Harmful Algae 48 (2015) 1–11

Present-day and future climate pathways affecting Alexandriumblooms in Puget Sound, WA, USA

Stephanie K. Moore a,*, James A. Johnstone b, Neil S. Banas b, Eric P. Salathe Jr.c

a University Corporation for Atmospheric Research, Joint Office for Science Support. Visiting Scientist at Northwest Fisheries Science Center, National Marine

Fisheries Service, NOAA, 2725 Montlake Blvd E, Seattle, WA 98112, United Statesb Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, 3737 Brooklyn Ave NE, Box 355672, Seattle, WA 98195, United Statesc School of Science, Technology, Engineering and Mathematics, University of Washington Bothell, 18115 Campus Way NE, Box 358538, Bothell, WA 98011,

United States

A R T I C L E I N F O

Article history:

Received 28 October 2014

Received in revised form 15 June 2015

Accepted 16 June 2015

Keywords:

Alexandrium

Harmful algae

HAB

Climate change

Puget Sound

A B S T R A C T

This study uses a mechanistic modeling approach to evaluate the effects of various climate pathways on

the proliferative phase of the toxin-producing dinoflagellate Alexandrium in Puget Sound, WA, USA.

Experimentally derived Alexandrium growth responses to temperature and salinity are combined with

simulations of the regional climate and Salish Sea hydrology to investigate future changes in the timing,

duration, and extent of blooms. Coarse-grid (100–200 km) global climate model ensemble simulations of

the SRES A1B emissions scenario were regionally downscaled to a 12-km grid using the Weather

Research and Forecasting model for the period 1969–2069. These results were used to: (1) analyze the

future potential changes and variability of coastal upwelling winds, and (2) provide forcing fields to a

Regional Ocean Model System used to simulate the circulation of the Salish Sea, including Puget Sound,

and the coastal ocean. By comparing circa-1990 and circa-2050 climate scenarios for the environmental

conditions that promote Alexandrium blooms, we disentangle the effects of three climate pathways: (1)

increased local atmospheric heating, (2) changing riverflow magnitude and timing, and (3) changing

ocean inputs associated with changes in upwelling-favorable winds. Future warmer sea surface

temperatures in Puget Sound from increased local atmospheric heating increase the maximum growth

rates that can be attained by Alexandrium during the bloom season as well as the number of days with

conditions that are favorable for bloom development. This could lead to 30 more days a year with bloom-

favorable conditions by 2050. In contrast, changes in surface salinity arising from changes in the timing

of riverflow have a negligible effect on Alexandrium growth rates, and the behavior of the coastal inputs

in the simulations suggests that changes in local upwelling will not have major effects on sea surface

temperature or salinity or Alexandrium growth rates in Puget Sound.

� 2015 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Harmful Algae

jo u rn al h om epag e: ww w.els evier .c o m/lo cat e/ha l

1. Introduction

Puget Sound is an estuary of 2230 km2 in Washington State,USA, and is part of the Salish Sea. Puget Sound supports extensiveshellfish harvesting for commercial, recreational, and subsistencepurposes. The value of the shellfish industry in Washington State isestimated at $108 million per year (based on 2008 and 2009 datacompiled by the Pacific Coast Shellfish Growers Association), andapproximately half of that value is from product grown in PugetSound. In fact, the economic value of shellfish aquaculture inPuget Sound alone exceeds that for all other western USA states

* Corresponding author. Tel.: +1 206 860 3327; fax: +1 206 860 3335.

E-mail address: stephanie.moore@noaa.gov (S.K. Moore).

http://dx.doi.org/10.1016/j.hal.2015.06.008

1568-9883/� 2015 Elsevier B.V. All rights reserved.

combined (i.e., $16 million in California, $3 million in Oregon, and$0.6 million in Alaska).

Puget Sound is also home to a number of harmful algal bloom(HAB) species that can produce toxins and contaminate shellfish.Arguably, the most concerning of these harmful algae aredinoflagellates in the genus Alexandrium that produce a suite ofpotent neurotoxins collectively called paralytic shellfish toxins(PSTs). The most potent of the PSTs is saxitoxin (STX). WhenAlexandrium blooms, PSTs can accumulate in shellfish to levels thatare unsafe for human consumption (>80 mg STX equivalents100 g�1 shellfish tissue). If contaminated shellfish are consumedby humans, gastrointestinal and neurological symptoms includingvomiting and muscle paralysis can develop and death can occurin extreme cases (Quayle, 1969; Kao, 1993). The species ofAlexandrium thought to be responsible for the production of these

S.K. Moore et al. / Harmful Algae 48 (2015) 1–112

toxins in Puget Sound has historically been identified asAlexandrium catenella (Whedon & Kofoid) Balech. This is synony-mous with Alexandrium tamarense Group I, a provisional speciesname proposed by Lilly et al. (2007). However, the nameAlexandrium fundyense has recently been proposed to replace allGroup I strains of the A. tamarense species complex (John et al.,2014). In light of the recent work by John et al. (2014) and to avoidfurther confusion over Alexandrium species name designations, wewill refer here only to the genus name Alexandrium.

The life history of Alexandrium involves a benthic resting cyststage that is used to overwinter, and a proliferative planktonicphase that is associated with blooms (Anderson, 1998). The typicalbloom season for Alexandrium in Puget Sound broadly spans fromJuly through November, but interannual variability is high (Mooreet al., 2009). To prevent contaminated shellfish from reaching themarket, public health managers conduct extensive monitoring toensure that shellfish are safe for consumption (Trainer et al., 2003).This program is remarkably effective at preventing illnesses in WA,especially illnesses arising from the consumption of commercialshellfish, but it is time consuming and costly. Advanced warning ofthe increased risk for Alexandrium blooms would allow managersto efficiently target locations and time periods for monitoring. Thefrequency of Alexandrium blooms appears to be sensitive toenvironmental conditions in Puget Sound, with bloom onsetfavored by a combination of variables that include warm surfacewater and air temperatures, low stream flow, weak winds andsmall tidal variability (Moore et al., 2009). When this combinationof environmental conditions occurs, a ‘window of opportunity’presumably exists for Alexandrium growth, increasing the risk forshellfish toxicity. Knowledge of when and where environmentalconditions favor bloom development in Puget Sound may informearly warning capabilities and have considerable benefits forpublic health and the Puget Sound shellfish industry.

During the typical annual bloom season for Alexandrium, thehydroclimate of Puget Sound reaches annual extremes in severalrelated variables, complicating the isolation of specific environ-mental mechanisms that increase risk for toxic bloom events. Inspring and early summer, Puget Sound is fed by freshwater flowssupplied by melting residual snowpack in the surroundingmountain ranges hundreds of kilometers inland. Freshwaterinputs reach their annual minimum in early fall, approximatelycoincident with the annual water and air temperature maximum(Moore et al., 2008b). Note that Puget Sound has a robust estuarinecirculation, in which the upper, fresher layer generally exitstoward the ocean and colder, saltier, nutrient-rich water is drawnin at depth, although the layer structure, strength of theoverturning circulation, and hence residence time vary signifi-cantly among subbasins (Cokelet et al., 1991; Sutherland et al.,2011; Banas et al., 2014). The deep inflow originates on thecontinental slope of the Pacific Northwest upwelling zone, via theJuan de Fuca Canyon, and its water properties are therefore set by acombination of coastal wind-driven upwelling (Hickey, 1989),enhancement of the upwelling circulation by canyon effects(Alford and MacCready, 2014; Connolly and Hickey, 2014), andtidal mixing over multiple entrance sills (e.g., Geyer and Cannon,1982). Surface water properties in Puget Sound are thus influencedby a range of factors both local (i.e., atmospheric heat fluxes,nearby rivers) and remote (i.e., ocean influence via the overturningcirculation and mixing, wind-driven intrusions of the Fraser Riverplume in summer) (Moore et al., 2008a,b; Banas et al., 2014).

PST concentrations in shellfish from inside Puget Sound firstexceeded the regulatory limit for human consumption in 1978(Rensel, 1993). Since then, shellfish harvest closures have occurredwith increasing frequency and geographic scope (Trainer et al.,2003). This increase in Puget Sound has occurred in conjunctionwith an apparent global trend, attributed to various causes,

including increased nutrient loads and climatic changes (e.g.,Hallegraeff, 1993, 2010; Anderson et al., 2002, 2012; Glibert et al.,2005; Moore et al., 2008c). The sensitivity of Alexandrium bloomsto environmental conditions that are modulated by the local andregional climate has led to the hypothesis that climate variabilityand change may contribute to the interannual differences andlong-term increase in shellfish toxicity in Puget Sound. Moore et al.(2010) investigated the influences of regional climate patterns onPST accumulation in blue mussels from Puget Sound, finding thatannual toxicity increased with the number of days exceeding 13 8C,a temperature threshold for accelerated growth and toxicity oflocal Alexandrium (Norris and Chew, 1975; Nishitani and Chew,1984). Warmer summers, and higher toxicity in shellfish, tended tocoincide with the positive phase of the Pacific Decadal Oscillation(PDO), the predominant mode of North Pacific sea surfacetemperature (SST) variability that correlates with warm anomaliesalong the west coast of North America (Mantua and Hare, 2002).No relationship was found between shellfish toxicity and the ElNino-Southern Oscillation (ENSO), likely due to its predominantinfluence on Puget Sound conditions in winter, outside of thetypical annual bloom season.

Long-term historical trends and future scenarios have also beenexamined for the window of opportunity for Alexandrium in PugetSound (Moore et al., 2011). This analysis was based on theenvironmental criteria that were found to increase the risk ofmajor toxic events during the period 1993–2007 (Moore et al.,2009). The seasonal window of opportunity was found to increasein duration over the 40-year study period from 1967 to 2006, withnotable step-like increases around 1978 and 1992. The 1978 in-crease closely followed a major shift in regional climate associatedwith the PDO and coincided with the first shellfish harvestingclosure inside Puget Sound from unsafe levels of PSTs (Ebbesmeyeret al., 1995; Trainer et al., 2003). These climatic changes were mostevident by coastal sea surface warming in the northeast Pacific,including the offshore regions near Puget Sound. Coastal SSTs alsoexert a strong influence on air temperatures further inland(Johnstone and Mantua, 2015), which correlate strongly withsurface water temperatures within the Sound (Moore et al., 2008a).Future scenarios for the window of opportunity were evaluatedusing global climate model (GCM) output. The duration of theseasonal window was projected to increase by �13 d yr�1 by theend of the 21st century, under the A1B scenario of moderategreenhouse gas emissions (Moore et al., 2011). Of the environ-mental conditions found to increase the risk of major toxic eventsin Puget Sound, warmer SSTs appear to be a strong driver of boththe historical and projected increases.

The models described above that have been used to evaluatethe risk of Alexandrium blooms in Puget Sound are based on a‘heuristic’ understanding of toxic events. Heuristic models, orcorrelational approaches, look for common patterns in theenvironment associated with past blooms to inform the predictionof future blooms. This approach is particularly useful when long-term time series of observations are available to develop and testthe model. However, the statistical relationships that heuristicmodels are based on do not necessarily reflect the ecological andoceanographic processes and interactions that are important fortoxic bloom development. Alternatively, mechanistic modelssimulate these important processes and better explain thedynamics of a system (modeling approaches for coastal marineharmful algae are reviewed in Anderson et al., 2014). A notableexample of this kind of approach is the coupled biological–physicalmodel that is used to predict blooms of Alexandrium fundyense inthe Gulf of Maine (McGillicuddy et al., 2005). Mechanistic modelsrequire specific knowledge of a number of physical and biologicalrates that govern the movement, growth, and toxicity of blooms.They can quantitatively evaluate the effects of changes in the

S.K. Moore et al. / Harmful Algae 48 (2015) 1–11 3

environment on bloom dynamics and facilitate a deeper under-standing of the system.

This study uses a mechanistic approach to model the potentialgrowth response (i.e., proliferative phase) of Alexandrium toclimate-driven changes in Puget Sound sea surface temperatureand salinity. We combine detailed experimentally derivedAlexandrium growth responses to environmental conditions withregional atmosphere and ocean model simulations to investigatefuture changes in the timing, duration, and extent of toxic bloomsin Puget Sound. By comparing present-day and future scenarios forthe environmental conditions that promote Alexandrium blooms,we disentangle the effects of three climate pathways: (1) increaseddirect heat input, (2) changing stream flow magnitude and timing,and (3) changing ocean inputs (associated with upwelling winds).

2. Methods

2.1. Climate model simulations

Climate simulations for the U.S. Pacific Northwest (PNW) weredeveloped with a regional climate model using two different GCMsto provide forcing at the boundaries. GCM resolution of 100–200 km is too coarse to resolve important terrain features thatcontrol the PNW and Puget Sound climate (Salathe et al., 2010). Byusing a regional climate model to dynamically downscale theglobal models, such features as coastally enhanced winds,orographic precipitation, and air–sea interactions over PugetSound are represented. Both regional simulations use the WeatherResearch and Forecasting (WRF; Skamarock et al., 2008) modeldeveloped at the National Center for Atmospheric Research(NCAR). This model includes advanced representations of cloudmicrophysics and land-surface dynamics to simulate the complexinteractions between atmospheric processes like precipitation andland surface characteristics such as snow cover and soil moisture.

The GCM simulations were performed as part of the ThirdClimate Model Intercomparison Project (CMIP3). One simulationis forced by the NCAR Community Climate System Model version3 (CCSM3; Collins et al., 2006) and the second is forced by asimulation of the fifth-generation Max Planck Institute, Hamburg,GCM (ECHAM5/MPI-OM; Roeckner et al., 1999; Marsland et al.,2003; Roeckner et al., 2003). For the present climate (1970–1999),we used a simulation of the 20th century with historical forcing;for the 21st century, we used a simulation with the Special Reporton Emissions Scenarios (SRES) A1B greenhouse gas emissionsscenario (Nakicenovic et al., 2000). The A1B is a medium-highemission scenario reflecting ‘‘business as usual’’ in the first half ofthe 21st century and greater mitigation in the second half(Nakicenovic et al., 2000). Relative to the recent RepresentativeConcentration Pathways used in the Fifth IPCC Assessment, A1Bshow similar temperature and precipitation changes to RCP6.0(CIG Technical Report http://cses.washington.edu/cig/reports.shtml#sok).

The two GCMs (CCSM3 and ECHAM5) are compared with a setof 19 GCMs in Mote and Salathe (2009), who show that bothmodels yield relatively small temperature and precipitation biascompared to observations for the 20th century climate of the PNW.ECHAM5 in particular simulates atmospheric circulations patternsover the North Pacific region quite well. Compared to the multi-model average for the Pacific Northwest, ECHAM5 projects a lowtemperature increase and a high precipitation increase whileCCSM3 projects a relatively warmer and drier future.

Downscaling simulations were performed using WRF with 12-km grid spacing to simulate 6-hourly atmosphere and land-surfaceparameters over the PNW as described in Salathe et al. (2015). Theregional model considerably improves the simulation of dailyand extreme weather as compared to the global model providing

boundary conditions (Zhang et al., 2009; Duliere et al., 2011).Downscaled simulations of CCSM3 (CCSM3-WRF) and ECHAM5(ECHAM5-WRF) were generated for the period 1969–2069.

2.2. Ocean model simulations

Output from the CCSM3-WRF simulation was used as input to athree-dimensional numerical ocean model of the Salish Sea,including Puget Sound, and the adjacent coastal zone (Fig. 1). Theocean model, called MoSSea (Modeling the Salish Sea), wasimplemented using the Regional Ocean Modeling System (ROMS;Shchepetkin and McWilliams, 2005) following a frameworkdeveloped by the University of Washington Coastal ModelingGroup that was recently used for higher-resolution simulations ofPuget Sound (Sutherland et al., 2011) and the outer coast (Giddingset al., 2015). The model includes realistic bathymetry; freshwaterinputs from the Fraser, Columbia, and 14 smaller Puget Soundrivers; realistic tides; and is forced by surface fields from WRF(wind, pressure, air temperature, relative humidity, rainfall, andshortwave and longwave radiation). Open boundary conditions areset by one-way nesting inside output from Global NCOM (NavyCoastal Ocean Model; Barron et al., 2006) for the year 2006. Themodel grid has 40 vertical layers using a stretched verticalcoordinate. Horizontal grid spacing is 700 m over Puget Sound,stretching out to 3 km over the continental shelf and slope.

This horizontal grid is a compromise between the Sutherlandet al. (2011) design, which had variable grid cell resolution down to280 m in Puget Sound and up to 3 km offshore, and the Giddingset al. (2015) design, which had 1.5 km grid cell resolution over thecontinental shelf and maintained this resolution (which barelyresolves the narrow channels of Puget Sound) throughout theSalish Sea. On the shelf, the reduction in resolution does cause adegradation of model skill relative to the Giddings et al. (2015)hindcast of 2005, but the degradation is relatively small. AverageWillmott skill scores (Willmott, 1982) for comparisons with tidallyaveraged temperature, salinity and major-axis velocity at coastalmooring stations and sea level observations are 0.75, 0.68, 0.69 inthe present model, compared with 0.79, 0.68, 0.79 in the Giddingset al. (2015) model (note that the Willmott skill score incorporatesboth correlation error and bias, where 0 indicates no skill and 1 aperfect model). The new model’s reduced resolution in PugetSound relative to Sutherland et al. (2011) likewise causes increasedbiases in bottom-water properties (�1 salinity, 2 8C in summer inMain Basin; approximately twice the error in Sutherland et al.,2011), presumably a result of insufficiently resolved mixing at thesills, together with inherited bias from NCOM. Nevertheless,the model shows significant skill (�0.7) at reproducing thenear-surface fields most relevant to Alexandrium growth anddistribution, as detailed below.

2.3. Future scenarios

The two downscaled GCM projections described above (CCSM3-WRF and ECHAM5-WRF) both project strong warming trends insurface air temperatures over the Puget Sound basin throughoutthe year (averaged 46.9–498 N and 122–123.258 W, 2 m above seasurface; Fig. 2). Changes over the coastal ocean are more modest(not shown). Model agreement and strong trends suggests thatwarming is very likely to occur (i.e., 2–3 8C per century in CCSM3-WRF). The two projections, however, do not present a consistentpicture of future trends in the winds that drive coastal upwellingand hence ocean inputs to Puget Sound. CCSM3-WRF projects a10–20% increase in the magnitude of May through Septemberupwelling-favorable (i.e., from the north) winds over the 100-yperiod 1969–2069 (Fig. 3A). In contrast, changes in monthly meanwinds are more variable in ECHAM5-WRF and generally small,

Fig. 1. (A) Ocean model domain encompassing the inland waters of the Salish Sea and the coastal ocean from Vancouver Island, British Columbia to central Oregon. (B) Map of

the southern Salish Sea showing the North, Hood Canal, Whidbey, Main and South Sound basins discussed in the text. The transect along which results are calculated in Figs. 5,

7, 8 is shown in blue, with distance markers every 50 km, reaching from the continental slope via the submarine Juan de Fuca canyon and the Strait of Juan de Fuca into Main

Basin and Sound. Distances are calculated from the middle of Admiralty Inlet, the double sill separating the Strait of Juan de Fuca from Puget Sound. The 100 and 200 m

isobaths are shown in gray.

S.K. Moore et al. / Harmful Algae 48 (2015) 1–114

except for a strong trend toward increased upwelling in Junespecifically (Fig. 3B), suggesting a change in the timing of theupwelling season as opposed to its intensity. This inconsistency ischaracteristic of the driving global projections as well. Among the21 GCMs analyzed by Wang et al. (2010), 50-y trends in upwelling-favorable winds for the PNW vary from a decline of �40% to anincrease of �60% relative to the 1980s mean (Fig. 3D). The meanand standard deviation for this global ensemble is +8 � 17%.

Fig. 2. Trends in surface air temperature over the Puget Sound basin (46.9–498 N, 122–1

simulations.

Changes in upwelling, therefore, need to be treated not as aprediction but as a sensitivity experiment. We constructed asimple experiment by selecting individual years from CCSM3-WRFwhose seasonal cycle of wind stress were closest to decadal meansfor the 1980s and the 2040s (1988 and 2047), which are shown inFig. 3C. By driving the regional ocean model with atmosphericforcing from these years, we can assess whether the effect of amoderate-to-high increasing trend in upwelling would be likely to

23.258 W) over 100 y (1969–2069) from the (A) CCSM3-WRF and (B) ECHAM5-WRF

Fig. 3. Trends in the north-south component of regional winds (45–488 N, 124–1278 W) during the upwelling season (April through October) over 100 y (1969–2069) from the

(A) CCSM3-WRF and (B) ECHAM5-WRF simulations. Negative values indicate upwelling-favorable wind from the north. (C) Monthly mean north-south winds for the years

selected from CCSM3-WRF to drive present-day (dashed) and future (dotted) simulations in ROMS. (D) Histogram of relative trends in upwelling over 50 y (change in

upwelling index relative to the present-day July mean) in the 21 GCMs (not downscaled) analyzed by Wang et al. (2010) (see Fig. 7 in that study, top panels, for source data).

S.K. Moore et al. / Harmful Algae 48 (2015) 1–11 5

significantly affect the environmental conditions that promoteAlexandrium blooms in Puget Sound compared with the effects oflocal heating and changing river inputs.

Seasonal cycles of riverflow for present-day and 2040sconditions were constructed by synthesizing several existingprojections that used similar methods. Columbia River flow wastaken from the results in Elsner et al. (2010), Fraser River flow wastaken from Shrestha et al. (2012), and flows for 14 Puget Soundrivers from Cuo et al. (2011). These studies were each based on anensemble of statistically downscaled CMIP3 global climate modelsused as input to a hydrologic model. For all basins, the mostimportant projected change between the present day and 2040s isa substantial reduction in mountain snowpack, which results inhigher precipitation-driven flows in winter and lower snow-melt-driven flows in summer.

We combined the present day riverflow seasonal cycle(interpolated from monthly means, without extreme events) withCCSM3-WRF year 1988 to construct our ‘‘present day’’ ROMSscenario, and the 2040s riverflow seasonal cycle with CCSM3-WRFyear 2047 to construct the future ROMS scenario analyzed below.Open-ocean boundary conditions (1298W, 438N, 518N: see Fig. 1)were taken from Global NCOM fields from 2006 and held constantbetween the present-day and future scenarios, since downscalingchanges in the ocean on the Northeast Pacific or whole CaliforniaCurrent System scale was beyond the scale of this study.

2.4. Alexandrium bloom risk maps

Alexandrium bloom risk maps were produced using thetemperature and salinity fields from the present-day and futureclimate simulations of the MoSSea ocean model. These fields wereapplied to a growth function for Alexandrium that was developedusing two local strains from Puget Sound (Bill et al., in review).Briefly, acclimated cells were incubated in small volume (30 mL)

batch cultures in a temperature gradient bar enclosed within acontrolled light box modified from Watras et al. (1982). Light wasset to a 14:10 h light:dark cycle. In vivo fluorescence was recordeddaily at 19 temperatures ranging from 5 to 28 8C and six salinitiesranging from 10 to 35. Specific growth rates (m) were calculated forthe exponential phase of growth for each temperature and salinitytreatment following the method of Brand et al. (1981). Since littleinteraction was observed in the Alexandrium growth responses totemperature and salinity, the growth function was calculated as afourth order polynomial fit to temperature with a variableadjustment for the salinity response. The growth function wasfitted to the averaged growth rates of the two isolates. Theresulting function for Puget Sound strains of Alexandrium is shownin Fig. 4 and indicates that growth is inhibited at temperaturesbelow 7–8 8C, but then increases with warmer temperatures.Above 10 8C, the growth response is slightly less sensitive totemperature and a broad range of temperatures between 10 and24 8C support optimal growth. Growth is strongly inhibited attemperatures above 25–26 8C, and growth ceases completely at28 8C. In general, Puget Sound Alexandrium are euryhaline andgrowth is sensitive to salinity changes only in the range of 10–15salinity. Salinities below 10 prevent growth, while constantmaximum responses are seen between 20 and 35 salinity.

Alexandrium bloom risk maps were produced for the present-day and future climate simulations in two ways: (1) by calculatingthe mean of the maximum growth rates attained over the durationof the bloom season using the growth function described in Billet al. (in review), and (2) by calculating the number of daysduring the bloom season when a growth rate of 0.25 d�1 ormore was attained (approximately half of the maximum growthrate attained by Puget Sound strains of Alexandrium; Bill et al.,in review). The first map indicates potential changes toAlexandrium growth rates during the bloom season as a functionof temperature and salinity, whereas the second indicates

Fig. 4. Alexandrium growth (m, day�1) as a function of temperature (8C) and salinity.

m = specific growth rate. [Bill et al., in reviewBill et al., in preparation]. The mean

and standard deviation of sea surface temperature and salinity in the North (*),

Whidbey (&), Main (~), and South (^) basins of Puget Sound from the present-day

(open symbols) and future (filled symbols) climate scenarios are plotted on the

growth function.

S.K. Moore et al. / Harmful Algae 48 (2015) 1–116

potential changes to the seasonal duration of bloom-favorabletemperature and salinity conditions. Both maps assume no otherlimitations to growth. Together these maps provide a powerfulassessment of potential changes in Alexandrium bloom risk inspace and time, because small changes in the growth rate maystill result in large changes in Alexandrium abundance dependingon the length of time that cells are exposed to more favorableconditions.

Fig. 5. Present-day seasonal cycle of surface temperature and surface salinity in Main Bas

D) the model. Observations are a composite of sampling 1989–2012 as described in th

3. Results

3.1. Puget Sound water properties

Seasonal cycles of SST and salinity along a transect down thelength of Puget Sound’s Main Basin into South Sound (see map inFig. 1) are shown in Fig. 5 for the present-day scenario, incomparison with observations composited from a number ofmonitoring programs. CTD (conductivity–temperature–depth)samples in the upper 5 m, within 5 km of the transect line, wereassembled from monitoring 1989–2012 by the WashingtonDepartment of Ecology, King County Water and Land ResourcesDivision, and biannual PRISM (Puget Sound Regional SynthesisModel) cruises. Observations are displayed as a composite seasonalcycle (Fig. 5A and C). Some non-negligible model biases are evident(particularly SST in winter), but the model reproduces along-channel and seasonal variability with significant skill(WS = 0.69 for temperature and 0.72 for salinity).

Seasonal cycles of SST and salinity, for the present and futurescenarios, averaged for the entire Puget Sound basin, are shown inFig. 6A and B. SST is projected to increase by 2 8C at the spring and fallmargins of the warm season and 3 8C at the peak of late summerwarmth (Fig. 6C). Projected changes in surface salinity (Fig. 6D) showthat the early summer minimum, produced by freshwater inputs,occurs �20 d earlier due to earlier snow melt runoff (Fig. 6B). Thischange in the timing of peak runoff leads to lower inputs later inthe summer, leading to more saline conditions in Puget Sound,particularly in those areas fed by rivers originating in the Cascades.

Changes in deep water temperature and salinity are mutedrelative to these changes in surface fields. The difference in depth-resolved temperature between the present and future scenarios,

in and South Sound (see Fig. 1 for transect line) in (A and C) observations and (B and

e text.

Fig. 6. Daily mean values of (A) sea surface temperature (SST) and (B) salinity for the present-day (dashed) and future (dotted) climate scenarios from the CCSM3-WRF-

MoSSea simulation. The changes in present-day and future (C) SST and (D) salinity are also shown. Data are smoothed using a 31-day running mean (�15 d).

S.K. Moore et al. / Harmful Algae 48 (2015) 1–11 7

averaged May–September, is shown in Fig. 7 for an extendedtransect line reaching from the continental slope, via Juan de Fucacanyon and the Strait of Juan de Fuca, into Main Basin and SouthSound (see Fig. 1). The �2 8C temperature increase discussed aboveis strongest at the surface. The inflowing deep layer shows adecrease of 0.1–0.2 8C in the future scenario, the signature ofintensified upwelling or upwelling drawing from a deeper offshoresource. The weak temperature increase at depth in Puget Soundcan be interpreted as the sum of these opposing trends: thepositive effect of local surface heating mixing down from above,and the negative effect of increased upwelling transported inthrough the Admiralty Inlet mixing zone.

Consistent with this vertical structure, changes in stratificationin the future scenario largely follow the surface fields (Fig. 8). Inboth Main Basin and the Strait of Juan de Fuca, top-to-bottomsalinity stratification (Fig. 8A and B) increases November to May,and decreases during early summer, the present-day riverflow

Fig. 7. Change in mean summer temperature (future scenario minus present-day scen

peak. Density stratification (Fig. 8C and D) largely follows salinitystratification, except that in Main Basin, increasing surfacetemperature cancels the salinity effect on stratification duringearly summer (Fig. 8C).

3.2. Alexandrium bloom risk maps

The SST and salinity fields from the downscaled CCSM3-WRF-MoSSea simulations were applied to the empirical growth functionderived by Bill et al. (in review) (see Fig. 4) and used to mapAlexandrium bloom risk in Puget Sound under present-day andfuture climate conditions. The simulated changes in HAB-favorableconditions were calculated over a conservatively longer thannormal Alexandrium bloom season (May–October) to capture anychanges that might arise from a temporal expansion of HAB-favorable conditions (Moore et al., 2011). Relative to the present-day, future conditions in Puget Sound are projected to support both

ario, averaged May through September) along the transect line shown in Fig. 1.

Fig. 8. (A and B) Top-to-bottom salinity stratification and (C and D) density stratification (kg m�3) over the seasonal cycle in present-day (dashed) and future (dotted) model

scenarios, averaged over (A and C) Main Basin and (B and D) Strait of Juan de Fuca along the transect line shown in Fig. 1.

S.K. Moore et al. / Harmful Algae 48 (2015) 1–118

higher growth rates and a longer bloom season for Alexandrium

(Fig. 9). Increases in growth rates attained during the bloom seasonwere small (up to 0.05 d�1) but fairly uniform throughout all ofPuget Sound (Fig. 9A) and generally consistent throughout theduration of the bloom season (not shown). Increases in theduration of HAB-favorable conditions were more variable in PugetSound’s basins (Fig. 9B). The largest increases of up to 30 more daysof HAB-favorable conditions are projected to occur in North basin(including the eastern end of the Strait of Juan de Fuca) and in thefinger inlets of South basin, with moderate increases in the Main

Fig. 9. Maps of the potential changes to (A) Alexandrium growth rates as a function of

temperature and salinity, and (B) the seasonal duration of bloom-favorable

conditions in Puget Sound. Changes are calculated as the difference between the

present-day and future scenarios over the duration of the bloom season (May

through October).

basin and Hood Canal. The duration of HAB-favorable conditions isprojected to remain unchanged or increase by only a few days inmost of Whidbey basin with the exception of Skagit Bay which isprojected to experience a large increase in HAB-favorable days.

Smoothed daily values of Alexandrium growth rates averagedover the entire Puget Sound basin are shown in Fig. 10 for thepresent-day and future climate simulations. Growth rates inwinter are similarly low (<0.2 d�1) for the two simulations, butthey begin to diverge early in the spring. Not only are growth rateshigher for the future climate simulation, but they begin to increase�30 d earlier in the year. This is consistent with the findings fromthe heuristic modeling approach of Moore et al. (2011), indicatingan earlier start to the bloom season in the future. This increase inAlexandrium growth in the future is primarily driven by increasingSST. The mean values of SST and salinity during the bloom seasonin North, Whidbey, Main and South Basins of Puget Sound from the

Fig. 10. Daily mean growth rate of Alexandrium for the present-day (dashed) and

future (dotted) climate scenarios from the CCSM3-WRF-ROMS simulation. The

growth rate is averaged over the Puget Sound basin (122–123.258 W, 46.9–498 N)

and data are smoothed using a 31-day running mean (�15 d).

S.K. Moore et al. / Harmful Algae 48 (2015) 1–11 9

present-day and future climate simulations mapped onto thegrowth function for Alexandrium from Bill et al. (in review) inFig. 4. Increases in SST of �2 8C occur in each of the basins whereasnegligible or no changes are simulated for salinity. But moreimportantly, the increases in SST occur in a region of the growthfunction that is sensitive to change resulting in an increase inAlexandrium growth rates. The negligible changes in salinitycombined with the euryhaline nature of Puget Sound Alexandrium

result in the simulated growth rates from this study beinginsensitive to projected changes in salinity.

4. Discussion

We have conducted an analysis of regional atmosphere andocean modeling simulations to understand the linkages amongregional climate and Puget Sound oceanography in projectedclimate change. By focusing on a reduced set of climate pathwaysthat are most likely to impact Alexandrium in Puget Sound, wemechanistically evaluate the relative importance of greenhousegas driven changes to warming, stream flow, and upwelling onthe timing, duration, and extent of toxic HABs. Future warmer SSTdue primarily to increased direct heat input is the strongest driverof increased Alexandrium bloom risk in Puget Sound. BecauseAlexandrium are euryhaline, changes in salinity arising from earliersnow melt runoff contribute far less to projected changes in bloomrisk. However, the empirical growth model for Alexandrium that weuse is a function of SST and salinity values only and does notconsider changes to stratification. Increased stratification of thewater column may indirectly favor swimming dinoflagellates suchas Alexandrium because reduced vertical mixing decreases thenutrient supply to the surface waters, reducing the overall growthand biomass of phytoplankton in the surface waters (Behrenfeldet al., 2006). The ability to vertically migrate to deeper nutrient-rich waters may provide these species with a competitiveadvantage (Tozzi et al., 2004). More work is needed to quantita-tively evaluate the response of Puget Sound Alexandrium tostratification in relation to the rest of the phytoplanktoncommunity so that the potential effects of future changes canbe simulated.

Similarly, it is likely that changes in the timing of springstratification will lead to changes in the timing of the springphytoplankton bloom throughout Puget Sound, with unknownconsequences for total primary production and seasonal succes-sion. Spring stratification affects bloom timing by partiallyregulating the degree of light limitation that phytoplanktonexperience, but light limitation is controlled by other, potentiallyvariable factors as well. Changes in dissolved and particulatematter concentrations in the rivers could change the underwaterlight field. Changes in cloud cover could affect surface lightavailability, and indeed, data from moored Oceanic RemoteChemical Analyzer (ORCA) buoys in Puget Sound (not shown:http://orca.ocean.washington.edu/data.html) suggest that vari-ability in surface light in February–March may be important tointerannual variability in bloom timing at some stations. Theseinterannual observations do not address, however, the question ofwhether we should expect climate-timescale trends in spring lightavailability in this region. The climate model scenarios evaluatedhere show no change in the seasonality of surface light (i.e.,downward shortwave radiation) over the Puget Sound basin (46.9–498 N, 122–123.258 W), as opposed to the strong shift in theseasonality of stratification documented above (Fig. 8).

Present-day bloom season SSTs in Puget Sound lie in the rangeof 12–20 8C (Moore et al., 2008b). This is within the zone ofsensitivity for Alexandrium growth, such that an increase intemperature results in an increase in growth rate. This is incontrast to the salinity range of 20–30 which lies considerably

above the zone of sensitivity for Alexandrium growth. And becausePuget Sound Alexandrium are euryhaline, growth is relativelyinsensitive to changes in salinities except for changes that occur inthe fresher range of �10–20 (Fig. 4). Hence, the changes inAlexandrium bloom risk shown in Fig. 9 are almost entirely drivenby changes in SST. Warming of SSTs in the basins of Puget Sound isfairly consistent at �2 8C, producing a similarly consistent averageincrease in Alexandrium growth rates of �0.05 d�1 over theduration of the bloom season (Fig. 9A). This is a relatively smallincrease in growth rate. In the context of the simulated changes inSST and salinity in Whidbey basin (Fig. 4), an increase in growthrate from 0.25 to 0.30 d�1 translates into a decrease in generationtime from 2.77 to 2.31 d. The simulated increase in the duration ofHAB-favorable conditions of up to 30 d may be more importantthan the relatively small changes in growth rate for Alexandrium

bloom development in the future. Assuming no other limitations togrowth, this could allow blooms to develop earlier in the yearcompared to the present-day, and the longer period of time thatcells are exposed to more favorable SST conditions may result inhigher magnitude blooms. The biggest increase in the duration ofHAB-favorable conditions occurred in North basin and the Strait ofJuan de Fuca. Even though SST warmed by approximately the sameamount in North basin compared to the other basins of PugetSound, present-day values of SST were relatively cooler. Therefore,the increase in SST in this basin resulted in considerably more dayswhen growth rates could exceed the optimal growth thresholdof 0.25 d�1 for Alexandrium.

It is important to remember that this assessment of Alexandrium

bloom risk considers temperature and salinity conditions only.Other aspects of Alexandrium bloom ecology and oceanography,such as the location of cyst ‘seed beds’, the timing of cystgermination, physical transport of cells via mixing and currents,nutrient availability, competition with other phytoplanktonspecies, grazing, and infection by parasites, are not considered.With this in mind, the occurrence of a temperature and salinitywindow in space and time that is favorable for Alexandrium growthdoes not necessarily mean that Alexandrium blooms will occur.Rather, that the potential exists for bloom conditions to develop.Further, this assessment assumes that the growth response ofAlexandrium to temperature and salinity will remain unchangedand does not consider the potential for Alexandrium to adapt tochanging ocean conditions over long time-scales. This means thatthe growth response measured in present-day Alexandrium mightbe different from the growth response of Alexandrium in the future.Nevertheless, because temperature is such a strong determinant ofmany biological rates, this initial assessment of greenhouse gas-driven changes in Alexandrium bloom risk provides insight intopotentially important climate pathways that are relevant forAlexandrium bloom development in Puget Sound.

An increase in upwelling intensity, of a magnitude close tothe median of existing GCM projections (Fig. 3), does not appearto have a major direct effect on Puget Sound surface watertemperatures compared with concomitant changes in localatmospheric heat fluxes (Fig. 7). Likewise, the direct effect ofincreased upwelling on salinity (on the order of 0.1) is smallcompared with the changes that accompany the projected shift inthe timing of local river inputs, and smaller still in comparison withthe salinity sensitivity of Alexandrium. Still, this does not mean thatwe should discount changing ocean inputs as a potential driver offuture change in Puget Sound more generally, for two reasons.First, even small variations in temperature and salinity at depthmay indicate large changes in biogeochemical properties. Forexample, Davis et al. (2015) summarized the observed deepnitrate-salinity relationship in Washington coastal waters with apiecewise-linear fit that suggests 0.1–0.2 change in salinity near 34(i.e., matching the scale of change below 50 m in Juan de Fuca

S.K. Moore et al. / Harmful Algae 48 (2015) 1–1110

canyon and Strait of Juan de Fuca in our model) is looselyequivalent to a change of 3–7 mmol N m�3 in incoming nitrateconcentration, if the salinity change is interpreted as a change inthe depth from which incoming water is drawn. This is on the orderof 10–20% of the total nutrient supply in the Salish Sea (Mackas andHarrison, 1997), potentially enough to have effects on primaryproduction as a whole.

Second, remote upwelling-favorable winds farther south in theCalifornia Current System are responsible for a first-order fractionof summer upwelling in the PNW, and for setting seasonal waterproperties at slope depth via the California Undercurrent (Battistiand Hickey, 1984; Connolly et al., 2014). Likewise, modulationsof gyre-scale transport patterns such as the North Pacific GyreOscillation (NPGO; Di Lorenzo et al., 2008) are known to beimportant to interannual and decadal variation in source waterproperties as well. These processes occur on scales not resolved bythe nested model setup used in this study. If future change inupwelling-favorable winds are coherent between the PNW and therest of the California Current System, then the fact that our modelsetup neglects remote wind effects means that we have likelyunderestimated the effect of changes in the winds on incomingwater properties. Still, our conclusion that changes in the ocean arelikely to have only minor impacts on surface temperature andsalinity within Puget Sound, but potentially much greater impactson estuarine biogeochemistry, is unchanged.

The results from this study have allowed us to mechanisticallyexamine the potential climate pathways that are most likely toinfluence HABs of Alexandrium in Puget Sound in the future. Futurewarmer Puget Sound SSTs clearly increase the maximum growthrates that can be attained by Alexandrium during the bloom seasonas well as the number of days with conditions that are favorablefor bloom development. Changes in surface salinity arising fromchanges in the timing of riverflow have a negligible effect onAlexandrium growth rates; however, the resulting shift towardearlier stratification has strong potential to alter the seasonal cycleof phytoplankton in Puget Sound as a whole, with unknownconsequences for Alexandrium. The behavior of the coastal inputsin the simulations suggest that changes in local upwelling willnot have major effects on SST and surface salinity in Puget Sound,although changes in the nutrient supply delivered to Puget Soundfrom upwelled waters are likely, with unknown effects onAlexandrium growth. Changing nutrient concentrations may affectAlexandrium growth directly, or changes in the relative concentra-tions of nutrients, such as the nitrogen to silica ratio, may influencethe structure of the phytoplankton community with implicationsfor Alexandrium.

Acknowledgments

The authors thank Nathan J. Mantua for guidance during theearly stages of this project. Thanks also to Corinne Bassin, NicholasLederer, and Samantha Siedlecki for their essential help with theocean simulations and observational database, and Sarah Giddingsfor help with model intercomparisons. This research was funded inpart by a grant from the NOAA Ecology and Oceanography ofHarmful Algal Blooms (ECOHAB) Program to the NOAA NorthwestFisheries Science Center and ECOHAB GRANT10443455 and NOAAgrant NA10NOS4780158 to the University of Washington. This isECOHAB contribution number 807.

References

Alford, M.H., MacCready, P., 2014. Flow and mixing in Juan de Fuca Canyon,Washington. Geophys. Res. Lett. 41, 1608–1615.

Anderson, C.R., Moore, S.K., Tomlinson, M.C., Silke, J., Cusack, C., 2014. Living withharmful algal blooms in a changing world: strategies for modeling and miti-gating their effects in coastal marine ecosystems. In: Ellis, J.T., Sherman, D.J.

(Eds.), Coastal and Marine Hazards, Risks, and Disasters, Hazards and DisastersSeries. Elsevier, pp. 495–561.

Anderson, D.M., 1998. Physiology and bloom dynamics of toxic Alexandrium species,with emphasis on life cycle transitions. In: Anderson, D.M., Cembella, A.D.,Hallegraeff, G.M. (Eds.), Physiological Ecology of Harmful Algal Blooms. Spring-er-Verlag, Berlin, pp. 29–48.

Anderson, D.M., Cembella, A.D., Hallegraeff, G.M., 2012. Progress in understandingharmful algal blooms: paradigm shifts and new technologies for research,monitoring, and management. Ann. Rev. Mar. Sci. 4, 143–176.

Anderson, D.M., Glibert, P.M., Burkholder, J., 2002. Harmful algal blooms andeutrophication: nutrient sources, composition, and consequences. Estuaries25 (4b), 704–726.

Banas, N.S., Conway-Cranos, L., Sutherland, D.A., MacCready, P., Kiffney, P., Plum-mer, M., 2014. Patterns of river influence and connectivity among subbasins ofPuget Sound, with application to bacterial and nutrient loading. EstuariesCoasts, http://dx.doi.org/10.1007/s12237-12014-19853-y (online).

Barron, C.N., Kara, A.B., Martin, P.J., Rhodes, R.C., Smedstad, L.F., 2006. Formulation,implementation and examination of vertical coordinate choices in the GlobalNavy Coastal Ocean Model (NCOM). Ocean Model. 11, 347–375.

Battisti, D.S., Hickey, B.M., 1984. Application of remote wind-forced coastaltrapped wave theory to the Oregon and Washington coasts. J. Phys. Oceanogr.14, 887–903.

Behrenfeld, M.J., O’Malley, R.T., Siegel, D.A., McClain, C.R., Sarmiento, J.L., Feldman,G.C., Milligan, A.J., Falkowski, P.G., Letelier, R.M., Boss, E.S., 2006. Climate-driventrends in contemporary ocean productivity. Nature 444 , http://dx.doi.org/10.1038/nature05317.

Bill, B.D., Moore, S.K., Hay, L.R., Anderson, D.M., Trainer, V.L., 2015. Effects oftemperature and salinity on growth of Alexandrium (Dinophyceae) isolatesfrom the Salish Sea. J. Phycol. (in review).

Brand, L.E., Guillard, R.R.L., Murphy, L.S., 1981. A method for the rapid and precisedetermination of acclimated phytoplankton reproduction rates. J. Plankton Res.3 (2), 193–201.

Cokelet, E.D., Stewart, R.J., Ebbesmeyer, C.C., 1991. Concentrations and ages ofconservative pollutants in Puget Sound. In: Puget Sound Research ‘91, PugetSound Water Quality Authority, pp. 99–108.

Collins, W.D., Bitz, C.M., Blackmon, M.L., Bonan, G.B., Bretherton, C.S., Carton, J.A.,Chang, P., Doney, S.C., Hack, J.J., Henderson, T.B., Kiehl, J.T., Large, W.G.,McKenna, D.S., Santer, B.D., Smith, R.D., 2006. The Community Climate SystemModel version 3 (CCSM3). J. Clim. 19, 2122–2143.

Connolly, T.P., Hickey, B.M., 2014. Regional impact of submarine canyons duringseasonal upwelling. J. Geophys. Res. C 119, 953–975.

Connolly, T.P., Hickey, B.M., Shulman, I., Thomson, R.E., 2014. Coastal trappedwaves, alongshore pressure gradients, and the California undercurrent. J. Phys.Oceanogr. 44, 319–342.

Cuo, L., Beyene, T.K., Voisin, N., Su, F., Lettenmaier, D.P., Alberti, M., Richey, J.E., 2011.Effects of mid-twenty-first century climate and land cover change on the hydrol-ogy of the Puget Sound basin, Washington. Hydrol. Process. 25, 1729–1753.

Davis, K.A., Banas, N.S., Giddings, S.N., Siedlecki, S.A., MacCready, P., Lessard, E.J.,Kudela, R.M., Hickey, B.M., 2015. Estuary-enhanced upwelling of marine nutri-ents fuels coastal productivity in the U.S. Pacific Northwest. J. Geophys. Res. C119, 8778–8799, http://dx.doi.org/10.1002/ 2014JC010248.

Di Lorenzo, E., Schneider, N., Cobb, K.M., Franks, P.J.S., Chhak, K.C., Miller, A.J.,McWilliams, J.C., Bograd, S.J., Arango, H., Curchitser, E.N., Powell, T.M., Riviere,P., 2008. North Pacific gyre oscillation links ocean climate and ecosystemchange. Geophys. Res. Lett. 35, L08607.

Duliere, V., Zhang, Y.X., Salathe Jr., E.P., 2011. Extreme precipitation and temperatureover the U.S. Pacific Northwest: a comparison between observations, reanalysisdata, and regional models. J. Clim. 24, 1950–1964, http://dx.doi.org/10.1175/2010jcli3224.1951.

Ebbesmeyer, C.C., Chiang, R., Copping, A., Erickson, G.M., Horner, R.A., Ingraham Jr.,W.J., Nishitani, L., 1995. Decadal covariations of Sequim Bay Paralytic ShellfishPoisoning (PSP) with selected Pacific Northwest environmental parameters. In:Robichaud, E. (Ed.), Puget Sound Research Proceedings 1. Puget Sound WaterQuality Authority. Olympia, Washington, pp. 415–421.

Elsner, M.M., Cuo, L., Voisin, N., Deems, J.S., Hamlet, A.F., Vano, J.A., Mickelson, K.E.B.,Lee, S.-Y., Lettenmaier, D.P., 2010. Implications of 21st century climate changefor the hydrology of Washington State. Clim. Change 102, 225–260.

Geyer, W.R., Cannon, G.A., 1982. Sill processes related to deep water renewal in afjord. Geophys. Res. Lett. 87, 7985–7996.

Giddings, S.N., MacCready, P., Hickey, B.M., Banas, N.S., Davis, K.A., Siedlecki, S.A.,Trainer, V.L., Kudela, R.M., Pelland, N.A., Connolly, T.P., 2015. Hindcasts of poten-tial harmful algal bloom transport pathways on the Pacific Northwest coast.J. Geophys. Res. C 119, 2439–2461, http://dx.doi.org/10.1002/2013JC009622.

Glibert, P.M., Anderson, D.M., Gentien, P., Graneli, E., Sellner, K.G., 2005. The globalcomplex phenomena of harmful algae. Oceanography 18 (2), 136–147.

Hallegraeff, G.M., 1993. A review of harmful algae blooms and their apparent globalincrease. Phycologia 32 (2), 79–99.

Hallegraeff, G.M., 2010. Ocean climate change, phytoplankton communityresponses, and harmful algal blooms: a formidable predictive challenge.J. Phycol. 46, 220–235, http://dx.doi.org/10.1111/j.1529-8817.2010.00815.x.

Hickey, B.M., 1989. Patterns and processes of circulation over the Washingtoncontinental shelf and slope. In: Landry, M.R., Hickey, B.M. (Eds.), CoastalOceanography of Washington and Oregon. Elsevier, pp. 41–115.

John, U., Litaker, W., Montresor, M., Murray, S., Brosnahan, M.L., Anderson, D.M.,2014. Proposal to reject the name Gonyaulax catenella (Alexandrium catenella)(Dinophyceae). Taxon 63 (4), 932–933.

S.K. Moore et al. / Harmful Algae 48 (2015) 1–11 11

Johnstone, J.A., Mantua, N.J., 2015. Atmospheric controls on northeast Pacifictemperature variability and change, 1900–2012. Proc. Natl. Acad. Sci. U. S. A.111 (40), 14360–14365, http://dx.doi.org/10.1073/pnas.131837111.

Kao, C.Y., 1993. Paralytic shellfish poisoning. In: Falconer, I.R. (Ed.), Algal Toxins inSeafood and Drinking Water. Academic Press, London.

Lilly, E.L., Halanych, K.M., Anderson, D.M., 2007. Species boundaries and globalbiogeography of the Alexandrium tamarense complex (Dinophyceae). J. Phycol.43, 1329–1338.

Mackas, D.L., Harrison, P.J., 1997. Nitrogenous nutrient sources and sinks in the Juande Fuca Strait/Strait of Georgia/Puget Sound estuarine system: assessing thepotential for eutrophication. Estuar. Coast. Shelf Sci. 44 (4), 1–21.

Mantua, N.J., Hare, S.R., 2002. The Pacific decadal oscillation. J. Oceanogr. 58, 35–44.Marsland, S.J., Haak, H., Jungclaus, J.H., Latif, M., Roske, F., 2003. The Max Planck

Institute global ocean/sea ice model with orthogonal curvilinear coordinates.Ocean Model. 5, 91–127.

McGillicuddy Jr., D.J., Anderson, D.M., Lynch, D.R., Townsend, D.W., 2005. Mecha-nisms regulating large-scale seasonal fluctuations in Alexandrium fundyensepopulations in the Gulf of Maine: results from a physical-biological model.Deep-Sea Res. Pt. II 52, 2698–2714.

Moore, S.K., Mantua, N.J., Hickey, B.M., Trainer, V.L., 2010. The relative influences ofEl Nino Southern Oscillation and Pacific Decadal Oscillation on paralytic shell-fish toxin accumulation in Pacific Northwest shellfish. Limnol. Oceanogr. 6 (55),2262–2274, http://dx.doi.org/10.4319/lo.2010.2255.2266.2262.

Moore, S.K., Mantua, N.J., Kellogg, J.P., Newton, J.A., 2008a. Local and large-scaleclimate forcing of Puget Sound oceanographic properties on seasonal to inter-decadal timescales. Limnol. Oceanogr. 53 (5), 1746–1758.

Moore, S.K., Mantua, N.J., Newton, J.A., Kawase, M., Warner, M.J., Kellogg, J.P., 2008b.A descriptive analysis of temporal and spatial patterns of variability in PugetSound oceanographic properties. Estuar. Coast. Shelf Sci. 80, 545–554, http://dx.doi.org/10.1016/j.ecss.2008.1009.1016.

Moore, S.K., Mantua, N.J., Salathe Jr., E.P., 2011. Past trends and future scenarios forenvironmental conditions favoring the accumulation of paralytic shellfishtoxins in Puget Sound shellfish. Harmful Algae 10 (5), 521–529, http://dx.doi.org/10.1016/j.hal.2011.1004.1004.

Moore, S.K., Mantua, N.J., Trainer, V.L., Hickey, B.M., 2009. Recent trends in paralyticshellfish toxins in Puget Sound, relationships to climate, and capacity forprediction of toxic events. Harmful Algae 8 (3), 463–477, http://dx.doi.org/10.1016/j.hal.2008.1010.1003.

Moore, S.K., Trainer, V.L., Mantua, N.J., Parker, M.S., Laws, E.A., Fleming, L.E., Backer,L.C., 2008c. Impacts of climate variability and future climate change on harmfulalgal blooms and human health. Environ. Health 7 (S4 (Suppl. 2)), S1182–S1184,http://dx.doi.org/10.1186/1476-1069X-1187.

Mote, P.W., Salathe Jr., E.P., 2009. Scenarios: future climate in the Pacific Northwest.In: McGuire Elsner, M., Littell, J., Whitely Binder, L. (Eds.), The WashingtonClimate Change Impacts Assessment: Evaluating Washington’s Future in aChanging Climate.The Climate Impacts Group, Center for Science in the EarthSystem, Joint Institute for the Study of the Atmosphere and Oceans, Universityof Washington, Seattle, Washington, pp. 21–43, Available at http://www.cses.washington.edu/db/pdf/wacciareport681.pdf.

Nakicenovic, N., Alcamo, J., Davis, G., Vries, B.D., Fenhann, J., Gaffin, S., Gregory, K.,Grubler, A., Jung, T.Y., Kram, T., Rovere, E.L.L., Michaelis, L., Mori, S., Morita, T.,Pepper, W., Pitcher, H., Price, L., Riahi, K., Roehrl, A., Rogner, H.-H., Sankovski, A.,Schlesinger, M., Shukla, P., Smith, S., Swart, R., Rooijen, S.V., Victor, N., Dadi, Z.,2000. In: Nakicenovic, N., Swart, R. (Eds.), IPCC Special Report on EmissionsScenarios. Cambridge University Press.

Nishitani, L., Chew, K.K., 1984. Recent developments in paralytic shellfish poisoningresearch. Aquaculture 39, 317–329.

Norris, L., Chew, K.K., 1975. Effect of environmental factors on growth of Gonyau-lax catenella. In: LoCicero, V.R. (Ed.), Proceedings of the First InternationalConference on Toxic Dinoflagellate Blooms, November, 1974, Boston, Mas-sachusetts. Massachusetts Science and Technology Foundation, Wakefield,MA, pp. 143–152.

Quayle, D.B., 1969. Paralytic shellfish poisoning in British Columbia. Fish Res. BoardCan. Bull. 168, 1–68.

Rensel, J., 1993. Factors controlling paralytic shellfish poisoning (PSP) in PugetSound, Washington. J. Shellfish Res. 12 (2), 371–376.

Roeckner, E., Bauml, G., Bonaventura, L., Brokopf, R., Esch, M., Giorgetta, M.,Hagemann, S., Kirchner, I., Kornblueh, L., Manzini, E., Rhodin, A., Schlese, U.,Schulzweida, U., Tompkins, A., 2003. The Atmospheric General CirculationModel ECHAM5-Part I: Model description. Tech. Rep. 349. Max-Planck-Institutfur Meteorologie, Hamburg, Germany.

Roeckner, E., Bengtsson, L., Feichter, J., Lelieveld, J., Rodhe, H., 1999. Transientclimate change simulations with a coupled atmosphere–ocean GCM includingthe tropospheric sulfur cycle. J. Clim. 12, 3004–3032.

Salathe Jr., E.P., Hamlet, A.F., Mass, C.F., Stumbaugh, M., Lee, S.-Y., Steed, R., 2015.Estimates of 21st century flood risk in the Pacific Northwest based on regionalscale climate model simulations. J. Hydrometeorol. 15, 1881–1899, http://dx.doi.org/10.1175/JHM-D-1113-0137.1171.

Salathe Jr., E.P., Leung, L.R., Qian, Y., Zhang, Y., 2010. Regional climate modelprojections for the State of Washington. Clim. Change 102, 51–75.

Shchepetkin, A.F., McWilliams, J.C., 2005. The regional oceanic modeling system(ROMS): a split-explicit, free-surface, topography-following-coordinate oceanicmodel. Ocean Model. 9, 347–404.

Shrestha, R.R., Schnorbus, M.A., Werner, A.T., Berland, A.J., 2012. Modellingspatial and temporal variability of hydrologic impacts of climate changein the Fraser River basin, British Columbia, Canada. Hydrol. Process. 26,1840–1860.

Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Duda, M.G., Huang,X.-Y., Wang, W., Powers, J.G., 2008. A Description of the Advanced Research WRFVersion 3. NCAR/TN-475 + STR, NCAR Technical Note. Mesoscale and MicroscaleMeteorology Division, National Center for Atmospheric Research, Boulder, CO,USA, pp. 113.

Sutherland, D.A., MacCready, P., Banas, N.S., Smedstad, L.F., 2011. A model study ofthe Salish Sea estuarine circulation. J. Phys. Oceanogr. 41, 1125–1143.

Tozzi, S., Schofield, O., Falkowski, P., 2004. Historical climate change and oceanturbulence as selective agents for two key phytoplankton functional groups.Mar. Ecol. Prog. Ser. 274, 123–132.

Trainer, V.L., Eberhart, B.-T.L., Wekell, J.C., Adams, N.G., Hanson, L., Cox, F., Dowell, J.,2003. Paralytic shellfish toxins in Puget Sound, Washington State. J. ShellfishRes. 22 (1), 213–223.

Wang, M., Overland, J.E., Bond, N.A., 2010. Climate projections for selected largemarine ecosystems. J. Mar. Syst. 79, 258–266.

Watras, C.J., Chisholm, S.W., Anderson, D.M., 1982. Regulation of growth in anestuarine clone of Gonyaulax tamarensis Lebour: salinity-dependent tempera-ture responses. J. Exp. Mar. Biol. Ecol. 62, 25–37.

Willmott, C.J., 1982. Some comments on the evaluation of model performance. Bull.Am. Meteorol. Soc. 63, 1309–1313.

Zhang, Y., Duliere, V., Mote, P.W., Salathe Jr., E.P., 2009. Evaluation of WRF andHadRM mesoscale climate simulations over the U.S. Pacific Northwest. J. Clim.22, 5511–5526.