Nonnative Spartina species (cordgrasses) are widely distributed along the West Coast of North America, but have not invaded all bays with susceptible habitat. We used drift cards to assess the patterns and rates of potential Spartina propagule dispersal by ocean currents from estuaries with significant populations of one or more Spartina species. Cards were released monthly for 1 yr from Willapa Bay, Washington; Humboldt Bay, California; and San Francisco Bay, California; with recovery information reported by volunteers. Recovery rates averaged 37% for all release sites. Cards were commonly recovered close to their bay of release but were repeatedly found hundreds of kilometers both north and south of their release location. Cards most generally traveled northward from the release sites. Cards from Humboldt and Willapa bays were commonly recovered along the British Columbia coast, particularly in the winter. Cards released from Humboldt Bay were found farthest from their release sites. One card from Humboldt Bay traveled 2,800 km to Kodiak Island, Alaska. The timing of seed production, combined with prevailing currents, puts bays currently uninfested by Spartina at risk of repeated propagule loading by ocean currents. A coordinated coast-wide strategy for eradication of all nonnative Spartina will be critical to the success of individual bay-wide eradication efforts.
Nomenclature: Common cordgrass, Spartina anglica C. E. Hubbard, dense-flowered cordgrass, Spartina densiflora Brongn., saltmeadow cordgrass, Spartina patens (Ait.) Muhl., smooth cordgrass, Spartina alterniflora Loisel
Management Implications: Nonnative estuarine cordgrasses (Spartina spp.) are invasive weeds of saltmarshes and open mudflats in many bays and estuaries of North America, Australia, China, India, New Zealand, and numerous European countries. In North America the largest core infestations established along the Pacific Coast during the last 150 yr at Willapa Bay, Washington; Humboldt Bay, California; and San Francisco Bay, California. Estuarine plants such as Spartina can disperse to new bays when propagule-laden wrack mats consisting of senescent stems are washed out to the ocean and then drift with the currents. Understanding this natural dispersal mechanism is critical to the success of a coast-wide Spartina eradication strategy because there is a high likelihood of secondary spread from existing infestations—especially those with large historic propagule production. Indeed, a marked increase in novel satellite populations in the last 12 yr suggests spread to new habitats is increasing. This study provides the first examination of the dispersal potential for weedy plant species within the near-shore ocean environment from these three estuaries. The recovery patterns of drift cards, released monthly over the course of 1 yr underscore increased propagule loading at areas proximal to existing infestations, but also highlight the ability of ocean currents to carry seeds great distances, well within known propagule viability ranges. Drift card recoveries spanned 2,044 km north and south of Willapa Bay, 3,304 km north and south of Humboldt Bay, and 661 km north and south of San Francisco Bay. The scope of potential ocean dispersal presented here reinforces the need for effective eradication programs in propagule source areas of not only Spartina spp., but also other weeds that can disperse along this same pathway such as dwarf eelgrass (Zostera japonica), perennial pepperweed (Lepidium latifolium), and Algerian sea lavender (Limonium ramosissimum).
Humans often play a role in long-distance dispersal of nonnative species (Mack and Lonsdale 2001; Suarez et al. 2001; Wonham and Carlton 2005), but after a population becomes established in a new range, natural dispersal mechanisms such as wind, water, or animals can facilitate further spread or secondary infestations (Moody and Mack 1988; Ridley 1930). Rare dispersal events via natural pathways might explain Reid's paradox, the disparity between real-world rates of spread and models' predictions based on average dispersal rates (Mack and Lonsdale 2001). Even small proportions of seeds traveling long distances could have significant implications for predicted dispersal rates. Understanding potential dispersal patterns can highlight areas at risk for invasion, improve early detection efforts, and facilitate management or eradication efforts (Higgins and Richardson 1999).
Dispersal of invasive cordgrasses (Spartina spp.) along the west coast of North America can occur through long-distance transport of propagules on ocean currents. Often considered ecological engineers, estuarine Spartina spp. successfully colonize low-gradient mudflats and existing salt marsh (Bouma et al. 2005; Crooks 2002; Odum and Odum 2003) in areas subject to tidal action and sheltered from wave action (Daehler and Strong 1996). High stem densities and thick rhizome mats increase sediment accumulation (Ranwell 1964; Sayce 1988; Thompson 1991), reduce shorebird foraging habitat (Goss-Custard and Moser 1988; Patten and O'Casey 2007; Stralberg et al. 2004), alter benthic communities (Chen et al. 2007; Levin et al. 2006; Neira et al. 2006), and might facilitate the establishment of other nonnative species (Carr and Dumbauld 2000).
Four nonnative species of Spartina are established in North America, Europe, Asia, South Africa, New Zealand, and Australia due to intentional plantings for marsh reclamation and forage, and through accidental transport of seed as a contaminant in shipments of other species (Crawford 2008; Mobberly 1956; Strong and Ayres 2009). Initial introductions of nonnative Spartina on the west coast of North America occurred 100 to 150 yr ago (Civille et al. 2005; Strong and Ayres 2009). A three-fold increase in novel infestations documented over the last 12 yr suggests the rate of spread to new habitats is increasing. By 1996, five infestation sites were documented, including: Spartina anglica C. E. Hubbard in Puget Sound, Washington; Spartina alterniflora Loisel. in Willapa Bay, Washington; S. alterniflora and Spartina patens (Ait.) Muhl. in the Siuslaw River estuary, Oregon; Spartina densiflora Brongn. in Humboldt Bay, California; and all four species in San Francisco Bay, California (Frenkel and Boss 1988; Daehler and Strong 1996). Hybrids formed between the introduced S. alterniflora and the native Spartina foliosa Trin. now dominate San Francisco Bay, where clones of either parent species are increasingly difficult to find (Ayres et al. 2008). Sixteen new infestations were documented between 1996 and 2008 in British Columbia (4), Washington (5), Oregon (2) and California (5) by November 2008 (Howard et al. 2007; WSDA 2009) (Figure 1).
Figure 1.
Confirmed infestations of one or more invasive Spartina species and bays/estuaries predicted to have susceptible habitat on the West Coast of North America (Daehler and Strong 1996; Howard et al. 2007; WSDA 2009).

Because Spartina seeds float (Huiskies et al. 1995) and require periods of wet, cool conditions in order to germinate successfully (Kittelson and Boyd 1997; Plyler and Carrick 1993; Plyler and Proseus 1996), natural dispersal via ocean currents can lead to secondary invasions (Ayres et al. 2004; Sayce et al. 1997). Spartina sets seed in the late summer/fall and then senescing stems and inflorescences form floating rafts of wrack in mid- to late fall. These materials float within an estuary and can be carried on outgoing tides into the near-shore ocean zone (Minchinton 2006; Sayce et al. 1997). Waterlogging can cause loss of buoyancy; however, large wrack mats were reported by commercial fishermen in the open ocean off the Washington coastline (M. Pfauth, personal communication). During the winter months, the poleward Davidson Current, driven by south winds, typically occurs near shore in the eastern Pacific, whereas in the summer and during periods of fair weather, the equatorward California Current, driven by north winds, is more common (Banas et al. 2004; Hickey 1989). Biber and Caldwell (2008) showed germination of S. alterniflora seed peaked after 4 mo of cold, wet stratification, but also that even after 9 to 10 mo of stratification, over 10% of developed seeds remain viable. Similarly, when placed in comparable storage conditions, 5% of S. anglica seed remained viable for 300 d and some as long as 4 yr (Hubbard 1970) and large proportions of S. densiflora seed remain viable for 4 to 7 mo (Bortolus 2006; Kittelson and Boyd 1997).
Buoyant drift bottles and cards have been used previously to investigate near-shore surface currents (Chew et al. 1962; Dewees and Strange 1984; Schwartzlose 1962), shoreline accumulation of floating materials and pollutants (Ebbesmeyer and Coomes 1993; Schwartzlose and Reid 1972), and the fate of sewage effluent (Crone et al. 1998). They have also been used to investigate the dispersal of abalone (Chambers et al. 2003; Tegner and Butler 1985) and nonnative mangroves in Hawaii (Allen and Krauss 2006), the effect of wind regimes on sea turtle strandings (Hart et al. 2006), and the connectivity of marine animal populations through larval dispersal to inform siting of Marine Protected Areas (Klinger and Ebbesmeyer 2001; Sauers et al. 2003).
We used drift cards to mimic wrack or floating seed of Spartina to study natural, ocean dispersal. Our goal was to assess the distance and velocity of propagule dispersal by ocean currents from core infestations of S. alterniflora, S. densiflora, and the hybrid Spartina foliosa × alterniflora on the west coast of North America. We examined spatial patterns of drift card recoveries and direction and distance from selected release sites to better understand ocean transport of materials, including Spartina wrack and seed from infested estuaries. We also compared our drift card results to previous reports of near-shore ocean dynamics off the west coast of North America.
Materials and Methods
Drift cards measured 15 by 10 by 0.3 cm (5.9 by 3.9 by 0.1 in) and were constructed from low-density (approximately 0.61 g cm−3 (0.35 oz in−3) plywood painted bright yellow with nontoxic paint to facilitate detection (City Signs, Inc., Poulsbo, WA). Biodegradable materials were used with the expectation that they would degrade within a few months in the marine environment. The cards were similar to cards used for a study of ocean currents near Oahu, HI (NOAA 2010). We assigned each batch of cards a unique number denoting release date and site. Text printed on the cards instructed people to report the date and location they found the card along with the card number. Reporting options included telephone (local and toll free), email and postal service. Phone and electronic reports of card recoveries were collected for 5 mo after the last release was conducted.
Drift card deployments, conducted between September 2004 and August 2005, entailed pitching 200 cards from a boat or shoreline once per month within the first 2 h of an ebb tide. Deployments spanned the weak 2004 to 2005 El Niño (Lee and McPhaden 2010). We selected release bays based on the presence of large scale Spartina infestations (hundreds to thousands of acres). Release sites were close to each bay mouth; in Willapa Bay (WB), they took place southeast of Tokeland (46.6958°N, 123.9599°W), in Humboldt Bay (HB) from the bay-ward side of a narrow jetty channel (40.7543°N, 124.2305°W); and in San Francisco Bay (SF) from the shoreline at Fort Point (37.8110°N, 122.4772°W) (Figure 2).
Figure 2.
Locations of all in-bay drift card recoveries from releases sites at (a) Willapa, (b) Humboldt and (c) San Francisco bays. Stars indicates release site.

We plotted coordinates of all recoveries using ArcMap™ 9.1 (ESRI, Redlands, CA) and classified each as either “in-bay” or “coastal.” “In-bay” recoveries occurred within the bay of release. “Coastal” recoveries include those reported from beaches and rocky shorelines, and those from embayments other than the card's original release location.
Because our primary interest was to assess risk of spread to distant locations, in-bay recoveries were used to measure total recovery rate, but were excluded from further analysis. For all coastal recoveries, we used Jenness Enterprises' Distance and Azimuth Matrix 2.0 (Jenness Enterprises, Flagstaff, AZ) to calculate the direction of travel and straight-line distance to each recovery location (Jenness 2005). These calculations were performed on unprojected data and based on the “Great Arc” method for calculating distance across a curved surface. We estimated velocity by assuming a straight-line of travel and no lag-time between actual stranding and recovery; both assumptions contribute to conservative estimates of velocity.
Results and Discussion
Of the 7,198 cards released in the three bays, 37.3% (2,683) were recovered by February 2006. Of these, 825 (30.7%) were found at in-bay locations and 1,858 (69.3%) were found in coastal areas (Table 1). Mean monthly retrieval rates were 41.8%, 43.8%, and 26.2% for WB, SF, and HB, respectively. In-bay recoveries regularly comprised over one-third of the total monthly SF recoveries (Figure 3). In WB, greater than 69% of the total recoveries in 4 different mo occurred within the bay. In HB just 1.1% of the 629 total recovered cards were found in the confines of the bay.
Figure 3.
Coastal (white) and in-bay (black) recoveries of drift cards released between September 2004 and August 2005 at the mouth of (a) Willapa, (b) Humboldt, and (c) San Francisco bays.

Table 1.
The number of drift cards recovered inside and outside the bay of release by bay, month of release, and location. A total of 200 cards were released each month from each bay.a

In-bay cards were repeatedly stranded in the same locations over multiple months of releases (Figure 2). The majority of SF in-bay recoveries were from the Berkeley/Alameda shoreline, due east of the deployment location at Fort Point, or along Chrissie Field Beach. WB in-bay recoveries often occurred at North Cove, Tokeland, and Bruceport Park. In both bays, repeated stranding sites typically already had established Spartina stands or were areas where Spartina wrack accumulated, but was habitat unsuitable for Spartina establishment and/or survival. After eliminating in-bay recoveries, the number of coastal cards recovered by bay of release was not significantly different (ANOVA, P = 0.951, α = 0.05, df = 2) and ranged from 596 to 640. All results discussed hereafter refer to coastal cards only.
Cards were frequently recovered less than 25 km away from release sites, often within 1 to 6 d after release (DAR). Cumulative return rates for all 12 releases were similar between bays immediately following releases and after approximately 6 mo, but showed some disparities in the intermediate time range (Figure 4). Nearly half of the total cards recovered were found within 10 DAR and 96 to 99% within 180 DAR. Around 45 DAR the patterns from WB and SF remain similar, with 82% and 84% of total cards recovered, compared to only 72% from HB releases. After 100 DAR, recoveries from SF releases slowed considerably (one card found every 2 wk) whereas reports of cards from both WB and HB continued to be reported every 4 to 5 d.
Figure 4.
Cumulative recovery of drift cards as a function of the number of days after release (DAR) from 12 monthly releases in three bays.

WB-released cards were predominantly transported to the north; 88% of total coastal recoveries occurred to the north of the release site. Beaches from Grayland north to Moclips, and the south-facing coast of Vancouver Island were frequent collection zones for drift cards (Figure 5). Eleven cards were found inside of Grays Harbor just north of WB; the majority of these were deployed in October 2004. Northward transport was greatest during winter months (December to February) when the one card was reported over 2,000 km from WB at Middleton Island (129 DAR). The highest velocity of the study was estimated at over 36 km d−1 (found at Carmanah Beach, Vancouver Island, British Columbia, Canada, 221 km, 6 DAR). Many winter-released cards were also rapidly transported to SE Alaska, and the Queen Charlotte Islands and northern tip of Vancouver Island. For all other seasons, distances and estimated velocities were considerably less (Table 2). The furthest southward transport occurred during spring months (March to May), with cards recovered in Oregon as far as Newport (185 DAR) and Florence (227 DAR), but just 10% of recovered cards were carried south of WB during this period. Summer-released (June to August) cards were not carried far from WB with a total range of less than 200 km.
Figure 5.
Recovery locations for drift cards grouped into fall (September to November), winter (December to February), spring (March to May), and summer (June to August) deployments and released from Willapa Bay, Washington; Humboldt Bay, California; and San Francisco Bay, California. Recovery locations were recorded for 5 mo past the last deployment in August 2005. Individual dots can represent more than one recovered drift card.

Table 2.
Recoveries of drift cards found north and south of three release sites grouped by season of release (F, fall; W, winter; Sp, spring; Su, summer) with the mean, maximum (max), and minimum (min) distances (dist) to recovery location in km, and estimated velocity (vel) in km d−1.

HB-released cards traveled both north and south in relatively even proportions over the course all 12 deployments (59% vs. 41% respectively), with card recoveries widely scattered south as far as Asilomar State Beach, Pacific Grove, CA (36.63°N, 45 DAR) and north to the Ruby Beach Trail in Washington (47.71°N, 140 DAR) (Figure 5). Winter releases were heavily skewed to the north, with 27 cards transported up to 24.5 km d−1 reported at Fort Canby State Park Beach, Washington (613 km, 25 to 26 DAR), and recoveries from the Queen Charlotte Islands (1,360 to 1,607 km, 67 to 256 DAR) and Vancouver Island in British Columbia (937 to 1,159 km, 52 to 62 DAR). The maximum distance reported from this study was one card from the December HB release that was transported over 2,800 km in 214 DAR to the Uganik Island lagoon on Kodiak Island, Alaska (57.82°N, 153.51°W). Seasonal mean velocities to the north were approximately three to four times those to the south during fall, winter, and spring releases for HB. Velocities were lower for summer releases when a greatly reduced range (< 55 km), similar to that observed for WB releases, occurred. Three cards from September 2004 and August 2005 HB deployments were recovered inside the Eel River estuary, 15 km to the south.
A majority (71%) of recovered SF-released cards were carried north of the bay mouth for all 12 deployments. Over 90% of the northbound cards were reported within 50 km of the release site, along the Marin Headlands and the Point Reyes National Seashore, whereas the remaining 10% were found at scattered locations up to Cape Arago, Oregon (632 km, 93 DAR) (Figure 5). Many of the southward recoveries were similarly concentrated close to the bay mouth with a few isolated reports from Moss Landing (140 km, 24 DAR) and Asilomar State Beach, Pacific Grove, CA (129 km, 10 to 27 DAR) within Monterey Bay. The maximum northward velocity (16 km d−1) of SF-released cards occurred in winter and summer deployments. The maximum southward rate of 14 km d−1 occurred following spring releases.
We demonstrated that rapid and long-distance coastal movement of Spartina on ocean currents is possible. We found average poleward velocities between 2 to 7 km d−1 (2.3 to 8 cm sec−1) and peak velocities of 24.5 to 36.9 km d−1 (28.4 to 42.7 cm sec−1). Velocities for the California Current have been reported between 4 to 50 cm sec−1 with typical mean speeds around 10 cm sec−1 (Hickey and Banas 2003; Jennings and Schwartzlose 1960; Lynn and Simpson 1987; Poulain and Niiler 1989; Reid and Schwartzlose 1962; Schwartzlose 1962, Schwartzlose and Reid 1972;). Our estimated mean velocities were within the known current velocities of the California Current, which suggests that our simple drift cards provided a realistic estimate of potential dispersal of Spartina. Drift cards were buoyant due to their construction from low-density (approximately 0.61 g cm−3) materials. Fall-gathered wrack of Spartina densiflora and S. foliosa × alterniflora were approximately one third less dense (0.38 and 0.41 g cm−3 respectively) than the cards; thus, wrack would therefore be comparatively more buoyant than the cards, encountering more wind-forcing and potentially travel farther than the drift cards suggest. Sayce et al. (1997) demonstrated that for S. alterniflora, large proportions of vegetative shoots were capable of floating for more than 2 mo, but seeds and inflorescences with ripe seed attached floated for less than 1 mo. Prior to control program successes, large wrack mats were regularly stranded on shorelines just outside Willapa Bay and San Francisco Bay (F. Grevstad and J. Graves, personal communication). During storm conditions, these mats would presumably break apart, reducing their rafting capability and causing seed to sink more rapidly (K. Sayce, personal communication). Thus, turbulent surface water might lessen the chances of propagules traveling long distance via ocean currents.
We observed evidence of transport from three bays sufficient to consistently convey floating material to nearby locations and often to distant sites via ocean currents. Releases at HB resulted in the greatest transport range, a span of 3,300 km of the coastline. Movement was predominantly northward from September through February when S. densiflora seed and wrack would be expected to be present in the bay. Seed production by this species is highly variable, but up to 79% of seeds from the Humboldt population are viable (Kittelson and Boyd 1997). S. densiflora was introduced in the 1800s to HB where it currently occupies 385 ha (951 ac); (Grazul and Rowland 2011). Thus, natural dispersal from HB on ocean currents could account for the subsequent appearance of S. densiflora in proximal water bodies such as the Mad River and Eel River delta, where a total of 269 ha are currently infested (Grazul and Rowland 2011), as well as in estuaries as far north as British Columbia. Our results suggest that HB could be a source of propagules for susceptible habitats as far north as southern Alaska. Other vectors, such as solid ballast in dredges, waterfowl, and recreational and fishing vessels, could also be important in S. densiflora dispersal.
Temperature, wave exposure, inundation periods, and salinity tolerances for S. densiflora might limit the northern spread of this species (Bortolus 2006; Castillo et al. 2000, 2005). In its native range, S. densiflora survives in wave-exposed areas of rocky shores and cobble beaches (Bortolus 2006), which suggests that undiscovered populations on the open coast, outside of embayments, could also serve as seed sources. Infestations on the open coast could allow the species to move along the coastline in a series of short-distance dispersal events.
Transport of drift cards from WB was also far-ranging (∼ 2,350 km) and similarly favored deposition to the north in the fall and winter months, a pattern especially pronounced December through February. We observed transport of multiple WB drift cards into neighboring Gray's Harbor, which reinforces the likelihood of inter-bay transport suggested by Strong and Ayres (2009). No cards were recovered within the Strait of Juan de Fuca, nor within Puget Sound; this might reflect that card releases did not coincide with the occasional surface current reversals known to occur in the western Strait of Juan de Fuca (Cannon 1983; Holbrook and Halpern 1982) or possible retention within the cyclonic Juan de Fuca Eddy, which forms just outside the entrance to the Strait (Hickey and Banas 2003). Since the height of the Spartina alterniflora infestation in WB in 2003 when it occupied 3440 ha (net), the population has been reduced to just 1 ha in 2011 (C. Phillips, personal communication). Thus, WB no longer serves as a large annual source of propagules for potential ocean transport or other pathways.
Releases from San Francisco showed the least coast-wise transport across most seasons, but similar directional patterns as the other two bays: predominant northward transport resulting from September through February releases, and balanced north and south movement during spring and summer. SF is one of the most invaded ecosystems in the world (Cohen and Carlton 1998; Ruiz et al. 2000), and secondary spread from this point of introduction is considered likely for multiple species on the west coast (Grosholz and Ruiz 1995; Petersen 2006; Wasson et al. 2001). Hybrid cordgrass (S. foliosa × alterniflora) comprises the majority of cordgrass cover in this estuary system (Ayres et al. 2004) and its spread from San Francisco threatens native West Coast populations of cordgrass with extinction through pollen swamping (Anttila et al. 1998; Ayres et al. 2008). Control efforts in SF have mirrored the success achieved in WB; the SF infestation has been reduced by 90% from its peak of 327 ha (net) in 2006 (Olofson Environmental Inc. 2012). Since surveys were begun in 2001, small populations of S. alterniflora, S. foliosa × alterniflora, and S. densiflora have repeatedly been documented in lagoons and bays as far north as Tomales Bay (P. Olofson, personal communication)
Estuaries are often viewed as isolated from one another, but ocean currents effectively link the entire near-shore environment, including salt marshes and intertidal flats. Effective eradication programs targeting Spartina spp. or other invasive estuarine organisms such as green crab [Carcinus maenas (Linnaeus, 1758)], dwarf eelgrass (Zostera japonica Aschers. and Graebn.), perennial pepperweed (Lepidium latifolium L.), and Algerian sea lavender [Limonium ramosissimum (Poir.) Maire.] in propagule source areas would inhibit their potential dispersal on ocean currents. Furthermore, our results suggest that individual bay-focused Spartina eradication efforts will be ineffectual in the long-term without integration into a coast-wide strategy, such as that proposed within the West Coast Governors' Agreement on Ocean Health (2008).
Acknowledgments
Funding for this project came from the San Francisco Estuary Institute, the Oregon Department of Agriculture, and the Washington Department Natural Resources. Collaborators essential to the success of the project included Dave Heimer, Les Holcomb, Russell Nunez, and Travis Haring from the Washington Department of Fish & Wildlife; Kristen Ward from the National Park Service Golden Gate National Recreation Area; and William Pinnix from the U.S. Fish & Wildlife Service—Arcata Fish and Wildlife Office. We also recognize the invaluable participation of hundreds of curious beachcombers, students, visitors, and residents who took the time to report their cards as well as the various organizations which encouraged awareness of the project, including: Stop Oregon Litter and Vandalism (SOLV); the Grass Roots Garbage Gang; the Coastal Observation and Seabird Survey Team (COASST); and Sightline Daily ( Tidepool.org). Thanks to Catherine de Rivera (Portland State University) for her review and feedback of this manuscript.