Long-term exposure to experimental light affects the ground-dwelling invertebrate community, independent of light spectra

Abstract

Our planet endures a progressive increase in artificial light at night (ALAN), which affects virtually all species, and thereby biodiversity. Mitigation strategies include reducing its intensity and duration, and the adjustment of light spectrum using modern light emitting diode (LED) light sources. Here, we studied ground-dwelling invertebrate (predominantly insects, arachnids, molluscs, millipedes, woodlice and worms) diversity and community composition after 3 or 4 years of continued nightly exposure (every night from sunset to sunrise) to experimental ALAN with three different spectra (white-, and green- and red-dominated light), as well as for a dark control, in natural forest-edge habitat. Diversity of pitfall-trapped ground-dwelling invertebrates, and the local contribution to beta diversity, did not differ between the dark control and illuminated sites, or between the different spectra. The invertebrate community composition, however, was significantly affected by the presence of light. Keeping lights off during single nights did show an immediate effect on the composition of trapped invertebrates compared to illuminated nights. These effects of light on species composition may impact ecosystems by cascading effects across the food web.

This article is part of the theme issue ‘Light pollution in complex ecological systems’.

1. Introduction

Anthropogenic light at night continues to increase with expansion of human populations and economic growth. Upwards emitted light at night increases by approximately 2% yr⁻¹, and globally, the area exposed to artificial light at night (ALAN) of greater intensity than natural light increases by a comparable percentage [1]. However, the total amount of light that organisms living on the surface of the Earth are exposed to may be increasing by 6–10% annually because of exposure to horizontally emitted light which is not detected by satellites [2]. Understanding of the ecological impact of light at night is steadily growing and suggests that virtually all species are affected [3,4]. ALAN has a particularly strong impact on invertebrates. The effects are variable, ranging from the disturbance of daily and seasonal rhythms [5,6], shifts in foraging behaviour [7], physiology and immune responses [8,9], and disruption to reproductive behaviours and physiology [10–12]. One of the most conspicuous and widely known effects is the attraction of flying insects to light at night [13,14]. This effect can have severe consequences for local insect densities [15] and may be an important driver behind global insect declines [16]. Less information is available on the impact of anthropogenic light on ground-dwelling arthropods. Available evidence suggests that these communities may respond differently [17,18]. For airborne species, night-time abundance around long-term light posts increased (particularly if the lights emitted a higher proportion of short-wavelength light), but abundance of ground-dwelling invertebrates decreased and their species composition was altered [18]. For flying insects, the influence of shorter wavelengths (ultraviolet, blue) on attraction has been demonstrated [19,20] and can have consequences for mate attraction [10,21,22] and seasonal timing [23]. Despite the studies highlighted above, our understanding of the broader impact of light on invertebrate species diversity and community composition remains relatively poor, and important questions remain unanswered. How ground-dwelling invertebrates respond to long-term exposure of light in natural (non-urban) habitat, and the degree to which variation in light spectrum influences species diversity is relatively untested. Given the reduced mobility of ground-dwelling compared to airborne invertebrates, it is unlikely that we would see direct ‘recovery’ responses if nocturnal light disturbance was removed. This is of interest, as long-term changes have long-lasting consequences for the food web by cascading to other trophic levels, but this remains to be tested. Here, we determined the effect of long-term experimental night-time illumination of a forest-edge habitat on the diversity and composition of invertebrate communities. Specifically, we aim to identify the relative contributions of the presence and spectrum of light. Lastly, we explore whether light-induced effects have a permanent character by testing the effect of a short-term temporal return of darkness.

2. Methods

(a) . Study site

Data were collected at eight experimental field sites in forest edge habitat in nature reserves throughout The Netherlands, the ‘Light on Nature’ sites. A full description can be found in Spoelstra et al. [24]. Briefly, each site has four transects placed perpendicular to a forest edge. Each transect is illuminated with a specific light spectrum by Philips Residium FGS224 (1xPL-L36W HFP) armatures placed on 4 m tall lampposts. Each transect has a unique light spectrum, either Philips Fortimo white, Clearsky green, or Clearfield red light (Philips, Amsterdam, The Netherlands), or is left dark. All lights emit full spectrum light; green lamps have an increased blue and reduced red light emission, and red lamps have an increased red and reduced blue emission. The green spectrum was initially designed to minimize impact on nocturnally migrating birds; the red spectrum was designed to potentially minimize the impact on nocturnal species which are relatively strongly attracted to the blue part of the spectrum or are generally sensitive to short wavelengths (bats, mice) [24] (see the electronic supplementary material, figure S1 for spectra). The four transects are 100 m long and contain five light posts, with two lights in each transect in the forest, one light at the forest edge and two lights in the adjacent open area (except for at two sites where transects have only three lights—two in the forest and one in the forest edge). Lights are—to date—on every night from sunset to sunrise since spring 2012, hence at the time of invertebrate sampling in July and August 2016, the experimental exposure was present for over 4 years, except for site 2 where the lights were on since spring 2013 (see the electronic supplementary material, table S2 for precise onset of nightly illumination of each site). The intensity of all light colours was equalized to 7.6 ± 0.3 lux at ground-level, comparable to average intensity under a regular streetlight [25] (see the electronic supplementary material, figure S1 for the spectral composition of the three different light sources).

(b) . Pitfall trapping

To trap ground-dwelling invertebrates (e.g. insects, arachnida, molluscs, millipedes, woodlice and worms), we placed pitfall traps in a standardized way at ground level, 4 m below the light post fixture and at 2 m distance from the base of the light post. This resulted in the full exposure of 7.3 ± 0.3 lux at the trap entrance. Traps were placed near the outer light post (July) and the outer two light posts (August; figure 1). For each trap, a PVC pipe (Ø 9 cm, 12.5 cm long) was dug in the ground at 2 m distance to the light post base. In each pipe a plastic funnel was placed above a 70 ml lidless plastic collection vial with 50 ml of 70% ethanol. To shield from rain and debris, a rectangular piece of corrugated plastic was placed at 10 cm above each pitfall.

Figure 1.

Schematic layout of one of the eight research sites (table 1). Each site always has (in this example from left to right) a green, white, red illuminated transect and dark control. The order of light colour treatment at transects was randomized within each site when sites were established in 2012 and 2013; lights are always on from sunset to sunrise since (see text). Open circles indicate the locations of pitfalls placed. In July 2016, the pitfalls near the outer light post (black open circles) were active, in August 2016 pitfalls near both the outer two light posts were active (black and white open circles, table 1). See [24] for detailed information on different research sites. (Online version in colour.)

Vials were placed in all pitfall pipes during the daytime (4 and 5 July 2016) and collected approximately 24 h later (5 and 6 July, respectively). We deployed traps again on 7, 8 and 9 August (2016) and collected vials on 8, 9 and 10 August. To test whether the observed effects of lights on ground-dwelling invertebrates is immediately reversible (i.e. has a mere direct character instead of a more permanent character), we left the lights off at all sites during the night between 5 and 6 July, and again between 8 and 9 August (table 1; defined with the variable lights on/off in statistical models applied).

Table 1.

Overview samples per date and per site. (In July, invertebrates were trapped with pitfalls under the outer light post in the forest (one pitfall × four transects × eight sites); in August, invertebrates were trapped with pitfalls under both light posts in the forest edge (two pitfalls × four transects × eight sites). Three samples were lost (from sites 1 and 8 on 6 July, and from site 5 on 9 August). During the night before the collection of pitfall content on 6 July and 9 August, the experimental illumination—normally on each night from sunrise to sunset—was left off to test whether effects in temporary darkness would persist. At two sites on 5, 6 and 7 August, pitfall content was collected at dawn and dusk in order to test for differences between nocturnal and diurnal collection of ground-dwelling arthropods.)

site	site name	5 July	6 July	8 August	9 August	10 August	total
lights during preceding night:		on	off	on	off	on
1	Lebretshoeve	4	3	8	8	8	31
2	Voorstonden	4	4	8	8	8	32
3	Radio Kootwijk	4	4	8	8	8	32
4	ASK 4	4	4	8	8	8	32
5	ASK 5	4	4	8	7	8	31
6	Klaterweg 1	4	4	8	8	8	32
7	Klaterweg 2	4	4	8	8	8	32
8	Hijkerveld	4	3	8	8	8	31
all		32	30	64	63	64	253

site	site name		5 August	6 August	6 August	7 August
samples collected dawn/dusk:			dusk	dawn	dusk	dawn
1	Lebretshoeve		8	8
3	Radio Kootwijk				7	8

In order to test for differences between nocturnal and diurnal presence of ground-dwelling arthropods, pitfall content was collected separately at dawn and dusk on 5, 6 and 7 August in a separate trapping effort at two sites (1 and 3; table 1).

(c) . Invertebrate identification

All invertebrates were classified to family level when possible; collembola, millipedes, centipedes, isopods, pseudoscorpions and harvestmen were classified to order. Acari, nematodes and collembola were excluded owing to the extreme abundances of these groups relative to other taxa.

(d) . Statistical analyses

In all models, we tested for an effect of light colour (dark, red, green and white) on dependent variables. We first calculated the effective number of species (ENS; which is the exponent of the Shannon diversity index H’) using the diversity routine in the R package BiodiversityR [26], and the local contribution to beta diversity (LCBD) using the LCBD.comp routine in the R package Adespatial [27]. Effects on these two parameters by light colour was then tested for in linear mixed effect models, with site and pitfall number as random terms using the lmer routine in the R package lme4 [28]. We first fitted an interaction between lights on or off and light colour or light presence, and without a significant effect of these interactions all data collected during nights with lights on and off were combined and light colour was fitted as a fixed effect. Without an effect of light colour, the procedure was repeated for light presence (the presence of either red, green and white light, all coded as ‘light’ and dark control coded as ‘dark’).

To test for community effects, we calculated Bray–Curtis distance matrices using the vegdist routine in the R package vegan [29] based on relative abundances of taxa in pitfall data which we calculated with the decostand routine in the vegan package.

We then tested for the effects of light colour and light presence on the variance in the distance matrices with a permutational MANOVA (999 permutations) using the adonis2 routine in the vegan package, pitfall identity (ID) nested in transect. To compare treatments within date and site, permutations were restricted for colour treatments within date and site (strata). Comparable to models exploring differences in diversity, we tested first for an interaction between light colour and lights-on. Without a significant effect of these interactions, data for lights on or off were combined in all models. For the parsimony principle, the inclusion of these depended on the goodness of fit based on the model Akaike information criterion (AIC) value. If significant, a post hoc test was done to compare differences between light colour using the pairwise.adonis2 routine in the pairwiseadonis R package [30], to obtain (corrected) p-values for pairwise comparisons.

3. Results

In total, we collected 32 samples at each of the eight sites (except three missing samples, see table 1 for a full overview). No rain was present during any of the trapping dates (see the electronic supplementary material, table S3 for weather data). A total of 8058 invertebrates of 107 taxa were trapped in the pitfalls. The majority of individuals belonged to insects (91.9%), followed by Arachnida (6.4%). Other taxa trapped comprised molluscs, millipedes, woodlice and worms (for a full list of taxa trapped, see the electronic supplementary material). In some samples, swarming insects (such as Nematocera, elongated flies) or social insects (such as Formicidae and ants) were highly abundant (in some cases representing more than half of individuals captured). We retained these high numbers in the data analysis as they did not affect statistical models.

There was no interaction between light colour treatment or light presence (the presence of either white, green and red light versus dark control) and lights on/off (whether the lights were on or off during invertebrate collection) on the ENS, or LCBD values. We therefore did not further distinguish in statistical models between data collected during the 24 h period in which the lights were on or off during the night. Light colour or light presence (any colour of light) was not a significant driver of ENS or LCBD values (table 2). For the LCBD values, only pitfall ID could be fitted as a random term because of model singularity. The date of collection significantly influenced the exponent of the Shannon diversity index: values were higher for the samples collected in August compared to the ones collected in July (table 2).

Table 2.

Selection of linear mixed effect models to identify effects of light colour and light treatment on the ENS and LCBD. (Models are always based on pitfall data collected at all eight sites (table 1). For the LCBD data, interaction models and models including date and site as random terms resulted in singular fits. The effect of light colour and light presence was not significant in any diversity model; the model with the best fit is printed in italics. ***p ≤ 0.001.)

diversity and LCBD models			AIC	loglik	chi sq	d.f.	p
exp H'	∼	light colour * lights on/off+1|site+1|pitfall ID+1|date	1095.5	−535.8	2.057	3	0.561
exp H'	∼	light colour+lights on/off+1|site+1|pitfall ID+1|date	1091.6	−536.8
exp H'	∼	light colour+1|site+1|pitfall ID+1|date	1090.3	−537.1	0.966	3	0.809
exp H'	∼	1|site+1|pitfall ID+1|date	1085.2	−537.6
exp H'	∼	light presence * lights on/off+1|site+1|pitfall ID+1|date	1089.5	−536.8	0.267	1	0.605
exp H'	∼	light presence+lights on/off+1|site+1|pitfall ID+1|date	1087.8	−536.9
exp H'	∼	light presence+1|site+1|pitfall ID+1|date	1086.5	−537.2	0.760	3	0.383
exp H'	∼	1|site+1|pitfall ID+1|date	1085.2	−537.6
exp H'	∼	date+1|site+1|pitfall ID	1076.0	−530.0	36.193	4	<0.0001***
exp H'	∼	1|site+1|pitfall ID	1104.2	−548.1
LCBD	∼	light colour+1|pitfall ID	932.8	−460.4	0.434	3	0.933
LCBD	∼	1|pitfall ID	927.2	−460.6
LCBD	∼	light presence+1|pitfall ID	929.2	−460.6	0.073	1	0.787
LCBD	∼	1|pitfall ID	927.2	−460.6
LCBD	∼	date+1|pitfall ID	926.7	−456.4	8.511	4	0.075
LCBD	∼	1|pitfall ID	927.2	−460.6

Comparable to the species diversity indices, we did not find an interaction between light colour and lights on/off for community composition. There was also no main effect of lights on/off and therefore data collected under dark (lights off) and illuminated (lights on) nights was treated equally in all statistical models. In the best model fitted, with light colour as a fixed effect, pitfall ID nested in transect and stratified by date and site, community composition was significantly explained by light treatment (table 3). In pairwise comparisons, differences in community composition between dark control on the one hand and either white and green light treatment on the other hand were significant, with most significant differences between dark and white, and dark and green light treatment. Although light colour treatment had a significant overall effect, this effect origins from the difference between dark, and white and green light as no statistical difference between any other pair of light treatment was detected (table 3; figure 2).

Table 3.

Permutation models (MANOVA, 999 permutations) on the effect of light colour treatment on the Bray–Curtis dissimilarity matrix of the arthropod communities, and post hoc pairwise comparisons on the light colour combinations for the model with the best goodness of fit. (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.)

community composition (Bray–Curtis distance matrix) models		AIC	F	d.f.	p
light colour * lights on/off+pitfall ID | transect, strata=date+site		−384.9959
	light colour*lights on/off		1.1422	3	0.263
	light colour		1.4905	3	0.030*
	lights on/off		0.8774	1	0.001***
	pitfall ID		2.5153	60	0.001***
light colour+pitfall ID | transect, strata=date+site		−387.1632
	light colour		1.4881	3	0.035*
	pitfall ID		2.3828	60	0.001***
post hoc test (light colour)	white versus dark		2.3780	1	0.016*
(light colour)	white versus red		0.5146	1	0.857
(light colour)	white versus green		0.5509	1	0.847
(light colour)	dark versus red		1.8044	1	0.074
(light colour)	dark versus green		2.6577	1	0.006**
(light colour)	red versus green		1.1068	1	0.332
analysis of a separate dataset with samples collected at dawn and dusk at two sites
dawn/dusk+pitfall ID | transect, strata=date+site		−48.93079
	day/night		8.7910	1	0.180
	pitfall ID		1.3369	15	0.180
light colour+pitfall ID | transect, strata=date+site		−36.73252
	light colour		0.9701	3	0.042*
	pitfall ID		0.9228	13	0.458

Figure 2.

Community composition differences by proportion of the most abundant taxa (all taxa that contribute more than 1% to the total number of individuals caught) under the experimental illumination (colours combined) and dark control conditions. (Online version in colour.)

Because of model singularity, we could not test whether the collection of pitfall content affected the exponent of Shannon diversity or LCBD in the data collected at dawn and dusk, separately at sites 1 and 3. However, the effect of emptying the pitfalls as dawn or dusk on the community composition could be tested and was not significant (table 3). As in the main dataset collected at all sites, the effect of light colour treatment was, however, significant.

4. Discussion

Our study shows that years of continued nightly experimental exposure to artificial light (ALAN) in previously natural habitat significantly influences the local ground-dwelling invertebrate community composition but has limited impact on species or beta diversity. In our experiment, the presence of light was a significant factor affecting local community composition, with pairwise differences between dark control and white light, and dark control and green light.

Anthropogenic light at night has been recognized as a biodiversity threat [31,32], and it is therefore important to assess the effect of local species diversity and how this contributes to overall diversity. There are relatively few studies exploring light-related impacts on biodiversity for invertebrates, but the effect of anthropogenic light at night on insect populations is well recognized [16]. Here, we did however not observe significant changes in diversity and LCBD, and we can only speculate why. First, effects on diversity in invertebrates at our sites may only become detectable after periods of longer than 4 years of experimental light exposure. Second, although the mobility of many ground-dwelling invertebrate species is relatively limited, illumination of larger surfaces (such as found in more urban areas) may have a greater effect on their diversity and redistribution—how light affects species spatially at different scales is an equally important but open question [33,34].

The significant community effect found in the PERMANOVA tests suggests that the exponent of the Shannon diversity and LCBD of invertebrates were probably preserved because while some taxa left, others arrived. Despite the retention of diversity, such changes are highly relevant as ecological networks define species interactions, movement, trophic interactions, evolution and ecosystem stability and thus disruption is likely to have consequences [35]. Community responses are complex, can be caused by many different pathways and may depend on a myriad of species and species interactions [36]. Studies on the community impact of light at night are relatively sparse, but effects have been for example been reported in riparian habitat for primary producers [37] and invertebrates [38], invertebrate communities in grassland, roadside vegetation, and marine environment [17,18,39,40], and moths in urban environment [41]. These studies show for example changes in predators and scavengers and that these changes are often highly taxon specific. Likewise, effects vary between airborne and ground-dwelling invertebrates. Less information is available on how light spectrum impacts community composition, but effects have been shown for plants [42] and moths [41]. Effects of light spectrum on community composition were expected as these have shown to differentially alter trophic interactions between many species [43–46]. Typically, the impact of light spectra on, for example, nocturnal insect and mammal species, is relatively short wavelength (blue light) biased [19,20,43], but effects between different spectra on community composition remain less clear. Although light treatment was a significant factor driving community composition, we could only identify significant effects between the dark control and other light treatments. Lastly, switching the light off for a single night in between the illuminated nights had an immediate effect on the community composition. This effect is probably owing to direct changes in the activity of ground-dwelling arthropods.

The effects of long-term exposure to light with different spectra observed here indicate that light colour may not be a solution to mitigate effects on community composition. Leaving lights off had immediate significant effects, which indicates a direct response of species. This does not necessarily mean a change in community composition such that it becomes comparable again to dark control; in our data, the light colour treatment had no significant effect on community composition in a subset of data collected after dark nights (p = 0.146). The absence of community composition effects may however be caused by the small data subset. How persistent changes in community composition are on the food web remains largely unknown, and hence future studies on the temporal aspects of the impact of light at night on community composition are needed [47]. Automated set-ups for the assessment of invertebrate activity can further be used to disentangle more direct effects on activity and actual presence of species. A better understanding of effects of light colour on invertebrate species diversity and community composition continues to be important to reduce impact of light at night.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

All data, statistical analysis code and metadata are deposited from the Dataverse repository: https://doi.org/10.34894/YVN4YJ and are available on request.

Data are also provided in the electronic supplementary material [48].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

K.S.: conceptualization, formal analysis, investigation, methodology, project administration, writing—original draft, writing—review and editing; S.T.: formal analysis, writing—review and editing; M.C.: data curation, writing—review and editing; Z.M.H.: data curation, investigation, writing—review and editing; M.E.V.: conceptualization, methodology, writing—review and editing; T.M.J.: conceptualization, methodology, writing—review and editing; G.R.H.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This publication is part of the project Light on Nature (11110 and 11110 V) of the research programme Applied and Engineering Sciences (AES) which is financed by the Dutch Research Council (NWO). The project is supported by Signify. G.R.H. was supported by an Australian Research Council grant (grant no. DP150101191) to T.M.J. and M.E.V.