Ecological effects of artificial light at night on wild plants

Plants use light as a source of both energy and information. Plant physiological responses to light, and interactions between plants and animals (such as herbivory and pollination), have evolved under a more or less stable regime of 24‐h cycles of light and darkness, and, outside of the tropics, seasonal variation in day length. The rapid spread of outdoor electric lighting across the globe over the past century has caused an unprecedented disruption to these natural light cycles. Artificial light is widespread in the environment, varying in intensity by several orders of magnitude from faint skyglow reflected from distant cities to direct illumination of urban and suburban vegetation. In many cases, artificial light in the night‐time environment is sufficiently bright to induce a physiological response in plants, affecting their phenology, growth form and resource allocation. The physiology, behaviour and ecology of herbivores and pollinators are also likely to be impacted by artificial light. Thus, understanding the ecological consequences of artificial light at night is critical to determine the full impact of human activity on ecosystems. Synthesis. Understanding the impacts of artificial night‐time light on wild plants and natural vegetation requires linking the knowledge gained from over a century of experimental research on the impacts of light on plants in the laboratory and glasshouse with knowledge of the intensity, spatial distribution, spectral composition and timing of light in the night‐time environment. To understand fully the extent of these impacts requires conceptual models that can (i) characterize the highly heterogeneous nature of the night‐time light environment at a scale relevant to plant physiology; and (ii) scale physiological responses to predict impacts at the level of the whole plant, population, community and ecosystem.


Introduction
Light plays a central role in the physiology and ecology of plants. Plants use light both as a resource, via photosynthesis, and as a source of information. The timing, intensity and spectral composition of natural cycles of light provide cues for regulating circadian rhythms, seasonal phenology and the expression of phenotypic variation, including growth form and resource allocation. These cycles have provided a strong selective pressure for millennia. Since the early 20th century, the widespread use of outdoor electric lighting associated with human settlements, industry and transport networks has altered these natural daily and seasonal cycles of light across much of the globe (Cinzano, Falchi & Elvidge 2001), not just in urban environments but also in natural ecosystems (Bennie et al. 2015a;Gaston, Duffy & Bennie 2015). Meanwhile, there is an increasing recognition of the value of frequently illuminated habitats such as roadside verges and hedgerows (Tikka, H€ ogmander & Koski 2001;Hovd & Skogen 2005;Cousins 2006;Le Viol et al. 2008;Hanley & Wilkins 2015), domestic gardens (Davies et al. 2009;Goddard, Dougill & Benton 2010) and urban ecosystems (Angold et al. 2006) for plant and pollinator biodiversity and maintenance of ecosystem services, including human health and well-being (Gaston, Avila-Jim enez & Edmondson 2013). It has been estimated that 238 000 hectares of road verge alone exist in Britain, more than twice the area of natural or semi-natural grassland in the wider countryside (Plantlife 2013). These habitats are often directly illuminated by artificial light throughout the night, or intermittently lit by vehicle headlights. However, even reviews of the ecological effects of roads seldom give much attention to artificial light at night (Spellerberg 1998;Trombulak & Frissell 2000;Coffin 2007; although see Blackwell, DeVault & Seamans 2015). Many other natural or seminatural habitats may be directly lit from artificial light sources at their margins when they are adjacent to human settlements, industry or roads. Even remote areas may be affected as measureable levels of artificial light reflected and scattered in the atmosphere ('skyglow') are detectable tens to hundreds of kilometres from urban centres, particularly under overcast conditions (Kyba et al. 2015). Cinzano, Falchi & Elvidge (2001) concluded that 'mankind is proceeding to envelop itself in a luminous fog', after calculating that some 18.7% of the World's terrestrial surface (excluding Antarctica) could be considered as experiencing light pollution. This figure is far higher for some countries -84.7% of the United Kingdom, for example, or 100% of the Netherlands.
Some measureable degree of artificial light at night is clearly the norm, rather than an exception, across the land surface of many developed countries, but such figures mask a high degree of spatial heterogeneity; while the scattered light from a city may illuminate natural vegetation to levels approaching that of moonlight, localized direct illumination from vehicle headlights or roadside street lighting will be several orders of magnitude brighter. Many natural or semi-natural habitats thus experience artificial light at night at levels congruent with those that cause physiological effects on plants. There is an extensive literature on the effects of artificial manipulations of light conditions on plants, dating from the 17th and 18th centuries (Hunt 1854;Darwin 1881), and it is well-established that light at night, even of short duration or at low intensities, can have marked physiological effects (Smith 1982). Surprisingly, despite a growing awareness of the impacts that disruption of natural light cycles may have on animals (Longcore & Rich 2004;Gaston & Bennie 2014) and plant-animal interactions (MacGregor et al. 2014), much less work has been published on the unintended ecological impacts of artificial light at night on wild plants and natural vegetation (although see Briggs 2006 for a review of potential physiological mechanisms). In this study, we argue that there is a pressing need for plant ecologists to define ecologically meaningful measures of artificial light in the natural environment and to develop understanding of the thresholds and dose-response relationships of light-sensitive processes in plants, particularly at low light intensities. Furthermore, we must improve understanding of the effects of low-intensity light at night in terms of ecological, as well as physiological, processes including competition, herbivory, pollination, reproduction and dispersal.
How much artificial light is there in the environment?
Understanding the ecological impacts of artificial light requires information on the intensity, spatial pattern, spectral distribution, duration and timing of the artificial light to which wild plants are exposed. This is not straightforward, as any comparison of light levels in the environment is somewhat complicated by high degrees of temporal and spatial heterogeneity, and the wide range of spectral power distributions of lighting systems in use . Furthermore, there is no single measure of light intensity that is suitable for quantifying the varying physiological effects of artificial light on plants. In physical studies, the intensity of solar energy on a surface (solar irradiance) is typically measured as an energy flux density with SI units of W m À2 . These radiometric units, which measure power from the electromagnetic spectrum (including light), are distinct from photometric units, which describe the measurement of light in terms of its perceived brightness. Hence, the visible light flux on a surface is described in terms of illuminance, rather than radiance, as not all wavelengths are visible to the human eye. In order to convert measures of irradiance to illuminance, the energy flux must be weighted by a luminosity function, which describes the sensitivity of human visual perception to different wavelengths of light between 400 and 700 nm (Sharpe et al. 2005). The SI unit of illuminance is lux (lx), equal to one lumen (lm) per metre squared, where a lumen is the unit of luminous flux, which is the amount of visible light emitted by a source. Figure 1 shows a range of illuminances measured in vegetation the UK. At the upper end of the scale, the leaves of trees adjacent to street lights or in the beam of car headlights may be exposed to thousands of lux while typical stable ground level illuminances of roadside vegetation (directly beneath street lights) are around 50 lux. Skyglow within cities has been recorded at levels of 0.1-0.5 lux (Eisenbeis 2006), and decreases with distance from central urban areas, with detectable levels up to 10s to 100s of km (Biggs et al. 2012).
Undesirable light at night, or light pollution, is often measured in other units. To measure the brightness of the night sky (including skyglow), astronomers often use the logarithmic magnitude scale, which was developed to compare the apparent brightness of stars; for example, Sirius, the brightest star in the sky, has magnitude À1.5 which is equivalent to an illuminance of 10 À5 lx. When used to describe the brightness of the sky due to light pollution, measurements are typically taken vertically for a small portion of the sky (Davies et al. 2013;Kyba et al. 2015) and expressed as brightness per solid angle of the sky dome (magnitude per square arcsecond). A direct conversion into a measure of illuminance is problematic, however, since skyglow is strongly anisotropic, usually much brighter towards the horizon in the vicinity of towns and cities.
While radiometric measurements are essential for calculations of the energy balance of a surface, and photometric measures are useful for assessing the relative brightness of light to the human eye, in isolation neither is an ideal measure for assessing the ecological impact of light, particularly where that light derives from sources with a range of different spectral signatures. In this study, we use photometric units such as lux (lx) as a standard unit of comparison for light measurementsthis has the advantage of allowing a direct comparison to levels of illuminance required for human activity. However, we stress that illuminance is only useful as a proxy for the biological effects of light when information on the spectral power distribution of light, and the action spectra of the response, is also available. In ecological studies, the physiological pathway, and hence the spectral sensitivity of the photoreceptors involved, will often be unknown. Wherever possible, sample spectra and/or descriptions of the type of light source should be provided.
Correlated colour temperature (CCT) is frequently used as a description of spectral power distribution for vision-lighting systems. The CCT of a light source is the temperature (K) of an ideal blackbody radiator that radiates light of a comparable hue to the source. As a description of the spectral power distribution, and hence the biological effects, of a light source CCT is only meaningful when the source approximates a blackbody, as is the case with the sun, flames or incandescent bulbs. For other light sources, CCT gives an approximation of the aesthetic appearance of white light (from 'cool' blue lights to 'warm' orange), but is a poor indicator of the spectrally dependent response of photoreceptors.
A final factor to consider when measuring light in ecological studies is the orientation, field of view and angular response of the sensor, compared to the geometry of lightsensitive plant organs. Typically, light sensors have a field of view approaching 180°and are cosine corrected, so that the measured flux approximates that on a planar surface; commonly they will be mounted in situ horizontally, effectively measuring the flux on a horizontal plane. In this case, light sources positioned horizontally from the sensor, for example at or near the horizon, will make a small contribution to the measured flux. This is appropriate for a flat surface such as a leaf parallel to the sensor plane. However, interception of light by the organs of a plant may occur at a range of angles, including horizontally. Depending on the nature of the process, incident light fluxes on a horizontal plane, light on a vertical plane or integrated flux on the surface of a sphere may be more appropriate.

How do plants sense light?
Since the 19th century, it has been known that a plant's response to light can be sensitive to wavelength. Hunt (1844Hunt ( , 1854 identified that the extreme red portion of the spectrum was associated with flowering and that blue light induced germination of seed, independently of the broad spectral range of light associated with the 'decomposition of carbonic acid' (photosynthesis). During the 20th century, several plant pigments involved in these and other responses were identified, and action spectra for several such photoreceptor pigments have been defined; for example cryptochrome has been shown to respond to light between 390 to 530 nm (violet to blue/ green; Ahmad et al. 2002); phototropins respond primarily to blue light (Christie 2007); and phytochrome primarily to red and/or far-red light, depending on the form (Casal, Candia & Sellaro 2014). In algae, a wide range of pigments spanning  the visible spectrum have been identified (Rockwell et al. 2014).
The best known action spectrum for plants is that of photosynthesis; photosynthetically active radiation (PAR), usually defined as light between 400 and 700 nm, has an action spectrum defined by the absorbance spectra of chlorophyll and carotenoids. Because photosynthesis is a quantum process, the flux of photons, rather than energy is appropriate, and it is usually quantified as a photosynthetic photon flux density (PPFD) with units of lmol photons m À2 s À1 . As short-wavelength photons carry more energy than those at longer wavelengths, and both the sensitivity of human vision and that of photosynthetic pigments vary considerably in their absorption at different wavelengths, the relationship between irradiance, illuminance and PPFD is strongly sensitive to the spectral distribution of the light source. The effect of artificial light at night on net photosynthesis in the environment is limited due to the low quantum flux densities associated with outdoor lighting when compared to daylight. While theoretically urban skyglow may be sufficient to induce a small photosynthetic response (Raven & Cockell 2006), in practice measureable effects on carbon fixation are likely to be limited to situations where leaves are in very close proximity to light sources (such as the canopies of trees around street lights), or when artificial lighting is introduced into naturally dark situations such as cave systems. In the latter case, the installation of lighting in caves for tourism is often sufficient to support Lampenflora, communities of algae, bryophytes and vascular plants solely reliant on electric lighting as an energy source (Lef evre 1974;Johnson 1979) .
In addition to the use of light as an energy source for photosynthesis, plants utilize a suite of other photoreceptors in order to sense information about their environment, the time of day and season of the year. The photosynthetic system itself is sensitive to light at night, providing a set of secondary pathways through which artificial light could influence carbon fixation. Poulin et al. (2014) showed that low levels of light from a high-pressure sodium street light, at an illuminance equivalent to that observed at the shore of an urban lakeside (6.6 lux), significantly changed several aspects of the photobiology of phytoplankton, including decreasing the intracellular chlorophyll a concentration and the number of Rubisco molecules per cell. In higher plants, light quality, even at low fluence rates, is known to affect physical characteristics of the photosystem, such as leaf stomatal density, as well as the opening of stomata (Smith 1982).
Many photoreceptors have been identified, of which three main groups have been characterizedphytochromes, cryptochromes and phototropins. Phytochromes, of which phyA and phyB play the dominant roles, are present in plants in two forms, P r and P fr . The P r form absorbs light following an absorbtion spectrum with a peak sensitivity at a wavelength of around 650-670 nm, in the red portion of the spectrum; on absorbing light, a molecule in the P r form converts into the P fr form. The P fr form has a different absorbtion spectrum, lacking such a distinct red peak but with another peak in the far-red, at 705-740 nm. On subsequently absorbing light (or in darkness), the P fr form is converted back to P r . Each phytochrome molecule thus acts as a 'switch', flipped from one form to another by different wavelengths of light. Any light source can be characterized by a 'phytochrome stationary state' (PSS; Sager & McFarlane 1997), an estimate of the relative proportion of far-red absorbing (P fr ) to total (P r + P fr ) phytochrome within tissues at equilibrium under a stable light source, calculated from the phytochrome cross-sectional area for each form and the spectral irradiance of the light source. Since daylight beneath the shade of a photosynthetic canopy is depleted in red light relative to far-red (Franklin & Whitelam 2005), the PSS is an indication of the degree of shading by other plants that a plant is experiencing. Phytochrome is also used to detect photoperiod and hence plays an important role in the initiation of phenological events such as budburst, flowering and senescence. Flowering plants are often characterized as falling into one of three groupsshort-day plants, in which flowering is initiated, advanced or promoted when a dark night-time period is sufficiently long to allow enough phytochrome in the P fr form to revert to P r ; long-day plants, in which the dark night-time period must be sufficiently short to increase the night-time levels of P r ; and day-neutral plants which have no detectable effect of darkness on their flowering phenology, but rely on other environmental cues. The detection of photoperiod in plants can be very sensitive - Rivera & Borchert (2001) provide evidence that in the tropics, where annual variation in day length is less than one hour, photoperiodism is still used as a seasonal cue for flowering in some tropical trees. Early experimental work showed that even brief exposure to red light at night can be sufficient effectively to interrupt the detection of an unbroken dark period, and prevent flowering in short-day plants, while subsequent exposure to far-red light can reverse the effects of the interruption (Borthwick et al. 1952). Intermittent light of certain wavelengths may thus be effective in shortening the perceived day lengthindeed, cycling pulses of lights for just six minutes every half hour can be as effective as continuous light (Runkle et al. 1998;Blanchard & Runkle 2010). Low levels of lights are often effective in inducing a response. Whitman et al. (1998) found that effects on flowering saturated at levels as low as < 0.05 and 0.4 mol m À2 s À1 PPFD from a broad-spectrum white light source, roughly equivalent to < 3.5-28 lux, and well within levels recorded in roadside vegetation. By contrast, Ishikawa et al. (2009) found a fluencedosage response in rice in which the degree to which flowering was suppressed increased with the light intensity supplied.
Detection of photoperiod in plants is not always, if ever, a simple function of the photoreversible forms of phytochrome, however. In the facultative long-day plant Arabidopsis thaliana, night interruptions from red, far-red or blue light are all effective in inducing flowering (Goto, Kumagai & Koornneef 1991). Cryptochromes, sensitive to light in the UV-A, violet and blue portions of the spectrum, act together with phytochromes to regulate the circadian clock, keeping daily rhythms set to a 24-h cycle of light and darkness. Photoperiodic control of flowering integrates both the circadian clock, and sensing of the length of the dark period; thus, there is an interaction between the detection of daily and seasonal cycles, and between photoreceptor systems. The cryptochrome/photolyase family of photoreceptors also plays critical roles in controlling a wide range of light-induced responses in germination, growth and development, and shade avoidance (Kami et al. 2010), as well as DNA repair (Fortunato et al. 2015). A further blue light-sensitive receptor protein, phototropin, is responsible for phototropism, the growth of plant organs towards a light source. Figure 3 shows the spectral power distribution of five types of lighting frequently used in outdoor street lighting, along with the relationship between PPFD (photosynthesis), the PSS (phytochrome) and the relative amount of blue light between 350 and 500 nm (cryptochrome and phototropins).
The detection of light in plants is complex, often relies on more than one physiological pathway and may have partial redundancy, so that processes induced by one photoreceptor system may also be induced or repressed by another system (Song, Ito & Imazuimi 2010). Furthermore, processes that rely on light-induced responses such as photoperiodism in one species or phenotype may be linked to other environmental cues in others (Basler & K€ orner 2012).

D I R E C T E F F E C T S
With the exception of a handful of original papers and the review of Briggs (2006), surprisingly few studies apply knowledge of the physiological effects of light on plants to examining the effects of exposure to artificial light in the environment. This is particularly surprising as experimental studies have shown that relatively low levels of light (Whitman et al. 1998) or light over a short duration (Runkle et al. 1998) can be effective in influencing the response of plants to photoperiodsuch levels are frequently observed under conditions of night-time lighting in the environment (Figs 1 and  2). Furthermore, the high red to far-red ratio of most common forms of outdoor lighting compared to sunlight make them particularly effective in disrupting detection of photoperiod through the phytochrome pathway, while blue-rich LED lighting, which is rapidly being adopted in many regions, emits strongly in the region of the spectrum to which cryptochrome and phototropins are sensitive (Fig. 3). In the context of light pollution, these observations suggest several key areas for research, namely -How widespread are ecologically significant physiological effects of artificial light on plants in the environment? What is the relative contribution of ambient low-level light (e.g. skyglow), continuous direct illumination (e.g. street lighting) and intermittent pulsed light (e.g. vehicle headlights)? And can ecological impacts be mitigated by selecting light sources with specific spectral characteristics?
Some of the earliest documented effects of artificial light at night on plants in urban environments concerned the retention of leaves on deciduous trees in the vicinity of street lights (Matzke 1936). Certain tree species appear to be more sensitive to direct illumination than others, but it is a common phenomena in several species planted as urban roadside trees. In North America, this phenomenon has been recorded in Populus canadensis, Salix fragilis (Matzke 1936), Populus nigra, Ulmus americana (Schroeder 1945) and Liquidambar styraciflua (Briggs 2006). In both Europe and North America, it is widely observed in Platanus species (Matzke 1936;Schroeder 1945;Briggs 2006;Fig. 4), and in Europe, it has been observed in Aesculus hippocastanum and Betula pendula variation in illuminance measured in roadside vegetation between sunset and sunrise on a rural main road with no fixed lighting. Peaks represent pulses of light from passing vehicles. Typically, light from these sources has a high degree of variability, but can reach much higher magnitudes than those under street lights (see Fig. 1).
(Sybille Schroer, pers. obs.). Artificial light may also have a measureable effect on spring budburst in urban trees; at the University of Exeter's Cornwall Campus, near Falmouth, U.K., budburst occurred up to two weeks earlier on branches of large-leaved lime trees in the vicinity of LED lighting (Fig. 5). In New Zealand, Metrosideros excelsa flowers more profusely next to street lights than between them (Jacqueline Beggs, pers. obs.) While such dramatic effects on budburst, flowering and leaf colouring and abscission are frequently observed in urban trees under direct illumination, very few scientific studies have described this phenomenon, and the effects on tree species that are rarely found in urban situations, and under lower levels of diffuse ambient light and on herbaceous species with less obvious phenological stages are poorly known. While these phenomena have been described predominantly in temperate trees, in seasonal tropical environments, budburst in trees may respond to changes in day length of 30 min or less (Rivera & Borchert 2001) and could be highly sensitive to artificial light at dawn and dusk. Such changes in leaf and flowering phenology may have significant effects on the health, survival and reproduction of plants. Matzke (1936) reported injury from low temperatures in leaves retained on urban trees under street lights. The timing of budburst and leaf fall of deciduous trees determines the exposure to frost damage of photosynthetic tissue (Hanninen 1991;Cannell 1997); it may also determine the exposure to fungal pathogens in spring (Marc ßais, Kavkova & Desprez-Loustau 2009) and to herbivores which time their emergence or activity to the emergence of leaves (Visser & Holleman 2001). The timing of budburst is often determined by interactions between temperature (both spring warmth and winter chilling) and photoperiod. Photoperiod is thought to be critical in late-successional species, which tend to trigger budburst later in the spring than early-successional species, which are triggered predominantly by temperature and insensitive to photoperiod (Basler & K€ orner 2012;K€ orner & Basler 2010). Cathey & Campbell (1975a,b) found that a wide variety of ornamental garden plants were sensitive to artificial light at night, testing a variety of different light types and inducing marked effects on flowering and growth rates at illuminance levels comparable to those at which garden plants are frequently exposed (< 5 lux). Both suppression and induction of  flowering, and enhanced and supressed growth, were found, depending on species, and the response was greatest under light sources with a high proportion of red light and a high red/far-red ratio (such as high-pressure sodium lighting). Effects of light pollution on agricultural crops have also been reported, although little scientific work has addressed the issue; Briggs (2006) reports that bright floodlights from an Ohio prison prevented normal development of Glycine max (soya beans) in nearby fields, and Sinnadurai (1981) reported that Zea mays (maize) crops alongside main roads in Accra, Ghana, grew rapidly but failed to flower after high-pressure sodium lamps were installed. Less obvious effects in wild species may go unnoticedin an experimental system, we have found that Lotus pedunculatus, a species in which flowering is known to be sensitive to day length (Forde & Thomas 1966), produced between 10 and 25% fewer flower heads under night-time artificial light treatments simulating street lighting (Bennie et al. 2015b).

D A R K R E C O V E R Y F R O M S T R E S S
A period of darkness can be crucial for repair and recovery from environmental stresses. Tropospheric ozone is a major global pollutant produced by photochemical reactions of precursors (nitrogen oxides and hydrocarbons) associated with road traffic (Munir, Chen & Ropkins 2012). Tropospheric ozone can have severe effects on vegetation (Ashmore 2005), including foliar injury. Plants at northern latitudes often display more signs of foliar injury, a phenomenon attributed to long day lengths impairing repair and defence processes . Experimental studies have shown that constant artificial light at night at levels reported at approximately 1 lmol m À2 s À1 PPFD (using a fluorescent lamp, equivalent to around 74 lux and within the range of exposure of roadside vegetation under street lighting; Fig. 1) significantly increased foliar injury due to ozone in three clover species ). This process is probably regulated by the phytochrome pathway as it can be induced by short periods of exposure to red light during the night (5-30 min) and reversed by subsequent exposure to far-red light (Eriksen et al. 2012). The combined effects of ozone damage and night-time illumination (including constant illumination from street lights or short bursts and high-intensity light from vehicle headlights) on vegetation near roads have not been studied.

P L A N T -A N I M A L I N T E R A C T I O N S
In addition to direct effects on plants, artificial light at night is likely to influence interactions between plants and animals, potentially in complex ways. For example, light has been shown to affect caterpillars of the moth Mamestra brassicae by decreasing both male caterpillar and pupal mass and reducing the duration of pupation in both sexes (van Geffen et al., 2014), as well as reducing activity and mating in Operpphtera brumata (van Geffen et al. 2015). Winter moths are important herbivores of deciduous trees, causing defoliation in severe outbreaks (Wesołowski & Rowi nski 2006). They typically show strong synchrony in egg hatching with spring budburst in host trees such as Quercus robur, but environmental change may disrupt this synchrony, leading to mismatches in timing between the insect and its host (Visser & Holleman 2001;Van Asch et al. 2007). Such mismatches in phenology are likely to occur when species respond individualistically to temperature and photoperiod. Disruption of seasonal light cues by artificial light, population level changes or aggregation of individuals could modify interactions including herbivory, pollination and seed dispersal. Both top-down and bottom-up effects are possible, although few studies have explicitly looked for evidence of such effects; in an experimental system, artificial light suppressed flowering in the legume Lotus pedunculatus, apparently leading to a subsequent suppression in numbers of the aphid Acyrthosiphon pisum feeding on flowering shoots (Bennie et al. 2015b). Street lighting can cause local aggregations of predatory invertebrates ), but the subsequent effects on herbivorous invertebrates and on plant species are unknown. Many plant species rely on nocturnal or crepuscular pollinators and are adapted to some extent for night-flying flower visitors (Baker 1961). Macgregor et al. (2014) review the potential effects of light pollution on pollination by nocturnal Lepidoptera, a group that are strongly attracted to artificial light (Eisenbeis 2006;van Langevelde et al. 2011;Warrant & Dacke 2011). The effects are potentially complexlight may affect Lepidoptera at the population level through effects on reproduction and mortality and modify their phenology, as well as by modifying behaviour. Where 'flight-to-light' behaviour occurs, and insect pollinators are attracted to light, it is unclear at present to what extent lit vegetation will have a higher concentration of active pollinators, or whether the presence of light will act as an ecological trap or have a disruptive effect, reducing pollination rates (Eisenbeis 2006;Macgregor et al. 2014). In tropical and subtropical regions, bats may be significant pollinators of certain groups of flowering plants and of particular importance in long-distance pollen dispersal (Fleming, Geiselman & Kress 2009), as well as being important in seed dispersal (Fleming & Heithaus 1981). Light can act as a barrier to movement for many species of bats (Stone, Jones & Harris 2009, effectively fragmenting landscapes and potentially acting as a barrier for gene flow or dispersal in plants (Lewanzik & Voigt 2014). Effects on pollination may not be limited to nocturnally pollinated plantsartificial light may disrupt the behaviour and demography of diurnal as well as nocturnal animals, and habitats around roads are increasingly recognized as key habitats for pollinators (Hanley & Wilkins 2015).

Summary
The effects of artificial light at night on wild plants and natural or semi-natural ecosystems are of potential ecological significance world-wide. However, since the review of potential physiological mechanisms by Briggs (2006), surprisingly little research has been published on the impacts of artificial light in the as an environmental pressure for plants. There is relatively little information available concerning how pervasive are the ecological impacts of artificial light at night, the relative contributions of diffuse, direct, constant or intermittent light sources and how any adverse effects may be mitigated. We suggest that plant ecologists should address this gap in knowledge by focusing on four key issues. First, the nighttime light environment of plants in urban, suburban and natural ecosystems needs to be better characterized. Plants are exposed to extremely high degrees of heterogeneity in terms of the intensity, spectral power, timing and duration of light to which they are exposed, which complicates our understanding of how widespread ecological effects may be. Secondly, our understanding of the physiological effects of light need to be expanded into an understanding of the ecological effects, and we should search for population and community-level impacts as well as impacts at the individual level. Thirdly, interactions between direct and indirect effects of artificial light, for example those mediated through herbivores, pollinators, pathogens or competitors, should be explored. Fourth, there is a need to put studies of the effects of artificial light into the context of other stressors such as eutrophication, climate change, invasive species, chemical pollution, habitat fragmentation and invasive species. Artificial light will rarely affect wild plants in isolation, and there is a need not only to understand the relative importance of light pollution among other anthropogenic drivers of change, but also to investigate the combined impact of multiple stressors; roadside vegetation, for example, may act as a corridor for the dispersal of non-native species (Tyser & Worley 1992) and will be exposed to chemical pollution such as tropospheric ozone in addition to being exposed to artificial light. There is considerable potential for making progress in understanding the impacts of this increasing, but poorly understood environmental pressure.  Fig. 5. Leaf out date (recorded when 50% or more of the buds nearest the apex have fully unfolded leaves) of five shoots from each of five mature small-leaved lime Tilia cordata trees adjacent to a footpath lit by white LED lighting on the university of Exeter's Penryn Campus in 2014 (one tree and lighting fixture illustrated on 1st April). Night-time illuminance levels were measured at each shoot on an overcast night at new moon prior to leaf unfolding. Each tree is represented by a different symbol, dashed lines represent regression lines for each individual tree, solid line represents overall regression line (significant at P < 0.01 with tree as a random effect). Photo: James Duffy.

Data accessibility
This manuscript does not use data.