factors controlling the seasonal distribution of pelagic ... · thus, sargassum may not act as an...

MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 599: 1–18, 2018https://doi.org/10.3354/meps12646

Published July 12

INTRODUCTION

Observations of pelagic Sargassum date back cen-turies (Dickson 1894), sparking debate over its ori-gins and life history for much of that time (Deacon1942). The species of the brown macroalgae (Phaeo-phyceae) Sargassum, S. fluitans and S. natans, formfloating aggregations or ‘rafts’ over much of the trop-ical and subtropical North Atlantic. Both species areholopelagic, having no attached benthic stage intheir life cycle, and reproduce solely vegetatively byfragmentation (Butler et al. 1983, Stoner 1983).

Changes in Sargassum abundance from the earlyto the late 20th century have been suggested but

© M. Brooks, V. Coles, R. Hood and Fisheries and Oceans Can-ada 2018. Open Access under Creative Commons by AttributionLicence. Use, distribution and reproduction are un restricted.Authors and original publication must be credited.

Publisher: Inter-Research · www.int-res.com

*Corresponding author: [email protected]

FEATURE ARTICLE

Factors controlling the seasonal distribution ofpelagic Sargassum

Maureen T. Brooks1,*, Victoria J. Coles1, Raleigh R. Hood1, Jim F. R. Gower2

1University of Maryland Center for Environmental Science, Horn Point Laboratory, PO Box 775, Cambridge, MD 21613, USA2Fisheries and Ocean Canada, Institute of Ocean Sciences, Sidney, BC V8L 4B2, Canada

OPENPEN ACCESSCCESS

Pelagic Sargassum contributes structural habitat to the sur-face waters of the tropical and subtropical Atlantic Ocean.

Photo: Victoria J. Coles

ABSTRACT: Pelagic Sargassum (S. fluitans and S.natans) is endemic to the tropical and subtropicalNorth Atlantic, where it provides habitat for a diverseand economically important ecosystem. Here, weinvestigate what controls the Sargassum seasonal dis-tribution using a coupled modelling approach thatintegrates output from a data-assimilating 1/12°HYCOM simulation, a 1/4° coupled HYCOM−biogeo-chemical model, and individual-based LagrangianSargassum growth models. Passively advected, buoy-ant particles with no Sargassum physiology aggregatein the central North Atlantic Subtropical Gyre atannual time scales and do not show distributions con-sistent with satellite observations of Sargassum. How-ever, at shorter time scales, advection alone canexplain up to 60% of the following month observeddistribution during some periods of the year. Connec-tivity between the tropical Atlantic and Sargasso Seais largely one-way, with the Sargasso Sea acting as a‘dead end’ for Sargassum. Adding growth, mortalityand a simple formulation of reproduction throughfragmentation to the passive advection of Sargassumparticles generates distributions that match observa-tions with 65 to 75% accuracy across all seasons.Incorporating both ocean circulation and Sargassumphysiology appears to be key in successfully reproduc-ing the seasonal distribution of biomass. We propose aconceptual model of the Sargassum seasonal cyclethat incorporates new information about a populationin the tropical Atlantic. Additionally, we suggest thatthe Gulf of Mexico and Western Tropical Atlantic areregions whose Sargassum populations may dispropor-tionately influence the basin-wide biomass.

KEY WORDS: Pelagic Sargassum · Macroalgae · Lagrangian transport

Mar Ecol Prog Ser 599: 1–18, 20182

have proven difficult to verify (Parr 1939, Stoner1983). Seasonal variability in Sargassum distributionand biomass coupled with seasonality in the sparseobservations led to this perceived decline (Butler &Stoner 1984). More recently, Sargassum distributionthroughout the annual cycle has been mapped viasatellite (Gower & King 2011, Gower et al. 2013,Wang & Hu 2016). The remote sensing derived distri-butions also suggest that changes in biomass andsouthern range expansion have occurred in recentyears (Gower et al. 2013, Wang & Hu 2016), and fieldobservations suggest this may be influencing speciesand morphotype composition in the Caribbean(Schell et al. 2015). In situ observations confirm thatSargassum natans occurs off the coast of Brazil, southof the presumed range (Széchy et al. 2012, Sissini etal. 2017). Franks et al. (2016) suggest this is related toa recirculation in the tropical gyre. These range andbiomass changes coincide with an increase in reportsof beaching events affecting fishing and tourism fromAfrica to the Caribbean (Franks et al. 2011, Smetacek& Zingone 2013), which prompted the developmentof a satellite-based Sargassum early warning systemfor the region (Webster & Linton 2013). Understand-ing what environmental factors are controlling recentvariability is difficult, however, without a betterunderstanding of what is driving variability in Sar-gassum distribution on seasonal time scales.

Remote sensing algorithms for Sargassum exist forsensors such as the Medium Resolution ImagingSpectrometer (MERIS) and Moderate Resolution Im-aging Spectroradiometer (MODIS) satellites be causefloating vegetation causes enhanced reflect ance inthe near-infrared (Gower et al. 2006, Hu et al. 2015).Satellite observations (Gower et al. 2006, Gower &King 2011) from 2002 through 2010 demonstrate aseasonal cycle in Sargassum distribution initiating inApril when integrated basin-wide biomass is at aminimum. High abundances occur in the Gulf ofMexico in the spring and early summer, in the GulfStream extension region in late summer and earlyfall, and in the southern Sargasso Sea in the winterand early spring (Fig. 1). High abundances are alsosuggested (but with little direct validation) in thetropical gyre in late spring through fall. Subsequentyears show the influence of the high biomass seen inthe western tropics (Gower et al. 2013, Wang & Hu2016).

Buoyant Sargassum is expected to passively followsurface currents and local eddy fields (Zhong et al.2012), although there are suggestions that inertialeffects acting directly on Sargassum rafts could causetheir trajectories to differ from that of the surround-

ing water and influence their distribution (Beron-Vera et al. 2015). However, pelagic Sargassum has aminimum doubling time of 9 to 13 d, depending onspecies (Hanisak & Samuel 1987), which is short rel-ative to advection time scales. Thus, Sargassum maynot act as an entirely passive tracer, since its distribu-tion can be influenced by growth and mortality inaddition to advection. Sargassum primary productionvaries across its range (Carpenter & Cox 1974), and isgenerally higher in neritic waters with higher nutri-ent availability (Lapointe 1995). Pelagic Sargassumunderpins a diverse ecosystem (Butler et al. 1983,Laffoley et al. 2011, Huffard et al. 2014) supporting awide range of fish species (Hoffmayer et al. 2005) andplaying a role in the migration of juvenile sea turtles(Carr & Meylan 1980, Witherington et al. 2012). Someevidence suggests that the macroalgae may even beresponsive to excretion of ammonium and phospho-rus from associated fish species (Lapointe et al. 2014).This raises the question of the extent to which theobserved seasonal pattern is a result of passiveadvection of Sargassum around the Atlantic, versusdue to growth and mortality of Sargassum rafts asthey encounter changing environmental conditions.

In this study we use a numerical model to investi-gate what controls the distribution and seasonalcycle of Sargassum in the North Atlantic and Gulf ofMexico. Physical circulation and Lagrangian particleadvection models are used to estimate the con -tribution of advection to the Sargassum seasonal dis-tribution. This is accomplished via analysis of particletrajectories, particle densities, and regional connec-tivity. The influence of Sargassum growth and mor-tality is examined by adding an algal physiologymodel, coupled to a biogeochemical model embed-ded within the circulation. The simulations suggestspatial and temporal variability in Sargassum growthregions strongly influences basin-wide seasonal bio-mass patterns.

METHODS

Data description and validation

Sargassum biomass is derived from satelliteimagery from the European Space Agency MERISsensor (Rast et al. 1999). A climatology of Sargassumbiomass is generated from monthly 1° griddedMERIS counts from 2002 to 2012, with an estimated1400 t (wet weight) of Sargassum per grid per MERIScount (Gower & King 2011, Gower et al. 2013). Theresulting distribution is smoothed with an adjustable-

Brooks et al.: Seasonal distribution of Sargassum

tension continuous curvature spline with a tensionfactor of 0.25 (Fig. 1). The climatology illustrates theseasonal changes in Sargassum biomass in the keyhabitat regions of the Gulf of Mexico and SargassoSea, as well as potentially high Sargassum abun-dance in the tropics.

Both MERIS and MODIS observations show whatappears to be a regime shift in Sargassum in the trop-ics, initiating with a high biomass event in 2011(Gower et al. 2013, Wang & Hu 2016). However,smaller amounts of Sargassum have been detected inthe region over the entire satellite record (Gower &

3

Fig. 1. Relative Sargassum biomass based on monthly satellite climatologies from Gower & King (2011), Gower et al. (2013).Each panel represents a different month, with the seasonal cycle initiating in the spring when integrated basin-wide biomass

is at its annual minimum


King 2011, Wang & Hu 2016). A second climatologyusing only the period 2002 to 2010 shows a similaroverall seasonal pattern and phenology, though withreduced tropical biomass. Thus, here we use the full2002 to 2012 climatology, which gives a robustdescription of the seasonality in the subtropics andalso includes the signal of the modern tropicalregime.

Only limited observations were available toGower & King (2011) for algorithm validation.Anomalies such as satellite detection of MaximumChlorophyll Index (MCI) signal in the Pacific out-side of the known range of Sargassum, and thelarge biomass in the tropics where there had beenfew direct observations of Sargassum until recently(Széchy et al. 2012) motivate further validation.The Sea Education Association has performedneuston tows from 1973 to 2010 (Siuda 2011). Inspring, low to no Sargassum is found in tows along65° W from 20 to 40° N and higher densities arefound south of 32° N and west of 65° W (Siuda2011). This pattern is consistent with the Maysatellite climatology (Fig. 1), which has the lowestdensities in the Sargasso Sea, with some moderatebiomass to the south and west. In fall, lowdensities are found in tows south of 30° N along 55to 60° W, and higher densities (0 to 3 g m−2) arefound between 30 and 40° N and extendingtowards the US East Coast from 55 to 70° W.This is broadly consistent with the September toNovember satellite climatologies, when moderateto high density regions of Sargassum, defined ashaving greater than 1% of maximum possible MCIsignal, extend over the largest area of the SargassoSea. This band of high biomass stretches from35° N to nearly 50° N, beyond the range of directobservations. Satellite detection of abundant Sar-gassum stretching north from the eastern Carib-bean to near 24° N from 60 to 65° W in Septemberis less consistent with the limited ship-based data,which suggests low Sargassum biomass in thislocation.

Within the Gulf of Mexico, higher satellite ob -served densities are seen in spring (March) andincrease to over half the area of the Gulf in July(Fig. 1) and August (not shown). Like the SargassoSea, the Gulf of Mexico also experiences seasonalperiods of low or undetectable Sargassum densities,but in January rather than May. In contrast, the Car-ibbean and central tropics (east of 60° W, south of15° N) have Sargassum year-round. In these regions,the extent of the moderate to high densities are thelargest in the late summer (September).

Physical model description

Daily output from a data-assimilating HybridCoordinate Ocean Model (HYCOM) simulation(Chas signet et al. 2009) is used to advect Lag -rangian particles. This is an eddy-resolving (1/12°~7 km resolution) global model run by the NavalResearch Laboratory (HYCOM.org GLBa0.08 expt_90.9). The model has 32 hybrid vertical layers, 11 ofwhich are fixed-depth in the upper 60 m of thewater column including a 1 m thick surface layer.The high surface vertical resolution is ideal for mod-eling Sargassum, whose buoyancy begins to dimin-ish at depths below 35 m and is fully compromisedbelow 120 m (Johnson & Richard son 1977). Surfaceforcing, including wind speed, wind stress, precipi-tation and heat flux is from the Navy OperationalGlobal Atmospheric Prediction System. The short-wave radiation forcing has an analytic diurnal cyclesuperimposed. Three-dimensional multivariate dataassimilation is performed via the Navy CoupledOcean Data Assimilation system (Cummings 2005,Cummings & Smedstad 2013).

The HYCOM particle-tracking code (Halliwell etal. 2003), based on particle advection schemes fromthe Miami Isopycnal Coordinate Ocean Model (Gar-raffo et al. 2001) is used to advect idealized Sargas-sum rafts. Daily instantaneous model velocities,interpolated to 4 h intervals using second-orderRunge−Kutta, are used as input for the Lagrangianparticle model, which treats particles as waterparcels with respect to inertial forces. Solutions areinsensitive to realistic values of diffusion (< 1% dif-ference in pairwise particle density at 60 d timescales with added horizontal turbulence velocityvariance of 4.63 × 10−6 m2 s−2). The code is modifiedfor this study to allow specification of float buoyancy(here set to 0.1 m s−1 following Johnson & Richardson1977). This is achieved by adding this rate of rise tothe vertical velocity after interpolation of the velocityfields from HYCOM to the particle location. Addi-tional modifications allow for running particles back-wards in time by reversing the time step and veloci-ties. All particles in this study are initialized in theuppermost meter to represent buoyant healthy Sar-gassum rafts. Experimental information, includingparticle release dates and tracking times are speci-fied in Table 1. To mitigate the effects of interannualvariability, we launch particles in a total of 8 differentmodel years, 2 at 1/12° resolution for physics-onlyexperiments and 6 at 1/4° resolution (Table 1). Wetrack individual particles for 2 yr after initialization.Following best practices for particle models of bio-


logical–physical interactions we quantitatively eval-uate the optimum particle number for model ex -periments (North et al. 2009). The number of modelparticles required to obtain stable statistics is deter-mined using the fraction of unexplained variance(FUV) method described by Simons et al. (2013). Par-ticle initializations range from 20 to 280 particlesdaily, randomly distributed over the domain of inter-est over 2 yr (Fig. 2). Particle density distribution(PDD) is calculated as the number of particles of agiven age in a 2° box divided by the total number ofparticles released. The FUV is calculated as 1 − r2 ofthe linear correlation coefficient between the PDD ateach intermediate total particle count and the finaltotal count of particles. At 60 d (Fig. 2a) FUV does notreach the acceptable threshold of 0.05 until there aremore than 30 000 particles. As particles approach ayear of deployment time variability decreases andFUV is


eled on Tricho desmium), 1 zooplankton, and 2 detri-tal compartments of different sizes and remineraliza-tion rates. All of the living compartments have fixedN:P ratios, while the detrital compartments areallowed variable stoichiometry. Model structure isadapted from the work of Fennel et al. (2006), withaddition of inorganic phosphorus, diazotrophy, andchanges to phytoplankton light response after Hoodet al. (2001) and Coles & Hood (2007). The Polar Sci-ence Center Hydrographic Climatology (Steele et al.2001) is used for nutrient initial conditions andboundary re laxation. Model equations are describedin the Supplement (www. int-res. com/ articles/ suppl/m599 p001 _ supp. pdf). Normalized root mean square(RMS) error between model fields and SeaWiFSmonthly chl a climatologies ranged from 0.009 to0.016 (O’Reilly et al. 2000, SeaWiFS Project 2003).These residuals are slightly lower than those for seasurface temperature.

Sargassum model description

A Sargassum model runs within each Lagrangianparticle, with each particle representing a super-individual aggregate of Sargassum. At each timestep, Sargassum is advected by the particle model,and then growth and mortality are calculated basedon ambient conditions at the particle location. Thisphysiology model represents a functional group ofpelagic Sargassum rather than either species specifically, so rates are selected from the range forboth S. fluitans and S. natans during parameter opti-mization. Sargassum super-individuals within eachLagrangian particle are modeled with a macroalgalframework that includes light (I), temperature (T),and nutrient conditions (N) based on the biogeo-chemical model.

(1)

Mortality (m) here is the aggregate effect of senes-cence and grazing, where μmaxS is Sargassum maxi-mum growth rate and m is Sargassum mortality rate.

Because the Lagrangian approach allows for track-ing the fate of discrete aggregations of Sargassum,modeled light availability accounts for the growth ofepiflora and epifauna as a function of super-individ-ual age:

(2)

with an exponential decay in light response atincreasing ages:

(3)

where aref is the reference age for Sargassum lightlimitation due to colonization by epiflora and epifauna.

Nutrient uptake is modeled as a Monod function:

(4)

where N is the limiting nutrient (NO3, NH4, or DIP).Growth experiments suggest little effect of temper-

ature on Sargassum growth above 18°C (Carpenter &Cox 1974), with reduced growth below this tempera-ture for both species (Hanisak & Samuel 1987). S.fluitans may experience low-temperature stressstarting at 24°C (Hanisak & Samuel 1987), howeverobservations of Sargassum below 18°C showed signsof distress and wilting (Winge 1923). For this reason,we implement a temperature dependence with athreshold for growth at 18°C.

(5)

The Sargassum model also includes pressure-induced sinking. Buoyancy of Sargassum floats isdiminished at relatively shallow depths, and they arefully compromised by excursions to 120 m (Johnson &Richardson 1977, Woodcock 1993). Any Sargassumparticle that descends below the maximum depth offloat integrity (zmax) becomes too dense to recoverand all biomass is lost via sinking.

We perform parameter optimization because sev-eral model parameters have few or no reported val-ues in the literature for pelagic Sargassum, andbecause we are using a Sargassum functional grouprather than explicitly modeling both species. A 3-wayoptimization tests 12 possible values for each param-eter (1728 total parameter combinations) to select theparameter values that minimize error with satellitefields. Each model solution is compared with obser-vations by binning both the satellite climatologiesand model output onto an identical 2° grid and com-puting the RMS difference. This RMS error is aver-aged over the model domain monthly. The optimalparameter values, used in all subsequent simula-tions, are shown in Table 2. Note that the nutrienthalf saturation value is likely biased to be lowbecause the biogeochemical model tends to haveslightly low surface nutrient concentrations.

Particle model validation

HYCOM model fields have been widely utilized instudies of particle dispersal (Coles et al. 2013, Put-

dd S

St

S m Sarg

arg ƒ(I) ƒ(T) ƒ(N) max arg= × × × × ×μ –

ƒ( ) (1– e ) ƒ( )– / kI ageI I= ×

ƒ( ) – /age age a= e ref

ƒ( )NN

k NSarg=

+

ƒ( ) ,,

T Temp TTemp T

= {


man & He 2013, Rypina et al. 2013, Stukel et al. 2014)and model validation studies have been publishedfor the global domain (Chassignet et al. 2003, 2009).Here we concentrate on how well the model particlescompare with the limited set of surface observationaldrifter data to ensure the surface currents relevant toSargassum dispersal are captured reasonably, and toidentify potential regions of divergence. Drifterobservations for drogued drifters within the modeldomain region are obtained from the Global DrifterProgram Drifter Data Assembly Center (Hansen &Poulain 1996). Model particles are launched to matchinitial drifter observations, with 7 particles initializedon the same day of the year and distributed within3 km of each observed launch. These particles aretracked for 1 yr (Fig. 3).

After 2 mo, the distributions of observed drifters(Fig. 3a) and model particles (Fig. 3b) differ by lessthan 1% particle density in any given 2° × 2° bin. Atthis time scale, there is slightly higher retention ofobserved drifters in the Gulf of Mexico and near theBahamas, but both model and observed drifters showthe highest densities in the northwest region of thedomain off of Georges Bank. Although the number ofobserved drifters with a full year of tracking data islow (n = 310), there is good agreement between themodel and observations (Fig. 3c). The bulk of thedrifters aggregate in the central gyre after 1 yr. Ob -served drifters show slightly greater spread, espe-cially in the region northwest of the Gulf Stream,where eddy activity in the model may be resolution-limited. However, because model particles areremoved when they go aground, particle density inthe model tends to be very low near to the coastline,which explains some of the discrepancy.

RESULTS AND DISCUSSION

The role of advection

If Sargassum is initially evenly distributed through-out the Atlantic at low densities below satellite detec-tion limits, as a result of winter mixing processes forexample, seasonal variations in advection could po -tentially generate aggregations similar to the observedseasonal patterns. Lagrangian particle experimentR-ISO (Table 1) tests this hypothesis.

Randomly seeded, surface-restricted particles arelaunched every day for 1 yr over the whole domain(Fig. 4a). Transport of particles from the tropicsthrough the Caribbean archipelago helps to maintainsome particle density in the Gulf of Mexico at 2 mo

7

Sym

bol

Par

amet

erV

alu

eU

nit

Lit

erat

ure

ran

ge

Sou

rce

μmax

SS

arg

assu

mm

axim

um

gro

wth

rat

e0.

1d

−1

0.02

9−0.

11L

apoi

nte

(19

86),

Lap

oin

te e

t al

. (20

14),

Han

isak

& S

amu

el (

1987

)m

Sar

gas

sum

mor

tali

ty r

ate

0.05

d−

125

−60

% o

f g

ross

C

arp

ente

r &

Cox

(19

74),

Lap

oin

te (

1995

)p

rod

uct

ivit

yI k

Sar

gas

sum

gro

wth

-irr

adia

nce

par

amet

er70

W m

−2

33−

78H

anis

ak &

Sam

uel

(19

87),

Lap

oin

te (

1995

)a r

efR

efer

ence

ag

e fo

r S

arg

assu

mli

gh

t li

mit

atio

n55

d*

Th

iel

& G

uto

w (

2005

)k

Sar

gN

Sar

gas

sum

nu

trie

nt

up

tak

e h

alf

satu

rati

on0.

012

mm

ol N

m−

3**

Lap

oin

te (

1986

), L

apoi

nte

(19

95),

Car

pen

ter

& C

ox (

1974

)T

ref

Min

imu

m t

emp

erat

ure

for

Sar

gas

sum

gro

wth

18°C

18−

24C

arp

ente

r &

Cox

(19

74),

Han

isak

& S

amu

el (

1987

)z m

axM

axim

um

dep

th b

efor

e S

arg

assu

mb

uoy

ancy

12

0m

35−

135

Joh

nso

n &

Ric

har

dso

n (

1977

)is

com

pro

mis

ed

*No

lite

ratu

re v

alu

es a

re r

epor

ted

for

th

is p

aram

eter

. In

itia

l es

tim

ates

for

par

amet

er o

pti

miz

atio

n w

ere

mad

e b

ased

on

max

imu

m a

ge

of o

ther

raf

tin

g m

acro

alg

aean

d t

he

gro

wth

rat

es o

f S

arg

assu

mep

iph

ytes

**N

o d

irec

t m

easu

rem

ents

are

rep

orte

d f

or t

his

par

amet

er; h

owev

er, s

ee r

efer

ence

s fo

r ex

amp

les

of g

row

than

d n

utr

ien

t co

nce

ntr

atio

ns

mea

sure

d u

nd

er v

aryi

ng

envi

ron

men

tal

con

dit

ion

s

Tab

le 2

. Sar

gas

sum

ph

ysio

log

y m

odel

par

amet

ers


time scales (Fig. 4b). After 6 mo (Fig. 4c), particledensity is reduced by half in the Gulf of Mexico. Par-ticle densities are also reduced at the northern andsouthern extents of the domain, due to loss of parti-cles from the domain to the north, and to inflow ofwaters devoid of particles from the south. Particlesare lost from the coast of Africa and the tropics due toupwelling and divergence in the Tropical Gyre andGuinea Dome (near 20° W, 15° N). After 1 yr, there isa strong tendency for particle aggregation in theNorth Atlantic Subtropical Gyre (Fig. 4d). At this

time scale, particle densities in the centralgyre have increased by an order of magnitude, while ecologically importantregions such as the Caribbean and Gulf ofMexico are almost completely devoid ofparticles. The particle distribution resem-bles that of surface plastics in the region(Law et al. 2010) due to convergentEkman flow, rather than the observedSargassum distribution.

We subsequently examine each of thefollowing assumptions: that Sargassum ishighly buoyant and surface-restricted,that Sargassum exists at low densitiesthroughout the domain, and that Sargas-sum behaves like a passive particle. Theassumption that Sargassum is surface-restricted is tested first. Sargassum aggre-gations are frequently observed driftingsubsurface. Small-scale, wind-driven fea-tures such as Langmuir cells can drivethem as deep as 100 m (Johnson &Richardson 1977), and wind speeds as lowas 4 m s−1 can result in subsurface Sar -gassum (Woodcock 1993). In experimentR-3D, the randomly initialized particlesare neutrally buoyant 3-dimensional particles, and experiment R-3DB added positive buoyancy of 0.1 m s−1. Both ofthese simulations resulted in particlesaggregating in the central gyre at longtime scales similarly to R-ISO (not shown).Particle density distributions for these 3experiments converged after 6 mo, withpairwise FUV < 0.05 after 1 yr. At longtime scales, particles initialized at the surface aggregate in the central gyre dueto Ekman transport whether they are surface-restricted, neutrally buoyant, orslightly positively buoyant.

To test whether the Sargassum seasonalcycle is dependent on initial conditions,

we initialize model particles based on the derivedsatellite climatologies. Particles are initialized daily ina randomly generated pattern within contours whereMCI > 1% of maximum. Simulations with surface-restricted (C-ISO), neutrally buoyant (C-3D) and posi-tively buoyant (C-3DB) particles initialized in accor-dance with monthly satellite climatologies of Sargassumall yield similar aggregated particle densities in thegyre at time scales of 6 mo or greater. Surface-constrained particles (Fig. 5) have slightly higher den-sities in the northern portion of the subtropical gyre

8

Fig. 3. Comparison of observed drifters and model particles. Density dis-tributions for (a) observed, and (b) modelled drifters after 60 d transittime. Black points are model and drifter launch positions. (c) Observed

(red) and modelled (blue) drifter locations after 1 yr transit time


than 3-D positively buoyant particles (not shown)since there are some losses from the surface layerswhen the 3-D particles experience winter mixing.However, all 3 experiments result in the same pat-tern of aggregation in the gyre despite their obser-vationally determined initial condition.

At shorter time scales, particles initialized in thisway more closely match observations, and weexamine monthly distributions to determine thedegree of short-term advective control on Sargas-sum. We track model particles which originatewithin a given monthly observed distribution overa 2 mo period to determine what fraction remainwithin the satellite derived contours of Sargassumobservations 2 mo later (e.g. Fig. 6b in Septem-ber). Observed GDP drifters (Hansen & Poulain1996) in the domain are also analyzed with anidentical method (Fig. 6a).

9

Fig. 4. Particle density distribution for randomly seeded surface particles. (a) Initial condition, and (b–d) distribution after (b) 60, (c) 180, and (d) 360 d particle transit time. The initial, randomly dispersed distribution ends up concentrated in the

central gyre at time scales between 60 and 360 d

Fig. 5. Effects of particle motion. Surface-constrained particles initialized according to monthly climatologies ofSargassum biomass aggregate in the central gyre after 1 yr


Match rate for particle location is calculated by countingparticles within boundaries defined by a 1 km buffer aroundthe 1% contour of MCI from the satellite climatology. Thismatch rate is normalized to the total number of drifters or par-ticles in the domain. The potential maximum match rate isexpected to be below 100% due to discrepancies in compar-ing individual years to a climatology, limitations of model res-olution, and small numbers of observed drifters.

Fig. 6. Match in September between observed drifters, model floats,and satellite observations after 60 d transit time. Only (a) observeddrifters and (b) model floats that originated in Sargassum-containingregions are shown. Contours are 1% of maximum MCI from Sargas-sum satellite climatology. Green points fall within the bounds of

observations, black points represent mismatch

Fig. 7. Percent match rate with satellite observations of Sargassumfor (a) observed drifters, and (b) modeled floats. Upper, dashed linesare drifters or floats that originate in regions where Sargassum isobserved. Lower, solid lines include all observed drifters (a), and allrandomly initialized floats (b). Both model and observations show apeak in % match in the fall, indicating high advective control on the

Sargassum distribution at that time

The match rate is bimodal for both ob -served and modelled drifters (Fig. 7), with apeak in late winter and a stronger peak inlate summer/early fall. Although the samplesize for observed drifters is small (fewer than10 drifters are present in Sargassum-denseareas in some mont hs), the overall pattern isconsistent with our Lagrangian particles.These match rates are much higher andmore strongly seasonally varying than thedistributions for all observed drifters in thedomain or for randomly distributed modelparticles (Fig. 7). Examining 3 mo andlonger time intervals continues to show adistinct peak in match rate until time scalesexceed 6 mo. Beyond that time period, theoverall match is low and variable. After a fullyear, only 15% of the particles initialized inSargassum-dense regions are still consistentwith Sargassum observations. Thus, initialSargassum distribution influences the finaldistribution for up to 6 mo.

The seasonality of the match rate indi-cates that processes other than advectionare in flu encing the Sargassum distribution. Ele vated match in the winter (months 2 to 4)is consistent with Sargassum experiencinglower temperatures and suppressed growth


(Hanisak & Samuel 1987), causing physical transportto explain more of the spatial distribution. A maxi-mum match of 59% of model particles (Fig. 7b) occursin the month of September, indicating another periodwhere advection plays a large role in determining theSargassum distribution. The subtropical water columnis strongly stratified in September, leading to low nu-trient conditions and reduced vertical mixing whichmake it more likely that surface transport rather thangrowth is controlling the Sargassum density. Periodsof minimum match rate in December and June indi-cate times when Sargassum physiology may be moreimportant in setting its distribution.

Gower & King (2011) proposed that the Sargassumseasonal cycle begins in the Gulf of Mexico inMarch-April with Sargassum growth advected out tothe Atlantic, where it eventually senesces about ayear later in the southern Sargasso Sea. To evaluatethis hypothesis, we track buoyant, 3-D particles ini-tialized in the Gulf of Mexico in experiment C-3DB,to quantify what fraction exit into the Atlantic within1 yr, and whether there is a seasonal pattern. Con-nectivity between the Gulf of Mexico and the Sar-gasso Sea (Fig. 8) is high, with 29% of particlesreaching the Atlantic within 1 yr. Particles are 5 timesas likely to make this passage within 3 mo of launchas they are at longer time scales (Fig. 8a). The sea-sonal peak in particles crossing into the Atlantic inAugust (Fig. 8b) is a result of the large Sargassumbiomass observed in the Gulf of Mexico in the springand early summer exiting the region at this 3 mo timescale. However, with less than 1/3 of Sargassum exit-ing the Gulf within 1 yr, there must be robust growthof Sargassum after leaving the Gulf or additionalseeding from other sources to match the densitiesseen in the northern Sargasso Sea in the fall.

However, the Sargassum originating in the Gulf ofMexico does not contribute to high density of Sargas-sum in the tropics. Trajectories of randomly seededsurface particles initialized in the Gulf of Mexico(Fig. 9a) and the Caribbean (Fig. 9b) remain exclu-sively north of 10° N after 1 yr. Sargassum originatingin the tropics (Fig. 9c) does reach the Caribbean andGulf of Mexico at this time scale. Even accounting fortransit times longer than 1 yr, Sargassum from theGulf of Mexico does not reach the tropics, since theconvergent North Atlantic Subtropical Gyre is essen-tially a dead end (Fig. 9d). Connectivity analysisshows that over 92% of particles launched in the Car-ibbean in the spring (that did not run aground) areadvected into the subtropical gyre after 1 yr. For thewestern and eastern Gulf of Mexico 54% and 81% ofparticles, respectively, are found in the subtropical

gyre after 1 yr, while the rest remain within the Gulf.None of these regions export particles to the tropicsat any time scale from 30 to 360 d. This suggests thata southern or eastern source of Sargassum is neces-sary to account for biomass in the tropics and Carib-bean. This is supported by the connectivity betweenthe Western Tropical Atlantic, which exports parti-cles to the tropics (77%), but also to the Gulf of Mex-ico (5%) and Caribbean (8%), with the remainderentering the subtropical gyre.

To better understand the connectivity betweenregions and to potentially highlight source regionsof Sargassum we conducted an experiment inwhich we ran particles backwards in time (C-BK).Particles are initialized using the Sargassummonthly climatology for each month, and run witha negative time step for 60 d (Fig. 10). The centraland eastern subtropical gyre is largely devoid ofparticles, confirming the pattern seen in Fig. 9dwhere particles tend to remain confined to thatregion. The highest densities of particles are foundalong the coast of Brazil and along the equator,suggesting again the potential for a southernsource for Sargassum in the tropics. These resultsare consistent with recent evidence of linksbetween the equatorial population and beachings

11

Fig. 8. Floats exiting the Gulf of Mexico. (a) Age and (b)month at exit, defined as crossing 81° W, of model floats initialized randomly in the Gulf of Mexico over 1 model yr


of Sargassum in the Caribbean (Franks et al. 2016).Of particular interest, our results suggest a pathwayfor connectivity between the putative large equato-rial population (Gower et al. 2013) and the Carib-bean and Gulf of Mexico. Even though relativelyfew particles follow this pathway, over 60% of theparticles in the Gulf and Caribbean track back tothis hypothetical tropical population.

The time scales associated with the travel ofthese particles indicate that equatorial Sargassumob served between December and February couldbe contributing to the peak in Sargassum biomassin the Gulf of Mexico between May and July.Match percent between particles run backwardsfor 2 mo and observed Sargassum distributionsshowed that 50% of particles track back to regionsthat do not match observations. This suggests thatlow densities of Sargassum, invisible to satellites,could contribute to blooms over much of the At -lantic as they are advected into regions with favor-able growth conditions.

12

Fig. 9. A subsample of trajectories over 1 yr for particles launched in (a) the eastern Gulf of Mexico, (b) the Caribbean Sea, (c) the central tropics, and (d) the subtropical gyre. Dashed boxes indicate particle launch region

Fig. 10. Particle density for 3-D particles initialized from Sar-gassum monthly satellite climatologies and run backwardsin time. This distribution integrates the seasonal cycle by

including all particles of age −60 d

Sargassum model

To determine the role Sargassum biology plays indriving its seasonal distribution, we apply the Sar-gassum super-individual model to each particle in a1/4° HYCOM simulation which is coupled with a bio-geochemistry model (simulation R-SAR). We conducta sensitivity analysis to understand which parame-ters have the largest impact on the Sargassum model.Single parameter model sensitivity is evaluated forall Sargassum model parameters (Table 2),focused on a range bounded by ± 10% ofreported values, where available. Mortal-ity, nutrient uptake half saturation, andreference age for light limitation are theparameters with the greatest impact onmodel solutions. The combined impact ofgrazing and mortality is about a 5% lossd−1, and Sargassum older than 55 d beginsto experience notable light limitation fromgrowth of epiflora and epifauna or fromloss of buoyancy.

For Sargassum growth experiments, par-ticles are initialized daily across the modeldomain for 6 yr. Nitrogen, phosphorus,light and temperature are sampled at eachdaily particle location and input into theSargassum physiology model. Particledensity maps are similar to the results from1/12° physics-only simulations indicatingthat reducing the horizontal model resolu-tion did not significantly alter the particledispersion characteristics at the spatiotem-poral scale of this analysis (Fig. 11a). Thematch between model (Fig. 11b) and ob-servations (Fig. 1d) is greatly improved byincluding Sargassum biomass from theSargassum physiology model in the parti-cle density calculation (Fig. 11c). Biomassis higher in and near regions of high ob-served Sargassum density, and dramati-cally reduced in the central gyre wherelow nutrient conditions are unfavorable forgrowth. Growth rates approached theirmaximum of 0.1 d−1 most frequently in thetropics in winter, suggesting a strongerrole for nutrients than for light and tem-perature in controlling Sargassum growth.The percent match between model andobservations (Fig. 12a) is higher when Sar-gassum physiology is included than forany other simulation. The match rate isalso consistently high throughout the year,

implying that Sargassum growth and mortality arelarge contributors to the observed distribution evenduring periods when advective control is also high.

Because observed Sargassum distribution ispatchy, we estimate the best possible model fit toobservations by comparing the observed interannualvariability to the monthly climatology. We use theRMS error between satellite derived Sargassum bio-mass for 10 individual years and the climatology tocharacterize observed variability (Fig. 12b). The

Brooks et al.: Seasonal distribution of Sargassum 13

Fig. 11. (a) Particle density in November without Sargassum growth. (b)Normalized Sargassum model biomass. (c) Difference between Sargas-sum model and observed normalized biomass. Model Sargassum bio-mass is consistent with observations, in contrast to the high abiotic

particle densities in the central gyre


mean RMS difference between the observations andclimatology is 0.44, while the RMS difference be -tween the model and climatology is 0.50. Variabilityin biomass in the tropics in the spring and summeraccounted for the highest differences between indi-vidual years of observations and the climatologies.For the model, RMS differences peaked slightly laterin the summer and early fall when Sargassum bio-mass is starting to diminish in the tropics. On anannual basis, the model biomass falls within therange of variability of the satellite observations.

Sargassum seed populations

This model still underestimates Sargassum bio-mass in the tropics over much of the year even whennutrient, temperature, and light conditions are favor-able. Biomass is overestimated elsewhere. Tropical

Sargassum has high growth rates in the model, but isadvected away too quickly to allow accumulation ofbiomass. One scenario that accounts for this missingSargassum is import from outside our model domain.Observations of pelagic Sargassum off the coast ofAfrica are relatively recent in the literature (Oyesiku& Egunyomi 2011, Gower & et al. 2013), but evidencefrom local fishing communities suggests it has beenpresent over at least several decades, although not atcurrent densities (Ackah-Baidoo 2013). A modestexport of Sargassum from a region such as the Gulf ofGuinea, combined with favorable growth conditions,could be contributing to the high biomass across thetropical Atlantic (Franks et al. 2016).

We assess the potential that export from one or sev-eral sub-regions of the Atlantic sustains the seasonaldistribution of Sargassum in a series of model seed-ing experiments. Particles are initialized daily in 17sub-regions with average area of 1.32 × 106 km2 inthe Sargasso Sea and along the coastlines on bothsides of the basin, at the same density as the whole-domain experiments for 6 yr. Notably, seeding in anyof the individual sub-regions yields better matchwith observed Sargassum distributions than initializ-ing across the entire domain (Fig. 13, gray bars),again suggesting that although Sargassum has thepotential to be dispersed throughout the Atlantic,these dispersed fragments do not appear to be driv-ing the seasonal cycle. The highest match betweenmodel and observations is obtained by seeding parti-cles in the western Gulf of Mexico, west of 90° W(RMS difference = 0.36), the southwestern tropicalAtlantic between 5° S to 10° N and 40° W to 60° W(RMS difference = 0.35), or in both of these regionssimultaneously (RMS difference = 0.34). A closerexamination within individual sub-regions showsthat while the basin-wide average agreement withobservations is similar when seeding in either thesouthwestern tropics or the western Gulf of Mexico,only the experiment that seeded in both simultane-ously reproduces the seasonal pattern in those loca-tions (Fig. 12a), but still underrepresents peak Sar-gassum biomass, especially in the Sargasso Sea.

Another possible driver for seasonal changes inSargassum biomass is vegetative reproduction. Aspelagic Sargassum and related species break, newgrowth is primarily initiated from the residual frag-ments at apical meristems (Tsukidate 1984, Hanisak& Samuel 1987). We expand the model to includevegetative propagation by tracking when each Sar-gassum particle dies, and resetting its biomass to asmall initial condition instead, simulating a smallfragment of the Sargassum mat breaking off at that

14

Fig. 12. Model performance. (a) Percent match between randomly initialized particles with (dashed line) and without(solid line) accounting for Sargassum growth. (b) Root meansquare (RMS) difference between model and satellite clima-tologies for the model (black line) and individual years ofsatellite data (shaded region indicates mean ± SD). Matchwith observations is high, consistent throughout the year, andgenerally within the bounds of observed biomass variability


location to start a new organism. Since rafting macro-algae can have expected lifetimes on the order ofmonths (Thiel & Gutow 2005), we apply this propaga-tion mode only to Sargassum particles over 1 mo old.This avoids over-representation of biomass in regionswith high mortality due to seasonal temperaturechanges or severe nutrient limitation. We apply vegetative propagation only to particles where theSargassum has died because fragmentation and

growth at other times is assumed to be aggregatedwith the other biomass in a given particle.

We repeat the 6-yr, full-domain model experimentswith vegetative propagation enabled. Simulationswith and without this reproductive strategy addedhave comparable RMS differences of 0.52 and 0.50respectively, when compared with the monthly satel-lite climatologies. However, without vegetative prop-agation the model has a negative bias in mean rela-tive Sargassum biomass of −10 to −40% of themaximum in the western Atlantic and Gulf of Mex-ico, while the bias with vegetative propagation ismuch smaller and more centered about the mean(−7% to +1.5%) (Fig. 13, white bars). Vegetativepropagation also yields increased match with observations in seeding experiments with sources ofSargassum in the Gulf of Mexico and southwestern

15

Fig. 13. (a) Root mean square (RMS) difference and (b) mean biasbetween model and observations for seeding and vegetativepro pagation experiments. Box-and-whiskers are aggregateresults from analysis within individual subregions. Boxesindicate median and interquartile ranges and whiskers de -lineate full ranges. Circles are the mean bias calculated overthe full model domain. Seeding in the 2 subregions reduceserror and mean bias, while vegetative propagation further

reduces mean bias

Fig. 14. Seeding and vegetative propagation experiments.(a) Seeding in the Gulf of Mexico and southwestern tropicsreproduces local patterns but underrepresents Sargassumbiomass elsewhere in the domain. (b) With vegetative prop-agation, Sargassum biomass in the Sargasso Sea can result

from advection and growth of nonlocal sources


tropics, while allowing increased accumulation ofbiomass in the Sargasso Sea (Fig. 14). Without vege-tative propagation, local patterns are reproduced butthe basin-wide distribution is unrealistic withoutseeding across the domain because of the differencebetween the mortality rate and the time scale ofadvection (Fig. 14a). The addition of vegetative prop-agation allows biomass accumulation in the Carib-bean and Sargasso Sea from nonlocal sources. Thus,the additive effects of vegetative propagation andlocalized sources of new Sargassum appear to be keyin accurately reproducing the Sargassum seasonalcycle.

CONCLUSIONS

At short time scales of 2 mo or less, advection alonecan be responsible for a 60% match between modelparticles and Sargassum observations, indicating thatadvection is a central element determining Sargassumdistributions in the Atlantic. However, at longer timescales, particles, whether surface, neutrally buoyant,or positively buoyant, aggregate in the subtropicalgyre. This aggregation occurs regardless of whetherparticles are initialized throughout the Atlantic oronly in regions with observed Sargassum. Thus, ad-vection is not sufficient to maintain the seasonal pat-tern of Sargassum biomass across multiple years. Thishighlights the significance of mortality, growth, andreseeding from putative source regions in maintainingSargassum’s seasonal distribution.

Results of seeding experiments with the Sargassumgrowth model highlight that the western Gulf ofMexico and the western tropics, especially south ofthe equator, appear to have a strong role in generat-ing the seasonal cycle. These regions are associatedwith 2 of the largest rivers discharging into theAtlantic, the Mississippi and the Amazon. Nutrientloading may enhance Sargassum growth locally inthese 2 regions, however even under the globallyhigh discharge of the Amazon, inorganic nitrogenfalls below detection limits about 200 km from theriver mouth (Weber et al. 2017). Neritic waters nearthe coastline have also been linked with highergrowth rates of Sargassum than in the central Sar-gasso Sea in observations (Lapointe et al. 2014). Inthis model, that growth is what allows for high Sar-gassum biomass even in regions that would other-wise have low accumulation due to the circulation.

It is also noteworthy that due to the circulation pat-terns in the Atlantic, we are unable to reproduce theseasonal biomass pattern without continuous seed-

ing in both key regions, as there are no purely advec-tive pathways for Sargassum from the Gulf of Mexicoto reach the tropics at biologically relevant timescales. The backwards-timestep particles and con-nectivity analysis presented here also support thehypothesis that Sargassum in the tropics has a south-ern source. Different morphological forms of S.natans appear to be dominant in the Sargasso Seaversus the Western Tropics (Schell et al. 2015), whichis consistent with this two-source hypothesis. Addi-tional observational studies would be helpful indetermining if Sargassum biomass is present in theproposed seed regions year-round, and whether thegenetic diversity of Sargassum fluitans and Sargas-sum natans supports this hypothesis.

The Western Tropical Atlantic source region is ofparticular interest due to its potential role in fuelingbeaching events in the Caribbean. Reverse-timemodeling and observed drifters suggest that Sargas-sum from recent beaching events in the eastern Car-ibbean and Brazil may have originated in the equato-rial region (Franks et al. 2011, 2016). Biomass in thetropics is highest in the summer, during the period ofeastward retroflection of the Amazon plume in theNorth Equatorial Counter Current. However, there isalso year-round connectivity between the Amazonplume and the Caribbean, with water dischargedfrom the Amazon in the spring having the highestprobability of reaching the Caribbean (Coles et al.2013). The connectivity analysis in this study is con-sistent with the results of Coles et al. (2013), and alsoshows that the connectivity from the tropics to theCaribbean is higher than any other potential sourcein the Atlantic. This provides a mechanism for adirect link between the elevated Sargassum biomassin the tropics in recent years and the increasedreports of Sargassum beaching in the Caribbean.

We propose a scenario where the Sargassum sea-sonal cycle begins in the spring with growth in thetropics and the western Gulf of Mexico, in agreementwith analysis of the satellite observations alone(Gower & King 2011). Biomass from the tropics getsadvected through the Caribbean to the Gulf, supple-menting local growth there into the summer. Thismodel highlights how connectivity between the Gulfof Mexico and the Sargasso Sea, in conjunction withdirect inputs from the population in the tropics, fuelsgrowth in the Sargasso Sea in summer and early fall.At this point temperature and light constrain growthat the northern extent of Sargassum’s range, andadvection works to aggregate biomass toward theSargasso Sea, where much of the biomass is exportedin the late winter. Regions of high Sargassum mortal-

16


ity in this model are consistent with observations ofSargassum on the sea floor (Schoener & Rowe 1970)and warrant further study as a possible locally impor-tant source of carbon export. This enhanced under-standing of the drivers of the Sargassum seasonalcycle should help inform management of fisheriesdependent on Sargassum habitat, and forecasting ofcostly Sargassum wash-up events in the Caribbeanand Gulf of Mexico.

Acknowledgements. The authors thank Lou Codispoti, JudyO’Neil, and Zulema Garraffo for insightful discussionsregarding this work. We also thank the 3 anonymous re -viewers whose comments greatly improved the manuscript.This research is part of the Blue Waters sustained-petascalecomputing project, which is supported by the National Sci-ence Foundation (awards OCI-0725070 and ACI-1238993)and the state of Illinois. Blue Waters is a joint effort of theUniversity of Illinois at Urbana-Champaign and its NationalCenter for Supercomputing Applications. This work was alsofunded in part by a University of Maryland Center for Envi-ronmental Science Horn Point Laboratory Bay and RiversFellowship. M.B. led the modeling effort in collaborationwith V.C. and R.H. Satellite data and interpretation was pro-vided by J.G. This manuscript is UMCES contribution #5501

LITERATURE CITED

Ackah-Baidoo A (2013) Fishing in troubled waters: oil pro-duction, seaweed and community-level grievances in thewestern region of Ghana. Community Dev J 48: 406−420

Beron-Vera FJ, Olascoaga MJ, Haller G, Farazmand M,Triñanes J, Wang Y (2015) Dissipative inertial transportpatterns near coherent Lagrangian eddies in the ocean.Chaos 25: 087412

Butler JN, Stoner AW (1984) Pelagic Sargassum: has its bio-mass changed in the last 50 years? Deep-Sea Res A 31: 1259−1264

Butler JN, Morris BF, Cadwallader J, Stoner AW (1983)Studies of Sargassum and the Sargassum community.Bermuda Biol Stn Spec Publ 22

Carpenter EJ, Cox JL (1974) Production of pelagic Sargas-sum and a blue-green epiphyte in the western SargassoSea. Limnol Oceanogr 19: 429−436

Carr A, Meylan AB (1980) Evidence of passive migration ofgreen turtle hatchlings in Sargassum. Copeia 1980: 366−368

Chassignet EP, Smith LT, Halliwell GR (2003) North Atlanticsimulations with the HYbrid Coordinate Ocean Model(HYCOM): impact of the vertical coordinate choice, ref-erence pressure, and thermobaricity. J Phys Oceanogr33: 2504−2526

Chassignet E, Hurlburt H, Metzger EJ, Smedstad O and oth-ers (2009) US GODAE: Global ocean prediction with theHYbrid Coordinate Ocean Model (HYCOM). Oceanog-raphy (Wash DC) 22: 64−75

Coles VJ, Hood RR (2007) Modeling the impact of iron andphosphorus limitations on nitrogen fixation in theAtlantic Ocean. Biogeosciences 4: 455−479

Coles VJ, Brooks MT, Hopkins J, Stukel MR, Yager PL, HoodRR (2013) The pathways and properties of the Amazon

River plume in the tropical North Atlantic Ocean. J Geo-phys Res Oceans 118: 6894−6913

Cummings JA (2005) Operational multivariate ocean dataassimilation. Q J R Meteorol Soc 131: 3583−3604

Cummings JA, Smedstad OM (2013) Variational data assim-ilation for the global ocean. In: Park SK, Xu L (eds) Dataassimilation for atmospheric, oceanic and hydrologicapplications (Vol II). Springer, Berlin, p 303−343

Deacon GER (1942) The Sargasso Sea. Geogr J 99: 16−28Dickson HN (1894) Recent contributions to oceanography.

Geogr J 3: 302−310Fennel K, Wilkin J, Levin J, Moisan J, O’Reilly J, Haidvogel

D (2006) Nitrogen cycling in the Middle Atlantic Bight: Results from a three-dimensional model and implicationsfor the North Atlantic nitrogen budget. Global Bio-geochem Cycles 20: GB3007

Franks JS, Johnson DR, Ko DS, Rubio GS, Hendon JR, Lay M(2011) Unprecedented influx of pelagic Sargassum alongCaribbean island coastlines during summer 2011. ProcGulf Caribb Fish Inst 64: 6−8

Franks JS, Johnson DR, Ko DS (2016) Pelagic Sargassum inthe tropical North Atlantic. Gulf Caribb Res 27: SC6-11

Garraffo ZD, Mariano AJ, Griffa A, Veneziani C, ChassignetEP (2001) Lagrangian data in a high resolution numericalsimulation of the North Atlantic. I: Comparison with in-situ drifter data. J Mar Syst 29: 157−176

Gower JFR, Hu C, Borstad G, King S (2006) Ocean colorsatellites show extensive lines of floating Sargassum inthe Gulf of Mexico. IEEE Trans Geosci Remote Sens 44: 3619−3625

Gower JFR, King SA (2011) Distribution of floating Sargas-sum in the Gulf of Mexico and the Atlantic Oceanmapped using MERIS. Int J Remote Sens 32: 1917−1929

Gower JFR, Young E, King S (2013) Satellite images suggesta new Sargassum source region in 2011. Remote SensLett 4: 764−773

Halliwell GR, Weisberg RH, Mayer DA (2003) A syntheticfloat analysis of upper-limb meridional overturning cir-culation interior ocean pathways in the tropical/subtrop-ical Atlantic. In: Goni GJ, Malanotte-Rizzoli P (eds) Inter-hemispheric water exchange in the Atlantic Ocean.Elsevier Oceanogr Ser 68. Elsevier, Amsterdam p 93−136

Hanisak MD, Samuel MA (1987) Growth rates in culture ofseveral species of Sargassum from Florida, USA. Hydro-biologia 151−152: 399−404

Hansen DV, Poulain PM (1996) Quality control and interpo-lations of WOCE-TOGA drifter data. J Atmos OceanTechnol 13: 900−909

Hoffmayer ER, Franks JS, Comyns BH, Hendon JR, WallerRS (2005) Larval and juvenile fishes associated withpelagic Sargassum in the Northcentral Gulf of Mexico.Proc Gulf Caribb Fish Inst 56: 259−269

Hood RR, Bates NR, Capone DG, Olson DB (2001) Modelingthe effect of nitrogen fixation on carbon and nitrogenfluxes at BATS. Deep-Sea Res II 48: 1609−1648

Hu C, Feng L, Hardy RF, Hochberg EJ (2015) Spectral andspatial requirements of remote measurements of pelagicSargassum macroalgae. Remote Sens Environ 167: 229−246

Huffard CL, von Thun S, Sherman AD, Sealey K, Smith KL(2014) Pelagic Sargassum community change over a40−year period: temporal and spatial variability. Mar Biol161: 2735−2751

Johnson DL, Richardson PL (1977) On the wind-inducedsinking of Sargassum. J Exp Mar Biol Ecol 28: 255−267

17

https://doi.org/10.1093/cdj/bst022https://doi.org/10.1063/1.4928693https://doi.org/10.1016/0198-0149(84)90061-Xhttps://doi.org/10.4319/lo.1974.19.3.0429https://doi.org/10.2307/1444022https://doi.org/10.1175/1520-0485(2003)033%3C2504%3ANASWTH%3E2.0.CO%3B2https://doi.org/10.5670/oceanog.2009.39https://doi.org/10.5194/bg-4-455-2007https://doi.org/10.1002/2013JC008981https://doi.org/10.1256/qj.05.105https://doi.org/10.1007/978-3-642-35088-7https://doi.org/10.2307/1788092https://doi.org/10.2307/1773694https://doi.org/10.1016/0022-0981(77)90095-8https://doi.org/10.1007/s00227-014-2539-yhttps://doi.org/10.1016/j.rse.2015.05.022https://doi.org/10.1016/S0967-0645(00)00160-0https://doi.org/10.1175/1520-0426(1996)013%3C0900%3AQCAIOW%3E2.0.CO%3B2https://doi.org/10.1007/BF00046159https://doi.org/10.1016/S0422-9894(03)80144-7https://doi.org/10.1080/2150704X.2013.796433https://doi.org/10.1080/01431161003639660https://doi.org/10.1109/TGRS.2006.882258https://doi.org/10.1016/S0924-7963(01)00015-Xhttps://doi.org/10.18785/gcr.2701.08https://doi.org/10.1029/2005GB002456


Laffoley Dd’A, Roe HSJ, Angel MV, Ardron J and others(2011) The protection and management of the SargassoSea: The golden floating rainforest of the Atlantic Ocean.Summary science and supporting evidence case. Sar-gasso Sea Alliance, IUCN, Washington, DC

Lapointe BE (1986) Phosphorus-limited photosynthesis andgrowth of Sargassum natans and Sargassum fluitans(Phaeophyceae) in the western North Atlantic. Deep-SeaRes A 33: 391−399

Lapointe BE (1995) A comparison of nutrient-limited pro-ductivity in Sargassum natans from neritic vs. oceanicwaters of the western North Atlantic Ocean. LimnolOceanogr 40: 625−633

Lapointe BE, West LE, Sutton TT, Hu C (2014) Ryther revis-ited: Nutrient excretions by fishes enhance productivityof pelagic Sargassum in the western North AtlanticOcean. J Exp Mar Biol Ecol 458: 46−56

Law KL, Morét-Ferguson S, Maximenko NA, ProskurowskiG, Peacock EE, Hafner J, Reddy CM (2010) Plastic accu-mulation in the North Atlantic subtropical gyre. Science329: 1185−1188

North EW, Gallego A, Petitgas P (eds) (2009) Manual of rec-ommended practices for modelling physical-biologicalinteractions during fish early life. ICES Coop Res Rep 295

O’Reilly JE, Maritorena S, O’Brien MC, Siegel DA and oth-ers (2000) SeaWiFS postlaunch calibration and validationanalyses, part 3. SeaWiFS Postlaunch Tech Rep Ser 11

Oyesiku OO, Egunyomi A (2011) Identification and chemi-cal studies of pelagic masses of Sargassum natans (Lin-naeus) Gaillon and S. fluitans (Borgessen) Borgesen(brown algae), found offshore in Ondo State, Nigeria. AfrJ Biotechnol 13: 1188−1193

Parr AE (1939) Quantitative observations on the pelagic Sar-gassum vegetation of the western North Atlantic. BullBingham Oceanogr Coll Yale Univ 6: 1−94

Putman NF, He R (2013) Tracking the long-distance disper-sal of marine organisms: sensitivity to ocean model reso-lution. J R Soc Interface 10: 20120979

Rast M, Bezy JL, Bruzzi S (1999) The ESA Medium Resolu-tion Imaging Spectrometer MERIS: a review of the in -strument and its mission. Int J Remote Sens 20: 1681−1702

Rypina II, Jayne SR, Yoshida S, Macdonald AM, Dou-glass E, Buesseler K (2013) Short-term dispersal ofFukushima-derived radionuclides off Japan: modelingefforts and model-data intercomparison. Biogeo-sciences 10: 4973−4990

Schell JM, Goodwin DS, Siuda ANS (2015) Recent Sargas-sum inundation events in the Caribbean: shipboardobservations reveal dominance of a previously rare form.Oceanography (Wash DC) 28: 8−10

Schoener A, Rowe GT (1970) Pelagic Sargassum and itspresence among the deep-sea benthos. Deep-Sea Res,Oceanogr Abstr Oceanogr Bibliogr Sect 17: 923−925doi.org/10.1016/0011-7471(70)90010-0

SeaWiFS Project (2003) SeaWiFS global monthly mapped9 km chlorophyll a. Ver. 1. https: //podaac.jpl.nasa.gov/dataset/SeaWiFS_L3_CHLA_Monthly_9km_R

Simons RD, Siegel DA, Brown KS (2013) Model sensitivityand robustness in the estimation of larval transport: a

study of particle tracking parameters. J Mar Syst119−120: 19−29

Sissini MN, de Barros Barreto MBB, Széchy MTMd, deLucena MB and others (2017) The floating Sargassum(Phaeophyceae) of the South Atlantic Ocean — likelyscenarios. Phycologia 56: 321−328

Siuda ANS (2011) Summary of Sea Education Associationlong-term Sargasso sea surface net data. Sargasso SeaAlliance Science Rep Ser 10

Smetacek V, Zingone A (2013) Green and golden seaweedtides on the rise. Nature 504: 84−88

Steele M, Morley R, Ermold W (2001) PHC: A global oceanhydrography with a high-quality Arctic Ocean. J Clim14: 2079−2087

Stoner AW (1983) Pelagic Sargassum: evidence for a majordecrease in biomass. Deep-Sea Res A 30: 469−474

Stukel MR, Coles VJ, Brooks MT, Hood RR (2014) Top-down, bottom-up and physical controls on diatom- diazotroph assemblage growth in the Amazon Riverplume. Biogeosciences 11: 3259−3278

Széchy MTMd, Guedes PM, Baeta-Neves MH, Oliveira EN(2012) Verification of Sargassum natans (Linnaeus) Gail-lon (Heterokontophyta: Phaeophyceae) from the Sar-gasso Sea off the coast of Brazil, western Atlantic Ocean.Check List 8: 638−641

Thiel M, Gutow L (2005) The ecology of rafting in the marineenvironment. I. The floating substrata. Oceanogr MarBiol Annu Rev 42: 181–264

Tsukidate J (1984) Studies on the regenerative ability of thebrown algae, Sargassum muticum (Yendo) Fensholt andSargassum tortile C. Agardh. Hydrobiologia 116-117: 393−397

Uppala SM, Kållberg PW, Simmons AJ, Andrae U and others(2005) The ERA-40 re-analysis. Q J R Meteorol Soc 131: 2961−3012

Wang M, Hu C (2016) Mapping and quantifying Sargassumdistribution and coverage in the Central West Atlanticusing MODIS observations. Remote Sens Environ 183: 350−367

Weber SC, Carpenter EJ, Coles VJ, Yager PL, Goes J, Mon-toya JP (2017) Amazon River influence on nitrogen fixa-tion and export production in the western tropical NorthAtlantic. Limnol Oceanogr 62: 618−631

Webster RK, Linton T (2013) Development and implementa-tion of Sargassum Early Advisory System (SEAS). ShoreBeach 81: 1−6

Winge Ø (1923) The Sargasso Sea, its boundaries and vege-tation. Rep Danish Oceanogr Exped 1908–1910 toMediterr Adjac Seas 3: 1−34

Witherington B, Hirama S, Hardy R (2012) Young sea turtlesof the pelagic Sargassum-dominated drift community: habitat use, population density, and threats. Mar EcolProg Ser 463: 1−22

Woodcock AH (1993) Winds subsurface pelagic Sargassumand Langmuir circulations. J Exp Mar Biol Ecol 170: 117−125

Zhong Y, Bracco A, Villareal TA (2012) Pattern formation atthe ocean surface: Sargassum distribution and the role ofthe eddy field. Limnol Oceanogr Fluids Environ 2: 12−27

18

Editorial responsibility: Claire Paris, Miami, Florida, USA

Submitted: November 22, 2017; Accepted: May 18, 2018Proofs received from author(s): June 30, 2018

https://doi.org/10.1016/0198-0149(86)90099-3https://doi.org/10.4319/lo.1995.40.3.0625https://doi.org/10.1016/j.jembe.2014.05.002https://doi.org/10.1126/science.1192321https://doi.org/10.5897/AJB2013.12335https://doi.org/10.1098/rsif.2012.0979https://doi.org/10.1080/014311699212416https://doi.org/10.5194/bg-10-4973-2013https://doi.org/10.5670/oceanog.2015.70https://doi.org/10.1016/0011-7471(70)90010-0https://doi.org/10.1016/j.jmarsys.2013.03.004https://doi.org/10.2216/16-92.1https://doi.org/10.1215/21573689-1573372https://doi.org/10.1016/0022-0981(93)90132-8https://doi.org/10.3354/meps09970https://doi.org/10.1002/lno.10448https://doi.org/10.1016/j.rse.2016.04.019https://doi.org/10.1256/qj.04.176https://doi.org/10.1007/BF00027708https://doi.org/10.15560/8.4.638https://doi.org/10.5194/bg-11-3259-2014https://doi.org/10.1016/0198-0149(83)90079-1https://doi.org/10.1175/1520-0442(2001)014%3C2079%3APAGOHW%3E2.0.CO%3B2https://doi.org/10.1038/nature12860

factors controlling the seasonal distribution of pelagic ... · thus, sargassum may not act as an...

Documents