How nighttime artificial light affects natural communities of plants and animals

A vast area of the earth is exposed to artificial light at night (ALAN) (Bennie, Davies, Duffy, Inger, & Gaston,; Falchi et al.,; Gaston, Duffy, Gaston, Bennie, & Davies,). Such light pollution arises from streetlights and a diversity of other sources typically associated with urban infrastructure. It takes two broad forms, direct light emissions that are typically experienced in close proximity to one or more lamps (e.g., in the vicinity of a streetlight) and skyglow that results from upwardly directed or reflected light emissions being scattered in the atmosphere and creating the familiar artificial brightening of the night sky. Direct light emissions are more intense than skyglow by two or more orders of magnitude (in terms of lux), but narrower in spatial extent by at least an order of magnitude because of the horizontal propagation of skyglow. Virtually all of the major terrestrial ecosystem types experience direct emissions (Bennie, Duffy, Davies, Correa‐Cano, & Gaston,) and thus also skyglow, with the extent of both forms of light pollution growing rapidly, including in previously less economically developed regions of the world (Bennie, Duffy et al.,; Kyba et al.,). 2014 2016 2014 2015 2015 2017

Although concern about the potential ecological impacts of ALAN has long been expressed (e.g. Cathey & Campbell,; Matzke,; Schroeder,; Verheijen,), fuelled by improved understanding of its regional and global extent, recent years have seen a dramatic increase in research to gain an understanding of these impacts (Davies & Smyth,; Gaston, Visser, & Hölker,). Foremost, this has focused on determining the physiological, behavioral and ecological responses of a number of individual species. This has revealed a range of different pathways by which impacts can occur, particularly through the effects of ALAN on light (and darkness) as a resource (e.g., for photosynthesis, for partitioning of organismal activity between day and night, for dark repair and recovery processes) and on light as a source of information to organisms (e.g., for circadian clocks and photoperiodism, for visual perception, for spatial orientation; Davies, Bennie, Inger, de Ibarra, & Gaston,; Gaston et al.,; Gaston, Davies, Nedelec, & Holt,). ALAN can affect a very wide diversity of species. These are already known to include microbes (Hölker et al.,), plants (Bennie, Davies, Cruse, & Gaston,; ffrench‐Constant et al.,), crustaceans (e.g., Moore, Pierce, Walsh, Kvalvik, & Lim,), insects (e.g., Altermatt & Ebert,; Bird & Parker,; Pacheco‐Tucuch, Ramirez‐Sierra, Gourbière, & Dumonteil,; van Geffen, van Grunsven, van Ruijven, Berendse, & Veenendaal,), spiders (Frank,; Heiling,), fish (Brüning, Hölker, Franke, Preuer, & Kloas,; Riley, Bendall, Ives, Edmonds, & Maxwell,), amphibians (Buchanan,), reptiles (e.g., Thums et al.,), birds (e.g., de Jong et al.,; Dwyer, Bearhop, Campbell, & Bryant,; Rodríguez et al.,), and mammals (e.g., Le Tallec, Perret, & Théry,; Lewanzik & Voigt,; Robert, Lesku, Partecke, & Chambers,). 1975 1936 1945 1960 2017 2015 2013 2015 2017 2015 2016 2016 1998 2016 2014 2012 2014 2009 1999 2015 2012 1993 2016 2015 2013 2017 2013 2014 2015

By contrast with the attention paid to the impacts on individual species, there remains rather limited understanding of how whole ecological communities respond to ALAN. Indeed, a key question must be whether community responses are likely to be predictable from single species responses or if a complementary approach is required. Similar discussions have advanced understanding of the impacts of other anthropogenic environmental impacts, such as climate change (Walther,) and chemical pollution (Maltby et al.,). 2010 2017

Often a mechanistic understanding of environmental impacts can be gained from model organisms, an approach that is widely used in biological science (e.g., Hedges,; Scholz et al.,). These individual responses can help to predict the impact on a wider range of taxa. However, while very useful to predict other species responses this approach misses an important building block of ecological communities: species interactions. Species in ecological communities form complex networks of direct and indirect interactions and these interactions drive ecological and evolutionary processes in ecosystems. As these networks of interactions are important to predict ecosystem stability and functioning (e.g., Thébault & Fontaine,), a community, rather than single species, approach is necessary for getting a predictive understanding of an overall ALAN impact. Considering the importance of these interaction networks, it has also been suggested to use this information about their structure for biological conservation (Tylianakis, Laliberté, Nielsen, & Bascompte,). Further, biodiversity–ecosystem functioning research has demonstrated a positive relationship between a range of ecosystem functions and species richness (Maestre et al.,; Scherber et al.,), something that would fail to emerge from single species studies. However, so far studies that look at the impact of ALAN on whole ecological communities are scarce (but see Bennie, Davies, Cruse, Inger, & Gaston,; Davies et al.,; Hölker et al.,; Knop et al.,; Lewanzik & Voigt,), and all of them are limited to certain target taxa or a specialized or particular kind of interaction, rather than looking at the whole community across taxa and functional groups. To get a full understanding of how ALAN impacts ecosystem functions and stability needs a community and ecosystem level approach. 2002 2008 2010 2010 2012 2010 2015 2015 2015 2017 2014

In this paper, we summarize current knowledge about community responses to ALAN and illustrate different pathways that can lead to these responses and their impact on ecosystem functioning and stability. Further, we suggest different approaches as to how best to advance research in this area to meet the outlined challenges.

As outlined above, we think that a focus on whole communities by looking at interaction networks and trait distributions will greatly improve understanding of ALAN effects on ecosystem functions (Figure 1). First steps in this direction have already been taken (see Bennie, Davies et al., 2015; Davies et al., 2015; Knop et al., 2017), but we need studies that cover different habitats and interfaces between habitats, that acknowledge ALAN as a diverse impact (see below), and that specifically measure functions and stability and how changes in communities bring these impacts about. We outline five different areas that we think will lead to a better understanding of how ALAN affects ecosystem functions. Empirical research using a combination of observation and experiments needs to target major players for ecosystem functions and measure the impact of ALAN on these functions. A focus should be on community responses in terms of changes in food web or network structure and/or distribution of traits. These changes also need to be linked to the wider community (e.g., nocturnal–diurnal components, as done for pollinators in Knop et al., 2017) to uncover knock on effects.A change in community structure observed under ALAN is potentially the result of one or both of two processes being influenced: community assembly and the structure of naïve communities (Figure 1). Experiments are an excellent tool to uncover the relative importance of these two processes leading to a realized community structure under ALAN. This can be done by looking at the assembly of newly created habitats vs. the naïve communities exposed to ALAN and communities in open vs. closed (caged) plots with the latter preventing immigration of individuals.Considering the large scale impact of ALAN we can expect some adaptations to this environmental pressure, such as a reduction in flight‐to‐light behavior in moths (Altermatt & Ebert, 2016), and this should have consequences for ecosystem functions and might reduce the negative impact of ALAN on these (such as reduction in pollination) to at least some degree. This needs a mechanistic approach by uncovering adaptations in species from ALAN exposed communities and the demonstration that these adapted species perform differently in lit habitats with improved ecosystem functions as compared to naïve exposed communities.It has become obvious that artificial light at night is a diverse impact with a range of different light intensities and spectra that differ in timings and can interact with different day length regimes and seasons. The significance of this variation has so far been little explored (but see Davies et al., 2017; Day, Baker, Schofield, Mathews, & Gaston, 2015; de Jong et al., 2016; Spoelstra et al., 2017). The distribution of ALAN from global to very patchy and small scale impact which also included other components such as variation in intensity and spectrum needs to be included in study designs. Particularly, we expect the impact of skyglow to be very different as compared to direct light emissions, such as in the immediate vicinity of a streetlight. While the latter will very much shift the distribution of organisms by movement (attraction and avoidance), the impact of the former happens on a much larger scale and there is no escape. This has to be targeted by different approaches in study design, for example, so that there is no avoidance possible for the organisms under skyglow or large scale ALAN impact—this needs to be reflected in studies by either caged communities or creating large scale light treatments. This impact will vary with mobility of the organisms (e.g., flying vs. crawling).

It would further be very interesting to compare ALAN impacts in temperate to tropical areas. One might argue that a response should be stronger in temperate areas, because organisms are known to respond to variation in daylength as this is a crucial trigger for certain stages of their lifecycle. But, evidence shows that tropical species are able to detect very small seasonal variations in day length (e.g., Hau, Wikelski, & Wingfield,; Rivera et al.,), and therefore it is possible that they are very sensitive to ALAN impact. 1998 2002 5 Finally, we need research that combines different human impacts to test for interactive effects (e.g. disentangle impact of urbanization) (Gaston et al.,). This will lead to a better understanding of how mitigation strategies that can potentially more easily target reduction in ALAN (Gaston, Davies, Bennie, & Hopkins,) will remove pressure from ecological communities. There have so far only been a few studies, for example, by combining ALAN and temperature increase to examine predator–prey interactions with implications for biological control under climate change scenarios (Miller et al.,), and combining light and sound to examine parasitism in frogs (McMahon, Rohr, & Bernal,) and singing behavior of robins (Fuller, Warren, & Gaston,). 2014 2012 2017 2017 2007