The impacts of new street light technologies: experimentally testing the effects on bats of changing from low-pressure sodium to white metal halide

Abstract

Artificial light at night is a major feature of anthropogenic global change and is increasingly recognized as affecting biodiversity, often negatively. On a global scale, newer technology white lights are replacing orange sodium lights to reduce energy waste. In 2009, Cornwall County Council (UK) commenced replacement of existing low-pressure sodium (LPS) high intensity discharge (HID) street lights with new Phillips CosmoPolis white ceramic metal halide street lights to reduce energy wastage. This changeover provided a unique collaborative opportunity to implement a before-after-control-impact field experiment to investigate the ecological effects of newly installed broad spectrum light technologies. Activity of the bat species Pipistrellus pipistrellus, P. pygmaeus and Nyctalus/Eptesicus spp. was significantly higher at metal halide than LPS lights, as found in other studies of bat activity at old technology (i.e. mercury vapour) white light types. No significant difference was found in feeding attempts per bat pass between light types, though more passes overall were recorded at metal halide lights. Species-specific attraction of bats to the metal halide lights could have cascading effects at lower trophic levels. We highlight the need for further research on possible ecosystem-level effects of light technologies before they are installed on a wide scale.

1. Introduction

Anthropogenic light pollution is increasing globally [1–4] and recognized as a biodiversity threat [5–8]. Artificial lighting has been shown to affect a range of species and their behaviours, including foraging [9,10], movement [11–14], reproduction [15–17] and communication [18]. Lighting is now recognized to have community-level impacts and affects key population parameters such as mortality and demography [6,19–22]. The increase in distribution and intensity of artificial lighting worldwide has coincided with changes in the spectral content of lighting. As they reach the end of their lifetime, older street lights such as traditional orange high-pressure sodium (HPS) and low-pressure sodium (LPS) lights are being replaced with broad spectrum, high brightness technologies such as light-emitting diodes (LEDs) and ceramic metal halide lights [5,19,23].

The switch to broad spectrum technologies is motivated by increased energy efficiency, a longer lifespan of the lights and therefore lower maintenance costs, as well as improved light quality and colour rendering for human vision [24]. New artificial lighting strategies are being developed worldwide, targeted at energy efficiency to meet international targets for CO₂ reduction [4]. Broad spectrum technologies are predicted to alter the balance of species interactions by creating disparity in effects across taxonomic groups [19]. Few studies, however, have assessed the ecological impacts of emerging broad spectrum lighting. Recently, experimental field studies with white LED street lights demonstrated species-specific effects on bats, causing spatial avoidance and delayed commuting times, with potential effects on foraging and energetics for relatively slow-flying species such as Myotis spp. and Rhinolophus hipposideros [11].

In 2009, Cornwall County (UK) Council's (CCC) operating lighting stock totalled 47 097 units, comprising 68% LPS and 32% HPS lights [25]. CCC implemented an ‘Invest to Save Project’ to replace all existing sodium street lights with new Philips CosmoPolis 2800 kelvin (K) white ceramic metal halide lights (45 and 60 W) to increase directionality of light and consequently reduce light spill, improve visual perception for humans and reduce energy consumption. In collaboration with CCC, we used this unique opportunity to conduct a novel field experiment to assess and compare the ecological impact of changing light types from LPS to 2800 K white metal halide lights on bat activity. We tested the hypothesis that changing lights would increase the activity of bat species that are attracted to insects around lights, because metal halide lights are broad spectrum and emit shorter wavelengths than LPS lights, hence potentially attracting more insects.

2. Material and methods

Experiments were conducted at existing street lights in the vicinity of three towns (Launceston, Penzance and St Austell) in the county of Cornwall, England between June and September 2010. Bat activity was measured at 13 paired sites using a before-after-control-impact design. Each pair consisted of two CCC-maintained street lights; for one of the pair (the experimental column) the LPS light was replaced with a 45 or 60 W metal halide light, and for the other (the control column) the LPS light was not changed. New metal halide lights, without UV filtering glass, were installed onto the existing columns at the same height and armature length. Philips CosmoPolis lights (CosmoWhite CPO-TW 45/60 W) are ceramic metal halide, broad spectrum white lights (figure 1) with a correlated colour temperature (CCT) of 2800 K. Mean time interval between recording bat activity at each pair was 49 days (range 14–77 days). Sites were chosen in predominantly suburban habitats, comprising mainly family houses with small gardens, and outlying areas where bat activity is often higher than in more urban areas [26]. Study sites were restricted to areas where LPS was the dominant light type at the start of the experiment and where metal halide lights were due to be installed that year. Columns within each pair were matched by habitat as far as possible. Light columns within each pair were located a mean distance of 1040 m (range 550–2180 m) apart to minimize the chances of counting the same bats at different sites. To compare typical light intensity between light types, illuminance (lux) was measured using a T-10 illuminance meter (Konica Minolta Sensing, Japan) 1 h after sunset, recorded approximately 1.5 m above ground height directly under four LPS and four metal halide lights in four suburban streets in Penzance.

Figure 1.

Spectral power distribution of low-pressure sodium (LPS) and Philips CosmoPolis white 60 W metal halide lights (source [5]). (Online version in colour.)

Bat activity was recorded simultaneously for each pair for two periods of three consecutive nights. The first period was when both lights in each pair were LPS lights, the second when the experimental column in that pair had been changed to a metal halide light. We waited a minimum of 7 days after the experimental column had been converted to a metal halide light before recording bat activity. Pairs were situated a mean of 48.7 km apart (range 0.99–102.6 km between experimental street light columns; range 0.76–102.6 km between control columns).

Bat activity (number of bat passes) was recorded using an AnaBat acoustic detector (either SD1 or AnaBat II, Titley Electronics, Ballina, New South Wales, Australia) attached to each light column on average 4.3 m (range 4.0–4.5 m) from the base of the lighting column. We recorded bat activity simultaneously at each paired site from 50 min before sunset on each of the three consecutive night until 50 min after sunrise. We used eight different AnaBat detectors set to a sensitivity of 8 and a division ratio of 8; they were randomly rotated between experimental and control columns to control for potential differences in equipment sensitivity. Hourly weather data were collected from Meteorological Office weather stations in Bodmin (N 50°50′ W 04°67′), Culdrose (N 50°08′ W 05°26′) and Plymouth (N 50°35′ W 04°12′) for the study sites in St Austell, Penzance and Launceston, respectively. Data are available in the electronic supplementary material.

(a). Echolocation call analyses

Echolocation calls were analysed in AnalookW 3.5t 2008 (Titley Electronics, Ballina, New South Wales, Australia). Bats were identified to five species (Pipistrellus pipistrellus, P. pygmaeus, P. nathusii, Nyctalus noctula and Plecotus auritus) and two species groups (Myotis spp. and Nyctalus/Eptesicus spp.) as described by Stone et al. [10,11] based on their diagnostic call parameters [27,28]. As it is not possible to count individual bats from their echolocation calls, relative bat activity was measured using the mean number of bat passes per night. We defined a single bat pass as occurring when the time between calls exceeded four times the inter-pulse interval [11]. Bat feeding activity was measured by calculating the ‘buzz ratio’ of diagnostic feeding buzzes relative to the number of bat passes [29,30].

(b). Statistical analyses

Repeated Measures Generalized Linear Models (RMGLZMs) and Repeated Measures General Linear Models (RMGLMs) were used to assess the effect of light type (LPS and metal halide) and weather (mean temperature, mean rainfall and mean wind speed) on bat activity (number of bat passes per night) for species or species groups in R (v. 3.0.1) [31]. Owing to the low number of pairs (n = 8) with sufficient numbers of bat passes for analysis, data from the control and experimental sites were analysed in separate models. RMGLZMs were used as they allowed us to fit all possible model structures for each species and select the best fitting model based on model assumptions and Akaike's information criteria (AIC).

Column was fitted as a random effect in each model to account for repeated measures within light columns (LPS and metal halide). As the three weather variables were highly correlated, they were reduced using principal components analysis (PCA) [32]. Scores from the first principal component were used in the model to account for the effect of weather (proportion of variance explained 0.58). We tested each dataset with different error structures and selected the best models by analysing the residuals for heteroscedasticity and non-normality. If model assumptions were met, models with different error structures were compared using AIC. Any non-significant effects were removed for model simplification [33]. To ensure that any observed increases in bat activity were not simply a function of young flying when we collected the data, we compared bat activity at control sites at the same time that the LPS lights were changed to metal halide at the experimental sites.

Separate models were constructed for individual species and species groups from both control and experimental sites. Zero-inflated Poisson models were used for Nyctalus/Eptesicus spp. activity data (using the glmmADMB package) [34] and Poisson models for P. pipistrellus, P. pygmaeus and Pipistrellus spp. activity data from experimental sites (using the nlme package) [35]. Pipistrellus pipistrellus and P. pygmaeus data recorded at control sites were transformed (ln + 1) and analysed using Gaussian RMGLMs, with the nlme package. Nyctalus/Eptesicus spp. activity data for control nights were analysed using a RMGLZM with Penalized Quasi-Likelihood and the Mass package [36]. Differences in bat feeding activity (buzz ratio) and the number of species (where Myotis spp. and Eptesicus/Nyctalus spp. were both counted as one species) were compared between light types using RMGLMs with log transformed data (ln + 1) and the nlme package, and zero-inflated Poisson models with the glmmADMB package.

3. Results

Mean light intensity at metal halide lights was 56.2 lux (s.d. = 13.4, range 39.1–66.9, n = 4) and 32.8 lux (s.d. = 4.8, range 26.7–38.1, n = 4) at LPS lights.

(a). Bat activity

Pipistrellus pipistrellus was the most common species recorded across sites (mean = 676.9 passes, s.d. = 835.5, range 4–2878), followed by Pipistrellus spp. (mean 73.1 passes, s.d. = 195.3, range 0–876), P. pygmaeus (mean = 20.2 passes, s.d. = 53.9, range 0–290), Nyctalus/Eptesicus spp. (mean = 16.5 passes, s.d. = 40.6, range 0–183), N. noctula (mean = 15.3 passes, s.d. = 39.8, range 0–178) and Myotis spp. (mean = 0.3 passes, s.d. = 0.5, range 0–2). Pipistrellus pipistrellus and N. noctula were recorded at both LPS and metal halide lights and every light column. Pipistrellus pygmaeus was recorded at both light types and every column except one. Plecotus auritus was rare (n = 6 passes in total) and only recorded under metal halide lights. Myotis spp. (n = 25 passes in total) and P. nathusii (n = 8 passes in total) were both recorded in all three Cornish towns.

Weather had a significant positive effect on P. pipistrellus activity in both control and experimental sites (experimental = 0.79, s.e. = 0.05, p ≤ 0.01; control = −0.43, s.e. = 0.09, p = 0.004). Bat activity and weather had a significant negative interaction for both experimental and control sites (experimental = −1.86, s.e. = 0.13, p ≤ 0.01; control = −1.16, s.e. = 0.21, p = 0.003).

Sufficient feeding activity data were obtained from seven control sites and eight experimental sites to compare the effects of light type on P. pipistrellus activity. No significant difference in feeding activity (buzz ratio) was found between light types (mean 0.05 before changeover (LPS), mean 0.05 after changeover (metal halide), −0.001, s.e. = 0.007, p = 0.863) or within control sites between the two sampling periods (mean 0.05 first period, mean 0.04 second period, 0.111, s.e. = 3.046, p = 0.97).

(b). Effects of changing light type

Eight of the 13 initial paired sites contained sufficient data for statistical analyses to test the effect of changing light treatment (LPS before changeover and metal halide after changeover) on bat activity (SM1). There were four 45 W and four 60 W metal halide lights. Sample sizes were sufficient to test the effect of light treatment on two species (P. pipistrellus and P. pygmaeus) and one species group (Nyctalus/Eptesicus spp.). There were significantly fewer bat passes under LPS (before changeover) than metal halide lights (after changeover) (figure 2) for P. pipistrellus (mean 414 bat passes LPS, mean 878 bat passes metal halide, −1.21, s.e. = 0.04, p ≤ 0.01), P. pygmaeus (mean 1 bat pass LPS, mean 11 bat passes metal halide, −2.67, s.e. = 0.47, p ≤ 0.01) and Nyctalus/Eptesicus spp. (mean 5 bat passes LPS, mean 13 bat passes metal halide, −0.69, s.e. = 0.21, p ≤ 0.05). There was a stronger effect for P. pygmaeus (table 1).

Figure 2.

Mean bat activity (number of bat passes) per species at experimental sites before (LPS lights) and after light changeover (metal halide lights). Error bars indicate standard errors of the mean.

Table 1.

Estimates (Est.) and standard errors (s.e.) of bat activity (number of bat passes) across control (LPS lights for both sampling periods) and experimental (LPS lights for the first sampling period, metal halide lights for the second period) treatments. Light type for control = the contrast between LPS and LPS; light type for experimental = the contrast between LPS and metal halide; PCAweather = PCA component comprising mean rainfall, mean temperature and mean wind speed.

	P. pipistrellus			P. pygmaeus			Nyctalus/Eptesicus spp.
	Est.	s.e.	p-value	Est.	s.e.	p-value	Est.	s.e.	p-value
control
intercept	2.42	0.50	0.002*	1.02	0.30	0.012*	1.60	0.65	0.04*
lighting type (LPS versus LPS)	0.004	0.11	0.968	−0.32	0.20	0.167	−0.22	0.22	0.36
PCAweather	0.43	0.09	0.004*	0.16	0.14	0.29	−0.19	0.10	0.11
LPS × PCAweather	−1.16	0.21	0.003*	−0.12	0.32	0.72	0.36	0.18	0.11
experimental
intercept	6.37	0.81	<0.01*	2.99	0.74	<0.001*	1.63	0.56	<0.01*
lighting type (LPS versus metal halide)	−1.21	0.04	<0.01*	−2.67	0.47	<0.001*	−0.69	0.21	<0.05*
PCAweather	0.79	0.05	<0.01*	—	—	—	0.12	0.14	0.38
LPS × PCAweather	−1.86	0.13	<0.01*	—	—	—	−0.58	0.27	0.03*

*p = 0.05.

For the control sites, there was no significant difference between the two sampling periods for P. pipistrellus (mean 718 bat passes first period, mean 697 bat passes second period, 0.004, s.e. = 0.11, p = 0.968), P. pygmaeus (mean 17 bat passes first period, mean 51 bat passes second period, −0.32, s.e. = 0.20, p = 0.167) and Nyctalus/Eptesicus spp. (mean 26 bat passes first period, mean 22 bat passes second period, −0.22, s.e. = 0.22, p = 0.36) activity (figure 3). Thus, differences in activity at metal halide lights were the result of changing light treatments rather than natural changes in bat activity between the two sampling periods (table 1).

Figure 3.

Mean bat activity (number of bat passes) at control sites where LPS lights were used for the first and second sampling periods. Error bars indicate standard errors of the mean.

There was no significant difference in the number of species recorded in the two sampling periods (table 2) (control sites, mean number of species first period 2.9, s.d. = 0.8, range = 2–4; mean number of species second period 3.4, s.d. = 1.1, range = 2–4; experimental sites mean number of species before changeover 2.9, s.d. = 0.4, range = 2–3; mean number of species after changeover 3.4, s.d. = 0.5, range = 3–4).

Table 2.

Estimates (Est.) and standard errors (s.e.) of number of bat species across control (LPS lights for both sampling periods) and experimental (LPS lights for the first sampling period, metal halide lights for the second period) treatments. Light type for control = the contrast between LPS and LPS; light type for experimental = the contrast between LPS and metal halide; PCAweather = PCA component comprising mean rainfall, mean temperature and mean wind speed.

	control			experimental
	Est.	s.e.	p-value	Est.	s.e.	p-value
intercept	0.621	0.043	0.000*	0.634	0.018	0.000*
lighting type	−0.054	0.050	0.332	−0.046	0.019	0.055
PCAweather	−0.013	0.026	0.649	−0.015	0.010	0.214
LPS × PCAweather	0.050	0.055	0.409	0.012	0.023	0.608

*p = 0.05.

4. Discussion

Activity of P. pipistrellus, P. pygmaeus and Nyctalus/Eptesicus spp. was significantly higher at white 2800 K metal halide lights compared with the orange LPS lights. Previous studies found higher bat densities and activity around old technology mercury-vapour white lights that also emit some light in the UV spectrum. Three times more bat activity (mostly P. pipistrellus) was recorded on roads lit by white street lights than those with orange or no street lights in England [10], and densities of Eptesicus nilssonii were five times higher in lit compared with dark areas in Sweden [37]. Such bats may benefit from the increased numbers of insects that are attracted to white lights, particularly those containing UV [23,38]. Mercury-vapour lights attract more insects than HPS lights [38], whereas LPS lights attract similar numbers of insects to those recorded on dark streets [39]. LPS are narrow spectrum lights with a CCT of around 1800 K [19], in contrast to our Philips ceramic metal halide lights which are broad spectrum with a CCT of around 2800 K. Metal halide lights have a peak wavelength around 550–600 nm (similar to LPS and HPS) and also contain smaller peaks around 400–420 nm (figure 1). As metal halide lights emit some UV light (less than 400 nm) [19], we would have expected to record higher bat feeding activity at experimental metal halide lights due to the higher abundance of insects; moths and Diptera in particular are attracted to the lights [38,40,41]. However, our results did not show increased P. pipistrellus feeding buzz ratios (buzzes/pass) at metal halide lights compared with LPS lights, which would be expected if the attraction of bats was due to increased feeding opportunities. Even though the buzz ratio did not differ between the light types, the higher activity of P. pipistrellus around metal halide lights suggests that more insects may have been attracted to them. We were only able to analyse feeding activity for one species, and recorded low levels of feeding activity overall. In addition, the level of insect attraction to lights is also a function of CCT, with lower CCT lights attracting fewer insects [42]. Further research is required to compare insect attractiveness to metal halide lights compared with other light types.

The impact of artificial lighting on bats is species-specific, with some species avoiding lit areas due to perceived predation risk. Species-specific responses are believed to be a function of flight morphology and echolocation. Fast-flying bats which typically forage in the open using echolocation pulses adapted for long distance ranging (including Pipistrellus, Nyctalus and Eptesicus species) are more common around street lights [10,39]. By contrast, slow-flying bats with echolocation and wing morphology adapted for cluttered environments (such as Rhinolophus, Myotis and Plecotus species) [43], may be less likely to exploit insects attracted to street lights due to light-dependent predation risk. Fast-flying species are better able to avoid predation by diurnal birds of prey, and emerge earlier from their roosts when light levels are relatively high [44]. Conversely, slow-flying species emerge later, and appear to have an innate intolerance of lit conditions, even when light levels are relatively low [44].

In this study Myotis spp. were rarely recorded at both control and experimental lights. This confirms previous research in which Myotis spp. avoided flight routes illuminated with HPS and white LED street lights [11,12,45], and avoided street lights within their foraging range in Canada and Sweden [39,46]. Rhinolophus hipposideros was not recorded despite the study area (Cornwall) being a stronghold for the species in Britain [47]. This suggests 2800 K white metal halide lights cause similar avoidance behaviour for Myotis spp. and R. hipposideros, as has been recorded with white LED lights [12]. There was no significant difference in bat species richness between LPS and metal halide lights. Because similar species were found at each light type, our results suggest that light-averse taxa are intolerant of both LPS and metal halide lights. Experimental studies are required to test the effect of metal halide lights on ‘light-averse’ species, such as Myotis spp. and R. hipposideros to confirm these predictions.

Our results suggest that large scale replacement of LPS lights by white metal halide lights could alter the balance of species distributions through species-specific impacts, in this case by attracting more P. pipistrellus, P. pygmaeus and Eptesicus/Nyctalus spp. While ‘light tolerant’ bat species may benefit from increased foraging on insects attracted to white street lights, altering the balance of predators and prey could have cascading ecosystem effects at lower tropic levels, with implications for ecosystem service provision. Bats feeding on tympanate moths may have a competitive advantage under white light as the moths' evasive behaviours are reduced, increasing their vulnerability to predation [48]. Increased mortality at white lights will have consequences for other species in the ecosystem that depend on these insects for food and pollination. The attraction of ‘light tolerant’ bats to white light may also increase competition with other ‘light-averse’ bat species that forage in dark areas nearby. Population declines of R. hipposideros in Switzerland may have been caused by competitive exclusion by P. pipistrellus, which was able to take advantage of the increased foraging opportunities provided by street lights [49].

5. Conclusion

We used a novel field experiment to demonstrate that the impact of street lighting on bats is species-specific and varies according to light type. In line with predictions based on flight behaviour and morphology, we recorded higher activity for open-space foragers at 2800 K white metal halide lights than at traditional LPS. We show that street lighting can alter patterns of bat species abundance and activity, which may in turn have cascading effects at the ecosystem level. Our results highlight the importance of adopting an ecosystem-level approach in understanding the ecological consequences of emerging light technologies. Further research is required to assess the ecological impacts of emerging light technologies at the community and ecosystem level. New lighting policies should give careful consideration towards the impact these new, energy-saving street lights may have on local biodiversity and ecological interactions before wide-scale installation.

Authors' contributions

E.L.S. conceived, designed and jointly supervised the study, conducted the data analysis and drafted the manuscript; A.W. is joint first author, assisted with study design and drafting of the manuscript, collected the field data, conducted acoustic analysis and secured the funding; G.J. and S.H. coordinated the study and helped draft the manuscript. All authors gave final approval for publication.

Funding statement

We thank the People's Trust for Endangered Species for funding.