bay of Fundy, Canada. MarEcol Prog Ser 487: 261273

Tittensor DP, Mora C, Jetz W, Lotze HK, Ricard D, VandenBerghe E, Worm B (2010) Global patterns and predictorsof marine biodiversity across taxa. Nature 466: 10981107

Urmy SS, Horne JK, Barbee DH (2012) Measuring the verti-cal distributional variability of pelagic fauna in MontereyBay. ICES J Mar Sci 69: 184196

Williams MJ, Ausubel J, Poiner I, Garcia SM and others(2010) Making marine life count: a new baseline for pol-icy. PLoS Biol 8: e1000531

Wingfield DK, Peckham SH, Foley DG, Palacios DM andothers (2011) The making of a productivity hotspot in thecoastal ocean. PLoS ONE 6: e27874

Worm B, Lotze HK, Myers RA (2003) Predator diversityhotspots in the blue ocean. Proc Natl Acad Sci USA 100: 98849888

183

Submitted: February 24, 2012; Accepted: May 23, 2012 Proofs received from author(s): July 20, 2013

http://dx.doi.org/10.1126/science.1113399http://dx.doi.org/10.1073/pnas.1333941100http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=22132156&dopt=Abstracthttp://dx.doi.org/10.1371/journal.pbio.1000531http://dx.doi.org/10.1093/icesjms/fsr205http://dx.doi.org/10.1038/nature09329http://dx.doi.org/10.3354/meps10384http://dx.doi.org/10.1007/s00227-007-0758-1http://dx.doi.org/10.1016/j.dsr2.2006.03.001http://dx.doi.org/10.3354/meps09597http://dx.doi.org/10.3354/meps07475http://dx.doi.org/10.3354/meps10372http://dx.doi.org/10.1093/icesjms/fss048http://dx.doi.org/10.1016/j.pocean.2012.08.005http://dx.doi.org/10.1016/j.pocean.2011.04.002http://dx.doi.org/10.1093/icesjms/fsr046http://dx.doi.org/10.3354/meps08513http://dx.doi.org/10.3354/meps10350http://dx.doi.org/10.1577/1548-8446(2001)026%3C0023%3ARDONGS%3E2.0.CO%3B2http://dx.doi.org/10.1016/j.biocon.2012.02.016http://dx.doi.org/10.1126/science.1067728http://dx.doi.org/10.1016/j.dsr2.2004.12.021http://dx.doi.org/10.3354/meps10387http://dx.doi.org/10.1029/2010GB003789http://dx.doi.org/10.3354/meps10217http://dx.doi.org/10.1016/j.dsr2.2006.03.004http://dx.doi.org/10.1890/10-1460.1

MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 487: 185200, 2013doi: 10.3354/meps10387

Published July 30

INTRODUCTION

Marine snow plays a critical role in the marine car-bon cycle, as a dominant component of carbon fluxfrom the surface ocean, a site of enhanced bacterialactivity, and often an important food source for zoo-plankton (Alldredge & Silver 1988, Smith et al. 1992,

Simon et al. 2002, Kirboe 2011). Knowledge of thevertical distribution of marine aggregates in thewater column, as well as of the factors that underliethese distributions, are critical to understanding thefunctions of aggregates in the ocean.

Recent studies have observed that thin layers ofaggregateswith concentrations several times back -

Inter-Research 2013 www.int-res.com*Email: [email protected]

Delayed settling of marine snow at sharp densitytransitions driven by fluid entrainment and

diffusion-limited retention

Jennifer C. Prairie1,2,3,*, Kai Ziervogel1, Carol Arnosti1, Roberto Camassa2,3, Claudia Falcon2,3, Shilpa Khatri2,3, Richard M. McLaughlin2,3, Brian L. White1,3,

Sungduk Yu1,3

1Department of Marine Sciences, 2Carolina Center for Interdisciplinary Applied Mathematics, Department of Mathematics, and 3Joint Fluids Lab, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, USA

ABSTRACT: Marine snow is central to the marine carbon cycle, and quantifying its small-scalesettling dynamics in different physical environments is essential to understanding its role in bio-geochemical cycles. Previous field observations of marine aggregate thin layers associated withsharp density gradients have led to the hypothesis that these layers may be caused by a decreasein aggregate settling speed at density interfaces. Here, we present experimental data on aggre-gate settling behavior, showing that these particles can dramatically decrease their settling veloc-ity when passing through sharp density transitions. This delayed settling can be caused by 2potential mechanisms: (1) entrainment of lighter fluid from above as the particle passes throughthe density gradient, and (2) retention at the transition driven by changes in the density of the par-ticle due to its porosity. The aggregates observed in this study exhibited 2 distinct settling behav-iors when passing through the density transition. Quantitatively comparing these different behav-iors with predictions from 2 models allow us to infer that the delayed settling of the first group ofaggregates was primarily driven by diffusion-limited retention, whereas entrainment of lighterfluid was the dominant mechanism for the second group. Coupled with theory, our experimentalresults demonstrate that both entrainment and diffusion-limited retention can play an importantrole in determining particle settling dynamics through density transitions. This study thus pro-vides insight into ways that delayed settling can lead to the formation of aggregate thin layers,important biological hotspots that affect trophic dynamics, and biogeochemical cycling in theocean.

KEY WORDS: Aggregation Thin layer Accumulation Biophysical coupling Hotspots

Resale or republication not permitted without written consent of the publisher

Contribution to the Theme Section Biophysical coupling of marine hotspots FREEREE ACCESSCCESS


ground levelsare common in coastal waters (Mac-Intyre et al. 1995, Alldredge et al. 2002, Prairie et al.2010), and are often associated with sharp densitytransitions (MacIntyre et al. 1995, Dekshenieks et al.2001, Prairie et al. 2010). These layers can have sig-nificant consequences for local carbon cycling, sinceaggregates may act as hot spots for bacterial activityor foraging zooplankton (Green & Dagg 1997, Kir-boe 2000, Ziervogel & Arnosti 2008, Ziervogel et al.2010). Some studies have suggested that a possibleexplanation for the formation of these layers is locallydecreased settling speeds as particles pass throughthese regions of sharp stratification (Derenbach etal. 1979, Alldredge & Crocker 1995, MacIntyre et al.1995).

Previous theoretical and experimental studies haveinvestigated the effect of sharp stratification on settling particles, and have shown that there are 2potential mechanisms that could explain a decreasein settling velocity for a particle falling through adensity transition. Studies with solid spheres havedemonstrated that a sphere passing through a den-sity transition will entrain lighter fluid from above ina thin boundary layer shell around the particle, andthe additional buoyancy can decrease its settlingvelocityin some cases even cause it to temporarilyreverse direction (Srdi -Mitrovi et al. 1999, Abaid etal. 2004, Camassa et al. 2009, 2010, Yick et al. 2009).This delayed settling due to entrainment of lighterfluid can be seen in Fig. 1A, a sequence of previouslypublished images from experiments with solid spheresalong with matching theory (Camassa et al. 2010).However, marine aggregates are extremely porous(usually containing >99.5% water by volume; Ploug& Passow 2007), providing for another potentialmechanism for delayed settling. The density of aporous particle will change depending on the fluidsurrounding it. Thus, a porous particle may en -counter a sharp density transition such that the parti-cle is initially too light to sink through the densitytransition; however, after denser fluid diffuses intothe particle, its density may increase sufficiently suchthat it can continue to sink. This diffusion-limitedretention of porous particles, depicted schematicallyin Fig. 1B, has been investigated theoretically (Mac-Intyre et al. 1995, Alldredge & Crocker 1995, Kindleret al. 2010) and experimentally using agarose spheres(Kindler et al. 2010).

This recent work with both solid and porous sphereshas provided important insight into the potential forsharp density gradients to lead to accumulations ofaggregates through delayed settling. However, thereis a lack of experimental work investigating whether

natural aggregates will exhibit the same behavior asartificial particles, since natural aggregates differfrom artificial spheres in important aspects includingcomposition, shape, density, and size. In addition, itremains unclear which of the proposed mechanismsfor delayed settling (i.e. entrainment of lighter fluidor diffusion-limited retention) dominates aggregatesettling behavior for different types of aggregatesin different environments.

Here, we present experimental results demonstrat-ing that natural aggregates formed in the laboratorycan significantly decrease their settling velocity asthey pass through sharp density transitions. Ourresults demonstrate 2 distinct settling behaviors; onesuggestive of diffusion-limited retention, and onesuggestive of entrainment of lighter fluid. Althoughour experimental conditions are directly comparableonly to a subset of field conditions, the generaldynamics underlying the observed behavior is appli-cable to a wide range of aggregates settling throughsharp density gradients. By comparing our resultswith those of previously developed models, we deter-mined the mechanism that plays a dominant roledepends on aggregate size, settling velocity, and thevertical density gradient. Lastly, we discuss theimplications for the delayed settling of aggregatesby each mechanism to form a layer that can act as ahot spot for foraging zooplankton and carbon re -mineral i zation.

186

Fig. 1. Two mechanisms for decreased settling velocity ofaggregates as they pass through sharp density transitions.(A) Lighter fluid is entrained around the sinking particlecausing it to slow as shown in a time series of 3 experimentalimages (along with comparisons to theory). Figure from Ca-massa et al. (2010). (B) The aggregate pauses at the densitytransition until diffusion of denser fluid into the porous ag-gregate increases its density allowing it to continue sinking.

Schematic adapted from figure in Kindler et al. (2010)

Prairie et al.: Delayed settling of marine snow

MATERIALS AND METHODS

Formation of aggregates

Aggregates were formed in a rotating acrylic cylin-drical tank (total volume ~4 l) filled with either riverwater or seawater and placed on a roller table. Thisapproach is widely used for aggregate formation inthe lab (Shanks & Edmondson 1989, Jackson 1994,Ziervogel & Forster 2005, Ploug et al. 2008, Ziervogel& Arnosti 2008, Ziervogel et al. 2010). We chose touse aggregates formed from both fresh water andseawater to observe the settling behavior of a widerrange of aggregate types with different densities andcompositions. Aggregates are derived from naturalorganic matter present in the source water, and river-ine and marine dissolved and particulate organicmatter differ in their sources and composition (Hedges& Oades 1997, Cauwet 2002), although the specificcomposition of the organic matter in the source waterwas not measured in this study.

River water was collected in October 2011 at themouth of the Tar River, North Caroline (NC) USA.Seawater was collected in December 2011 from thepier at the University of North Carolina Institute ofMarine Sciences, located in Morehead City, NC. Par-ticles were visible in both water samples prior toroller table experiments, which were started in thelaboratory at room temperature shortly after samplecollection. Rotation speed of the tank was set to3.5 rpm, which proved suitable for forming aggre-gates useable in settling experiments. Macroscopicaggregates formed within the first day of both exper-iments. Aggregates were incubated in roller tanks for7 d (seawater) and 17 d (river water); the time of incu-bation reflected the length of time required to formaggregates that were stable enough to use in settlingexperiments. At the end of the incubation, rollertanks were placed upright, and single aggregatesthat settled to the bottom of the tank were individu-ally removed via volumetric pi pette from the tankwater (hereafter referred to as aggregate ambientwater), and analyzed separately for their settlingbehavior. Four settling experiments were conductedin totalone with freshwater aggre gates (Expt 1)and 3 with seawater aggregates (Expts 2, 3, and 4).

Measuring aggregate size

Aggregates used for calculation of aggregate den-sities as well as aggregates used in the 2-layer settling experiments were first measured by micro -

scopy. Individual aggregates were placed on top of amillimeter square grid in a Petri dish with water ofdensity approximately equal to that of their ambientwater. Aggregates were then photographed with adigital microscope (Model 26700-300, Aven), pro -ducing images of the 2-dimensional projection ofthe aggregate (Fig. 2A). Images were processedusing MATLAB to determine the cross-sectional area(Fig. 2B), which was used to calculate the equivalentspherical diameter for each aggregate, i.e. the crosssectional area was assumed to represent that of asphere with an equivalent cross-sectional area (seeTable 2). Since the aggregates are irregularly shapedand are most likely to lie such that their largest cross-section is the area imaged, the estimates of equiva-lent spherical diameter for each aggregate are likelyoverestimated. In addition, the irregular shape of theaggregates has consequences for their surface areato volume ratio, which would be much larger than anequivalently sized sphere. The implications of thesefactors on the delayed settling behavior of the ag -gregates are discussed below when comparing theexperimental delayed settling behavior to models.

Calculating aggregate densities

Aggregate densities were calculated by measuringthe sinking velocity of 5 to 12 representative aggre-gates per experiment in water of homogenous den-sity (given in Table 1) approximately equal to that oftheir ambient water (~0.998 g cm3 for fresh wateraggregates and ~1.025 g cm3 for marine ag -gregates). All water densities were measured using aDMA 35 Portable Density Meter (Anton Paar). Aftermeasuring their sizes as described above, in dividualaggregates were gently transferred by pi pette to arectangular tank with a base 18 18 cm and a heightof 32 cm. The path of the aggregate as it settled inthe tank was recorded using a Pike F-100B camera

187

Fig. 2. (A) Microscope image of the aggregate shown on a1 mm2 grid. (B) Same image as (A) with perimeter outlinedto illustrate how cross-sectional area is approximated to

derive equivalent spherical diameter of aggregates


(Allied Vision Technologies) recording at a rate be -tween 25 and 33 frames s1 (but that remained con-stant within an experiment). Sin king velocity wascalculated from the vertical displacement (capturedat the recording rate of the camera) and then an aver-age sinking velocity (U) was calculated over at least3 continuous seconds. This was used to estimateaggregate density (a) using the following equation(Batchelor 1967, Ploug et al. 2008):

(1)

where g is the acceleration due to gravity, f is thedensity of the fluid, CD is the drag coefficient, and dis the equivalent spherical diameter as measuredfrom the microscope images. The drag coefficientwas calculated using the following empirical draglaw (White 1974):

(2)

for Re > 0.5 where Re is the Reynolds number calcu-lated as:

(3)

where is the kinematic viscosity of water (1.19 102 cm2 s1 at 15C; Ploug et al. 2008). Since theabove equations assume spherical particles, aggre-gates used to estimate aggregate density were chosen to be as spherical as possible. Individualdensities for all aggregates measured were thenaveraged to obtain an average a for each experi-ment (Table 1). Since aggregates are porous, theirdensities depend on the fluid in which they aremeasured; the aggregate densities reported hereare estimated in water approximately the density ofthe ambient water in which they were originallyformed (see Table 1).

Two-layer aggregate settling experiments

After an average aggregate density was calculated,different aggregates from the same roller table batchwere observed as they settled through a 2-layerwater column with a sharp density transition in themiddle. The tank used in these experiments had asquare base (30 30 cm) and had a height of 61 cm.The tank was filled to approximately 20 cm with saltwater. This salt water, hereafter defined as the bottom layer fluid, varied in density from experimentto experiment but was always more dense thanthe water in which the aggregates were formed(Table 1). After the bottom layer fluid was still, waterwith density approximately equal to that of theaggregates ambient waterdefined as the top layerfluidwas carefully poured on top of the bottomlayer fluid through a diffuser initially soaked with toplayer fluid. The diffuser floats at the surface of thewater column; it is ~2 cm thick and is constructed offoam and sponge to slow down the flow of top layerfluid as it is introduced into the tank in order to createa sharp density transition between the top layer andbottom layer fluid. The thickness of the density tran-sition was not measured for these experiments but isestimated from previous application of this stratifica-tion method (Abaid et al. 2004, Camassa et al. 2009)to be between 1 and 2 cm thick. Variations in thisthickness will affect the sharpness of the density gra-dient, which is likely to affect aggregate settlingbehavior, although future work investigating thisdependence is needed to determine the exact effect.The vertical density difference is calculated as thedifference between the density of the bottom layerfluid (BL) and the density of the top layer fluid (TL)and ranged between 0.0103 and 0.0416 g cm3. Foreach experiment, the initial excess density of theaggregates in the bottom layer was calculated as a BL and is given in Table 1.

Ug d

C( )= 43 a f

f D

CD Re Re.

.= +

++24 6

10 4

0 5

Re = dU

188

Expt Average aggregate Density of water Density Density No. of Aggregate no. density (g cm3) for aggregate density of top of bottom aggregates excess

measurement layer layer density (g cm3) (g cm3) (g cm3) (g cm3)

1 1.0098 0.0064 (n = 12) 0.9985 0.9985 1.0401 5 0.03032 1.0598 0.0102 (n = 5) 1.0261 1.0248 1.0404 7 0.01943 1.0606 0.0145 (n = 10) 1.0256 1.0249 1.0352 12 0.02544 1.0599 0.0112 (n = 10) 1.0253 1.0250 1.0460 15 0.0139

Table 1. Aggregate sinking experiments. Aggregate density was calculated using Eq. (1). Average aggregate density is shownover n aggregates along with the standard deviation. The subsequent columns provide the density of the homogenous waterused to cal culate aggregate densities, the densities of the top and bottom layer of the 2-layer settling experiments, andthe number of aggregates observed in the 2-layer settling experiments. Aggregate excess density in bottom layer gives the

difference between the average aggregate density and the density of the bottom layer


Aggregates were transferred gently by pipette intothe 2-layer water column one at a time to observetheir settling behavior. Aggregate settling behaviorwas recorded using the Pike camera, recording at aframe rate that was constant within runs, but variedamong runs between 12 and 25 frames s1. Record-ings were conducted with the room dark and thetank lit from the sides, using light-emitting diode(LED) strips attached to 2 panels that spanned theheight and width of the tank. Although the LED pan-els did introduce some heat laterally into the tank,measurements indicate that the rise in temperaturewas slight (


cases, settling velocity in the bottom layer was likelyunderestimated. Measured settling velocities for eachaggregate are reported in Table 2; it is important tonote that these measured values are not expected tomatch theoretical sinking velocity values (as calcu-lated from Eq. 1) given the wide range of aggregateshapes observed in these experiments, and that anaverage aggregate density was used for each experi-ment.

Two metrics were calculated to determine theextent to which the aggregates settling velocity de -

creased within the density transition. The normalizedminimum settling velocity (NMSV) was calculated asthe minimum value of the smoothed settling velocity(MSV) divided by the settling velocity in the bottomlayer. The delayed settling time (DST) was calculatedas the length of time that the aggregates smoothedsettling velocity was less than 90% of the settlingvelocity in the bottom layer. This definition for DSTwas chosen because it quantified the time of delayedsettling of the aggregate independent of its settlingvelocity; the threshold of 90% was chosen to obtain a

190

Expt Aggregate Aggregate equivalent Settling velocity Settling velocity NMSV DST Reno. no. spherical diameter in top layer in bottom layer (s) in top layer

(cm) (cm s1) (cm s1)

1 1 0.206 0.62 0.42 0.01 244.2 10.7 2 0.658 0.75 0.48 0.15 600.6 41.3 3 0.144 0.73 0.47 0.01 191.4 8.84 0.101 0.51 0.40 0 104.6 4.35 0.137 0.52 0.41 0 149.8 5.9

2 1 0.089 1.48 0.87 0.65 6.6 11.0 2 0.150 1.39 1.09 0.62 6.9 17.4 3 0.122 1.23 0.89 0.57 11.2 12.6 4 0.138 1.01 0.83 0.45 12.5 11.7 5 0.071 0.97 0.79 0.67 6.6 5.86 0.064 0.89 0.55 0.54 13.0 4.87 0.096 1.22 0.81 0.58 9.5 9.9

3 1 0.077 0.50 0.50 0.30 18.9 3.22 0.051 0.39 0.38 0.51 9.2 1.73 0.062 0.55 0.53 0.54 9.9 2.84 0.056 0.44 0.43 0.40 13.4 2.15 0.112 0.88 0.82 0.63 8.5 8.36 0.108 0.70 0.66 0.48 15.0 6.47 0.098 1.09 0.98 0.79 3.5 9.08 0.138 1.06 0.95 0.74 1.0 12.3 9 0.105 0.85 0.77 0.67 7.9 7.510 0.166 1.56 1.34 0.89 1.1 21.6 11 0.116 0.86 0.80 0.64 9.2 8.412 0.110 0.92 0.85 0.71 7.5 8.5

4 1 0.100 0.57 0.54 0.08 45.4 4.82 0.136 1.28 1.07 0.52 9.3 14.6 3 0.083 0.73 0.66 0.35 23.2 5.14 0.088 0.94 0.74 0.58 8.9 7.05 0.131 1.06 0.86 0.32 13.6 11.6 6 0.111 1.45 0.82 0.75 8.6 13.5 7 0.098 0.76 0.65 0.23 25.1 6.38 0.110 1.22 1.06 0.58 7.2 11.3 9 0.083 0.65 0.60 0.31 18.4 4.510 0.084 0.92 0.59 0.23 24.0 6.511 0.131 1.42 0.96 0.59 9.3 15.6 12 0.105 1.09 0.96 0.46 10.2 9.613 0.104 0.92 0.83 0.47 14.6 8.114 0.102 1.36 1.16 0.66 6.7 11.6 15 0.096 1.07 0.90 0.55 9.4 8.6

Table 2. For each aggregate from the four 2-layer sinking experiments, experimental results are reported, including equiva-lent spherical diameter, settling velocity in the top and bottom layer, normalized minimum settling velocity (NMSV), delayed

settling time (DST), and Reynolds number (Re) in the top layer


positive time scale even for aggregates thatdecreased their velocity only slightly while limitingthe effect of noise. These metrics were calculated foreach aggregate individually and compared to aggre-gate size and vertical density difference.

RESULTS

Experimental results

Data of settling behavior for all 39 aggregatesobserved in the 2-layer aggregate settling experi-ments are given in Table 2. It is important to note thataggregates that form in roller tanks are generallydenser than those formed in nature, and thus typi-cally exhibit much higher settling velocities thanaggregates observed in situ (Shanks 2002). Thus,aggregates used in this study may be more represen-tative of dense nearshore aggregates (which containmore mineral material) rather than offshore aggre-gates formed from phytoplankton blooms.

All aggregates demonstrated a settling velocityminimum when passing through the density transi-tion (Table 2). However, the extent to which theaggregates decreased their velocity varied betweenaggregates, as well as among experiments.

The most striking differences in settling behavioroccurred between aggregates in Expt 1 and aggre-gates in the other 3 experiments (Fig. 4). The 5aggregates in Expt 1 all demonstrated a VMSV lessthan or equal to 0 (within the resolution of ourimages) when passing through the density transition.Although some of the aggregates had a negativeNMSV, all except for Aggregate 2 were within theerror (due to video recording noise) of having a 0minimum settling velocity. By contrast, the NMSV of

the aggregates in Expts 2, 3, and 4 ranged from 0.08to 0.89, with a mean value of 0.53 (Fig. 4A). This dif-ference be tween Expt 1 and the other 3 experimentsis also evident in the DST, which ranged be tween104.6 and 600.6 s for Expt 1 (mean of 258.1 s), andbetween 1.0 and 45.4 s for Expts 2, 3, and 4 (mean of11.9 s) (Fig. 4B). The large spread in DSTs observedwithin a given experiment is expected because of thewide range of aggregate shapes and sizes used in theexperiments. Although aggregates from each of the 4experiments represent a range of sizes, it is importantto note that the largest aggregate in Expt 1 was morethan 3 times larger (by equivalent spherical diame-ter) than any of the other aggregates, and likewise itsdelayed settling time was much longer and not rep-resentative of the rest of the aggregates tested. Thedifferent settling behaviors are exemplified in plotsof aggregate vertical location and settling velocityover time for an aggregate from Expt 1 (Fig. 5A,B)and for an aggregate from Expt 4 (Fig. 5C,D).

Differences also occurred in the manner in whichDST varied with aggregate equi valent spherical dia -meter, which ranged between 0.051 cm and 0.658 cmfor all experiments (mean of 0.121 cm) (Table 2). DSTfor aggregates in Expt 1 showed a strong positive lin-ear relationship with equivalent spherical diameter(r2 = 0.988, p < 0.001; Fig 6A). However, the other3 experimentswhen grouped togethershoweda negative but not significant re lationship betweenDST and aggregate equivalent spherical dia meter(Fig. 6B). When Expts 2, 3, and 4 were consideredseparately, only Expt 3 showed a significant re -lationship at the p = 0.05 level (r2 = 0.398, p = 0.028).

Although many experimental parameters differedamong the experiments, one particularly notable dif-ference between Expt 1 and the other 3 experimentswas the initial excess density of the aggregates in

191

Fig. 4. Mean (A) normalized minimum settling velocity and (B) delayed settling time for aggregates in each experiment. Error bars = standard error


the bottom layer (Table 1). The aggregates in Expt 1were less dense than the bottom layer fluid initially(a negative excess density) while the aggregates inExpts 2, 3, and 4 were more dense than the bottomlayer fluid (a positive excess density).

Comparison of experimental results to models

The observed relationships between aggregate sizeand DST (Fig. 6A & B) can be used to determine thedominant mechanism of delayed settling for each experiment by comparing the relationships with theoretical predictions of particle settling behaviorfrom 2 different models. The first modelhereafterreferred to as the porous particle modelillustratesthe settling behavior of a porous particle as it passesthrough a sharp density transition. A brief descriptionof the model is given here, but specific details of thismodelongoing in developmentwill be published

elsewhere (R. Camassa, S. Khatri, R. M. McLaughlin,J. C. Prairie, B. L. White, S. Yu unpubl. data). The settling behavior in this model does not include anyentrainment of fluid around the particle; thus, any decrease in particle settling velocity at the densitytransition is strictly due to diffusion-limited retention.The model assumes: (1) the Re is much smaller than 1,(2) a spherical particle whose size is not changing intime, (3) a density gradient formulated as an errorfunction, (4) the ambient density gradient does notdiffuse during the total time that it takes the sphere tosettle through the transition region, (5) the radius ofthe sphere is small compared to variations in the ambient density profile, and (6) the time scale of dif-fusion into the sphere is small relative to the particlesettling time scale. Under these assumptions, themodel can be formulated as a force balance betweenthe drag forces acting on the particle and buoyancyforces exerted by the fluid on the particle. The diffu-sion equation is solved dependent upon the location

192

Fig. 5. Settling behavior of 2 aggregates through a sharp density transition. (A) Aggregate vertical location over time for Ag-gregate 3 from Expt 1. (B) Smoothed aggregate settling velocity over time for the same aggregate as in (A). (C) Aggregate verticallocation over time for Aggregate 5 from Expt 4. (D) Smoothed aggregate settling velocity over time for same aggregate as in(C). In plots (B) and (D), the dashed line shows the minimum settling velocity (MSV, not normalized) and the dotted line repre-sents 90% of the terminal velocity in the bottom layer, which was used to find delayed settling time (DST, shown with brackets)


of the sphere to determine how much salt has dif -fused into the fluid within the sphere and thereforedetermines the weight of the sphere. Also, the buoy-ancy force is dependent upon the ambient fluid den-sity surrounding the particle. By solving the diffusionequation analytically, we have developed a nonlinearintegro-differential equation for the position of theparticle as a function of time, which is then solved using numerical methods. Parameters in the modelinclude density of the top layer (TL), density of thebottom layer (BL), diffusivity coefficient for salt (S),the particle porosity (P), and the density of the solidfraction of the particle (solid).

This model was run for particle diameters rangingfrom 0.04 to 0.7 cm and the DST was calculated foreach particle diameter in the same way as for theexperimental results. These preliminary simulationsdemonstrate a positive correlation between particlediameter and DST that was fit with a quadratic poly-nomial (r2 > 0.999; Fig. 6C). The model was formu-lated at low Re (the Stokes flow regime) since thereis an analytical formula for the drag force in thiscase (Batchelor 1967). However, simulations using anempirical drag law for larger Re numbers, the Whitedrag law (White 1974), showed no significant differ-ences in the observed trends presented here while

193

Fig. 6. (A) Aggregate equivalent spherical diameter vs. delayed settling time (DST) for all aggregates from Expt 1. Linear re-gression shown as dotted line (y = 851.1x + 46.2; p < 0.001, r2 = 0.988). (B) Aggregate equivalent spherical diameter vs. DST forall aggregates from Expts 2 (s), 3 (), and 4 (n). Linear regression was not significant when all data was grouped together at p= 0.05. For each experiment calculated separately, only Expt 3 had a significant linear regression (y = 99.2x + 18.7; p = 0.028,r2 = 0.398). (C) DST for different particle diameters from the porous particle model. Model was run with parameters: density ofthe top layer (TL) = 0.998 g cm3, density of the bottom layer (BL) = 1.04 g cm3, particle porosity (P) = 0.98, density of the solidfraction of the particle (solid) = 1.3 g cm3, and diffusivity coefficient for salt (S) = 1.5 105 cm2 s1. Dashed line shows quad-ratic fit (y = 10747x2 2.4; p < 0.001, r2 > 0.999). (D) DST for different particle diameters from the solid particle entrainmentmodel. Model was run with parameters: TL = 1.36 g cm3, BL = 1.38 g cm3, and particle density (particle) = 1.48 g cm3. Dashedline shows a fit of the form y = 1/x (y = 21.2(1/x) 19.9; p < 0.001, r2 = 0.995). The different curves shown in (C) and (D) demon-strate that delayed settling due to diffusion limited retention results in a positive (quadratic) relationship between DST and

aggregate size while entrainment of lighter fluid results in a negative relationship between DST and aggregate size


having a minor decrease in the magnitude of the DST(


DST in Fig. 6B are likely due to the large range inaggregate shapes observed in these experiments.However, the non-zero minimum settling velocitiesfor all aggregates in Expts 2, 3, and 4 and the rela-tively short DSTs (as compared to Expt 1) further sug-gest that entrainment of lighter fluid is more impor-tant than diffusion in these cases. As for Expt 1, thediscrepancies in DST values between the solid parti-cle entrainment model and Expts 2, 3, and 4 can beexplained by the model assumption of a sphericalparticle, lack of diffusion in the model, and the modelparameterization.

DISCUSSION

Predicting the dominant mechanism for delayedsettling at density transitions

The sinking behavior of aggregates through sharpdensity transitions observed in both the experimentaldata and model simulations provides insight into thecircumstances under which the decreased settlingvelocity at the transition is primarily driven by dif -fusion-limited retention (Fig. 1B) or entrainment oflighter fluid (Fig. 1A). When the density of the aggre-gate in the top layer fluid is less than the density ofthe bottom layer fluid, the aggregate will have to stopat the density transition until sufficient diffusion intothe particle allows it to continue to sink. However,when the density of the aggregate in the top layerfluid is greater than the density of the bottom layer

fluid, the aggregate can continue settling throughthe density transition but experiences a decrease insettling velocity because of fluid entrained from thetop layer. Thus, the curve

BL = a (4)

divides 2 broad regimes: above this curve, diffusion-limited retention controls the particles delayed set-ting at the density transition, although there may beimportant interactions with the effects due to entrain-ment; below this curve, diffusion is no longer im -portant and the dominant mechanism for delayedsettling is entrainment of lighter fluid.

The 2 regimes described by Eq. (4) are shown inFig. 7A. For the 4 experiments conducted, the set-tling behavior regime can be identified easily fromthe initial excess density of the aggregate in the bot-tom layer (Table 1). When this excess density is posi-tive, the density of the aggregate is greater than thedensity of the bottom layer fluid, thus placing it in theentrainment regime. Likewise, when this excess den-sity is negative, the aggregates fall in the diffusion-limited retention regime. It is important to note thatthe regimes represented in Fig. 7 simply depict rela-tive effects of one mechanism compared to the other,rather than an absolute trend. For example, in theentrainment regime, moving further away from thedividing line indicates that the effect of entrainmentbecomes stronger relative to the effect of diffusion-limited retention; it does not indicate that the effectof entrainment becomes stronger in an absolutesense. In fact, in a study of solid spheres settling

195

Fig. 7. Two aggregate sinking behavior regimes for an aggregate passing through a sharp density transition. (A) Aggregatedensity vs. bottom layer density. The curve (Eq. 4) separates the regime dominated by diffusion-limited retention (above thecurve) and the regime dominated by entrainment of lighter fluid (below the curve). (d) Expt 1, (s) Expt 2, () Expt 3, and (n)Expt 4. (B) Aggregate sinking velocity vs. normalized layer density. Several curves of Eq. (6) are shown for different values ofaggregate radius (R): R = 0.025 cm (dashed line), R = 0.05 cm (solid line), and R = 0.1 cm (dotted line). For each curve, the region above the curve represents the regime dominated by diffusion-limited retention and the region below the curve

represents the regime dominated by entrainment of lighter fluid


through density transitions, the effect of entrainmentwas observed to become stronger as the density ofthe sphere became closer to the density of the bottomlayer fluid (R. Camassa, R. M. McLaughlin, A. Vaidyaunpubl. data).

Moreover, although one mechanism can be domi-nant in certain cases, both mechanisms are always atwork to some extent, and interactions between themcan alter the delayed settling behavior. Particularlyfor the diffusion-limited retention regime, althoughdiffusion is necessary for the particle to increase itsdensity and thus continue to sink, the effects of en-trainment have been shown to be important in manycases, particularly in accurately predicting delayedsettling times (R. Camassa, S. Khatri, R. M. McLaugh-lin, J. C. Prairie, B. L. White, S. Yu unpubl. data).

The 2 regimes can also be described as a functionof settling velocity, which is more easily measured inthe field than aggregate density. By combining Eq.(4) with Eq. (1) which provides an estimate ofaggregate density (in top layer fluid) as a function ofsettling velocity and aggregate size the followingcurve is obtained:

(5)

Eq. (5) can be rewritten so that a normalized meas-ure of the vertical density difference (BL TL)/TL isexpressed in terms of aggregate settling velocity andradius:

(6)

For given aggregate radii, a curve of this normal-ized vertical density difference vs. settling velocitycan be calculated; in the same way as for Eq. (4),above the curve represents delayed settling prima-rily driven by diffusion-limited retention while belowthe curve represents delayed settling primarily drivenby entrainment of lighter fluid. Fig. 7B illustrates several curves of the normalized vertical density dif-ference vs. aggregate settling velocity for differentaggregate radii. As before, both mechanisms can actto delay settling and interact with each other.

Implications of delayed settling for aggregate thinlayer formation

The results of this study have shown that aggre-gates derived from natural waters can exhibit sig -nificantly decreased settling velocities when passingthrough sharp density transitions because of 2 differ-

ent mechanisms: diffusion-limited retention and en -trainment of lighter fluid. In some cases, aggregatesdemonstrated only a modest decrease in settlingvelocity (with a minimum settling velocity as large as~89% of the settling velocity in bottom layer), whilein other cases, aggregates came to a complete stop.The DST reached up to ~600 s in the most dramaticcase observed; however, in the majority of cases thedecrease in settling velocity was on the order of~10 s, representing a relatively short-term eventgiven the total time required for an ag gregate to sinkout of the surface ocean. However, the importance ofthis phenomenon is not the delay experienced by asingle aggregate in its journey to depth, but rather,the resulting accumulation of aggregates at thedepth of the density transition.

Even modest decreases in aggregate settling speedat sharp density transitions can result in significantlocal increases in aggregate abundance. Using arough approximationonly considering the effectsof settling speedthe local increase in aggregateabundance will equal the inverse of the NMSV. Forexample, a NMSV of 0.5 would result in a doublingin local aggregate concentration, even if the delayedsettling time of an individual aggregate is only ~10 s.This model is overly simplistic, since it suggests thatin cases where the NMSV equals 0, there would bean infinite buildup of aggregates. Including theeffects of vertical diffusivity, as done in previousmodels of plankton thin layer formation (Stacey et al.2007, Birch et al. 2008, Prairie et al. 2011), will pro-vide a balance between the accumulation caused bythe delayed settling and the dissipating effects ofmixing, thus presenting a more realistic idea ofhow these mechanisms can affect vertical aggregatedistributions.

The intensity of the resulting aggregate layer willdepend strongly on whether the delay in aggregatesettling is driven primarily by diffusion-limited reten-tion or entrainment of lighter fluid. As demonstratedby this study, the decrease in settling velocity is morepronounced as well as longer-lasting for aggregatesundergoing diffusion-limited retention. Thus, layersformed by this mechanism will demonstrate higherincreases in aggregate concentrations compared tolayers formed by entrainment of lighter fluid. How-ever, even in the case where entrainment dominates,aggregate layers can form.

The persistence of aggregate layers would largelybe determined by the persistence of the density tran-sition that originally caused the decrease in settlingvelocity. If the sharp density gradient is transient,then the aggregate layer would also be expected to

BL TL D TL= +3

8

2U CgR

BL TL

TL

D = 38

2U CgR

196


be ephemeral. However, in many cases sharp densitygradients can persist for days or significantly longerin the ocean (Dekshenieks et al. 2001), thus allowinglong-lived aggregate layers to form.

Aggregate layers have important effects on foodweb dynamics, acting as hot spots for zooplanktonforaging. Studies have demonstrated that marinesnow can be an important food source for a diversevariety of zooplankton, including protists, copepods,other mesozooplankton, and larval invertebrates(Alldredge 1972, Lampitt et al. 1993, Steinberg 1995,Shanks & Walters 1996, 1997, Green & Dagg 1997,Artolozaga et al. 2002). Aggregate layers formedthrough delayed settling may further enhance localzooplankton abundance and grazing, explaining thefinding that up to 70% of aggregate carbon can bedegraded by invertebrate grazers before an aggre-gate sinks out of the surface ocean (Kirboe 2000).Laboratory experiments have shown that severaltaxa of zooplankton are able to seek out and remainin regions of high food concentration (Jakobsen &Johnsen 1987, Tiselius 1992, Menden-Deuer & Grn-baum 2006). For marine snow, zooplankton may beable to use chemical sensing to detect sinking parti-cles (Jackson & Kirboe 2004, Kirboe 2011). Fieldobservations have demonstrated that the distribu-tions and grazing rates of zooplankton and perhapshigher trophic levels such as small fish are oftenassociated with planktonic layers (McManus et al.2003, Benoit-Bird et al. 2009, 2010, Menden-Deuer &Fredrickson 2010). However, there is still a need forstudies of trophic interactions in aggregate layers inthe field, since most studies of zooplankton grazingon aggregates have focused on zooplankton asso -ciations with single particles (Green & Dagg 1997,Shanks & Walters 1997, Jackson & Kirboe 2004).Continuing advances in optical and acoustic techno -logy may allow for in situ measurements of the settling behavior and vertical distributions of aggre-gates concurrent with grazing rates and distributionsof zooplankton and fish, providing a better mecha-nistic understanding of the role of delayed settling inaggregate layer formation and its impact on trophicinteractions.

Aggregate layers may also be a hotspot for bacter-ial activity and organic matter remineralization.Experimental and theoretical studies have suggestedthat bacteria may preferentially find and colonizesinking aggregates (Kirboe et al. 2001, 2002, Kir-boe & Jackson 2001, Stocker et al. 2008). In addition,bacterial activity and enzymatic hydrolysis rates onaggregates can be several orders of magnitudehigher than that of aggregate-free water (Grossart et

al. 2007, Ziervogel & Arnosti 2008, Ziervogel et al.2010), demonstrating that aggregates not only repre-sent regions of enhanced bacterial abundance butalso active carbon remineralization. In aggregatelayers formed by delayed aggregate sinking, theeffect of aggregates as remineralization hot spotsmay be further strengthened. Given the potential foraggregate layers as hotspots for zooplankton forag-ing and bacterial activity, the presence of these lay-ers may act to reduce local carbon flux, with implica-tions for larger-scale carbon cycling.

Applicability to aquatic ecosystems

The experimental results presented here representa small fraction of natural conditions found in aquaticecosystems; however, the basic observation of de -layed aggregate settling at sharp density transitionsmay be widely applicable to many regions of theworlds oceans since the general dynamics under -lying the delayed settling behavior will remain thesame. The vertical density differences tested in thecurrent study (with density differences of ~1.025 to~1.046 g cm3 for the marine aggregates tested) aremuch greater and over shorter vertical distances thanwould be observed in most coastal regions or theopen ocean (density for the entire water column willtypically range from ~1.02 to ~1.029 g cm3). How-ever, similar aggregate settling behavior may beobserved at much weaker density gradients since theeffect of density gradients on aggregate settlingbehavior is controlled not only by the sharpness ofthe gradient but also by the density of the aggregate.Since aggregates used in this study were formed inroller tanks, they are likely much denser than aggre-gates found in the open ocean and other regions out-side of very nearshore environments (Alldredge &Gotschalk 1988, Shanks 2002). For example, for nat-ural marine aggregates from the San Pedro andSanta Barbara Basins, Alldredge & Gotschalk (1988)measured a median aggregate density of 1.02502 gcm3 (equal to 0.00014 g cm3 higher than the ambi-ent seawater). Thus, natural, less-dense aggregatesin these systems are likely to exhibit significantlydecreased settling velocities even with weaker den-sity gradients than those tested in this study.

In some stratified aquatic environments, the den-sity differences tested in this study are representativeof naturally occurring gradients. In estuaries, forexample, the transition from fresh water to seawatercan occur over a relatively small vertical distancewhich can retain particles for extended periods of

197


time, resulting in estuarine turbidity maxima, ashas been observed in the Columbia River, USA, estu-ary (Crump et al. 1999). Under these conditions, car -bon remineralization can be enhanced by particle-attached bacteria (Crump et al. 1999). Furthermore,deep hypersaline basins, such as those in the Gulf ofMexico, the Red Sea, and the Mediterranean Sea,can exhibit up to an order of magnitude verticalchange in salinity over very short depth scales (Poly-menakou et al. 2007, Tribovillard et al. 2009). Inthese extreme environments, trapped particles at thehalocline can affect the distribution, composition, andabundance of both organic matter and the microbialcommunity (LaRock et al. 1979, Polymenakou et al.2007, Tribovillard et al. 2009). In estuaries as well ashypersaline basins, the intense vertical density dif-ferences would make delayed settling by aggregatesmore likely to be driven by diffusion-limited reten-tion, resulting in minimum settling velocities closeto zero and resulting in very long particle residencetimes. These settling behavior attributes explainthe intense particle accumulations often observedat these sites (Crump et al. 1999, Tribovillard et al.2009).

Finally, it is important to consider whether the den-sity gradients in question are driven by changes insalinity or temperature. In some parts of the ocean,vertical salinity variations can cause sharp pycno-clines (most notably in estuaries and hypersalinebasins as discussed previously), thus mimicking theexperiments presented here. However, most often inthe upper ocean, temperature gradients are responsi-ble for vertical changes in density. Since the molecu-lar diffusion of heat occurs at a rate nearly 2 ordersof magnitude faster than that of salt (Gargett etal. 2003), aggregates settling through temperature-driven density gradients will likely exhibit decreasesin settling velocity that are much more short-livedthan those observed in this study.

CONCLUSIONS

These experimental results demonstrate that de -layed settling of natural aggregates at density transi-tions can occur by 2 different mechanismsentrain-ment of lighter fluid and diffusion-limited retention.The time scale and extent to which an aggregatedecreases its velocity at the density transition de -pends on the dominant mechanism at work. A first-order prediction of the conditions in which entrain-ment vs. diffusion-limited retention will dominateshows that the governing mechanism depends on the

aggregate size, settling speed, and strength of thedensity gradient. However, since in many inter -mediate cases both mechanisms are important, futurework exploring the interactions between these 2mechanisms is needed. In addition, although wedemonstrate decreased settling velocity across arange of aggregate sizes and conditions, furtherexperimental work is needed to understand how settling behavior of natural aggregates varies withchanging aggregate and environmental parameters,including porosity, aggregate density, aggregate size,and density gradient. Lastly, investigation on whetherthe settling behavior of aggregates (especially inhighly concentrated regions) is affected by fluid inter -actions with other aggregates will provide a more com -plete understanding of aggregate settling dynamics.

The mechanisms for delayed settling of aggregatesacross density transitions illustrated here haveimportant implications for both trophic dynamics andcarbon cycling. Even modest decreases in aggregatesettling velocities can

Documents

Marine Ecology Progress Series 487:176