See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/325724554

Rapid assessment of lamp spectrum to quantify ecological effects of light at

night

Article in Journal of Experimental Zoology Part A Ecological and Integrative Physiology · June 2018
DOI: 10.1002/jez.2184

CITATIONS READS

156 2,823

6 authors, including:

Travis Longcore Airam Rodríguez

University of California, Los Angeles Museo Natural de Ciencias Naturales CSIC
148 PUBLICATIONS 5,540 CITATIONS 126 PUBLICATIONS 2,365 CITATIONS

SEE PROFILE SEE PROFILE

Jay Penniman Lorna Herf

University of Hawaiʻi at Mānoa 2 PUBLICATIONS 156 CITATIONS
14 PUBLICATIONS 748 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Travis Longcore on 24 January 2019.

The user has requested enhancement of the downloaded file.

Received: 15 December 2017 Revised: 30 April 2018 Accepted: 1 May 2018

DOI: 10.1002/jez.2184

RESEARCH ARTICLE

Rapid assessment of lamp spectrum to quantify ecological

effects of light at night

Travis Longcore1 Airam Rodríguez2 Blair Witherington3 Jay F. Penniman4

Lorna Herf5 Michael Herf5

1 University of Southern

California, Los Angeles, California Abstract

2 Estación Biológica de Doñana CSIC, Sevilla, For many decades, the spectral composition of lighting was determined by the type of lamp, which
Spain also influenced potential effects of outdoor lights on species and ecosystems. Light-emitting diode
3 Disney's Animals, Science and Environment,
(LED) lamps have dramatically increased the range of spectral profiles of light that is economi-
Lake Buena Vista, Florida cally viable for outdoor lighting. Because of the array of choices, it is necessary to develop meth-
4 Pacific Cooperative Studies Unit, University of
ods to predict the effects of different spectral profiles without conducting field studies, especially
Hawaii at Manoa, Honolulu, Hawaii
because older lighting systems are being replaced rapidly. We describe an approach to predict
5 f.lux Software LLC, Los Angeles, California
responses of exemplar organisms and groups to lamps of different spectral output by calculat-
Correspondence
Travis Longcore, University of Southern
ing an index based on action spectra from behavioral or visual characteristics of organisms and
California, Watt Hall 204, Los Angeles, CA lamp spectral irradiance. We calculate relative response indices for a range of lamp types and light
90089, USA. sources and develop an index that identifies lamps that minimize predicted effects as measured
Email: [email protected]
by ecological, physiological, and astronomical indices. Using these assessment metrics, filtered
yellow-green and amber LEDs are predicted to have lower effects on wildlife than high pressure
sodium lamps, while blue-rich lighting (e.g., K ≥ 2200) would have greater effects. The approach
can be updated with new information about behavioral or visual responses of organisms and used
to test new lighting products based on spectrum. Together with control of intensity, direction, and
duration, the approach can be used to predict and then minimize the adverse effects of lighting
and can be tailored to individual species or taxonomic groups.

KEYWORDS
action spectrum, behavioral response, light pollution, phototaxis

1 INTRODUCTION economically viable for widespread deployment and their spectral

characteristics were limited. For example, low pressure sodium lamps,
It has long been known that artificial night lighting affects wildlife with nearly all emissions in the yellow/orange at 589 nm became the
through attraction and disorientation (Allen, 1880), and recent lamp of choice around astronomical observation sites and near sea tur-
research has documented the extent of the adverse consequences tle nesting beaches because both night sky observation and sea turtle
of artificial night lighting to include, for example, plant phenology orientation benefit from a narrow-band light in the longer wavelengths
(Somers-Yeates et al., 2016), predator–prey relations (Minnaar, Boyles, (Witherington, 1992). Other lamps were similarly deployed in differ-
Minnaar, Sole, & McKechnie, 2015), circadian rhythms (Dominoni, ent situations and consequently most studies of ecological effects are
2015), and nocturnal rest and recovery (Gaston, Bennie, Davies, & on these types—low-pressure sodium, high-pressure sodium, metal
Hopkins, 2013). Importantly, light attraction and disorientation results halide, and mercury vapor (although this lamp type has largely been
in direct mortality of many groups of insects (Eisenbeis & Hänel, 2009), phased out) (Eisenbeis & Eick, 2011; Rich and Longcore, 2006). In the
birds (Longcore et al., 2012), including seabirds (Rodríguez et al., past decade, however, light-emitting diode (LED) lamps have become
2017b), and sea turtles (Salmon, 2003), contributing to species decline economically viable, bringing a range of new spectral characteris-
(Fox, 2013; Wilson et al., 2018). The degree of influence of outdoor tics to the marketplace (Boyce, Fotios, & Richards, 2009; Gaston,
electric lighting is determined by the direction, intensity, duration, and 2013) along with concerns about their differential effects on wildlife
spectrum of the lights (Gaston, Davies, Bennie, & Hopkins, 2012; Long- species (Davies, Bennie, Inger, de Ibarra, & Gaston, 2013; Gaston,
core and Rich, 2017). For many years, only a handful of lamp types were 2013).

J. Exp. Zool. 2018;1–11. wileyonlinelibrary.com/journal/jez 

c 2018 Wiley Periodicals, Inc. 1
2 LONGCORE ET AL .

In the early days of commercial LEDs for outdoor lighting, full spec- behavioral response of hatching sea turtles to longer wavelengths of
trum light was achieved through coating a blue LED with a phosphor, light (Witherington, 1992) has become the basis to limit the permis-
which produced light across the visual spectrum (Hecht, 2012). These sible spectral characteristics of lights on and near nesting beaches in
lamps had a high correlated color temperature (CCT), indicating a high many jurisdictions. Such regulations to minimize adverse effects of
proportion of blue and violet in the emissions, as a result of the under- lighting on nature are always compromises and usually driven by the
lying blue LED. This blue hue became more dramatic as the phosphor species or species group with regulatory protection in a particular sit-
aged. Many in the general public and scientific community may have uation.
developed the perception that all light from LEDs was a “cool” white The current challenge for conservationists is that assessing the
(high CCT) at this time. Technological innovation in the LED indus- effects of different spectral distributions on wildlife in experimental or
try has, however, been rapid, because the energy savings from LEDs field situations is time consuming and an increasing number of lamp
are so attractive that replacement lamp types that address a range of types are being developed, while jurisdictions are making decisions
color spectrum specifications have been developed (Dudley, Erkintalo, about replacement of aging fixtures every day (Hecht, 2016). Once
& Genty, 2015). While earlier efforts to develop LEDs with lower color such decisions are made, new lamps will be in place for years to come.
temperatures came with a penalty of less efficiency, by 2015, LEDs Tools are therefore needed to assess the potential adverse effects of
at 2700 K and 3000 K were commercially available that matched the newly developed lights compared with existing technologies in a rapid
energy efficiency of 5000 K lamps. Furthermore, the development of manner and in a way that allows tradeoffs between adverse effects on
different colors of LEDs and different filtering technologies has led to wildlife and human needs to be compared. In this paper, we assemble
a range of different spectral signatures for lamps that are all economi- a series of spectral response curves from the literature and a series
cally competitive in terms of energy efficiency. of spectral emission curves for established and new outdoor lighting
Conservation scientists need to keep up with the changing array of sources, develop a standardized index that weights the spectral output
outdoor lighting options to provide guidance to officials and managers by the response curves, provide a matrix of lighting performance mea-
around the world who are faced with the obvious economic choice of sures (e.g., color rendering index, correlated color temperature, Star
switching to high-efficiency lighting such as LEDs (Hecht, 2016). Such Light Index), and present these results on a website that can be peri-
a switch can be catastrophic for the effects on other species, or it odically updated to serve as a clearinghouse for this information.
can be a benefit, depending on the spectrum, duration, direction, and
intensity of the new lamps (Gaston et al., 2012; Longcore et al., 2015;
Rodríguez, Dann, & Chiaradia, 2017a). The same applies to sky glow 2 METHODS
(Kinzey et al., 2017). Some ecologists have voiced generic concerns
about LEDs in general, questioning whether they pose a risk across We obtained spectral power distribution curves for a wide range of
the board (Pawson and Bader, 2014; Stone, Jones, & Harris, 2012), and lamp types and calculated indices representing the degree of over-
noting the unfortunate “rebound effect” in which more efficient light- lap with a series of spectral response curves for different organisms.
ing leads to deployment of even more light (Kyba et al., 2017; Kyba, Following recommendations of the Bureau International des Poids et
Hänel, & Hölker, 2014). Similar concerns about the adverse effects Mesures (BIPM), action spectra are dimensionless, while spectral irra-
of the rapid spread of full spectrum LED lighting are voiced by dark diance is measured in 𝜇W⋅cm–2 ⋅nm–1 , from which we calculate the
sky advocates (Bierman, 2012). The spectrum of light used will greatly weighted sum across wavelengths (BIPM, 2006, Appendix 3, Section
affect the amount of scattering of light at different distances from a 2). We treat spectral response curves like action spectra even if they
source (Kinzey et al., 2017). The extent of these effects depends in part do not meet the high standards for a true action spectrum (Björn,
on the spectral characteristics of the LEDs used, and many opportuni- 2015). Species response curves were converted from photons to spec-
ties are available to evaluate the performance of the wide array of LED tral power (𝜇W⋅cm–2 ⋅nm–1 ) because organismal responses are depen-
spectral configurations, such as investigating multiple spectral configu- dent on the number of photons, not the energy of the light (Johnsen,
rations of 2700 K LEDs to reduce attraction of flying insects (Longcore 2012) while light is frequently measured with power units.
et al., 2015) or comparing LEDs of different color temperatures (Eisen- Spectral power distributions were obtained in 𝜇W⋅cm–2 ⋅nm–1 and
beis & Eick, 2011). resampled to 1 nm increments from 350 nm (well in the ultravio-
Differences between the spectral response curve for human vision let, which is still the visual spectrum for some insects) (Menzel &
(both photopic and scotopic) and the visual sensitivity and measured Greggers, 1985) through 780 nm to encompass the full range of vision
behavioral responses of animals indicate an opportunity to configure for organisms. Spectral response curves were normalized to 1 at the
outdoor lighting that avoids sensitive regions of the spectrum while maximal value, and multiplied by the emissions at each wavelength and
providing needed visibility for humans. For example, many insects are then summed over all wavelengths, yielding three metrics.
attracted to shorter wavelengths (blue, violet, and ultraviolet) more
than longer wavelengths (Eisenbeis, 2006; Eisenbeis & Hänel, 2009). 1. A standard “effective irradiance” metric, computed by multiplying
Light sources that have low blue and shorter wavelength emissions spectral irradiance at each wavelength by the spectral response
attract fewer insects (Cleve, 1964; Eisenbeis & Eick, 2011; Eisenbeis (“actinic power”). (BIPM, 2006, Appendix 3 and CIE, 2007)
& Hänel, 2009; Menzel & Greggers, 1985) and consequently, fewer
bats that forage on insects (Stone, Harris, & Jones, 2015). The lower Eeff = ∫ E𝜆 Si (𝜆) d𝜆,
LONGCORE ET AL . 3

where E𝜆 represents the source spectral irradiance and Si is the TA B L E 1 Lamps and spectral output curves included in study, by
actinic spectrum. type, correlated color temperature (CCT), and color rendering index
(CRI)
2. The actinic power per lux (the human photopic response, V(𝜆)):
Lamp/Standard Type CCT CRI
∫ E𝜆 Si (𝜆) d𝜆 D65 (Daylight) Natural 6504 100
Elux = .
∫ E𝜆 V (𝜆) d𝜆
CIE Illuminant A Lighting Standard 2856 100
The resulting measurement is thereby standardized in terms of the Kerosene Oil Combustion 1913 99
effect on each species per lux produced by the lamp and can be Full moon Natural 4134 98
referred to as the taxonomic (e.g., turtle, salmon) action factor of Philips TL950 Fluorescent 4684 96
the light source (CIE, 2014). SORAA Vivid LED 4965 93
3. To allow comparison across species, we scaled the action factor rel- CFL Greenlite 13 W Fluorescent 2892 81
ative to the response that would be elicited by daylight. Philips AmbientLED LED 2601 81

Elux (E) LLT Telescope Light Filtered LED 1908 81

aD65 = .
Elux (D65) 3000K LED LED 3262 80
OCTRON 32 W Fluorescent 4012 79
The resulting values indicate the increase of effects on species rel-
Metal Halide 70W Metal Halide 3071 79
ative to sunlight for each additional lux. A metric indexed to day-
LSG Good Night 2016 LED 2266 76
light allows actinic response metrics to be compared across species,
even when the “shape” of the action spectra varies. LEDway Streetlight LED 6270 75
CW 54W

This approach allows comparison across lamp types and for differ- City of Los Angeles LED 4310 73
Streetlight
ent intensities by isolating the effect of spectrum. These methods fol-
LED VBLFL-855-4-40 LED 4663 70
low the overall approach of Aubé, Roby, and Kocifaj (2013) and the rec-
Cosmopolis 60W Metal Halide 2879 66
ommendations of the BIPM (2006) and CIE (2014).
Yard Blaster LED 4164 64
We used measured spectral distributions for mercury vapor, metal
halide, high pressure sodium, low pressure sodium, incandescent, PC Amber Cree PC Amber LED 1717 59

phosphor-coated amber LED, and 3000 K LED from Elvidge, Keith, AEL 75W PC Amber LED 1743 58

Tuttle, and Baugh (2010). We also obtained spectral power distribu- CWES 74 WW CW7 Filtered LED 2448 54
tions for three filtered LED systems (warm white LED with integrated Iwasaki 60W Mercury Vapor 3757 53
filter) from C&W Energy Solutions, a filtered LED from LED Living MH MASTER HPI-T Metal Halide 3808 51
Technology (LLT) and three lamps used in an experiment with attrac- Plus 400W/645
E40 1SL
tion of shearwaters to light (Rodríguez et al., 2017a; Table 1; Figure 1).
CWES 74 WW CW10 Filtered LED 2096 49
For the species responses, we used spectral response curves devel-
CWES Anna's Light Filtered LED 1193 26
oped for a range of organisms, including insects, sea turtles, and birds
HPS SON-T High Pressure Sodium 1947 18
(Table 2). Some response curves represent behavioral responses to
400W/220 E40 1SL
light of different wavelengths (e.g., moths and hatchling sea turtles)
150 W HPS High Pressure Sodium 2059 17
while others represent the visual sensitivity of the eyes of the organ-
18 W LPS Low Pressure Sodium 1810 −44
isms or physiological response (photosynthesis). For visual sensitivity
curves, we used log10 transformed values, which were then normal-
ized, because perceptual responses to visual cues are widely seen to
be on a log scale as suggested by Stevens’ power law (Stevens, 1961) and M/P ratio (melanopic/photopic ratio), using the spreadsheet from
and its application to sensory phenomena in insects (Ruchty, Roces, & Lucas et al. (2014).
Kleineidam, 2010). We then calculated the ratio of the actinic power of each lamp per
To evaluate the potential effect of each lamp on night sky pollu- lux of output compared to a D65 standard. This measurement com-
tion, we calculated the Star Light Index proposed by Aubé et al. (2013) pares the effect on each species response or light pollution metric of
using the spreadsheet provided as an electronic supplement, which an additional lux of each lamp type, compared with an additional lux
tracks human scotopic vision. We also calculated indices to evaluate of daylight (the D65 standard). We also calculated ratio of the actinic
the effect of spectrum on Rayleigh scattering, which would be preva- power of each lamp compared with the total power of the lamp. This
lent near cities, and Mie scattering, which would predominate in indi- measurement indicates how much of the energy output of the lamp will
rect skyglow >80 km from city centers (Aubé, 2015; Luginbuhl, Boley, affect each species or light pollution metric.
& Davis, 2014; see Figure 2). To illustrate the tradeoffs between minimizing effects on different
Finally, we calculated photometric indices for each light source groups of wildlife and optimizing performance for outdoor lighting,
that are important to lighting engineers and end users. These include we calculated mean values for each lamp, consisting of: 1) animal
the correlated color temperature (CCT), color rendering index (CRI), response by taxonomic group (insect mean, sea turtle mean, Newell's
4 LONGCORE ET AL .

FIGURE 1 Spectral power distributions of light sources investigated. The five panels are in order of decreasing CRI from top left to lower middle

TA B L E 2 Organismal response spectra

Taxon Response Format Notes and Source

Moths (Lepidoptera) Behavioral Digitized by CIE (Cleve, 1964)
Bee (Hymenoptera) Behavioral Digitized by CIE (Menzel & Greggers, 1985)
Insects (Class Insecta) Behavioral Modeled Composite metric for all Insecta
(Donners et al., 2018)
Green turtle hatchlings (Chelonia Behavioral Digitized (Witherington, 1992)
mydas)
Green turtle adults (Chelonia mydas) Visual sensitivity Digitized (Midolo, 2011) See also (Levenson,
Eckert, Cognale, Deegan, &
Jacobs, 2004)
Loggerhead hatchlings (Caretta Behavioral Digitized (Witherington, 1992)
caretta)
Juvenile Atlantic salmon (Salmo Visual sensitivity Digitized (Hawryshyn, Ramsden, Betke, &
salar) Sabbah, 2010)
Newell's shearwater (Puffinus Visual sensitivity Digitized (Reed, 1986)
newelli)
Photosynthesis (Plantae) Physiological Digital (DIN, 2016)

shearwater, juvenile salmon, or the mean of all four), 2) Star Light This approach is necessary to account for the –44 CRI of low
Index, 3) melatonin suppression, and 4) visual performance. For visual pressure sodium lamps so that all values of the index range 0–1.
performance, we assumed that CRI greater than 75 was acceptable We calculated which lamps performed best as an average of the
and assigned values as follows: four categories, running the average once for each of the organis-
mal responses (to match a scenario where that species or species
group was most important) and for all organismal responses with a
⎛CRI > 75 ⇒ 1 ⎞
weight of 1 for each of the major taxonomic groups. For compar-
If ⎜ (75 − CRI) ⎟.
⎜else ⇒ 1− ⎟ ison with a ranking that considers only environmental factors, we
⎝ 150 ⎠
LONGCORE ET AL . 5

FIGURE 2 Response curves that can be used to estimate influence of light sources [Color figure can be viewed at wileyonlinelibrary.com]

calculated performance for each lamp in the same manner but without each additional lux of light from a low pressure sodium lamp has 20%
incorporating CRI. of the effect on moths as would an additional lux of daylight, while an
To test this approach with experimental data, we compared the additional lux of a mercury vapor lamp would have 72% of the effect of
results of the light hazard for shearwaters in an experiment compar- an additional lux of daylight (Table 4).
ing light attraction of short-tailed shearwaters for metal halide, high The tested lamp types ranged in CRI from –44 (low pressure
pressure sodium, and 4536 K LED lamps (Rodríguez et al., 2017a). sodium) to 99, and CCT from 1193 (Anna's light) to 6270 (LEDway
We modeled relative attraction using the same approach of general- Streetlight). CCT and CRI were significantly but not strongly corre-
ized linear mixed models with night as a random factor and actinic lated (95% CI = 0.10–0.73). The variation in relative actinic power for
power, lamp type, brightness, and CCT each in separate models as an lamps varied most for juvenile salmon (range, 0.15–1), substantially
independent factor. We compared models using Akaike's Information for insects (range, 0.33–1.16) and sea turtles (range, 0.38–1.02), and
Criterion and visualized the fit using scatterplots. Pearson Product- least for Newell's shearwaters (range, 0.65–1). For three of the four
Moment Correlation between responses and photometric indices, species groups tested, narrow band lamps with restricted emissions
and all other statistics were calculated using JMP Pro 13 (SAS, Inc., in the shorter wavelengths had the lowest actinic power relative to
Cary, NC). daylight. Only for Newell's shearwater did one narrow spectrum lamp
All of the calculations and visualization of the intersection of light (CWES Anna's Light) score higher than full spectrum lamps (Figure 3).
spectrum and human and animal response curves can be viewed at a Composite assessments that gave equal weight to a wildlife group
website (https://github.com/herf/ecological) that will be updated with response, melatonin suppression, and Star Light Index showed low-
new lamp spectra and response curves and will allow users to submit est effects for lamps with low emissions in the shorter wavelengths
spectra for analysis. (Figure 4a), with low pressure sodium showing the lowest impacts.
When CRI was included as a factor, low pressure sodium lamp did
not perform as well (Figure 4b), despite low actinic power for wildlife,
3 RESULTS because of its low CRI. Instead, PC Amber and two filtered LEDs scored
lowest overall.
Actinic power as a percent of total power describes the amount of Correlations between photometric values for lamps and resulting
energy from each lamp spectrum that affects the various species light pollution effects were positive and strongest for CCT and both
and photometric indices. For some lamps this proportion is rela- melanopic effect and Star Light Index, positive but weak for CRI and
tively high for most action spectra, and for some species responses other metrics and modestly strong and positive for CCT and equally
the proportion is high for most lamps (Table 3). For example, a high weighted wildlife effects (Table 5). Most importantly to our approach,
proportion of the power from all lamp types is calculated to influ- although CCT has a high correlation with the aggregate wildlife effects
ence loggerhead hatchlings, while few lamps concentrate their power (95% CI = 0.57–0.90), the correlation between CRI and wildlife effects
in the areas of the spectrum most attractive to juvenile salmon is lower (95% CI = 0.43–0.86). The same is true for nearly all of the
(Table 3). individual responses; CCT predicts wildlife effects more than CRI, with
Actinic power per lux compared with daylight calculates the effect higher CCT values more likely to have higher effects on the wildlife
on species of increasing or decreasing illumination (in lux). For example, assessed in this study than higher CRI values.
6 LONGCORE ET AL .

TA B L E 3 Actinic power as a percent of total power for each of the taxonomic-specific responses

Green Green
Insect turtle turtle Loggerhead
Light source Photosynthesis Moth Bee index behavior visual behavior Salmon Shearwater
D65 56 43 22 30 50 66 78 30 60
A 44 22 12 11 23 45 62 14 41
Kerosene Oil 34 13 4.9 4.6 12 31 47 10 25
Full moon 53 31 19 19 37 57 74 20 54
TL950 65 42 27 26 52 72 90 26 75
SORAA Vivid 65 43 25 27 51 71 88 27 70
LLT Telescope Light 61 26 14 13 19 61 90 11 69
CFL Greenlite 13 W 58 38 30 27 40 72 91 22 81
Philips AmbientLED 61 30 20 17 31 65 90 15 72
3000K LED 57 35 25 24 39 67 87 18 73
OCTRON 32 W 62 43 32 29 52 74 91 28 81
Metal Halide 70W 56 37 25 27 39 68 87 22 73
Ceramic Metal Halide 56 37 25 27 39 68 87 22 73
70 W
LSG Good Night 2016 62 30 18 17 27 66 93 13 75
LEDway Streetlight 65 45 32 28 61 75 91 31 79
CW 54W
Los Angeles LED 64 41 29 26 51 72 91 27 77
Cosmopolis 60W 58 38 24 27 41 70 90 21 79
Yard Blaster 56 47 28 37 53 75 87 29 76
PC Amber Cree 61 25 12 13 17 61 92 11 73
AEL 75W 61 25 12 13 17 61 91 10 72
CWES 74 WW CW7 58 28 23 18 27 66 93 10 80
Iwasaki 60W 41 30 17 24 29 51 65 20 48
CWES 74 WW CW10 59 27 18 16 22 64 93 10 78
CWES Anna's Light 64 23 3.8 8.7 8.4 59 92 10 71
150 W HPS 57 30 15 19 29 65 89 14 82
LPS 18 W 55 28 13 20 25 68 95 8.7 97

The reanalysis of shearwater grounding data shows that actinic although it is not a significant issue for most LEDs used for outdoor
power per lux provides at least an equally valid model (AICc 546.83, lighting.
effect 95% CI 3.69–61.84) as a categorical analysis with lamp type The approach described here establishes appropriate units for mea-
(AICc 547.59, LED effect 95% CI –1.07 to 0.45, MH 95% CI 0.20– suring ecological responses to light that are consistent with interna-
1.72) (Figure 5). The model for CCT had a higher AICc (549.13) tional standards and thereby provides a basis for comparison that is
with an effect 95% CI intersecting 0, while the model for bright- replicable and testable. Quantification of actinic power can be used to
ness had a still higher AICc (551.44) and a 95% CI for effect also develop hypotheses to test in the field, such as the comparison of lamp
intersecting 0. types undertaken by Rodríguez et al. (2017a) that we revisited. Fur-
thermore, it allows the rapid and easily updatable comparison of new
lamp types so that the most promising spectral configurations for a par-
4 DISCUSSION ticular situation can be identified and tested in the field.
Our approach is, however, only as accurate as the action spec-
Our effort extends the approach presented by Aubé et al. (2013) to tra and as applicable as the number of different species groups for
develop a method to calculate indices for any organismal response which action spectra are available. These response curves are scat-
to lighting spectrum assuming equal visual light intensity to humans. ted in the literature and although many physiological response curves
These calculations can be easily repeated and updated with additional could be calculated from, for example, peak opsin sensitivities (Davies
organismal response curves or with additional lighting products. We et al., 2013), behavioral response curves derived from field and labora-
included the ultraviolet part of the spectrum because many other light tory tests are more rare. In at least one instance (loggerhead sea tur-
sources do include ultraviolet and it is important for animal responses, tle hatchlings) there may be behavioral response differences between
LONGCORE ET AL . 7

TA B L E 4 Actinic power per lux of each lamp type, compared with a lux of daylight (D65)

Green Green
Insect turtle turtle Loggerhead
Light source Photosynthesis Moth Bee index behavior visual behavior Salmon Shearwater
D65 1 1 1 1 1 1 1 1 1
A 1 0.639 0.681 0.482 0.588 0.865 1.010 0.587 0.867
Kerosene Oil 1.360 0.673 0.494 0.340 0.558 1.050 1.340 0.754 0.924
Full moon 0.922 0.704 0.821 0.597 0.72 0.841 0.917 0.642 0.874
TL950 0.827 0.691 0.858 0.611 0.736 0.774 0.815 0.618 0.876
SORAA Vivid 0.927 0.793 0.891 0.711 0.822 0.860 0.894 0.720 0.920
LLT Telescope Light 0.772 0.425 0.458 0.306 0.275 0.660 0.818 0.259 0.812
CFL Greenlite 13 W 0.573 0.487 0.746 0.490 0.445 0.606 0.648 0.410 0.748
Philips AmbientLED 0.716 0.464 0.593 0.375 0.408 0.648 0.756 0.33 0.785
3000K LED 0.647 0.522 0.714 0.515 0.497 0.655 0.714 0.392 0.778
OCTRON 32 W 0.632 0.573 0.847 0.556 0.599 0.648 0.670 0.534 0.773
Metal Halide 70W 0.656 0.568 0.732 0.576 0.512 0.673 0.723 0.481 0.788
Ceramic Metal Halide 0.656 0.568 0.732 0.576 0.512 0.673 0.723 0.481 0.788
70 W
LSG Good Night 2016 0.696 0.431 0.526 0.343 0.343 0.625 0.743 0.284 0.779
LEDway Streetlight 0.715 0.645 0.900 0.574 0.748 0.697 0.713 0.629 0.800
CW 54W
Los Angeles LED 0.688 0.579 0.782 0.510 0.614 0.657 0.700 0.545 0.771
Cosmopolis 60W 0.603 0.519 0.644 0.518 0.485 0.622 0.668 0.415 0.764
Yard Blaster 0.646 0.701 0.821 0.783 0.686 0.729 0.717 0.624 0.816
PC Amber Cree 0.718 0.387 0.361 0.273 0.223 0.613 0.768 0.232 0.792
AEL 75W 0.711 0.383 0.366 0.274 0.225 0.609 0.762 0.229 0.785
CWES 74 WW CW7 0.542 0.342 0.539 0.309 0.283 0.530 0.624 0.178 0.695
Iwasaki 60W 0.771 0.731 0.806 0.822 0.613 0.817 0.869 0.715 0.827
CWES 74 WW CW10 0.581 0.342 0.446 0.285 0.246 0.540 0.653 0.186 0.715
CWES Anna's Light 0.876 0.414 0.131 0.221 0.129 0.681 0.898 0.266 0.898
150 W HPS 0.529 0.368 0.365 0.335 0.307 0.517 0.593 0.243 0.705
LPS 18 W 0.375 0.254 0.221 0.254 0.193 0.393 0.462 0.112 0.615

populations of the same species (Fritsches, 2012), meaning that cau- Any approach to reduce ecological effects of lights must keep inten-
tion should be used in universally applying action spectra. The emer- sity to a minimum and can then perhaps further reduce adverse effects
gence of highly configurable outdoor lighting demonstrates the need through tuning of the spectrum used.
for research to produce more action spectra and to compile them in We also note that the influence of lamps of different spectra will
a repository. This is a central research need from experimental zool- be affected by atmospheric conditions that influence the amount and
ogists to provide the information necessary for lighting designers and nature of reflection and scattering of light (Aubé, Kocifaj, Zamorano,
especially regulators to act quickly in response to new lighting tech- Lamphar, & de Miguel, 2016; Kyba, Ruhtz, Fischer, & Hölker, 2011). Our
nologies. Peak opsin sensitivity provides a first pass on behavioral wildlife response assessments do not include any shifts in spectral dis-
responses, and indeed, behavioral response curves can be calibrated tribution of light that would result from scattering in the atmosphere
from opsin response curves (Donners et al., 2018). Workers in the field and therefore are most relevant to situations where direct effects are
and with captive animals should, however, prioritize research to obtain being evaluated (e.g., local attraction and disorientation). Additional
behavioral response information for sensitive species and to test the calculations could be added to our approach to address different prop-
generalizable patterns in responses within clades where visual systems agation patterns of light under varying weather conditions.
are conserved. Our use of CRI as a metric for performance of lamps for human
We are aware of the limitations of using spectral information that vision should not be taken as a blanket endorsement of CRI as an
may only be applicable within a certain range of intensity values. excellent metric, which it is not (Galadí-Enríquez, 2018). It is, how-
Some species respond to spectrum differently depending on its inten- ever, widely understood and used in the lighting design community
sity (Wiltschko, Stapput, Thalau, & Wiltschko, 2010). Also, mitigation and therefore provides a means to incorporate human design prefer-
schemes that depend on spectrum can be undermined by brightness. ences into a composite metric of lighting performance. Furthermore,
8 LONGCORE ET AL .

F I G U R E 3 Relative modeled impact on insects, sea turtles, shearwaters, and juvenile salmon per additional lux from different light spectra com-
pared with a D65 (6500 K) standard. Colors indicate CCT from low (orange) to high (blue) [Color figure can be viewed at wileyonlinelibrary.com]

F I G U R E 4 Nighttime light performance index balancing Star Light Index, melatonin suppression, and a wildlife impact score (a) and incorporating
CRI (b) for equal lux from different light spectra compared with a D65 (6500K) standard. Lower values indicate lower predicted impacts and greater
CRI. Colors indicate CCT from low (orange) to high (blue) [Color figure can be viewed at wileyonlinelibrary.com]

this approach can be updated to use other metrics as desired by an end worthwhile to balance those adverse effects against desirable charac-
user. teristics of outdoor lighting for human use. Lamps that perform well
As a conservation tool, our assessments assume that it is a valuable in this assessment would represent a conservation compromise—no
approach to minimize the intersection between the wavelengths that light on a sea turtle nesting beach, on a penguin colony, or on the route
affect sensitive wildlife species and the output of lamps and that it is a fledgling seabird takes to the sea would be optimal, but if there is
LONGCORE ET AL . 9

TA B L E 5 Pearson's product moment correlation between CCT, CRI, Star Light index, Melanopic response, and average wildlife response. Above
diagonal, correlation estimates. Below diagonal, 95% confidence intervals

CCT CRI Star light index Melanopic Wildlife

CCT – 0.48 0.94 0.94 0.78
CRI 0.10–0.73 – 0.64 0.67 0.71
Star Light Index 0.87–0.97 0.40–0.84 – 1.00 0.85
Melanopic 0.87–0.97 0.33–0.83 0.99–1.00 – 0.85
Wildlife 0.57–0.90 0.43–0.86 0.69–0.93 0.69–0.93 –

to be a light nearby, minimizing the wavelengths in the part of the available to reduce the ecological effects from light pollution that are in
spectrum to which turtles or seabirds are most sensitive is prefer- place around the world are focused predominantly on the direction and
able (Rodríguez et al., 2017b, 2018), so long as intensity is also min- intensity of lighting; very few legislators saw the dramatic change in
imized. Such hierarchical minimizing approaches might ignore other color on the technological horizon. Those jurisdictions that have taken
more complete solutions such as embedded roadway lighting, which steps to use energy efficient lighting with a spectrum designed to min-
provides guidance to drivers and virtually no light on nearby beaches imize adverse environmental effects have been motivated mostly by
(Bertolotti & Salmon, 2005), but they do provide guidance for reduc- particular species protection laws (e.g., the Endangered Species Act in
ing adverse effects from existing lighting infrastructure, which will be the United States) and by the economic considerations associated with
replaced with full-spectrum lights in the absence of guidance from astronomical observatories.
ecologists and consideration of wildlife responses. The State of Florida requires that new coastal construction limit
Given the rapid pace of replacement of street and other outdoor lighting near beaches to sources that emit wavelengths only greater
lighting motivated by energy savings (Hecht, 2016), an approach to than 560 nm to protect sea turtles. Our calculations suggest that sev-
minimize the adverse effects of lighting through choice of spectrum eral of the filtered LEDs that we assessed would be less attractive to
that is endorsed by conservation scientists is desperately needed. Laws hatchling sea turtles than existing HPS lamps, but none of the filtered

F I G U R E 5 Analysis of birds grounded from Rodríguez et al. (2017a), comparing Actinic Power per Lux with CCT, brightness, and lamp type as
explanatory variables
10 LONGCORE ET AL .

lamps meets the 560 nm cutoff. This raises the interesting regulatory BIPM. (2006). The international system of units (SI). Sèvres, France: Bureau
question of whether it might be acceptable to modify the strict 560 nm International des Poids et Mesures.

cutoff in favor of a whole-spectrum assessment that we have proposed Björn, L. O. (2015). Action spectroscopy in biology. Photobiology (pp. 85–96).
New York, NY: Springer.
here, which would lead to approving lamps for street and outdoor light-
ing (e.g., at ports) that we predict would be less disruptive to turtles, Boyce, P., Fotios, S., & Richards, M. (2009). Road lighting and energy saving.
Lighting Research & Technology, 41, 245–260.
increase color rending when replacing existing HPS, and save signif-
CIE. (2014). Technical Note: Relating photochemical and photobiological
icant energy. Of course, to fully address outdoor light management,
quantities to photometric quantities. CIE TN 02:2014. Vienna, Austria:
additional techniques to control light intensity, direction, and duration International Commission on Illumination.
would need to be employed (Longcore and Rich, 2017), such as use of
Cleve, K. (1964). Der Anflug der Schmetterlinge an künstliche Lichtquellen.
shields, baffles, and louvers to reduce spill light (Mizon, 2002). Mitteilungen der Deutschen Entomologischen Gesellschaft, 23, 66–76.
Decision-making power for new lighting types is often vested in Davies, T. W., Bennie, J., Inger, R., de Ibarra, N. H., & Gaston, K. J. (2013). Arti-
street lighting agencies and departments of transportation. When reg- ficial light pollution: Are shifting spectral signatures changing the bal-
ulations exist to control lighting to reduce harms to certain species, ance of species interactions? Global Change Biology, 19, 1417–1423.

these agencies must comply with relevant laws. They also answer to DIN. 2016. DIN 5031-10: Optical radiation physics and illuminating engi-
public opinion on the aesthetics of lighting, as has been shown for many neering - Part 10: Photobiologically effective radiation, quantities, sym-
bols and action spectra. Berlin, Germany: Author [German Institute for
LED projects around the USA that have raised the ire of local residents
Standardization].
because the high CCT lamps produce significant glare and were dis-
Dominoni, D. M. (2015). The effects of light pollution on biological rhythms
pleasing to residents (Hecht, 2016). For those governmental actors try- of birds: An integrated, mechanistic perspective. Journal of Ornithology,
ing to balance considerations for wildlife, the night sky, and safety, clear 156, 409–418.
advice on spectrum is needed to navigate the many available choices. Donners, M., van Grunsven, R. H. A., Groenendijk, D., van Langevelde, F.,
This information is also necessary for regulators facing these issues. Bikker, J. W., Longcore, T., & Veenendaal, E. M. (2018). Colours of attrac-
tion: A general model for insect phototaxis. Journal of Experimental Zool-
ogy A, https://doi.org/10.1002/jez.2188

ACKNOWLEDGMENTS Dudley, J. M., Erkintalo, M., & Genty, G. (2015). Environment, Wildlife and
LED Illumination. Optics and Photonics News, 26, 42–47.
CW Energy Systems provided seed funding for this investigation but
Eisenbeis, G. (2006). Artificial night lighting and insects: Attraction of
did not have any part in the research or reporting of the results. We
insects to streetlamps in a rural setting in Germany. In C. Rich & T. Long-
thank Tim Robinson for productive discussions of this approach and B. core (Eds.), Ecological consequences of artificial night lighting (pp. 281–
M. Seymoure for constructive comments on a draft of this manuscript. 304). Washington, D.C.: Island Press.
AR was supported by a Marie Curie International Outgoing Fellow- Eisenbeis, G., & Eick, K. (2011). Studie zur Anziehung nachtaktiver Insekten
ship (330655 FP7-PEOPLE-2012-IOF) and a Juan de la Cierva contract an die Straßenbeleuchtung unter Einbeziehung von LEDs [Attraction of
nocturnal insects to street lights—A study of lighting systems, with con-
from the Spanish Ministry of Economy, Industry and Competitiveness
sideration of LEDs]. Natur und Landschaft, 86, 298–306.
(IJCI-2015-23913).
Eisenbeis, G., & Hänel, A. (2009). Light pollution and the impact of artifi-
cial night lighting on insects. In M. J. McDonnell, A. K. Hahs, & J. Breuste
(Eds.), Ecology of cities and towns: A comparative approach (pp. 243–263).
ORCID Cambridge, England: Cambridge University Press.
Travis Longcore http://orcid.org/0000-0002-1039-2613 Elvidge, C. D., Keith, D. M., Tuttle, B. T., & Baugh, K. E. (2010). Spec-
Airam Rodríguez http://orcid.org/0000-0001-7882-135X tral identification of lighting type and character. Sensors, 10, 3961–
3988.
Fox, R. (2013). The decline of moths in Great Britain: A review of possible
REFERENCES causes. Insect Conservation and Diversity, 6, 5–19.
Allen, J. A. (1880). Destruction of birds by light-houses. Bulletin of the Nuttall Fritsches, K. A. (2012). Australian loggerhead sea turtles do not avoid yel-
Ornithological Club, 5, 131–138. low. Marine and Freshwater Behaviour and Physiology, 45, 79–89.
Aubé, M. (2015). Physical behaviour of anthropogenic light propagation Galadí-Enríquez, D. (2018). Beyond CCT: The spectral index system as a
into the nocturnal environment. Philosophical Transactions of the Royal tool for the objective, quantitative characterization of lamps. Journal of
Society B Biological Sciences, 370, 20140117. Quantitative Spectroscopy and Radiative Transfer, 206, 399–408.
Aubé, M., Kocifaj, M., Zamorano, J., Lamphar, H. A. S., & de Miguel, A. S. Gaston, K. J. (2013). A green light for efficiency. Nature, 497, 560–561.
(2016). The spectral amplification effect of clouds to the night sky radi-
Gaston, K. J., Bennie, J., Davies, T. W., & Hopkins, J. (2013). The ecologi-
ance in Madrid. Journal of Quantitative Spectroscopy and Radiative Trans-
cal impacts of nighttime light pollution: A mechanistic appraisal. Biology
fer, 181, 11–23.
Reviews, 88, 912–927.
Aubé, M., Roby, J., & Kocifaj, M. (2013). Evaluating potential spectral
Gaston, K. J., Davies, T. W., Bennie, J., & Hopkins, J. (2012). Reducing the eco-
impacts of various artificial lights on melatonin suppression, photosyn-
logical consequences of night-time light pollution: Options and develop-
thesis, and star visibility. PLoS One, 8, e67798.
ments. Journal of Applied Ecology, 49, 1256–1266.
Bertolotti, L., & Salmon, M. (2005). Do embedded roadway lights protect sea
Hawryshyn, C., Ramsden, S., Betke, K., & Sabbah, S. (2010). Spectral and
turtles? Environmental Management, 36, 702–710.
polarization sensitivity of juvenile Atlantic salmon (Salmo salar): Phy-
Bierman, A. (2012). Will switching to LED outdoor lighting increase sky logenetic considerations. Journal of Experimental Biology, 213, 3187–
glow? Lighting Research & Technology, 44, 449–458. 3197.
LONGCORE ET AL . 11

Hecht, J. (2012). Changing the lights: Are LEDs ready to become the market Reed, J. R. (1986). Seabird vision: Spectral sensitivity and light-attraction
standard? Optics and Photonics News, 23, 44–50. behavior [Dissertation]. Madison, WI: University of Wisconsin. 1–
Hecht, J. (2016). The early-adopter blues. IEEE Spectrum, 53, 44–50. 190

Johnsen, S. (2012). The optics of life: A biologist's guide to light in nature. Prince- Rich,C. & Longcore,T. (2006). Ecological consequences of artificial night lighting
ton, England: Princeton University Press. (pp. 458). Washington, D.C.: Island Press.

Kinzey, B., Perrin, T. E., Miller, N. J., Kocifaj, M., Aubé, M., & Lamphar, Rodríguez, A., Dann, P., & Chiaradia, A. (2017a). Reducing light-induced
H. S. 2017. An investigation of LED street lighting's impact on sky glow. mortality of seabirds: High pressure sodium lights decrease the fatal
Richland, Washington: U.S. Department of Energy (Contract DE-AC05- attraction of shearwaters. Journal of Nature Conservation, 39, 68–
76RL01830). 72.

Kyba, C., Hänel, A., & Hölker, F. (2014). Redefining efficiency for outdoor Rodríguez, A., Holmes, N. D., Ryan, P. G., Wilson, K-J., Faulquier, L., Murillo,
lighting. Energy & Environmental Sciences, 7, 1806–1809. Y., …Corre, M. L. (2017b). Seabird mortality induced by land-based arti-
ficial lights. Conservation Biology, 31, 986–1001.
Kyba, C. C. M., Kuester, T., de Miguel, A. S., Baugh, K., Jechow, A., Hölker, F.,
…Guanter, L. (2017). Artificially lit surface of Earth at night increasing Rodríguez, A., Holmberg, R., Dann, P., & Chiaradia, A. (2018). Pen-
in radiance and extent. Science Advances, 3, e1701528. guin colony attendance under artificial lights for ecotourism. Jour-
nal of Experimental Zoology Part A Ecology and Integrative Physiology.
Kyba, C. C. M., Ruhtz, T., Fischer, J., & Hölker, F. (2011). Cloud coverage acts https://doi.org/10.1002/jez.2155
as an amplifier for ecological light pollution in urban ecosystems. PLoS
One, 6, e17307. Ruchty, M., Roces, F., & Kleineidam, J. (2010). Detection of minute tem-
perature transients by thermosensitive neurons in ants. Journal of
Levenson, D. H., Eckert, S. A., Cognale, M. A., Deegan, II, J. F., & Jacobs, G. Neurophysiology, 104, 1249–1256.
H. (2004). Photopic spectral sensitivity of green and loggerhead sea tur-
tles. Copeia, 2004, 908–914. Salmon, M. (2003). Artificial night lighting and turtles. Biologist, 50, 163–
168.
Longcore, T., Aldern, H. L., Eggers, J. F., Flores, S., Franco, L., Hirshfield-
Yamanishi, E., …Barroso, A. M. (2015). Tuning the white light spectrum Somers-Yeates, R., Bennie, J., Economou, T., Hodgson, D., Spalding, A., &
of light emitting diode lamps to reduce attraction of nocturnal arthro- McGregor, P. K. (2016). Light pollution is associated with earlier tree
pods. Philosophical Transactions of the Royal Society B Biological Sciences, budburst across the United Kingdom. Proceedings. Biological Sciences,
370, 20140125. 283, 20160813.

Longcore, T., & Rich, C. 2017. Artificial Night Lighting and Protected Lands: Stevens, S. S. (1961). To honor Fechner and repeal his law. Science, 133, 80–
Ecological Effects and Management Approaches (Revised August 86.
2017). Natural Resource Report NPS/NRSS/NSNS/NRR—2017/1493. Stone, E. L., Harris, S., & Jones, G. (2015). Impacts of artificial lighting on
Fort Collins, Colorado: National Park Service. 1–51. bats: A review of challenges and solutions. Mammalian Biology, 80, 213–
Longcore, T., Rich, C., Mineau, P., MacDonald, B., Bert, D. G., Sullivan, L. M., 219.
…Drake, D. (2012). An estimate of avian mortality at communication Stone, E. L., Jones, G., & Harris, S. (2012). Conserving energy at a cost to
towers in the United States and Canada. PLoS One, 7, e34025. biodiversity? Impacts of LED lighting on bats. Global Change Biology, 18,
Lucas, R. J., Peirson, S. N., Berson, D. M., Brown, T. M., Cooper, H. M., 2458–2465.
Czeisler, C. A., …O'Hagan, J. B. (2014). Measuring and using light in the Wilson, J. F., Baker, D., Cheney, J., Cook, M., Ellis, M., Freestone, R.,
melanopsin age. Trends in Neurosciences, 37, 1–9. …Hodgers, D. (2018). A role for artificial night-time lighting in long-
Luginbuhl, C. B., Boley, P. A., & Davis, D. R. (2014). The impact of light source term changes in populations of 100 widespread macro-moths in UK
spectral power distribution on sky glow. Journal of Quantitative Spec- and Ireland: A citizen-science study. Journal of Insect Conservation , 1–8.
troscopy and Radiative Transfer, 139, 21–26. https://doi.org/10.1007/s10841-018-0052-1

Menzel, R., & Greggers, U. (1985). Natural phototaxis and its relationship Wiltschko, R., Stapput, K., Thalau, P., & Wiltschko, W. (2010). Directional ori-
to colour vision in honeybees. Journal of Comparative Physiology A, 157, entation of birds by the magnetic field under different light conditions.
311–321. Journal of the Royal Society Interface, 7, rsif20090367.

Midolo, N. (2011). The suggested effect of lighting on green turtles based Witherington, B. E. (1992). Behavioral responses of nesting sea turtles to
on a proposed V(𝜆) curve. Lighting Magazine, 31, 36–41. artificial lighting. Herpetologica, 48, 31–39.

Minnaar, C., Boyles, J. G., Minnaar, I. A., Sole, C. L., & McKechnie, A. E. (2015).
Stacking the odds: Light pollution may shift the balance in an ancient
predator–prey arms race. Journal of Applied Ecology, 52, 522–531. How to cite this article: Longcore T, Rodríguez A, Witherington
Mizon, B. (2002). Light pollution: Responses and remedies. London, England: B, Penniman JF, Herf L, Herf M. Rapid assessment of lamp spec-
Springer.
trum to quantify ecological effects of light at night. J Exp Zool.
Pawson, S., & Bader, M.-F. (2014). LED lighting increases the ecological 2018;1–11. https://doi.org/10.1002/jez.2184
impact of light pollution irrespective of color temperature. Ecology Appli-
cations, 24, 1561–1568.

View publication stats