Intertidal habitats provide important feeding areas for migratory shorebirds. Anthropogenic developments along coasts can increase ambient light levels at night across adjacent inter-tidal zones. Here, we report the effects of elevated nocturnal light levels upon the foraging strategy of a migratory shorebird (common redshank Tringa totanus) overwintering on an industrialised estuary in Northern Europe.
To monitor behaviour across the full intertidal area, individuals were located by day and night using VHF transmitters, and foraging behaviour was inferred from inbuilt posture sensors. Natural light was scored using moon-phase and cloud cover information and nocturnal artificial light levels were obtained using geo-referenced DMSP/OLS night-time satellite imagery at a 1-km resolution.
Under high illumination levels, the commonest and apparently preferred foraging behaviour was sight-based. Conversely, birds feeding in areas with low levels of artificial light had an elevated foraging time and fed by touch, but switched to visual rather than tactile foraging behaviour on bright moonlit nights in the absence of cloud cover. Individuals occupying areas which were illuminated continuously by lighting from a large petrochemical complex invariably exhibited a visually based foraging behaviour independently of lunar phase and cloud cover.
We show that ambient light levels affect the timing and distribution of foraging opportunities for redshank. We argue that light emitted from an industrial complex improved nocturnal visibility. This allowed sight-based foraging in place of tactile foraging, implying both a preference for sight-feeding and enhanced night-time foraging opportunities under these conditions. The study highlights the value of integrating remotely sensed data and telemetry techniques to assess the effect of anthropogenic change upon nocturnal behaviour and habitat use.
Coastal areas are under increasing pressure from human activity and development worldwide. The demand for the construction of industrial complexes, hotels and residential properties throughout the coastal zone is driven primarily by access to the sea. The rate and magnitude of coastal development is increasing on a global scale, often elevating the nocturnal illumination of biologically productive areas. The negative effects of artificial lighting on orientation amongst animals as they move between the sea and the land have been well documented (le Corre et al. 2002; Longcore & Rich 2004; Tuxbury & Salmon 2005). Its effects, however, upon ecological processes, such as predator–prey interactions, are less well understood.
The intertidal zone along exposed coastlines and estuaries provides essential habitat for migrating shorebirds. Many shorebirds also spend the winter on the intertidal, duly establishing body reserves before making the return migration to summer breeding grounds at high latitudes. For those shorebirds overwintering in temperate regions, diurnal feeding alone has been reported to be insufficient to balance energy budgets (Goss-Custard 1969; Turpie & Hockey 1993; Zwarts et al. 1996). This is not only due to increased metabolic demands during adverse weather but also because foraging time is restricted by short day length and fluctuating tides (Goss-Custard et al. 1977; Mouritsen 1994). To feed efficiently throughout the diel cycle, some species of shorebirds switch their foraging behaviour between high and low ambient light levels (Robert, McNeil & Leduc 1989; McNeil, Drapeau & Goss-Custard 1992). During the day, these birds employ visual information to guide bill position when detecting prey (i.e. visual foraging). At night, however, birds switch to using tactile information on prey, obtained by probing or sweeping the bill in soft substrates (Piersma et al. 1998; Martin & Piersma 2010). As sight-based foraging behaviour allows birds to search and select for the most profitable prey items (Goss-Custard 1969; Robert & McNeil 1989), it is generally preferred when natural light levels are adequate (Pienkowski 1983; Robert, McNeil & Leduc 1989).
In this study, the influence of nocturnal light levels on night foraging in shorebirds was examined using an individual-based approach. Posture-sensitive VHF radio transmitters were used to collect quantitative measurements of behaviour and habitat use of free-ranging common redshank Tringa totanus. This provided a non-disruptive method to study animal behaviour and habitat use over extensive and inaccessible tidal flats, even under poor visibility, darkness and inclement weather which would be highly challenging under a conventional visual-based approach. Measuring light levels using hand-held photo-detectors at each shorebird location would have been unreliable as human presence would likely have disturbed the foraging birds. We utilised geo-referenced night-time satellite imagery data obtained from the U.S. Air Force Defence Meteorological Satellite Program (DMSP), with the Operational Linescan System (OLS) to quantify artificial illumination under the night sky (Elvidge et al. 1997; Imhoff et al. 1997; Sutton et al. 1997; Cinzano, Falchi & Elvidge 2001). Night-time satellite data provided by DMSP/OLS were captured on cloudless nights under very low or no lunar illuminance and transformed into values of night brightness (Elvidge et al. 1999; Letu et al. 2010); an archive of global data sets is available online. Although this is the first use of these data known to us for an animal behavioural study, DMSP/OLS maps have been applied to investigate electrical power consumption (Elvidge et al. 1999; Letu et al. 2010), predict offshore fishing vessel density (Waluda et al. 2002; Kiyofuji & Saitoh 2004) and to monitor biomass burning and gas flares (Elvidge et al. 1997).
Using these tools, the relative contribution of ambient light to foraging persistence and behaviour in redshank was examined throughout the middle reaches of the industrialised Forth estuary in Eastern Scotland, U.K. The method allowed us to test whether individuals took advantage of artificial light, in a similar manner to moonlight, whilst determining whether night foraging was contingent on moonlight levels rather than on the influence of the lunar cycle on prey behaviour (Milsom, Rochard & Poole 1990). We hypothesised that the positive effects of light may not simply open up opportunities with respect to extending the daily foraging period, but it may also allow individuals to switch foraging modalities (e.g. from touch to sight) and thus make nocturnal foraging relatively more efficient.
Materials and methods
Capture and Transmitter Attachment
Between 14 November 2008 and 15 February 2009, we captured juvenile 20 redshank (under licence) by mist netting at five high-tide roosts on the Forth estuary, Scotland (56oN, 03oW; Fig. 1). Individuals were aged according to plumage characteristics (Prater, Marchant & Vuorinen 1977) and fitted with backpack-mounted transmitters weighing 2·4 g (model Pip2 with Ag393 cell battery; Biotrack U.K. Ltd, Wareham, Dorset, UK), fixed to a sheet of gauze and attached to a small area of clipped feathers on the lower back using cyanoacrylate glue (Loctite® Super Glue; Henkel Consumer Adhesives, Cheshire, UK; Warnock & Warnock 1993). To avoid overloading birds and to minimise the risk of premature detachment, each transmitter weighed no more than 2% of an individual's body mass (mean = 1·48%, max = 1·95%). Every individual fitted with a transmitter was also given a unique combination of Darvic plastic colour-rings to aid subsequent visual identification in the field.
Calibration of Posture Sensors
Each transmitter was fitted with a tilt-switch posture sensor designed to alter the transmitter-pulse rate (hereafter referred to as ‘pulse rate’) when the bird tilted its body when feeding. As the bird changed from a standing (head up) to a feeding (head down) position, the pulse rate switched instantaneously from a slow pulse rate of 45 pulses per minute (ppm), to a fast pulse rate of up to 65 ppm. The angle of the tilt-switch was pre-set in a front-to-back position at 0o to the long axis of the transmitter. This angle was chosen based on measurements taken from taxidermic mounts and from field trials conducted between 2 December 2007 and 28 February 2008. In these trials, we attached transmitters to nine birds where the switch was positioned at 5o intervals between 350o and 10o relative to the long axis of the transmitter to establish the optimum transmitter-angle in which to collect information on foraging activities. The pulse rate was recorded manually by counting the number of pulses over a 1-min period (Fig. S1a).
Between 17 November 2008 and 19 March 2009, observations of individual bird behaviour and records of transmitter-pulse rates were made concurrently. Clearly, it was only possible to make these detailed observations during daylight hours, when a tagged individual was close to the shore. The inferred behaviour based on pulse rate was calibrated for each bird using cumulative instantaneous behaviour sampling. Over a 2-min session, we recorded the behaviour of the bird at 5-s sample intervals, timed using a Korg MA-30 hand-held metronome (Korg U.K. Ltd, Buckinghamshire, UK). Individuals were identified in the field by their colour-ring combinations, and their behaviours recorded as walking, pecking, resting, vigilant, preening, agonistic and flying. No sweeping behaviour was noted during this study. Behaviours were grouped for some analyses according to whether the bird was ‘roosting’ (resting and preening) or ‘foraging’ (walking and pecking). The pulse rate was then compared with the proportion of all sample points on which the behaviour pattern was occurring. As instantaneous sampling is not suitable for recording discrete events of short duration (Martin & Bateson 1993), vigilant, flying and agonistic behaviours were excluded from all analyses.
In our calibration tests, transmitter-pulse rate proved very effective at discriminating between roosting and foraging birds (F1,32 = 47·96, P <0·001; Fig. S1b). There was a positive linear relationship with regard to pulse rate and the proportion of time spent pecking (r2 = 0·59, F1,24 = 35·23, P <0·001; Fig. S1c), and a negative linear relationship with respect to pulse rate and time spent walking (r2 = 0·60, F1,24 = 36·29, P <0·001; Fig. S1d). Based on these calibration measurements, birds in which pulse rate was below 49 ppm were classed as ‘roosting’ and those between 49 and 65 ppm were ‘foraging’.
All tracking of bird behaviour was with a Televilt model RX-81 receiver (Televilt/TVP Positioning AB, Lindesberg, Sweden) at the 173·000–173·999 MHz range with a hand-held three-element Yagi antenna. In order to fix behaviour to a location on the estuary, we located individuals via triangulation, by taking two to three bearings, obtained as quickly as possible to reduce the probability that a focal bird moved between each bearing (mean ± SD: 29 ± 21 min; White & Garrott 1990). When birds were not located visually, three bearings were taken from vantage points along the river and the locations of tagged redshank were triangulated using Location Of A Signal positioning software (LOAS™, Ecological Software Solutions, Sacramento, California, USA; Taft, Sanzenbacher & Haig 2008). A single transmitter-pulse rate was recorded when taking each bearing, the first of which was used for estimating the individual's behaviour for that location.
Detailed searches for tagged redshank were conducted throughout the estuary on 6 days per week and divided into two morning (06.00–12.00 h), two afternoon (14.00–20.00 h) and two night searches (21.00–03.00 h) per week. The order in which sites were visited within the specified times was alternated to provide a representative sample of bird movements throughout the tidal cycle, across spring and neap tides and thus across lunar cycles. Tracking commenced on the day following release and ended when the transmitter's signal was no longer detected after several extensive searches of the estuary. Of the 20 transmitters despatched on redshank in winter 2008–2009, we obtained tracking and behaviour data from 13 juvenile birds covering 724 bird-radio-days (Table S1).The data from these 13 birds were used in our analysis. The signals from five transmitters were lost soon after release (presumably to birds leaving the estuary) and two posture sensors failed mid-way through our study; data from these birds were removed from all analyses. No tagged individuals ‘lost’ during the study period were re-identified during subsequent observations of the estuary. Of the 239 behavioural samples that were obtained during the period of study, 136 (43%) were taken at night, that is after nautical dusk and before nautical dawn. All searches of the estuary conducted on ‘moonlit’ nights occurred after moonrise and before moonset; when the moon was above the horizon.
Tidal data were obtained from the British Oceanographic Data Centre (BODC) collected by the tidal gauge site located at Leith docks, Edinburgh (55·59oN, 03·10oW). Temperature data were supplied by the Meteorological Office from the weather station at Edinburgh Gogarbank (55·93oN, 03·35oW), and nautical twilight dawn and dusk times were obtained from the Astronomical Applications Department, U.S. Naval Observatory. Anthropogenic light imaging data were collected by the U.S. Air Force Defense Meteorological Satellite Program (DMSP; National Geophysical Data Center of the National Oceanographic and Atmospheric Administration, Boulder, Colorado, USA; http://www.ngdc.noaa.gov/dmsp/dmsp.html). Artificial light radiance is reported as W cm−2 sr−1 µm−1 following calibration by Elvidge et al. (1999). The geo-referenced artificial night sky brightness raster layer was supplied at a 1 km2 grid resolution, and the radiance value was extracted for each location using ARCGIS 10 (ESRI, Redlands, CA, USA).
Diurnal and nocturnal home ranges for each individual were generated using the minimum convex polygons technique in ARCGIS. Minimum convex polygons (MCPs) are defined as the smallest convex polygon that incorporates all locations; they have been used extensively in ecological studies of habitat use. The mean radiance value for each diurnal and nocturnal MCP, and the proximity of the MCP centroid (i.e. the median longitude and latitude for all the location data) to the industrial complex at Grangemouth (Fig. 1) was extracted using ARCGIS.
A generalised linear mixed-effect model (GzLMM) with binomial errors was used to investigate the environmental factors driving diurnal and nocturnal foraging in tagged common redshank. In this model, the probability that an individual bird was foraging was taken as the response variable. Investigation of the factors that might influence foraging behaviour were analysed using a general linear mixed-effect model (GLMM) with the original transmitter-pulse rate data as the response variable (see details above). Day|night (=day vs. night), moon phase [new (=waning crescent-waxing crescent) vs. full (=waxing gibbous – waning gibbous)), cloud cover (clear (0–3 oktas) vs. cloudy (4–8 oktas)), artificial light radiance (W cm−2 sr−1 µm−1) were included in our models as fixed effects. The interactions between day|night × moon phase × cloud cover and day|night × artificial light radiance × cloud cover were included in our initial model to investigate the relationship between ambient light and the potential additive effects of cloud cover. Tide height (m) and day length (hours) and their two-way interactions with day|night were also included in the GzLMM as fixed effects. Statistical analysis of telemetry data can be problematic as successive behaviours from the same individuals may lead to data points that are non-independent. Because behaviours were repeated measures from the same individuals, we ensured that at least one full tidal cycle (>12 h) had passed between observations and included Bird ID within our models as random effects. Unimportant variables were removed from models by stepwise deletion: only variables significant at the 5% level were retained in the final model. A stepwise procedure with sequential deletion of variables was chosen as it is known to be a suitably conservative method of variable selection (Murtaugh 2009). Mixed models were constructed using the ‘lme4’ package (Bates & Maechler 2009), and all analysis was conducted in the R programming language (R Development Core Team 2010).
Effect of Ambient Light Levels on Day vs. Night Foraging Time
The proportion of birds classified as foraging by day or by night differed (mean day (SE) = 0·46 (0·19); mean night (SE) = 0·28 (0·27)). In the GzLMM, significant interactions were found between day|night × moon phase × cloud (REML, χ21 = 6·07, P =0·014) and day|night × artificial light radiance in predicting the proportion of time spent foraging by individual redshank (REML, χ21 = 10·67, P =0·001). Day length and temperature had no significant effect on the proportion of time spent foraging by day or by night (REML, χ21 = 0·05, P =0·829 and REML, χ21 = 0·05, P =0·818). As expected, there was no effect of moon phase and cloud cover on diurnal foraging behaviour (Fig. 2). At night, the proportion of time tagged individuals spent foraging on cloudy nights was similar to that in the daytime; however, the moon appeared to influence behaviour on clear nights: birds spent a lesser proportion of time foraging on nights when there was a new moon, and a greater proportion of time foraging on brighter nights when there was a full moon (Fig. 2).
When day and night data were analysed independently, there was no relationship between artificial light radiance and the probability that a bird was foraging during daytime (REML, χ21 = 0·93, P =0·335; Fig. 3a). At night, however, individuals occupying well-lit areas spent a greater proportion of time foraging than those occupying darker areas of the estuary (REML, χ21 = 5·07, P =0·024; Fig. 3b). Home range size was smaller by night than by day (t12 = 2·32, P =0·041; Fig 3c). There was no significant difference in the mean artificial radiance (t12 = 1·95, P =0·078), or in the proximity of polygon centroids to the industrial complex at Grangemouth (t12 = −0·67, P =0·518).
Effect of Ambient Light Levels on Day vs. Night Foraging Behaviour
Using the raw transmitter-pulse rate data for foraging birds only, significant interactions were found between day|night × moon phase × cloud (REML, χ21 = 4·73, P =0·029). Artificial light radiance had no significant effect on transmitter-pulse rate, however, either as a single term or within any higher-level interactions (P >0·05). Birds foraging at night had a transmitter-pulse rate which was on average 4·26 ppm faster than those foraging by day (Fig. 4).
Birds foraging on nights with a new moon and little cloud cover had the fastest transmitter-pulse rate (mean (SE) = 60·23 ppm (1·56); Fig. 5). Artificial light radiance had no significant effect on pulse rate, and no interactions were found between artificial light radiance × day|night (REML, χ21 = 0·44, P =0·508). There was, however, some evidence that pulse rate was influenced by artificial light: the effect of clear moonless nights on transmitter-pulse rate was most pronounced in darker areas of the estuary, where the pulse rate was 3·05 ppm (SE 2·05) faster than the average for the estuary on a clear moonless night. In addition, three of four of the birds which showed little change (or a slight decrease) in transmitter-pulse rate between day and night occupied well-lit areas in and around the petrochemical complex (Fig. 3b). The fourth moved into a relatively well-lit intertidal area located under a road bridge to feed by night, an area which was generally avoided by day.
Studies into the effects of artificial light pollution on animal behaviour are uncommon and generally document negative effects (le Corre et al. 2002; Longcore & Rich 2004; Tuxbury & Salmon 2005). Here, we show that artificial light generated from a large industrial site situated upon an estuary in Eastern Scotland, significantly altered the foraging strategy of common redshank. The 24 h light emitted from lamps and flares created, in effect, a perpetual full moon across the local inter-tidal area. Consistent with the study hypothesis, greater nocturnal illumination increased the potential for birds to forage for longer periods at night and to use a potentially more effective foraging behaviour to locate prey. Our findings parallel observational studies on wetland birds under wholly natural light, including many wildfowl, which similarly take advantage of moonlight to increase foraging opportunities (e.g. Pienkowski 1983; Robert, McNeil & Leduc 1989; Sitters 2000; Tinkler, Montgomery & Elwood 2009). Empirical studies have shown that foraging periods utilising a visual over a touch-based behaviour provides a higher net food intake for inter-tidal feeding birds (Lourenço et al.2008; Santos et al. 2010), and we argue that the effects of artificial lighting upon the Forth estuary enhanced feeding opportunities for redshank.
The tagged redshank did not select illuminated areas for feeding. Rather, they maintained a high degree of site-fidelity to the area where they were captured and tagged (Fig. 3c). Redshanks are known to live within a small home range on their winter feeding grounds (Furness & Galbraith 1980; Burton 2000; Rehfisch, Insley & Swann 2003), where individuals are known to redistribute according to social status and habitat quality (Cresswell 1994). If feeding opportunities were enhanced under artificial lighting, then relatively higher densities of redshank might occur within localities of artificially elevated ambient light levels (Sutherland 1983). Similarly, the usage of these favoured areas may be age-dependent; whereby juveniles are excluded by the more experienced adults (Cresswell 1994). It was not possible to record bird density during the night, because at low tide, the birds would forage up to 2 km from the shore and were mixed with other species of shorebird. Furthermore, poor visibility made it impossible to distinguish between adult and juvenile birds out on the mudflats. The logistical constraints on visual observations coupled with the sensitivity of redshank to human presence made behavioural observation through VHF telemetry the only reliable method by which nocturnal activities could be monitored throughout the home range of known-age birds. Future studies should seek to use these techniques to investigate intraspecific responses of individual birds to ambient light levels according to social status.
Studies into the effect of light levels on wildlife behaviour have hitherto measured ambient light directly using a hand-held photo-detecting diode (e.g. Gorenzel & Salmon 1995; Tuxbury & Salmon 2005). These measurements would not have been possible in the present study because the birds regularly occupied inaccessible mudflats, moreover, behaviour would have been altered by any human approach. Consequently, direct measurements of light would be biased towards localities close to the shoreline and would have resulted in an erroneous association between measured ambient light levels and the behaviour exhibited by the bird. Our approach, which utilised remotely sensed DMPS/OLS light radiance values gathered by satellite, provided a novel measure of ambient light levels throughout the entire estuary at a broad-scale resolution. We found that birds inhabiting areas with higher light radiance values foraged for significantly longer periods than birds inhabiting areas with lower radiance values. Similarly, the proportion of birds foraging on cloudy nights was elevated regardless of lunar illuminance (Fig. 2), presumably due to the reflection of artificial light sources with increased cloud cover (Kyba et al. 2011). While there was evidence of a relationship between foraging behaviour and DMSP/OLS light radiance values, this was not statistically significant. This may be attributed to low-intensity fine-scale point sources of light undetected by the satellites, the absence of truly dark areas in an urbanised estuary as well as undetected factors inherent to such a complex ecological system.
Observational studies have shown that nocturnal foraging may be less efficient than daytime foraging. This is because shorebirds are unable to detect the presence of the most profitable prey visually, making them less selective in prey choice under darkness (Goss-Custard 1969; Sutherland 1982). Foraging visually also allows shorebirds to scan for prey and handle prey at the same time, also increasing foraging efficiency (Evans 1976). Those shorebirds foraging by touch, however, are unlikely to locate more than one prey at a time. In this study, our tagged redshank exhibited a transmitter-pulse rate which implied visual foraging under a full moon, but this was not apparent when the full moon was obstructed by cloud cover. Similarly, tagged redshank inhabiting artificially illuminated areas exhibited a pulse rate indicative of visual foraging and operated independently of lunar phase. We argue that this allowed redshank to detect and select the most profitable prey.
When considering the trophic chain on a larger scale, redshank are predated by raptors by day (Whitfield 2003; Cresswell & Whitfield 2008), while owls and mammalian carnivores (including foxes Vulpes vulpes) can pose a serious threat at night (Page & Whitacre 1975; Sitters et al. 2001; Piersma et al. 2006). Tawny owls Strix aluco, short-eared owls Asio flammeus and red foxes were all observed hunting in areas frequented by the tagged redshank (R.G. Dwyer personal observation). While ambient light levels were found to improve visibility for sight-based foraging, increased night-time lighting could also affect predator and prey relationships. We suggest that higher light levels would not provide a significant advantage for nocturnal ambush predators with excellent night vision, perhaps even imposing a penalty, but could provide improved predator detection by their shorebird quarry. For example, some birds will choose to roost in areas of high night-time illumination in response to predation risk from owls (Gorenzel & Salmon 1995). The location of the eyes in many species of shorebird results in a blind area to the rear of the birds' head, rendering the bird vulnerable to predation when employing tactile foraging (Martin & Piersma 2010), and as a consequence, there will be a trade-off between being vigilant for predators or searching for prey items (Cresswell & Whitfield 2008). Shorebirds will often increase levels of vigilance in areas of poor visibility (Metcalfe 1984; Burger & Gochfeld 1991), and redshank may therefore be able to reduce vigilance time and also increase nocturnal feeding potential when foraging under elevated ambient light levels. Furthermore, tagged redshank in our study spent a similar amount of time foraging during the day regardless of the degree of nocturnal illumination in their foraging area. This provides further evidence that rather than a trade-off against predator disturbance, increased foraging under nocturnal illumination reflected improved foraging opportunities through enhanced visibility.
Estuarine and coastal areas throughout the world are being developed for industry, agriculture, mariculture, leisure and housing. These are critical habitat for shorebirds, yet the extent to which light pollution may impact upon their ecology has received little attention. Our study has shown that localised artificial illumination affects foraging behaviour in redshank in a manner similar to elevated natural light. We suggest that in estuaries close to major urban and industrialised regions, artificial illumination should be considered as an important environmental factor driving nocturnal habitat selection, foraging behaviour and potentially the structuring of animal communities. The integration of VHF radio transmitters with inbuilt posture sensors and illumination data from DMSP/OLS satellites provided an insight into the relationship between behaviour and ambient light levels which would have been otherwise logistically challenging. The study provides a framework for future investigators to assess the impact of artificial light upon animal behaviour and predator–prey relationships.
We would like to thank J. Calladine (BTO Scotland) for help with fieldwork, M. Witt for his guidance with ARCGIS and S. Walls and B. Cresswell (Biotrack) for their advice on telemetry equipment. We also thank A. Broderick and J. Quinn for helpful discussions, H. Sitters and an anonymous referee for detailed suggestions which strengthened the article. DMB thanks Transport Scotland for financial support.