Filtered Streetlights Attract Hatchling Marine Turtles
ABSTRACT
On many nesting beaches, hatchling marine turtles are exposed to poled street lighting that disrupts their ability to crawl to the sea. Experiments were done to determine how hatchlings responded to street lighting transmitted through 2 filters that excluded the most disruptive wavelengths (those < 530 nm; those < 570 nm). Filtered lighting, however, also attracted the turtles though not as strongly as an unfiltered (high-pressure sodium vapor) lighting. Filtering is therefore of limited utility for light management, especially since other alternatives (such as lowering, shielding, or turning off unnecessary lighting; use of dimmer lights embedded in roadways) are more effective.
Hatchling sea turtles emerge from their nests at night (Bustard 1967; Mrosovsky 1968; Witherington et al. 1990) and crawl toward the sea. This behavior, known as “seafinding” (Parker 1922; Daniel and Smith 1947; Carr and Ogren 1960; Ehrenfeld and Carr 1967; Mrosovsky 1972), is based upon 2 orientation cues. Hatchlings crawl toward the brightest area (typically, the seaward horizon) using a positive phototaxis (Mrosovsky 1972; Mrosovsky and Kingsmill 1985). Hatchlings also detect regions elevated above the horizon (such as a tall dune and its associated vegetation). Turtles crawl away from the dune and toward the beach that presents a lower, flatter, horizon (Salmon et al. 1992; Witherington 1992).
Artificial lighting disrupts seafinding orientation. Bright luminaires on land attract turtles so that they crawl toward the lights and away from the ocean (“misorientation” or “light trapping”; Verheijen 1958, 1985). When sources of artificial light are less attractive, hatchlings may show “disorientation”, or an inability to maintain a directional crawl. This response probably occurs when hatchlings simultaneously respond to natural cues and artificial lighting, but cannot orient toward either stimulus (Tuxbury and Salmon 2004).
At night, most marine turtle hatchlings will crawl toward visible light; the shorter wavelengths (violet, blue) are especially attractive (Mrosovsky and Carr 1967; Mrosovsky and Shettleworth 1968). However, this response varies with wavelength among the species. Witherington (1992) and Witherington and Bjorndal (1991a) used monochromatic lights as stimuli in laboratory experiments showing that green turtle (Chelonia mydas), olive ridley (Lepidochelys olivacea), and hawksbill (Eretmochelys imbricata) hatchlings were attracted to wavelengths between 350 and 600 nm (ultraviolet to yellow). Loggerheads (Caretta caretta), however, were attracted to wavelengths between 350 and 500 nm (ultraviolet to green) and were either indifferent to, or repelled by, wavelengths between 530 and 700 nm (green-yellow to red).
These results led to the hypothesis that lights containing only green-yellow to red wavelengths (530–700 nm) would not attract loggerhead hatchlings or interfere with their orientation. Assuming this hypothesis was correct, the Florida Power and Light Company (FPL) installed light filters (orange acrylic sheets) in hundreds of pole-mounted streetlights bordering coastal roadways in South Florida. Without the filters, these lights disrupt seafinding because they are frequently visible at nest sites, and because their high-pressure sodium vapor (HPS) luminaires emit wavelengths shorter (as well as longer) than 530 nm (Witherington and Bjorndal 1991b). Orange filters exclude the shorter but transmit the longer light wavelengths.
Filtered lighting consists of a spectrum of longer (orange to red) wavelengths. Monochromatic light within these wavelengths evokes either indifference or aversion from loggerhead hatchlings tested under laboratory conditions (Witherington 1992). The present study therefore had 2 goals. The first was to determine if loggerheads showed indifference and/or aversion when presented with a spectrum of orange or red (filtered HPS) wavelengths. The second goal was to determine how other species (represented here by green turtle hatchlings), attracted to longer monochromatic light wavelengths than loggerheads, responded to the same stimuli.
METHODS
Hatchlings
Hatchling loggerhead turtles and green turtles were obtained during the 2000 nesting season from previously marked nests at Coral Cove Beach in Palm Beach County, Florida (26°57′N, 80°05′W). These hatchlings were used in arena experiments conducted nearby at the Marinelife Center, Juno Beach, Florida. Hatchlings during the 2001 season were obtained from nests relocated to a hatchery at the Hillsboro Club, Broward County, Florida (26°18′N, 80°05′W). These turtles were used in T-maze experiments conducted nearby at Florida Atlantic University in Boca Raton, Florida. All turtles were transported from the collection site in polystyrene foam containers that allowed for air exchange but were covered with a black cloth to exclude light. Experiments were done in air-conditioned, dark rooms rendered lightproof by sealing all windows with black plastic sheeting.
During both years, hatchlings were collected in the afternoon of the evening when they were scheduled to emerge. Date of emergence was estimated by adding 55 days to the egg deposition date. Turtles were used only if they were captured within 15 cm of the sand surface, and if their plastron was “flat” (indicating they were ready to emerge and developmentally competent to migrate offshore). To ensure genetic diversity, hatchlings from at least 2 nests were used each evening, and all tests were repeated over 2 or more evenings.
Before testing, turtles were held in a dark room for several hours at ambient temperatures (27°–30°C), until dusk. They were then briefly (ca. 15 min) exposed to dim light and slightly cooler temperatures to stimulate locomotor activity (Bustard 1967; Mrosovsky 1968). Each turtle was used in a single trial, then released later that evening on a nearby dark beach, in accordance with state guidelines (Florida Department of Environmental Protection 1996).
Light Measurements
Light was measured (in fractions of watts) with a radiometer (Model 351; United Detector Technology, Baltimore, MD) that had a uniform response between 400 and 700 nm. An Optec stellar photometer (Optec, Inc, Lowell, MI; Model 351 with a 16° angle of acceptance; range of 300-1100 nm; peak sensitivity at 520 nm) was used to determine relative street light radiance at coastal roadways. Both the radiometer and stellar photometer were calibrated against a 500-nm light of known intensity across a 3 decade range of light amplitudes. Measurements from both instruments were then converted to a common scale of absolute radiance (in photons·cm−2·s−1 at 500 nm).
Arena Experiments
The arena was a circular horizontal platform used to determine how green turtle and loggerhead hatchlings responded to HPS and to filtered HPS light (Fig. 1). It was made of rough-textured plywood, painted light brown (sand) in color. It contained 60 cloth-lined pockets at the periphery sufficiently large to trap each turtle that had crawled there from the arena center (where each turtle was released). Crawling vectors for each turtle were measured by the angle between the light source, the center of the arena, and the pocket. The pocket in front of a light was arbitrarily designated as 0°.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
The light (an attenuated 70-W HPS light enclosed in a small wooden box) was located 1.87 m from the arena center and elevated 40 cm above the platform surface (Fig. 1). Light escaped from the box through a small pinhole opening made in aluminum flashing. This configuration matched the elevation and radiance of poled streetlights observed from the location of several nests. Filters were attached to the box over the pinhole. Two identical light boxes were placed near the arena about 90° apart, but during each experiment only 1 light was turned on. The light beam from each box was aimed down at the arena center but could be seen from any location within the arena.
Turtles were exposed to 4 treatments: 1) light turned off (turtles crawl in complete darkness), 2) HPS light on, 3) HPS light transmitted through a 2422 amber filter (excludes wavelengths < 530 nm), and 4) HPS light transmitted through a NLW red filter (excludes wavelengths < 570 nm; Fig. 2). Both filters were made of dyed plastic sheets manufactured by the General Electric Lighting Corporation (Lexington, KY).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
Equal numbers of green turtles and loggerheads from several nests were exposed to each treatment until a target sample size (n = 30 turtles/treatment) was achieved for each species. A trial began with the release in the arena center of 5 hatchlings (to simulate the typically simultaneous emergence of several turtles from one nest) contained inside a black cloth bag and ended after all of the turtles had fallen into a pocket. Crawling progress was monitored using a video camera and monitor (suspended out of sight of the turtles, above the arena). In the “no light” treatment, records were videotaped under infrared illumination.
Because filtering excluded some wavelengths, filtered radiance was lower than unfiltered (HPS) radiance. The intensities of the light stimuli used (as measured at the arena center) were (in photons·cm−2·s−1): HPS, 12.0 × 1012; 2422 filter, 7.0 × 1012; and NLW, 7.0 × 1012. The 2422 intensity fell within the range of values measured in the field from several filtered 70-W cut-off fixtures, mounted 60 m distant from nests on 10-m-tall street poles.
T-Maze Experiments
This apparatus was used to study the response of the turtles to lighting in choice situations (Fig. 3). Turtles crawled down the runway toward a white plate that reflected light from a single or paired source. Once it reached the “T”, the turtle turned either to the right or left. Twenty-five hatchlings from 2 or more nests were used in each trial. Because green turtle hatchlings were less abundant, only loggerheads were used in these tests.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
The same light boxes were used to house each light. One box was placed on each side of the maze and positioned so that its illumination was only directly visible to turtles that had crawled to the end of the runway (Fig. 3). Light intensity (as measured at the position of the “T”) of the HPS light was adjusted to match an unfiltered 70-W HPS streetlight located either 40 or 60 m distant from a nest.
The turtles were exposed to 4 treatments: 1) a single HPS light from either the right or left side of the maze; 2) a single filtered HPS (2422 or NLW) light from either the right or left side; 3) paired HPS and filtered HPS (2422 or NLW) lights (as in Fig. 3) presented from opposite sides of the T-maze; and 4) paired filtered HPS (2422 and NLW) lights presented from opposite sides of the T-maze. In the paired light treatments, each treatment was replicated 4 times: with 1 of the 2 luminaires (HPS or the 2422 filtered HPS light) presented at full intensity, or with their intensity reduced by 1, 2, or 3 log units using neutral density filters (made from layered plastic hardware cloth).
In control experiments, hatchlings were exposed to pairs of lights (HPS or filtered HPS) adjusted to an identical radiance. These tests were done to confirm that no variable other than light was responsible for deviations from an expected 50:50 turning ratio.
Statistics
Crawl vectors for each turtle in a single arena treatment were used to calculate a second-order group mean angle and r-vector (measure of dispersion). Rayleigh tests (Zar 1999) were employed to determine whether groups were significantly oriented (p ≤ 0.05).
The number of hatchlings that turned right or left was recorded for each T-maze treatment. The null hypothesis of a 50:50 turning distribution was rejected when that distribution resulted in a p ≤ 0.05 (by a binomial test; Sokal and Rohlf 1995).
RESULTS
Arena Experiments
Neither the loggerheads (Fig. 4) nor the green turtles (Fig. 5) showed significant orientation when tested in darkness. Turtles exposed to a HPS or to a filtered (2422 or NLW) HPS light were strongly attracted to the stimulus.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
T-Maze Experiments
None of the control turtles showed a distribution of turns that deviated statistically from an expected 50:50 ratio. Hatchlings were significantly attracted to a single HPS light at the 40- and 60-m radiance levels (Table 1). Both types of filtered lighting (orange 2422, red NLW) failed to result in a significant attraction at the higher (40-m) radiance level. However, at the lower (60-m) radiance, the turtles were attracted to each light source. Attraction to filtered lighting was weaker (76% of the turtles to the 2422; 84% to the NLW) than attraction to HPS lighting (96%; Table 1).
More turtles turned toward the HPS light when it was paired with either a 2422 (Fig. 6) or a NLW (Fig. 7) filtered light. When HPS radiance was reduced by 2 log units, the number of turtles that turned toward each light did not differ statistically (Figs. 6 and 7). When HPS radiance was reduced by 3 log units, more turtles turned toward the 2422 filtered than the HPS light (Fig. 6). This trend was also evident when the HPS light was paired with a NLW light, though the probabilities just missed significance (Fig. 7).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
Turning tendencies shown in response to paired filtered lights were statistically equal (Fig. 8). When the 2422 source was reduced in radiance by 3 log units, more turtles turned toward the NLW light (Fig. 8).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[255:FSAHMT]2.0.CO;2
DISCUSSION
Response to HPS and Filtered Lighting
Our arena experiments showed that in the absence of lighting, green turtle and loggerhead hatchlings did not show significant group orientation. However, in the presence of lighting the turtles of both species crawled toward the source (Figs. 4 and 5). We conclude that both HPS and filtered HPS lighting attract sea turtle hatchlings.
Our T-maze experiments (Figs. 6 and 7) indicate that HPS lighting is more attractive to loggerheads than filtered HPS lighting. Two variables might account for these results. The first is light intensity because initially (in the 0:0 tests), the HPS light was slightly brighter than the filtered HPS light with which it was paired. A second possibility is that the HPS light was more attractive because its spectral composition included some shorter light wavelengths. Intensity was eliminated as a factor by repeating the tests after HPS radiance was reduced by 1 or more log units below the radiance of the filtered light. In response, the turtles either continued to orient preferentially toward the dimmer HPS source, or showed no significant orientation toward either light (Figs. 6 and 7). We conclude that in our experiments, the spectral composition of the HPS light made that stimulus more attractive to the hatchlings than filtered lighting.
Amber (2422) and red (NLW) filters were designed for use with pole-mounted HPS streetlights on roadways adjacent to nesting beaches. In a field experiment, the 2422 filter proved effective with adult nesting female loggerheads (Pennell 2000). However, field experiments with loggerhead hatchlings produced equivocal results (Cowan and Salmon 1998) because of nightly variation in other sources of artificial lighting (skyglow from nearby communities). Because this lighting could not be controlled, it was impossible to distinguish between responses caused by filtered lighting and responses to changes in background illumination. These problems led us to do further testing in a laboratory setting where extraneous sources of illumination could be excluded.
HPS lighting at both a higher (40-m) and lower (60-m) intensity attracted the turtles, but the hatchlings were attracted to filtered lighting only at a lower (60-m) radiance level (Table 1). In addition, the NLW light at 60 m was apparently more attractive to the turtles than the 2422 filtered light at 60 m (Table 1). Yet the NLW light excluded a larger proportion of the wavelengths around 530 nm that elicit “indifference”, while leaving present those wavelengths (> 570 nm) that elicit “aversion” (Witherington 1992).
The explanation for these responses may center on how hatchling loggerheads respond to different intensities, rather than wavelengths, of light. In previous experiments, Witherington (1992) found that responses such as “attraction”, “indifference”, and “aversion” were elicited at relatively high (perhaps photopically mediated) light levels. At lower light levels (perhaps mediated scotopically), all wavelengths of monochromatic light were attractive to hatchling loggerheads. We hypothesize that the 40-m light stimulus was sufficiently intense to permit wavelength discrimination (and indifference or aversion), whereas the 60-m light stimulus was not (and therefore attracted the turtles).
These results reveal some of the complexities associated with using filtered lighting as a management tool. Filtered lighting may be unattractive to hatchlings when they emerge from their nests because the light source is in close proximity, and therefore more intense. But as the turtles crawl away from the light (and toward the sea), the light source decreases in perceived intensity and could, as a consequence, become attractive. To properly assess the impact of filtered lighting on turtles at any location, then, hatchlings must be exposed to the entire range of light intensities they encounter as they crawl from their nests (and the light) toward the sea.
Management Implications
Ideally, filtered HPS lighting should have no effect on the orientation of hatchling sea turtles. Our results show, however, that hatchlings can, under some circumstances, be attracted to filtered lighting. Other problems are also associated with the use of filtered lighting.
First, responses to filtered lighting probably vary, depending upon the species. The 2422 and NLW filters were developed primarily for use near loggerhead rookeries and are based upon the unique response of loggerhead hatchlings to light wavelengths (Witherington 1992). The few tests that have been done with leatherbacks (Dermochelys coriacea) and green turtles (Cowan and Salmon 1998; Tuxbury and Salmon 2004) suggest that these species respond differently even to the longer wavelengths transmitted by these filters.
Second, the filters currently in use may represent the best technology that can be used with HPS luminaires, which for economic reasons are preferred for street lighting. Excluding any more of the shorter wavelengths transmitted by HPS luminaires may reduce luminance levels below levels required for roadway safety (as mandated by the Florida Department of Transportation [Scott Stephens, Florida Power and Light Co, pers. comm.]). These standards were set by engineers to provide minimum levels of illumination for motorists. However, a variance from these standards can be obtained if the roadway custodian accepts liability for accidents and installs warning signs to notify motorists of poor lighting conditions (Ecological Associates, Inc 1998). Currently, roadways in Florida with lighting that affects nesting beaches are being identified, and new standards are being determined for lighting roadways. Whether filtered lighting can meet those standards remains to be determined.
Third, there are better alternatives for managing coastal roadway lighting. One promising technology is the use of light-emitting diodes placed in the pavement itself (“embedded” roadway lighting). These lights produce far less illumination than streetlights and confine that light to the roadway itself (where it is needed). Field tests were recently done at a coastal roadway where embedded lighting was installed. Their illumination could not be detected at the beach either by humans, their instruments, or by loggerhead hatchlings. Turtles crawled toward the sea when the embedded lights were on and when they were turned off. However, when the poled HPS streetlights were turned on, orientation dispersion (and in some tests, mean angle) were affected (Bertolotti and Salmon, 2005).
Fourth, filtered lighting is a “half-way technology” (Frazer 1992) because it fails to eliminate the cause of the problem (light scatter to the beach); rather, it seeks to alter the impact of that light by modifying spectral output. The only proven methods of light management, however, are to turn off or redirect lighting so that it is no longer visible at the beach (Witherington and Martin 2000).
On the other hand, there are circumstances where filtered lighting might be useful. Hatchling loggerheads are less likely to crawl toward visible lighting if a tall, dark landward silhouette is present (Witherington et al. 1994; Tuxbury and Salmon 2004). Because filtered lighting is less attractive to the turtles, it might be used to illuminate roadways without affecting seafinding, even if some lighting escapes to the beach. However, before such a modification is made permanent, tests must be done at these sites to confirm that the turtles exposed to both filtered lighting and tall silhouettes will complete a seaward crawl.
Filtered lighting may also be beneficial at locations where the public believes that lights prevent crime and/or reduce roadway accidents (Witherington and Martin 2000). Filtered lighting at such a site has 3 benefits. It has a favorable psychological impact on users, and (for the turtles) reduces light intensity while transmitting less attractive spectra to the environment.
New technologies must be explored to determine their potential for reducing the impact of artificial lighting on wildlife. Initially, light filters were a new technology promoted by their manufacturer (General Electric Lighting Corporation) as a simple method for converting harmful, attractive lights into those that were “turtle friendly”. These claims; however, were made in the absence of adequate testing. Having now completed testing, we conclude that at the present time filtered lighting is potentially beneficial only under special, and unfortunately somewhat limited, circumstances.
The arena: light boxes were 90° apart and designed to mimic the radiance of a 70-W HPS streetlight. Boxes were positioned 40 cm (17°) above the arena surface and 1.87 m from the center of the arena.
Transmission characteristics of the General Electric Lighting Company filters. The orange-colored 2422 filter excludes wavelengths < 530 nm; the red-colored NLW filter excludes wavelengths < 570 nm.
The T-maze (overhead view). Hatchlings were released at *, crawled toward the light reflecting barrier at the end of the runway, then turned left or right toward one of the lights.
Response of loggerheads to a 70-W HPS light at 0° (top of each circle). In different treatments, the light is either turned off (A) or on (B–D). When on, it passes through a filter in C and D. Sample size is 30 turtles per treatment. Line length is proportional to the number of turtles orienting in each direction. A, group mean angle; r, dispersion; n.s., no significant group orientation.
Arena experiments with green turtles. Format and sample size, as in Fig. 4.
T-Maze experiments that show the percentage of hatchlings turning toward a HPS light when it is paired with a filtered (2422) HPS light. The 2 lights are initially presented at intensities comparable to a street light at a distance of 40 m (top graph) or 60 m (lower graph) from the turtle (0:0, left side of each graph). In 3 additional treatments, the HPS light is reduced in intensity by 1 (−1:0), 2 (−2.0), or 3 (−3:0) log units while the filtered light remains unchanged in intensity. n = 25 different hatchlings in each treatment. Points falling on or above the upper, or on or below the lower dashed lines are significant statistical departures (at p ≤ 0.05 level) from a 50:50 ratio (by a binomial test).
T-Maze experiments that show the percentage of hatchlings turning toward the HPS light when it is paired with a filtered (NLW) HPS light. Format as in Fig. 6.
T-Maze experiments that show the percentage of hatchlings turning toward a 2422 filtered HPS light when it is paired with a NLW filtered HPS light. Format as in Fig. 6 except that the 2422 filtered light is reduced in intensity using neutral density filters while the NLW light is left unchanged in intensity.
