- Email: [email protected]
Interpreting the spatio-temporal patterns of sea turtle strandings: Going with the flow
B I O L O G I CA L C O N S E RVAT I O N
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/biocon
Short communication
Interpreting the spatio-temporal patterns of sea turtle strandings: Going with the flow Kristen M. Hart*, Peter Mooreside, Larry B. Crowder Nicholas School of the Environment and Earth Sciences, Duke University Marine Lab, 135 Duke Marine Lab Road, Beaufort, NC 28516, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Knowledge of the spatial and temporal distribution of specific mortality sources is crucial
Received 12 February 2004
for management of species that are vulnerable to human interactions. Beachcast carcasses
Received in revised form
represent an unknown fraction of at-sea mortalities. While a variety of physical (e.g., water
22 October 2005
temperature) and biological (e.g., decomposition) factors as well as the distribution of ani-
Accepted 24 October 2005
mals and their mortality sources likely affect the probability of carcass stranding, physical
Available online 15 December 2005
oceanography plays a major role in where and when carcasses strand. Here, we evaluate the influence of nearshore physical oceanographic and wind regimes on sea turtle stran-
Keywords:
dings to decipher seasonal trends and make qualitative predictions about stranding pat-
Strandings
terns along oceanfront beaches. We use results from oceanic drift-bottle experiments to
Sea turtles
check our predictions and provide an upper limit on stranding proportions. We compare
Loggerhead
predicted current regimes from a 3D physical oceanographic model to spatial and temporal
Caretta caretta
locations of both sea turtle carcass strandings and drift bottle landfalls. Drift bottle return
Kemp’s ridley
rates suggest an upper limit for the proportion of sea turtle carcasses that strand (about
Lepidochelys kempii
20%). In the South Atlantic Bight, seasonal development of along-shelf flow coincides with
Physical oceanography
increased numbers of strandings of both turtles and drift bottles in late spring and early
Drift bottles
summer. The model also predicts net offshore flow of surface waters during winter – the
South Atlantic Bight
season with the fewest relative strandings. The drift bottle data provide a reasonable upper bound on how likely carcasses are to reach land from points offshore and bound the general timeframe for stranding post-mortem (< two weeks). Our findings suggest that marine turtle strandings follow a seasonal regime predictable from physical oceanography and mimicked by drift bottle experiments. Managers can use these findings to reevaluate incidental strandings limits and fishery takes for both nearshore and offshore mortality sources. Ó 2005 Elsevier Ltd. All rights reserved.
1.
Introduction
Estimating mortality is an important component of demographic analyses, having added consequence for long-lived, threatened and endangered vertebrate species with delayed
maturity (Crowder et al., 1994; Crouse, 1999). However, calculating mortality rates and assigning cause of death are particularly challenging for stranded marine animals because of their dynamic, aquatic environment. Previously, analyses of carcass landfall patterns have been performed for marine
* Corresponding author: Current address: US Geological Survey, Center for Coastal and Watershed Studies, 600 Fourth Street South, St. Petersburg, FL 33701, USA. Tel.: +1 727 803 8747x3046; fax: +1 727 803 2032. E-mail addresses: [email protected], [email protected] (K.M. Hart). 0006-3207/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2005.10.047
284
B I O L O G I C A L C O N S E RVAT I O N
mammals (Piatt and Ford, 1996; Garshelis, 1997; McLellan et al., 2002), seabirds (Bibby and Lloyd, 1977; Hyrenbach et al., 2001), and sea turtles (Murphy and Hopkins-Murphy, 1989; Epperly et al., 1996). Still, despite widespread collection and investigation of sea turtle strandings data for over 25 years (Sea Turtle Stranding and Salvage Network), there is no clear way to interpret trends in sea turtle stranding numbers. Sea turtle populations are imperiled worldwide, and our work corroborates the recommendations of the Turtle Expert Working Group (1998) to investigate landfall patterns of stranded turtles. The majority of sea turtle strandings involve individuals that died at sea due to natural or anthropogenic causes such as encounters with fishing gear (NRC, 1990); however, most carcasses show no evidence of cause of death (Sis and Landry, 1992; Turtle Expert Working Group, 1998). Because carcasses decompose while entrained in currents and eddies, the number of recorded sea turtle strandings likely represents a minimum estimate of mortality (Murphy and Hopkins-Murphy, 1989; Epperly et al., 1996). However, the relationship between turtle mortality at sea and observed strandings on shore is still poorly understood. Here, we use turtle and drift bottle data sets to decipher trends in strandings to reveal new insights into the probable locations of mortality sources and the probability of stranding as a function of spatial location. We evaluate nearshore transport of turtle carcasses relying on strandings data (recorded from 1995 to 1999 along the North Carolina (NC) coast) in conjunction with a physical oceanographic model (Werner et al., 1999). We assess these results in the context of extensive drift bottle release experiments conducted in the 1960s and 1970s (Harrison et al., 1967; Bumpus, 1973). Moreover, we develop reasonable bounds for the numbers of turtles that may be dead at sea based on beach strandings within our study area, as well as the general timeframe for and probability of carcass stranding events. Discerning seasonality of stranding patterns and how the number of turtles stranded on the beach is related to the number of carcasses at sea will help sea turtle managers develop more accurate estimates of mortality rates. This research provides the first attempt to identify seasons when turtles are, and are not, likely to strand. It also provides the first evidence to sea turtle managers that incidental stranding limits (ISLs) need to be recalculated on a seasonal basis, in light of oceanographic conditions. Such information could then be applied in fisheries management to create more accurate time and area closures for fisheries with historic bycatch of sea turtles. Although coastal water circulation tends to be local and hence difficult to predict primarily due to shoreline geography and bathymetry (S. Lozier, personal communication), recent advances in coastal current modeling made by Werner et al. (1999) allowed us to predict near-shore surface currents in the South Atlantic Bight (SAB). In our analysis, we compare turtle stranding patterns to modeled currents; the inference derived from the model helps to clarify the relationship between at-sea turtle mortality and on-shore turtle strandings. Population biology and physical oceanography have, until recently, been scientific fields with little crossover. Our goal in this paper is to examine how oceanographic and wind conditions potentially affect the flow of carcasses at sea, in hopes
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
that we may be better able to interpret perceived trends in strandings numbers, not only for sea turtles, but also for other marine animals. This paper provides a starting point for more robust analyses and demonstrates that stranding research requires knowledge of a species’ distribution as well as an understanding of seasonal physio-oceanic processes.
1.1.
Case study: NC sea turtle strandings
Loggerhead (Caretta caretta) and Kemp’s ridley (Lepidochelys kempii) sea turtles are typically the most common species to strand on ocean-facing beaches in NC. Among the five marine turtle species that reside in or migrate through US coastal waters, C. caretta and L. kempii are classified under the US Endangered Species Act as threatened and endangered, respectively (NRC, 1990). Despite their current regulatory status, NC Loggerheads are part of the ‘‘northern’’ subpopulation that is declining (NMFS, 2001). Incidental death via shrimp trawling has been cited as the most important source of anthropogenic mortality for sea turtles in US waters (NRC, 1990), but scientists have yet to quantify whether turtle strandings in one region are influenced by local fishing effort in another. Ideally, attempts to investigate stranding patterns should account for factors that affect the initiation and duration of carcass buoyancy (e.g., turtle size, carcass decomposition rate, water temperature, and presence of scavengers) and the probability of carcass landfall (e.g., direction, intensity, and seasonality of prevailing winds, surface and near-bottom current regimes, lunar tides, and the spatial proximity of mortality sources to shore) (NRC, 1990; Crowder et al., 1995; Epperly et al., 1996; Lewison et al., 2003). However, this information is not always available or known. Oceanic conditions that produce nearshore currents could facilitate the stranding of drifting turtle carcasses (Crowder et al., 1995), and hence partially explain the increased number of strandings observed during certain seasons in the northwestern Atlantic (e.g., spring; Amos, 1989). Similarly, winter wind regimes may initiate net offshore flow in shelf waters, thus precluding or reducing carcass landfall (Epperly et al., 1995a). This paper represents the first published attempt to integrate physical oceanography with sea turtle strandings data to examine the possible forcing effect of nearshore marine processes and dynamics.
2.
Materials and methods
By combining several novel yet complementary data sets (Table 1), we took a multi-faceted approach to understand and interpret patterns in sea turtle strandings.
2.1.
Marine turtle stranding data
We obtained sea turtle stranding records from 1995 to 1999 for NC, sorted it by month, and filtered the spatially referenced data to include strandings only occurring on ocean-facing beaches. Through v2 analyses (Steel et al., 1997), we tested the null hypothesis of uniform stranding distributions within two cuspate bays, Raleigh Bay and Onslow Bay, during May and June, the months when the highest cumulative number
B I O L O G I C A L C O N S E RVAT I O N
285
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Table 1 – Sources, spatio-temporal settings, and applications of data used to analyze seasonal oceanographic and stranding patterns Source #
Data type (sorted by month)
References
Setting (year(s) of data used and region) Chesapeake Bight (Cape Henlopen, DE to Cape Hatteras, NC), 1960–1970, with focus on 1963–1964 Atlantic Continental Shelf (25–44°N, 64 to 81°W), 1961–1964 Diamond Shoals Navigation Light (offshore from Cape Hatteras, NC), 1995–1999 South Atlantic Bight (Cape Romain, SC to Cape Hatteras, NC), 1999 estimates NC shoreline, 1995–1999 with focus on ocean-facing beaches
1
Surface drift bottle returns
Harrison et al. (1967)
2
Ballasted and surface drift bottles
Bumpus (1973)
3
Wind direction and magnitude
National Data Buoy Center
4
Nearshore oceanographic processes, real topography and bathymetry Sea turtle strandings
Werner et al. (1999)
5
R. Boettcher (unpublished data)
Seasonal parameter of interest % Recovery to shore
% Recovery to shore
Input into Werner et al. (1999) oceanographic model Estimates of nearshore water movements Observed point locations over time
Table 2 – Monthly and seasonal sea turtle strandings (all species) on inshore and ocean-facing beaches in North Carolina, 1995–1999 Season
Sp
Sp
Sp
Su
Su
F
MF
MF
W
W
W
W
Total
Month/year
M
A
M
J
J
A
S
O
N
D
J
F
1995 1996 1997 1998 1999
2 9 17 2 20
14 8 14 7 37
80 120 70 29 82
93 103 78 67 120
46 46 56 56 42
20 57 24 41 44
20 23 21 13 20
23 39 32 30 28
9 32 51 56 104
26 23 20 34 56
11 35 17 20 24
3 10 8 3 28
347 505 408 358 605
Total
121
189
770
921
605
439
227
388
641
542
227
112
5182
Note that these numbers do not include incidentally captured turtles (observations of a direct interaction between a sea turtle and a lawfully conducted human activity such as dredging or commercial/recreational fishing). Abbreviations are as follows: Sp, spring; Su, summer; F, fall; MF, Mariner’s fall; W, winter.
of strandings occur (Table 2). For this analysis, we divided each bay into approximate halves by finding the straight-line distance between capes, located the midpoint of that line, and drew a perpendicular line to the coast. We then counted the number of strandings within the eastern and western halves.
2.2.
Drift bottle data
We reevaluated previously conducted oceanic drift bottle experiments (Harrison et al., 1967; Bumpus, 1973) to infer recovery rates of both surface and bottom drift bottles over time and space. Harrison et al. (1967) released surface drift bottles (n = 11,052) in one deployment at distances inside the Chesapeake Bight (CB), a 300 + km extent of the southern Mid-Atlantic Bight (MAB) stretching between Cape Henlopen, Delaware (DE) and Cape Hatteras, NC. Similarly, Bumpus (1973) released 165,566 ballasted and surface drift-bottles (e.g., eight ounce (237 ml) soda bottles) from 1960 to 1970 at 78 locations widely distributed over the continental shelf (25 to 44°N, 64 to 81°W). For the purpose of calculating a bottle stranding rate, we evaluated the probability of bottles making landfall from various points offshore.
2.3.
Wind data
Wind speed magnitude (m/s) and direction (degrees clockwise from north) were recorded hourly via anemometer at a Coastal-Marine Automated Network (C-MAN) station, located at 35.15°N, 75.30°W off Cape Hatteras, NC. To attain representative wind values, we downloaded 15 years (encompassing the years for which we had sea turtle strandings records) of hourly wind data from the station via the National Data Buoy Center website (http://www.ndbc.noaa.gov), transformed winds into vector format, and then resolved each wind’s vector into its corresponding u (easterly) and v (northerly) components. The magnitudes of each wind component, originally recorded as velocities (m/s), were transformed into forces (Pascals) using a specialized Matlab script (Blanton et al., 1985). The resulting easterly and northerly wind stress components were then converted from Pascals into Dynes/ cm2. We calculated the stress magnitude (r) via vector addition of the u and v directional components by means of the Pythagorean theorem (sqrt[(u2) + (v2)] = r). Similarly, by taking the arctangent of v/u, we determined wind stress direction (h). After obtaining the model’s output for True North (0°) and True East (90°) winds, we used these directions
286
B I O L O G I C A L C O N S E RVAT I O N
2.4.
Oceanographic model
The underlying structure of the Werner et al. (1999) oceanographic model is a finite element matrix, serving as the domain for the South Atlantic Bight Recruitment Experiment (SABRE) flow-field computer model that is capable of estimating water current paths (Werner et al., 1999). Each node in the matrix was designed to respond to a wind stress applied to the water’s surface. Node responses are visually displayed as vectors, indicating the hypothetical flow of the water body under prescribed conditions. The spatial domain of the matrix incorporates the continental shelf and slope of the MAB and SAB. We focused exclusively on the model domain in proximity to NC.
3.
Results
3.1.
Sea turtle strandings
Filtering all NC sea turtle strandings data for those that occurred along ocean-facing beaches during 1995–1999 (Table 2) produced point locations for over 1300 Loggerhead and Kemp’s ridley turtles combined (Fig. 1). Whereas almost half of these strandings occurred in May and June, less than 8% occurred from December to March, and more than 75% of the observed strandings were Loggerheads. The numbers and spatial locations of stranded carcasses varied monthly and seasonally but patterns were consistent from year to year. During late spring and early summer (May and June), disproportionate numbers of carcasses stranded along the eastern stretches of both Onslow and Raleigh Bays. Inside Onslow Bay, over four times as many turtles stranded along its eastern half compared to the western side, with chi-square results v2 = 22.93, v2critical ¼ 3:84 and p < 0.05. Similarly, for Raleigh Bay, we also found a significant difference in stranding patterns with chi-square results v2 = 4.83, v2critical ¼ 3:84 and p < 0.05, though it was not as prominent as for Onslow Bay.
3.2.
Drift bottle data
Bumpus (1973) demonstrated low recovery rates of surface drift bottles released during 1961–1964 from November to February within the MAB and SAB. Likewise, Harrison et al. (1967) found that drifter returns from June 1963 to October 1964 demonstrated monthly variation, with the lowest returns during the winter when less than 1% (16 out of 2592) of the bottles released over the entire study area were recovered. Harrison et al. (1967) hypothesized that northwesterly winds, prevailing from late November into February, moved waters offshore and to the south. Harrison et al. (1967) discovered that during the rest of the year, large numbers of bottle recoveries most often coincided with periods of onshore winds, and, regardless of monthly variation, the majority of drift bottles were recovered within approximately two weeks of release. Furthermore, percent recovery of surface drift bottles was inversely related to distance from shore (Figs. 2 and 3). This relationship was strongest during the summer, moderate during spring and fall, and weakest during the winter (Fig. 2).
50
40 Bottles recovered (%)
to approximate the model’s response for any given wind direction, without having to run the model separately for each different wind direction.
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
October-63 January-64 April-64 June-64
30
20
10
0 0
50
100
150
Distance released from shore (km)
Fig. 2 – Recovery of surface drift bottles released inside the Chesapeake Bight (modified from data presented in Harrison et al., 1967). Corresponding N and r2 values were: 636 and 0.925 (October 1963), 618 and 0.617 (January 1964), 660 and 0.925 (April 1964), and 660 and 0.965 (June 1964).
300 250 200
C. caretta
150
L. kempii
100 50 0 jan feb mar apr may jun
jul aug sep oct
nov dec
Month
Fig. 1 – Ocean-beach strandings of Loggerhead (C. Caretta) and Kemp’s ridley (L. Kempii) sea turtles in NC, 1995–1999 (derived from unpublished data provided by R. Boettcher). Error bars depict one standard error.
Landfall probability (%)
# of stranded turtles
350
60 50 40 30 20 10 0 1
5
10
20
30
40
50
60
70
80
90
100
Distance from coast (km)
Fig. 3 – Stranding probability of bottles making landfall according to distance released from shore (derived from data presented in Harrison et al., 1967).
B I O L O G I C A L C O N S E RVAT I O N
287
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Table 3 – Monthly averaged wind stress magnitude and direction values at Diamond Shoals (Cape Hatteras, NC) from 1985 to 1999 Season Spring Spring Spring Summer Summer Fall Mariner’s fall Mariner’s fall Winter Winter Winter Winter
Month
Magnitude (dynes/cm2)
Direction (degrees)
Direction (bearing)
Sample size (# of hourly recordings)
March April May June July August September October November December January February
0.049 0.052 0.025 0.055 0.194 0.010 0.008 0.061 0.039 0.172 0.127 0.113
123.4 92.9 86.7 47.9 48.3 39.5 218.8 197.2 138.1 141.9 127.3 139.3
E–SE E E NE NE N–NE S–SW S–SW SE SE E–SE SE
11,114 11,426 11,758 11,370 11,126 9746 10,041 10,278 10,198 10,884 11,360 9923
Wind stress direction is presented in degrees clockwise from true North.
3.3.
Wind data
Inspection of averaged monthly winds at Diamond Shoals from January to December 1985–1999 (Table 3) revealed that consistently, southeastern and east–southeastern winds predominated from November to February (Winter). General wind direction gradually shifted to the east during March, April and May (Spring), and a northeastern wind pattern developed in June and July (Summer), followed by a north– northeastern transition period in August (Fall). South–southwestern winds persisted in September and October (Mariner’s fall). However, wind variability was high at the monthly scale; standard deviations for monthly wind stress components (data not shown) were sometimes as large, or larger than, their corresponding mean values.
3.4.
Oceanographic model
Vector flow fields (generated by using mean wind stress values and running the physical oceanographic model for each month) represented residual changes in surface water movement predicted at model nodes in accordance with monthly wind patterns. Model responses off the NC coast were as follows, divided into four seasonal wind regimes and five seasons, as originally described by Weber and Blanton (1980) for the SAB:
3.4.1.
Winter (November–February)
Surface water regimes differed north and south of Cape Hatteras. South of the Cape, the model predicted offshore movement of surface waters. North of Cape Hatteras, however, the along-shelf component persisted through the winter months, where net water movement assumed a southward track.
3.4.2.
Spring (March–May)
Modeled conditions in March favored the development of northward along-shelf flow south of Cape Hatteras. The net water flow north of Cape Hatteras reversed direction from the winter months, also taking a northward course. The model also predicted concentrated onshore flow along the eastern edges of the Carolina Bays.
3.4.3.
Summer (June–July) and fall (August)
The northward along-shelf surface flow persisted along the entire coast during these three months. In addition, net flow magnitude intensified throughout the region, peaking in July.
3.4.4.
Mariner’s fall (September–October)
In September, along-shelf flow reversed direction from its summer and fall course. This change in net direction was maintained through October, and later dissipated in November with the onset of offshore flow.
4.
Discussion
The lowest number of sea turtle strandings (Fig. 1 and Table 2) and recoveries of surface drift bottles (Fig. 2) consistently occurred during Winter conditions (e.g., January) with offshore flow. Thus, we conclude that there appears to be a predictable pattern of strandings-favorable seasons and strandingsunfavorable seasons, regardless of whether turtles or bottles are the stranded objects. The climatology for 1995–1999 is qualitatively similar to the long-term climatology for the area so strandings and recoveries for the time period under consideration are comparable to other years and decades. In continental shelf dynamics, although the magnitude of the along-shelf component can be characteristically greater than that of the cross-shelf component (Hare et al., 1999), net offshore flow may contribute to the decreased numbers of strandings and recoveries recorded in Winter months by transporting floating carcasses or bottles away from shore. According to oceanographic model results, such offshore surface flow was most prevalent south of Cape Hatteras from November to February, which corresponded to the season with the lowest number of recorded strandings in the study site. Stranding incidence increased during the Spring (from March to May), when the model predicted the development of the along-shelf current. However, during the winter north of Cape Hatteras, an established southern alongshore current was present despite few strandings. Alternative explanations for decreased strandings recorded in Winter in the study site include, but are not limited to, seasonal turtle migrations towards warmer waters (Coles and Musick, 2000), colder sea
288
B I O L O G I C A L C O N S E RVAT I O N
surface temperatures (which may prolong drowning times for air-breathing poikilotherms (Murphy and Hopkins-Murphy, 1989)), decreased trawling effort or spatially displaced fishing effort, and a generally landward shift of the Gulf Stream during Winter months. Our modeled results elucidate that there are certain months and seasons when strandings are a likely proxy for local marine turtle bycatch, and they support the claim of Epperly et al. (1996). These researchers argued that the number of sea turtle carcasses stranded in proximity to the southern flounder trawl fishery, which operates offshore in winter, was not a reliable indicator of turtle mortality at-sea because turtles killed in Winter in this fishery would not likely appear on land as stranded individuals. Rather, turtle carcasses would more likely be entrained in offshore-flowing currents and swept out to sea. Additional supporting evidence for this comes from aerial surveys, incidental take data from commercial fishers, and satellite tracking studies. Such studies have shown that sea turtles are relatively more abundant near convergence zones (Lutcavage et al., 1997; Polovina et al., 2000), which are farther offshore during Fall and Winter (Murphy and Hopkins-Murphy, 1989; NRC, 1990; Epperly et al., 1995b). This zone is close to the western boundary of the Gulf Stream (Hoffman and Fritts, 1982), and along the shelf break near the 200 m isobath of the MAB (Lutcavage et al., 1997). Thus an absence of sea turtles in coastal waters may very well result in fewer strandings recorded in Winter (NRC, 1990). However, regardless of the time of year, an absence of stranded carcasses may (or may not) coincide with an absence of turtle mortality sources. For example, shrimp fishing activity may begin seasonally during strandings-favorable conditions, implicating local shrimpers if strandings are high, whereas flounder trawl fisheries may take turtles that ever seldom strand due to strandings-unfavorable conditions in winter (Epperly et al., 1995a; NRC, 1990). This has important consequences for managers working to implement fishery management strategies that include reducing sea turtle bycatch (i.e., those fisheries operating closest to strandings locations may not be the cause of the strandings). In addition to seasonal concordance with sea turtle strandings, bottle recovery data from Harrison et al. (1967) suggest that strandings are strongly and negatively related to increased distance released from shore. This relationship was strongest during the Summer, moderate during Spring and Fall, and weakest during the Winter (Fig. 2). This predictable pattern of strandings-favorable and strandings-unfavorable seasons was present regardless of whether turtles or bottles were the stranded objects under examination; again, the lowest recoveries for both sea turtle carcasses and surface drift bottles consistently occurred during winter conditions (e.g., January) with offshore flow. Although direct comparisons between drifting bottles and sea turtle carcasses can be made, the difference in size may make stranded turtles easier to detect on shore, larger surface area of floating turtles could provide greater area upon which wind stress may directly act, and the organic composition of carcasses increases their susceptibility to air and water temperature effects (i.e., facilitating or inhibiting bacterial growth and activity) or to disarticulation by scavengers, both of which may considerably alter time spent afloat. Despite these potential biases,
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
the landfall probabilities extrapolated from drift bottle experiments may have direct relevance to sea turtle management. We maintain that drift bottle returns can provide an upper bound to describe how far sea turtle carcasses could theoretically travel and how likely those carcasses are to make landfall from points offshore. To recreate the stranding probability of drift bottles by distance in the CB, we extrapolated values from log equations derived from data published by Harrison et al. (1967). Drift bottles released at a distance of 20 km had approximately 21% chance of making landfall (Fig. 3), implying that only bottles released very close to shore had the highest probability of stranding. Additionally, Harrison et al. (1967) found that the majority of drift bottles recovered in their study were found within two weeks of the release date. We use these estimates to put into context the percentage of turtles dead at-sea that might strand (about 20%) and the timeframe for stranding (within two weeks). Intuitively, objects released farther from shore should have a lower probability of making landfall compared to objects released closer to shore; we use drift bottle return data (i.e., after Bumpus, 1973) to go further and suggest that, on average, the number of carcasses stranded on ocean-facing beaches may represent, at best, approximately 20% of the total number of available carcasses at sea (Fig. 3). Managers should heed these complementary lines of evidence that indicate that only turtles killed very close to the shore (i.e., within 10 s of kms) may be likely to strand, even in stranding-favorable conditions. Furthermore, if carcasses are going to strand, then, according to the results from other drift bottle experiments (e.g., Harrison et al., 1967), they would probably do so within two weeks after achieving buoyancy. This information indicates that ISLs need to be reevaluated because incidental take of sea turtles from offshore fisheries may not be reflected by onshore stranding events. We are confident that the physical oceanographic model qualitatively predicts the behaviour of near-shore currents in which turtle carcasses could be entrained, especially shoreward of the 40-m isobath. However the model cannot predict the specific locations of stranding events or mortality sources. But by using monthly flow fields predicted by the model and by relying on empirical evidence from surface drift bottle experiments, we can describe large-scale seasonal patterns in coastal waters of NC. Our comparative approach allows us to address one of the fundamental issues in strandings research; that is, how did the stranded animal get to its position on land? Because strandings data differ across species and ocean basins, our interpretation of NC sea turtle strandings data may not characterize strandings trends elsewhere. Our intention is to encourage managers to consider nearshore physical oceanographic processes when interpreting spatial and temporal trends in strandings of marine animals, especially when implicating fisheries whose bycatch of sea turtles may or may not be realized. For example, in NC, sea turtle carcasses may be more likely to strand from April to October along ocean-facing beaches due to the development of an along-shelf current parallel to the coast. Likewise, carcasses seem least likely to strand from November to February due to the presence of offshore flow. Moreover, relative to their recorded stranding position on the beach within the study area, sea turtles are more likely to have been killed in areas to the
B I O L O G I C A L C O N S E RVAT I O N
south from April to August and to the north from September to October due to the net direction of along-shelf flow. By providing a crude estimate of the spatial direction of carcass transport, the Werner et al. (1999) model could be used to investigate potential up- or down-stream mortality sources during periods of sustained along-shelf flow. If we are better able to predict overlap of sea turtle distribution and mortality sources and we have an understanding of how oceanographic and wind conditions affect the flow of carcasses at sea, we may be better able to interpret perceived trends in strandings numbers, not only for sea turtles, but also for other marine animals.
Acknowledgments Brian Blanton graciously provided assistance with model runs and Matlab, and Patrick Halpin, Caterina D’Agrosa, and Ari Friedlaender provided guidance and assistance with GIS. Ruth Boettcher provided strandings data, and Sheryan Epperly, Victoria Thayer, and Lesley Thorne commented on earlier drafts of the manuscript. We thank Francisco Werner for providing access to the physical oceanographic model. We acknowledge the National Marine Fisheries Service, Pew Charitable Trusts, National Oceanographic Partnership Program, Sloan Foundation’s ‘‘Census of the Sea’’, and the Edna Bailey Sussman Foundation for support. We dedicate this paper to Dean Bumpus whose massive empirical work on coastal currents may contribute to better protection of sea turtles and other endangered marine species. This work was part of the Master’s requirement for P.M. at Duke University.
R E F E R E N C E S
Amos, A., 1989. Trash, debris, and human activities: potential hazards at sea and obstacles to Kemp’s ridley sea turtle nesting. In: Caillouet, C., Jr., Landry, A., Jr., (Eds.), Proceedings from the First International Symposium on Kemp’s Ridley Sea Turtle Biology. Texas A&M University, Galveston, p. 42. 105pp. Bibby, C.J., Lloyd, C.S., 1977. Experiments to determine the fate of dead birds at sea. Biological Conservation 12, 295–309. Blanton, J., Schwing, F., Weber, A., Pietrafesa, L., Hayes, D., 1985. Wind stress climatology in the South Atlantic Bight. In: Atkinson, L., Menzel, D., Bush, K. (Eds.), Oceanography of the Southeastern US Continental Shelf, Coastal and Estuarine Sciences, vol. 2. American Geophysical Union, pp. 23–32. Bumpus, D., 1973. A description of the circulation on the continental shelf of the east coast of the United States. In: Warren, B. (Ed.), Progress in Oceanography, vol. 6. Pergamon Press, Oxford, pp. 111–157. 193. Coles, W., Musick, J., 2000. Satellite sea surface temperature analysis and correlation with sea turtle distribution off North Carolina. Copeia 2, 551–554. Crouse, D.T., 1999. The consequences of delayed maturity in a human-dominated world. In: J.A. Musick (Ed.). Life in the Slow Lane: Ecology and Conservation of Long-lived Marine Animals. American Fisheries Society Symposium 23, Bethesda, MD, pp. 195–202. Crowder, L.B., Crouse, D.T., Heppell, S.S., Martin, T.H., 1994. Predicting the impact of turtle excluder devices on loggerhead sea turtles. Ecological Applications 4, 437–445.
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
289
Crowder, L., Hopkins-Murphy, S., Royle, J., 1995. Effects of turtle excluder devices (TEDs) on loggerhead sea turtle strandings with implications for conservation. Copeia 4, 773–779. Epperly, S., Braun, J., Chester, A., Cross, F., Merriner, J., Tester, P., 1995a. Winter distribution of sea turtles in the vicinity of Cape Hatteras and their interactions with the summer flounder trawl fishery. Bulletin of Marine Science 56 (2), 547–568. Epperly, S., Braun, J., Veishlow, A., 1995b. Sea turtles in North Carolina waters. Conservation Biology 9 (2), 384–394. Epperly, S., Braun, J., Chester, A., Cross, F., Merriner, J., Tester, P., Churchill, J., 1996. Beach strandings as an indicator of at sea mortality of sea turtles. Bulletin of Marine Science 59 (2), 289–297. Garshelis, D.L., 1997. Sea otter mortality estimated from carcasses collected after the Exxon Valdez oil spill. Conservation Biology 11 (4), 905–916. Hare, J., Quinlan, J., Werner, F., Blanton, B., Govoni, J., Forward, R., Settle, L., Hoss, D., 1999. Larval transport during winter in the SABRE study area: results of a coupled vertical larval behaviour – three-dimensional circulation model. Fisheries Oceanography 8 (2), 57–76. Harrison, W., Norcross, J., Pore, N., Stanley, E., 1967. Circulation of shelf waters off the Chesapeake Bight. Environmental Science Services Administration (ESSA) Professional Paper 3, US Department of Commerce, Washington, DC, 82pp. Hoffman, W., Fritts, T., 1982. Sea turtle distribution along the boundary of the Gulf Stream current off eastern Florida. Herpetologica 38 (3), 405–409. Hyrenbach, K.D., Baduini, C.L., Hunt Jr., G.L., 2001. Line transect estimates of short-tailed Shearwater (Puffinus tenuirostris) mortality in the Southeastern Bering Sea: 1997–99. Marine Ornithology 29, 27–34. Lewison, R.L., Crowder, L.B., Shaver, D.J., 2003. The impact of turtle excluder devices and fisheries closures on loggerhead and Kemp’s ridley strandings in the western Gulf of Mexico. Conservation Biology 17 (4), 1089–1097. Lutcavage, M., Plotkin, P., Witherington, B., Lutz, P., 1997. Human impacts on sea turtle survival. In: Musick, J., Lutz, P. (Eds.), The Biology of Sea Turtles. CRC Press, Boca Raton, pp. 387–409. 432pp. McLellan, W.A., Friedlaender, A.S., Mead, J.G., Potter, C.W., Pabst, D.A., 2002. Analysing 25 years of bottlenose dolphin (Tursiops truncates) strandings along the Atlantic coast of the USA: do historic records support the coastal migratory stock hypothesis? Journal of Cetacean Research and Management 4 (3), 1–8. Murphy, T., Hopkins-Murphy, S., 1989. Sea Turtle and Shrimping Interactions: A Summary and Critique of Relevant Information. Center for Marine Conservation, Washington, DC. 52pp. National Marine Fisheries Service, 2001. Stock Assessments of Loggerhead and Leatherback Sea Turtles. Technical Memorandum NMFS-SEFSC-455. National Oceanic and Atmospheric Administration, Miami. National Research Council, 1990. Decline of Sea Turtles: Causes and Prevention. National Academy Press, Washington, DC. Piatt, J., Ford, R., 1996. How many seabirds were killed by the Exxon Valdez oil spill? In: Rice, S., Spies, R., Wolfe, D., Wright, B. (Eds.). Proceedings of the Exxon Valdez Oil Spill Symposium. American Fisheries Society Symposium 18, Bethesda, pp. 712–719, 931pp. Polovina, J.J., Kobayashi, D.R., Parker, D.M., Seki, M.P., Balazs, G.H., 2000. Turtles on the edge: movement of loggerhead turtles (Caretta caretta) along oceanic fronts, spanning longline fishing grounds in the central North Pacific, 1997–1998. Fisheries Oceanography 9, 71–82.
290
B I O L O G I C A L C O N S E RVAT I O N
Sis, R., Landry, A., 1992. Postmortem changes in the turtle. In: Proceedings of the 23rd Annual International Association for Aquatic Animal Medicine (IAAAM), Hong Kong. IAAAM, San Leandro, California, pp. 17–19. Steel, R., Torrie, J., Dickey, D., 1997. Principles and Procedures of Statistics: A Biometrical Approach, third ed. McGraw-Hill, New York. 666pp. Turtle Expert Working Group, 1998. An Assessment of the Kemp’s Ridley (Lepidochelys kempii) and Loggerhead (Caretta caretta) Sea Turtle Populations in the Western North Atlantic. Technical
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Memorandum NMFS-SEFSC-409. National Oceanic and Atmospheric Administration, Miami. Weber, A., Blanton, J., 1980. Monthly mean wind fields for the South Atlantic Bight. Journal of Physical Oceanography 10 (8), 1256–1263. Werner, F., Blanton, B., Quinlan, J., Luettich Jr., R., 1999. Physical oceanography of the North Carolina continental shelf during the fall and winter seasons: implications for the transport of larval menhaden. Fisheries Oceanography 8 (2), 7–21.
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/biocon
Short communication
Interpreting the spatio-temporal patterns of sea turtle strandings: Going with the flow Kristen M. Hart*, Peter Mooreside, Larry B. Crowder Nicholas School of the Environment and Earth Sciences, Duke University Marine Lab, 135 Duke Marine Lab Road, Beaufort, NC 28516, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Knowledge of the spatial and temporal distribution of specific mortality sources is crucial
Received 12 February 2004
for management of species that are vulnerable to human interactions. Beachcast carcasses
Received in revised form
represent an unknown fraction of at-sea mortalities. While a variety of physical (e.g., water
22 October 2005
temperature) and biological (e.g., decomposition) factors as well as the distribution of ani-
Accepted 24 October 2005
mals and their mortality sources likely affect the probability of carcass stranding, physical
Available online 15 December 2005
oceanography plays a major role in where and when carcasses strand. Here, we evaluate the influence of nearshore physical oceanographic and wind regimes on sea turtle stran-
Keywords:
dings to decipher seasonal trends and make qualitative predictions about stranding pat-
Strandings
terns along oceanfront beaches. We use results from oceanic drift-bottle experiments to
Sea turtles
check our predictions and provide an upper limit on stranding proportions. We compare
Loggerhead
predicted current regimes from a 3D physical oceanographic model to spatial and temporal
Caretta caretta
locations of both sea turtle carcass strandings and drift bottle landfalls. Drift bottle return
Kemp’s ridley
rates suggest an upper limit for the proportion of sea turtle carcasses that strand (about
Lepidochelys kempii
20%). In the South Atlantic Bight, seasonal development of along-shelf flow coincides with
Physical oceanography
increased numbers of strandings of both turtles and drift bottles in late spring and early
Drift bottles
summer. The model also predicts net offshore flow of surface waters during winter – the
South Atlantic Bight
season with the fewest relative strandings. The drift bottle data provide a reasonable upper bound on how likely carcasses are to reach land from points offshore and bound the general timeframe for stranding post-mortem (< two weeks). Our findings suggest that marine turtle strandings follow a seasonal regime predictable from physical oceanography and mimicked by drift bottle experiments. Managers can use these findings to reevaluate incidental strandings limits and fishery takes for both nearshore and offshore mortality sources. Ó 2005 Elsevier Ltd. All rights reserved.
1.
Introduction
Estimating mortality is an important component of demographic analyses, having added consequence for long-lived, threatened and endangered vertebrate species with delayed
maturity (Crowder et al., 1994; Crouse, 1999). However, calculating mortality rates and assigning cause of death are particularly challenging for stranded marine animals because of their dynamic, aquatic environment. Previously, analyses of carcass landfall patterns have been performed for marine
* Corresponding author: Current address: US Geological Survey, Center for Coastal and Watershed Studies, 600 Fourth Street South, St. Petersburg, FL 33701, USA. Tel.: +1 727 803 8747x3046; fax: +1 727 803 2032. E-mail addresses: [email protected], [email protected] (K.M. Hart). 0006-3207/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2005.10.047
284
B I O L O G I C A L C O N S E RVAT I O N
mammals (Piatt and Ford, 1996; Garshelis, 1997; McLellan et al., 2002), seabirds (Bibby and Lloyd, 1977; Hyrenbach et al., 2001), and sea turtles (Murphy and Hopkins-Murphy, 1989; Epperly et al., 1996). Still, despite widespread collection and investigation of sea turtle strandings data for over 25 years (Sea Turtle Stranding and Salvage Network), there is no clear way to interpret trends in sea turtle stranding numbers. Sea turtle populations are imperiled worldwide, and our work corroborates the recommendations of the Turtle Expert Working Group (1998) to investigate landfall patterns of stranded turtles. The majority of sea turtle strandings involve individuals that died at sea due to natural or anthropogenic causes such as encounters with fishing gear (NRC, 1990); however, most carcasses show no evidence of cause of death (Sis and Landry, 1992; Turtle Expert Working Group, 1998). Because carcasses decompose while entrained in currents and eddies, the number of recorded sea turtle strandings likely represents a minimum estimate of mortality (Murphy and Hopkins-Murphy, 1989; Epperly et al., 1996). However, the relationship between turtle mortality at sea and observed strandings on shore is still poorly understood. Here, we use turtle and drift bottle data sets to decipher trends in strandings to reveal new insights into the probable locations of mortality sources and the probability of stranding as a function of spatial location. We evaluate nearshore transport of turtle carcasses relying on strandings data (recorded from 1995 to 1999 along the North Carolina (NC) coast) in conjunction with a physical oceanographic model (Werner et al., 1999). We assess these results in the context of extensive drift bottle release experiments conducted in the 1960s and 1970s (Harrison et al., 1967; Bumpus, 1973). Moreover, we develop reasonable bounds for the numbers of turtles that may be dead at sea based on beach strandings within our study area, as well as the general timeframe for and probability of carcass stranding events. Discerning seasonality of stranding patterns and how the number of turtles stranded on the beach is related to the number of carcasses at sea will help sea turtle managers develop more accurate estimates of mortality rates. This research provides the first attempt to identify seasons when turtles are, and are not, likely to strand. It also provides the first evidence to sea turtle managers that incidental stranding limits (ISLs) need to be recalculated on a seasonal basis, in light of oceanographic conditions. Such information could then be applied in fisheries management to create more accurate time and area closures for fisheries with historic bycatch of sea turtles. Although coastal water circulation tends to be local and hence difficult to predict primarily due to shoreline geography and bathymetry (S. Lozier, personal communication), recent advances in coastal current modeling made by Werner et al. (1999) allowed us to predict near-shore surface currents in the South Atlantic Bight (SAB). In our analysis, we compare turtle stranding patterns to modeled currents; the inference derived from the model helps to clarify the relationship between at-sea turtle mortality and on-shore turtle strandings. Population biology and physical oceanography have, until recently, been scientific fields with little crossover. Our goal in this paper is to examine how oceanographic and wind conditions potentially affect the flow of carcasses at sea, in hopes
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
that we may be better able to interpret perceived trends in strandings numbers, not only for sea turtles, but also for other marine animals. This paper provides a starting point for more robust analyses and demonstrates that stranding research requires knowledge of a species’ distribution as well as an understanding of seasonal physio-oceanic processes.
1.1.
Case study: NC sea turtle strandings
Loggerhead (Caretta caretta) and Kemp’s ridley (Lepidochelys kempii) sea turtles are typically the most common species to strand on ocean-facing beaches in NC. Among the five marine turtle species that reside in or migrate through US coastal waters, C. caretta and L. kempii are classified under the US Endangered Species Act as threatened and endangered, respectively (NRC, 1990). Despite their current regulatory status, NC Loggerheads are part of the ‘‘northern’’ subpopulation that is declining (NMFS, 2001). Incidental death via shrimp trawling has been cited as the most important source of anthropogenic mortality for sea turtles in US waters (NRC, 1990), but scientists have yet to quantify whether turtle strandings in one region are influenced by local fishing effort in another. Ideally, attempts to investigate stranding patterns should account for factors that affect the initiation and duration of carcass buoyancy (e.g., turtle size, carcass decomposition rate, water temperature, and presence of scavengers) and the probability of carcass landfall (e.g., direction, intensity, and seasonality of prevailing winds, surface and near-bottom current regimes, lunar tides, and the spatial proximity of mortality sources to shore) (NRC, 1990; Crowder et al., 1995; Epperly et al., 1996; Lewison et al., 2003). However, this information is not always available or known. Oceanic conditions that produce nearshore currents could facilitate the stranding of drifting turtle carcasses (Crowder et al., 1995), and hence partially explain the increased number of strandings observed during certain seasons in the northwestern Atlantic (e.g., spring; Amos, 1989). Similarly, winter wind regimes may initiate net offshore flow in shelf waters, thus precluding or reducing carcass landfall (Epperly et al., 1995a). This paper represents the first published attempt to integrate physical oceanography with sea turtle strandings data to examine the possible forcing effect of nearshore marine processes and dynamics.
2.
Materials and methods
By combining several novel yet complementary data sets (Table 1), we took a multi-faceted approach to understand and interpret patterns in sea turtle strandings.
2.1.
Marine turtle stranding data
We obtained sea turtle stranding records from 1995 to 1999 for NC, sorted it by month, and filtered the spatially referenced data to include strandings only occurring on ocean-facing beaches. Through v2 analyses (Steel et al., 1997), we tested the null hypothesis of uniform stranding distributions within two cuspate bays, Raleigh Bay and Onslow Bay, during May and June, the months when the highest cumulative number
B I O L O G I C A L C O N S E RVAT I O N
285
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Table 1 – Sources, spatio-temporal settings, and applications of data used to analyze seasonal oceanographic and stranding patterns Source #
Data type (sorted by month)
References
Setting (year(s) of data used and region) Chesapeake Bight (Cape Henlopen, DE to Cape Hatteras, NC), 1960–1970, with focus on 1963–1964 Atlantic Continental Shelf (25–44°N, 64 to 81°W), 1961–1964 Diamond Shoals Navigation Light (offshore from Cape Hatteras, NC), 1995–1999 South Atlantic Bight (Cape Romain, SC to Cape Hatteras, NC), 1999 estimates NC shoreline, 1995–1999 with focus on ocean-facing beaches
1
Surface drift bottle returns
Harrison et al. (1967)
2
Ballasted and surface drift bottles
Bumpus (1973)
3
Wind direction and magnitude
National Data Buoy Center
4
Nearshore oceanographic processes, real topography and bathymetry Sea turtle strandings
Werner et al. (1999)
5
R. Boettcher (unpublished data)
Seasonal parameter of interest % Recovery to shore
% Recovery to shore
Input into Werner et al. (1999) oceanographic model Estimates of nearshore water movements Observed point locations over time
Table 2 – Monthly and seasonal sea turtle strandings (all species) on inshore and ocean-facing beaches in North Carolina, 1995–1999 Season
Sp
Sp
Sp
Su
Su
F
MF
MF
W
W
W
W
Total
Month/year
M
A
M
J
J
A
S
O
N
D
J
F
1995 1996 1997 1998 1999
2 9 17 2 20
14 8 14 7 37
80 120 70 29 82
93 103 78 67 120
46 46 56 56 42
20 57 24 41 44
20 23 21 13 20
23 39 32 30 28
9 32 51 56 104
26 23 20 34 56
11 35 17 20 24
3 10 8 3 28
347 505 408 358 605
Total
121
189
770
921
605
439
227
388
641
542
227
112
5182
Note that these numbers do not include incidentally captured turtles (observations of a direct interaction between a sea turtle and a lawfully conducted human activity such as dredging or commercial/recreational fishing). Abbreviations are as follows: Sp, spring; Su, summer; F, fall; MF, Mariner’s fall; W, winter.
of strandings occur (Table 2). For this analysis, we divided each bay into approximate halves by finding the straight-line distance between capes, located the midpoint of that line, and drew a perpendicular line to the coast. We then counted the number of strandings within the eastern and western halves.
2.2.
Drift bottle data
We reevaluated previously conducted oceanic drift bottle experiments (Harrison et al., 1967; Bumpus, 1973) to infer recovery rates of both surface and bottom drift bottles over time and space. Harrison et al. (1967) released surface drift bottles (n = 11,052) in one deployment at distances inside the Chesapeake Bight (CB), a 300 + km extent of the southern Mid-Atlantic Bight (MAB) stretching between Cape Henlopen, Delaware (DE) and Cape Hatteras, NC. Similarly, Bumpus (1973) released 165,566 ballasted and surface drift-bottles (e.g., eight ounce (237 ml) soda bottles) from 1960 to 1970 at 78 locations widely distributed over the continental shelf (25 to 44°N, 64 to 81°W). For the purpose of calculating a bottle stranding rate, we evaluated the probability of bottles making landfall from various points offshore.
2.3.
Wind data
Wind speed magnitude (m/s) and direction (degrees clockwise from north) were recorded hourly via anemometer at a Coastal-Marine Automated Network (C-MAN) station, located at 35.15°N, 75.30°W off Cape Hatteras, NC. To attain representative wind values, we downloaded 15 years (encompassing the years for which we had sea turtle strandings records) of hourly wind data from the station via the National Data Buoy Center website (http://www.ndbc.noaa.gov), transformed winds into vector format, and then resolved each wind’s vector into its corresponding u (easterly) and v (northerly) components. The magnitudes of each wind component, originally recorded as velocities (m/s), were transformed into forces (Pascals) using a specialized Matlab script (Blanton et al., 1985). The resulting easterly and northerly wind stress components were then converted from Pascals into Dynes/ cm2. We calculated the stress magnitude (r) via vector addition of the u and v directional components by means of the Pythagorean theorem (sqrt[(u2) + (v2)] = r). Similarly, by taking the arctangent of v/u, we determined wind stress direction (h). After obtaining the model’s output for True North (0°) and True East (90°) winds, we used these directions
286
B I O L O G I C A L C O N S E RVAT I O N
2.4.
Oceanographic model
The underlying structure of the Werner et al. (1999) oceanographic model is a finite element matrix, serving as the domain for the South Atlantic Bight Recruitment Experiment (SABRE) flow-field computer model that is capable of estimating water current paths (Werner et al., 1999). Each node in the matrix was designed to respond to a wind stress applied to the water’s surface. Node responses are visually displayed as vectors, indicating the hypothetical flow of the water body under prescribed conditions. The spatial domain of the matrix incorporates the continental shelf and slope of the MAB and SAB. We focused exclusively on the model domain in proximity to NC.
3.
Results
3.1.
Sea turtle strandings
Filtering all NC sea turtle strandings data for those that occurred along ocean-facing beaches during 1995–1999 (Table 2) produced point locations for over 1300 Loggerhead and Kemp’s ridley turtles combined (Fig. 1). Whereas almost half of these strandings occurred in May and June, less than 8% occurred from December to March, and more than 75% of the observed strandings were Loggerheads. The numbers and spatial locations of stranded carcasses varied monthly and seasonally but patterns were consistent from year to year. During late spring and early summer (May and June), disproportionate numbers of carcasses stranded along the eastern stretches of both Onslow and Raleigh Bays. Inside Onslow Bay, over four times as many turtles stranded along its eastern half compared to the western side, with chi-square results v2 = 22.93, v2critical ¼ 3:84 and p < 0.05. Similarly, for Raleigh Bay, we also found a significant difference in stranding patterns with chi-square results v2 = 4.83, v2critical ¼ 3:84 and p < 0.05, though it was not as prominent as for Onslow Bay.
3.2.
Drift bottle data
Bumpus (1973) demonstrated low recovery rates of surface drift bottles released during 1961–1964 from November to February within the MAB and SAB. Likewise, Harrison et al. (1967) found that drifter returns from June 1963 to October 1964 demonstrated monthly variation, with the lowest returns during the winter when less than 1% (16 out of 2592) of the bottles released over the entire study area were recovered. Harrison et al. (1967) hypothesized that northwesterly winds, prevailing from late November into February, moved waters offshore and to the south. Harrison et al. (1967) discovered that during the rest of the year, large numbers of bottle recoveries most often coincided with periods of onshore winds, and, regardless of monthly variation, the majority of drift bottles were recovered within approximately two weeks of release. Furthermore, percent recovery of surface drift bottles was inversely related to distance from shore (Figs. 2 and 3). This relationship was strongest during the summer, moderate during spring and fall, and weakest during the winter (Fig. 2).
50
40 Bottles recovered (%)
to approximate the model’s response for any given wind direction, without having to run the model separately for each different wind direction.
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
October-63 January-64 April-64 June-64
30
20
10
0 0
50
100
150
Distance released from shore (km)
Fig. 2 – Recovery of surface drift bottles released inside the Chesapeake Bight (modified from data presented in Harrison et al., 1967). Corresponding N and r2 values were: 636 and 0.925 (October 1963), 618 and 0.617 (January 1964), 660 and 0.925 (April 1964), and 660 and 0.965 (June 1964).
300 250 200
C. caretta
150
L. kempii
100 50 0 jan feb mar apr may jun
jul aug sep oct
nov dec
Month
Fig. 1 – Ocean-beach strandings of Loggerhead (C. Caretta) and Kemp’s ridley (L. Kempii) sea turtles in NC, 1995–1999 (derived from unpublished data provided by R. Boettcher). Error bars depict one standard error.
Landfall probability (%)
# of stranded turtles
350
60 50 40 30 20 10 0 1
5
10
20
30
40
50
60
70
80
90
100
Distance from coast (km)
Fig. 3 – Stranding probability of bottles making landfall according to distance released from shore (derived from data presented in Harrison et al., 1967).
B I O L O G I C A L C O N S E RVAT I O N
287
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Table 3 – Monthly averaged wind stress magnitude and direction values at Diamond Shoals (Cape Hatteras, NC) from 1985 to 1999 Season Spring Spring Spring Summer Summer Fall Mariner’s fall Mariner’s fall Winter Winter Winter Winter
Month
Magnitude (dynes/cm2)
Direction (degrees)
Direction (bearing)
Sample size (# of hourly recordings)
March April May June July August September October November December January February
0.049 0.052 0.025 0.055 0.194 0.010 0.008 0.061 0.039 0.172 0.127 0.113
123.4 92.9 86.7 47.9 48.3 39.5 218.8 197.2 138.1 141.9 127.3 139.3
E–SE E E NE NE N–NE S–SW S–SW SE SE E–SE SE
11,114 11,426 11,758 11,370 11,126 9746 10,041 10,278 10,198 10,884 11,360 9923
Wind stress direction is presented in degrees clockwise from true North.
3.3.
Wind data
Inspection of averaged monthly winds at Diamond Shoals from January to December 1985–1999 (Table 3) revealed that consistently, southeastern and east–southeastern winds predominated from November to February (Winter). General wind direction gradually shifted to the east during March, April and May (Spring), and a northeastern wind pattern developed in June and July (Summer), followed by a north– northeastern transition period in August (Fall). South–southwestern winds persisted in September and October (Mariner’s fall). However, wind variability was high at the monthly scale; standard deviations for monthly wind stress components (data not shown) were sometimes as large, or larger than, their corresponding mean values.
3.4.
Oceanographic model
Vector flow fields (generated by using mean wind stress values and running the physical oceanographic model for each month) represented residual changes in surface water movement predicted at model nodes in accordance with monthly wind patterns. Model responses off the NC coast were as follows, divided into four seasonal wind regimes and five seasons, as originally described by Weber and Blanton (1980) for the SAB:
3.4.1.
Winter (November–February)
Surface water regimes differed north and south of Cape Hatteras. South of the Cape, the model predicted offshore movement of surface waters. North of Cape Hatteras, however, the along-shelf component persisted through the winter months, where net water movement assumed a southward track.
3.4.2.
Spring (March–May)
Modeled conditions in March favored the development of northward along-shelf flow south of Cape Hatteras. The net water flow north of Cape Hatteras reversed direction from the winter months, also taking a northward course. The model also predicted concentrated onshore flow along the eastern edges of the Carolina Bays.
3.4.3.
Summer (June–July) and fall (August)
The northward along-shelf surface flow persisted along the entire coast during these three months. In addition, net flow magnitude intensified throughout the region, peaking in July.
3.4.4.
Mariner’s fall (September–October)
In September, along-shelf flow reversed direction from its summer and fall course. This change in net direction was maintained through October, and later dissipated in November with the onset of offshore flow.
4.
Discussion
The lowest number of sea turtle strandings (Fig. 1 and Table 2) and recoveries of surface drift bottles (Fig. 2) consistently occurred during Winter conditions (e.g., January) with offshore flow. Thus, we conclude that there appears to be a predictable pattern of strandings-favorable seasons and strandingsunfavorable seasons, regardless of whether turtles or bottles are the stranded objects. The climatology for 1995–1999 is qualitatively similar to the long-term climatology for the area so strandings and recoveries for the time period under consideration are comparable to other years and decades. In continental shelf dynamics, although the magnitude of the along-shelf component can be characteristically greater than that of the cross-shelf component (Hare et al., 1999), net offshore flow may contribute to the decreased numbers of strandings and recoveries recorded in Winter months by transporting floating carcasses or bottles away from shore. According to oceanographic model results, such offshore surface flow was most prevalent south of Cape Hatteras from November to February, which corresponded to the season with the lowest number of recorded strandings in the study site. Stranding incidence increased during the Spring (from March to May), when the model predicted the development of the along-shelf current. However, during the winter north of Cape Hatteras, an established southern alongshore current was present despite few strandings. Alternative explanations for decreased strandings recorded in Winter in the study site include, but are not limited to, seasonal turtle migrations towards warmer waters (Coles and Musick, 2000), colder sea
288
B I O L O G I C A L C O N S E RVAT I O N
surface temperatures (which may prolong drowning times for air-breathing poikilotherms (Murphy and Hopkins-Murphy, 1989)), decreased trawling effort or spatially displaced fishing effort, and a generally landward shift of the Gulf Stream during Winter months. Our modeled results elucidate that there are certain months and seasons when strandings are a likely proxy for local marine turtle bycatch, and they support the claim of Epperly et al. (1996). These researchers argued that the number of sea turtle carcasses stranded in proximity to the southern flounder trawl fishery, which operates offshore in winter, was not a reliable indicator of turtle mortality at-sea because turtles killed in Winter in this fishery would not likely appear on land as stranded individuals. Rather, turtle carcasses would more likely be entrained in offshore-flowing currents and swept out to sea. Additional supporting evidence for this comes from aerial surveys, incidental take data from commercial fishers, and satellite tracking studies. Such studies have shown that sea turtles are relatively more abundant near convergence zones (Lutcavage et al., 1997; Polovina et al., 2000), which are farther offshore during Fall and Winter (Murphy and Hopkins-Murphy, 1989; NRC, 1990; Epperly et al., 1995b). This zone is close to the western boundary of the Gulf Stream (Hoffman and Fritts, 1982), and along the shelf break near the 200 m isobath of the MAB (Lutcavage et al., 1997). Thus an absence of sea turtles in coastal waters may very well result in fewer strandings recorded in Winter (NRC, 1990). However, regardless of the time of year, an absence of stranded carcasses may (or may not) coincide with an absence of turtle mortality sources. For example, shrimp fishing activity may begin seasonally during strandings-favorable conditions, implicating local shrimpers if strandings are high, whereas flounder trawl fisheries may take turtles that ever seldom strand due to strandings-unfavorable conditions in winter (Epperly et al., 1995a; NRC, 1990). This has important consequences for managers working to implement fishery management strategies that include reducing sea turtle bycatch (i.e., those fisheries operating closest to strandings locations may not be the cause of the strandings). In addition to seasonal concordance with sea turtle strandings, bottle recovery data from Harrison et al. (1967) suggest that strandings are strongly and negatively related to increased distance released from shore. This relationship was strongest during the Summer, moderate during Spring and Fall, and weakest during the Winter (Fig. 2). This predictable pattern of strandings-favorable and strandings-unfavorable seasons was present regardless of whether turtles or bottles were the stranded objects under examination; again, the lowest recoveries for both sea turtle carcasses and surface drift bottles consistently occurred during winter conditions (e.g., January) with offshore flow. Although direct comparisons between drifting bottles and sea turtle carcasses can be made, the difference in size may make stranded turtles easier to detect on shore, larger surface area of floating turtles could provide greater area upon which wind stress may directly act, and the organic composition of carcasses increases their susceptibility to air and water temperature effects (i.e., facilitating or inhibiting bacterial growth and activity) or to disarticulation by scavengers, both of which may considerably alter time spent afloat. Despite these potential biases,
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
the landfall probabilities extrapolated from drift bottle experiments may have direct relevance to sea turtle management. We maintain that drift bottle returns can provide an upper bound to describe how far sea turtle carcasses could theoretically travel and how likely those carcasses are to make landfall from points offshore. To recreate the stranding probability of drift bottles by distance in the CB, we extrapolated values from log equations derived from data published by Harrison et al. (1967). Drift bottles released at a distance of 20 km had approximately 21% chance of making landfall (Fig. 3), implying that only bottles released very close to shore had the highest probability of stranding. Additionally, Harrison et al. (1967) found that the majority of drift bottles recovered in their study were found within two weeks of the release date. We use these estimates to put into context the percentage of turtles dead at-sea that might strand (about 20%) and the timeframe for stranding (within two weeks). Intuitively, objects released farther from shore should have a lower probability of making landfall compared to objects released closer to shore; we use drift bottle return data (i.e., after Bumpus, 1973) to go further and suggest that, on average, the number of carcasses stranded on ocean-facing beaches may represent, at best, approximately 20% of the total number of available carcasses at sea (Fig. 3). Managers should heed these complementary lines of evidence that indicate that only turtles killed very close to the shore (i.e., within 10 s of kms) may be likely to strand, even in stranding-favorable conditions. Furthermore, if carcasses are going to strand, then, according to the results from other drift bottle experiments (e.g., Harrison et al., 1967), they would probably do so within two weeks after achieving buoyancy. This information indicates that ISLs need to be reevaluated because incidental take of sea turtles from offshore fisheries may not be reflected by onshore stranding events. We are confident that the physical oceanographic model qualitatively predicts the behaviour of near-shore currents in which turtle carcasses could be entrained, especially shoreward of the 40-m isobath. However the model cannot predict the specific locations of stranding events or mortality sources. But by using monthly flow fields predicted by the model and by relying on empirical evidence from surface drift bottle experiments, we can describe large-scale seasonal patterns in coastal waters of NC. Our comparative approach allows us to address one of the fundamental issues in strandings research; that is, how did the stranded animal get to its position on land? Because strandings data differ across species and ocean basins, our interpretation of NC sea turtle strandings data may not characterize strandings trends elsewhere. Our intention is to encourage managers to consider nearshore physical oceanographic processes when interpreting spatial and temporal trends in strandings of marine animals, especially when implicating fisheries whose bycatch of sea turtles may or may not be realized. For example, in NC, sea turtle carcasses may be more likely to strand from April to October along ocean-facing beaches due to the development of an along-shelf current parallel to the coast. Likewise, carcasses seem least likely to strand from November to February due to the presence of offshore flow. Moreover, relative to their recorded stranding position on the beach within the study area, sea turtles are more likely to have been killed in areas to the
B I O L O G I C A L C O N S E RVAT I O N
south from April to August and to the north from September to October due to the net direction of along-shelf flow. By providing a crude estimate of the spatial direction of carcass transport, the Werner et al. (1999) model could be used to investigate potential up- or down-stream mortality sources during periods of sustained along-shelf flow. If we are better able to predict overlap of sea turtle distribution and mortality sources and we have an understanding of how oceanographic and wind conditions affect the flow of carcasses at sea, we may be better able to interpret perceived trends in strandings numbers, not only for sea turtles, but also for other marine animals.
Acknowledgments Brian Blanton graciously provided assistance with model runs and Matlab, and Patrick Halpin, Caterina D’Agrosa, and Ari Friedlaender provided guidance and assistance with GIS. Ruth Boettcher provided strandings data, and Sheryan Epperly, Victoria Thayer, and Lesley Thorne commented on earlier drafts of the manuscript. We thank Francisco Werner for providing access to the physical oceanographic model. We acknowledge the National Marine Fisheries Service, Pew Charitable Trusts, National Oceanographic Partnership Program, Sloan Foundation’s ‘‘Census of the Sea’’, and the Edna Bailey Sussman Foundation for support. We dedicate this paper to Dean Bumpus whose massive empirical work on coastal currents may contribute to better protection of sea turtles and other endangered marine species. This work was part of the Master’s requirement for P.M. at Duke University.
R E F E R E N C E S
Amos, A., 1989. Trash, debris, and human activities: potential hazards at sea and obstacles to Kemp’s ridley sea turtle nesting. In: Caillouet, C., Jr., Landry, A., Jr., (Eds.), Proceedings from the First International Symposium on Kemp’s Ridley Sea Turtle Biology. Texas A&M University, Galveston, p. 42. 105pp. Bibby, C.J., Lloyd, C.S., 1977. Experiments to determine the fate of dead birds at sea. Biological Conservation 12, 295–309. Blanton, J., Schwing, F., Weber, A., Pietrafesa, L., Hayes, D., 1985. Wind stress climatology in the South Atlantic Bight. In: Atkinson, L., Menzel, D., Bush, K. (Eds.), Oceanography of the Southeastern US Continental Shelf, Coastal and Estuarine Sciences, vol. 2. American Geophysical Union, pp. 23–32. Bumpus, D., 1973. A description of the circulation on the continental shelf of the east coast of the United States. In: Warren, B. (Ed.), Progress in Oceanography, vol. 6. Pergamon Press, Oxford, pp. 111–157. 193. Coles, W., Musick, J., 2000. Satellite sea surface temperature analysis and correlation with sea turtle distribution off North Carolina. Copeia 2, 551–554. Crouse, D.T., 1999. The consequences of delayed maturity in a human-dominated world. In: J.A. Musick (Ed.). Life in the Slow Lane: Ecology and Conservation of Long-lived Marine Animals. American Fisheries Society Symposium 23, Bethesda, MD, pp. 195–202. Crowder, L.B., Crouse, D.T., Heppell, S.S., Martin, T.H., 1994. Predicting the impact of turtle excluder devices on loggerhead sea turtles. Ecological Applications 4, 437–445.
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
289
Crowder, L., Hopkins-Murphy, S., Royle, J., 1995. Effects of turtle excluder devices (TEDs) on loggerhead sea turtle strandings with implications for conservation. Copeia 4, 773–779. Epperly, S., Braun, J., Chester, A., Cross, F., Merriner, J., Tester, P., 1995a. Winter distribution of sea turtles in the vicinity of Cape Hatteras and their interactions with the summer flounder trawl fishery. Bulletin of Marine Science 56 (2), 547–568. Epperly, S., Braun, J., Veishlow, A., 1995b. Sea turtles in North Carolina waters. Conservation Biology 9 (2), 384–394. Epperly, S., Braun, J., Chester, A., Cross, F., Merriner, J., Tester, P., Churchill, J., 1996. Beach strandings as an indicator of at sea mortality of sea turtles. Bulletin of Marine Science 59 (2), 289–297. Garshelis, D.L., 1997. Sea otter mortality estimated from carcasses collected after the Exxon Valdez oil spill. Conservation Biology 11 (4), 905–916. Hare, J., Quinlan, J., Werner, F., Blanton, B., Govoni, J., Forward, R., Settle, L., Hoss, D., 1999. Larval transport during winter in the SABRE study area: results of a coupled vertical larval behaviour – three-dimensional circulation model. Fisheries Oceanography 8 (2), 57–76. Harrison, W., Norcross, J., Pore, N., Stanley, E., 1967. Circulation of shelf waters off the Chesapeake Bight. Environmental Science Services Administration (ESSA) Professional Paper 3, US Department of Commerce, Washington, DC, 82pp. Hoffman, W., Fritts, T., 1982. Sea turtle distribution along the boundary of the Gulf Stream current off eastern Florida. Herpetologica 38 (3), 405–409. Hyrenbach, K.D., Baduini, C.L., Hunt Jr., G.L., 2001. Line transect estimates of short-tailed Shearwater (Puffinus tenuirostris) mortality in the Southeastern Bering Sea: 1997–99. Marine Ornithology 29, 27–34. Lewison, R.L., Crowder, L.B., Shaver, D.J., 2003. The impact of turtle excluder devices and fisheries closures on loggerhead and Kemp’s ridley strandings in the western Gulf of Mexico. Conservation Biology 17 (4), 1089–1097. Lutcavage, M., Plotkin, P., Witherington, B., Lutz, P., 1997. Human impacts on sea turtle survival. In: Musick, J., Lutz, P. (Eds.), The Biology of Sea Turtles. CRC Press, Boca Raton, pp. 387–409. 432pp. McLellan, W.A., Friedlaender, A.S., Mead, J.G., Potter, C.W., Pabst, D.A., 2002. Analysing 25 years of bottlenose dolphin (Tursiops truncates) strandings along the Atlantic coast of the USA: do historic records support the coastal migratory stock hypothesis? Journal of Cetacean Research and Management 4 (3), 1–8. Murphy, T., Hopkins-Murphy, S., 1989. Sea Turtle and Shrimping Interactions: A Summary and Critique of Relevant Information. Center for Marine Conservation, Washington, DC. 52pp. National Marine Fisheries Service, 2001. Stock Assessments of Loggerhead and Leatherback Sea Turtles. Technical Memorandum NMFS-SEFSC-455. National Oceanic and Atmospheric Administration, Miami. National Research Council, 1990. Decline of Sea Turtles: Causes and Prevention. National Academy Press, Washington, DC. Piatt, J., Ford, R., 1996. How many seabirds were killed by the Exxon Valdez oil spill? In: Rice, S., Spies, R., Wolfe, D., Wright, B. (Eds.). Proceedings of the Exxon Valdez Oil Spill Symposium. American Fisheries Society Symposium 18, Bethesda, pp. 712–719, 931pp. Polovina, J.J., Kobayashi, D.R., Parker, D.M., Seki, M.P., Balazs, G.H., 2000. Turtles on the edge: movement of loggerhead turtles (Caretta caretta) along oceanic fronts, spanning longline fishing grounds in the central North Pacific, 1997–1998. Fisheries Oceanography 9, 71–82.
290
B I O L O G I C A L C O N S E RVAT I O N
Sis, R., Landry, A., 1992. Postmortem changes in the turtle. In: Proceedings of the 23rd Annual International Association for Aquatic Animal Medicine (IAAAM), Hong Kong. IAAAM, San Leandro, California, pp. 17–19. Steel, R., Torrie, J., Dickey, D., 1997. Principles and Procedures of Statistics: A Biometrical Approach, third ed. McGraw-Hill, New York. 666pp. Turtle Expert Working Group, 1998. An Assessment of the Kemp’s Ridley (Lepidochelys kempii) and Loggerhead (Caretta caretta) Sea Turtle Populations in the Western North Atlantic. Technical
1 2 9 ( 2 0 0 6 ) 2 8 3 –2 9 0
Memorandum NMFS-SEFSC-409. National Oceanic and Atmospheric Administration, Miami. Weber, A., Blanton, J., 1980. Monthly mean wind fields for the South Atlantic Bight. Journal of Physical Oceanography 10 (8), 1256–1263. Werner, F., Blanton, B., Quinlan, J., Luettich Jr., R., 1999. Physical oceanography of the North Carolina continental shelf during the fall and winter seasons: implications for the transport of larval menhaden. Fisheries Oceanography 8 (2), 7–21.