Effects of climate change on Arctic arthropod assemblages and distribution PhD thesis

Rikke Reisner Hansen

Academic advisors: Main supervisor Toke Thomas Høye and co-supervisor Signe Normand

Submitted 29/08/2016

Data sheet

Title: Effects of climate change on Arctic arthropod assemblages and distribution

Author University: Aarhus University

Publisher: Aarhus University – Denmark URL: www.au.dk

Supervisors:

Assessment committee: Arctic arthropods, climate change, community composition, distribution, diversity, life history traits, monitoring, species richness, spatial variation, temporal variation

Date of publication: August 2016

Please cite as: Hansen, R. R. (2016) Effects of climate change on Arctic arthropod assemblages and distribution. PhD thesis, Aarhus University, Denmark, 144 pp.

Keywords:

Number of pages: 144

PREFACE………………………………………………………………………………………..5

LIST OF PAPERS……………………………………………………………………………….6

ACKNOWLEDGEMENTS……………………………………………………………………...7

SUMMARY……………………………………………………………………………………...8

RESUMÉ (Danish summary)…………………………………………………………………....9

SYNOPSIS……………………………………………………………………………………....10 Introduction……………………………………………………………………………………...10 Study sites and approaches……………………………………………………………………...11 Arctic arthropod community composition…………………………………………………….....13 Potential climate change effects on arthropod composition…………………………………….15 Arctic arthropod responses to climate change…………………………………………………..16 Future recommendations and perspectives……………………………………………………...20 References………………………………………………………………………………………..21

PAPER I: High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap……………………………………………………………………………….27

PAPER II: Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities……………………………………………………………………………………..37

PAPER III: Diversity and composition changes along a continentality gradient in the Arctic……………..55

PAPER IV: Synchronous seasonal dynamics in composition, richness and turnover of arthropod assemblages across the Arctic…………………………………………………………………..97

PAPER V: Habitat-specific effects of climate change on a low-mobility Arctic spider species…………...122

PAPER VI: High-Arctic butterflies become smaller with rising temperatures……………………………..132

PAPER VII: Phenology of high-Arctic butterflies and their floral resources: Species-specific responses to climate change………………………………………………………………………………….136

As if some little Arctic flower, Upon the polar hem, Went wandering down the latitudes, Until it puzzled came To continents of summer, To firmaments of sun, To strange, bright crowds of flowers, And birds of foreign tongue! I say, as if this little flower To Eden wandered in – What then? Why, nothing, only Your inference therefrom!

- Emily Dickinson 1860

PREFACE

This thesis is the realization of three years’ work at the Department of Bioscience, Aarhus University, during which I have been stationed mostly at the Arctic Research Center, but also at the section for Biodiversity and conservation, Kalø. The fieldwork, which formed a large part of the thesis, was conducted in Nuuk, South-west Greenland during the summer of 2013.

The thesis was supervised by Toke Thomas Høye with Signe Normand as co-supervisor, and I was also fortunate to be a part of a larger scientific collaboration with researchers from Canada, USA and Norway.

The thesis is comprised of two main parts. An introduction of the general conceptual framework, preceding a summary of the results of the scientific papers, included in the thesis. Part two consists of seven papers. Paper I and II have been published in Polar Biology and PeerJ respectively. Paper III is in preparation for BMC Ecology and paper IV in preparation for Ecography. Paper V – VII have been published in Polar Biology, Biology Letters and Current Zoology, respectively.

In this synopsis, I will first describe the patterns of global climate change and its amplification in the Arctic, which will set the scene for the main subject of my thesis; Overall impacts of climate change at the species level and at the community level. I will present some of the recent research in this field and contrast it to my research on diversity and compositional changes in Arctic arthropods. Life history traits such as phenology, reproduction and body size have been heavily impacted by climate change. I will present some of the results from recent research in this field and relate changes in life history traits to changes in distributional patterns (i.e. species occurrences). The Arctic is a large heterogeneous biome, and some of the results may therefor only be applicable in a narrow geographic context. I have attempted to distinguish between the patterns that can be generalized into an overall Arctic context and the patterns that are unique for Greenland. Finally, I round up the synopsis with a conclusion and finishing comments of future recommendations based on the findings of my PhD.

LIST OF PAPERS

PAPER I: Hansen RR, Hansen OLP, Bowden JJ, Normand S, Bay C, Sørensen JG, and Høye TT. 2016a. High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap. Polar Biology:1-10 DOI: 10.1007/s00300-016-1893-2.

PAPER II: Hansen RR, Hansen OLP, Bowden JJ, Treier UA, Normand S, and Høye TT. 2016b. Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities. Peerj 4:e2224 DOI: 10.7717/peerj.2224.

PAPER III: Hansen OLP, Hansen RR, Bowden JJ, Normand S, Treier U and Høye TT. In prep. Diversity and composition changes along a continentality gradient in the Arctic. BMC Ecology.

PAPER IV: Hansen RR, Asmus A, Bowden JJ, Loboda S, Iler AM, CaraDonna PJ, Hansen OLP, Normand S, Loonen MJJE, Schmidt NM, Aastrup P and Høye TT. In prep. Synchronous seasonal dynamics in composition, richness and turnover of arthropod assemblages across the Arctic. Ecography.

PAPER V: Bowden JJ, Hansen RR, Olsen K, and Høye TT. 2015b. Habitat-specific effects of climate change on a low-mobility Arctic spider species. Polar Biology 38:559-568 DOI: 10.1007/s00300-014-1622-7.

PAPER VI: Bowden JJ, Eskildsen A, Hansen RR, Olsen K, Kurle CM, and Høye TT. 2015a. High-Arctic butterflies become smaller with rising temperatures. Biology Letters 11 DOI: 10.1098/rsbl.2015.0574.

PAPER VII: Høye TT, Eskildsen A, Hansen RR, Bowden JJ, Schmidt NM, and Kissling WD. 2014. Phenology of high-arctic butterflies and their floral resources: Species-specific responses to climate change. Current Zoology 60:243-251

ACKNOWLEDGEMENTS

The past three years work, now culminating with the production of this thesis, has been exciting, but also challenging years and I could not have achieved this without the support of family, friends and colleagues.

Some of the challenges, I had help overcoming with tremendous support from my supervisor Toke Thomas Høye and co-supervisor Signe Normand. Both offered me sound scientific guidance during all phases of my PhD, and helped me to understand all aspects of the publication process. I would particularly like to thank Toke Thomas Høye for enthusiastically believing in my results, and encouraging me to pursue my goals. I am grateful to both for showing me that pursuing a scientific career does not need to come at the expense of family, but that career and family can act in concert and strengthen the outcome in the end.

My colleagues both at Kalø and Arctic Research Center as well as colleagues visiting from abroad, have made the years fly bye, not only by offering scientific sparring at a qualified and interdisciplinary level, but particularly by adding pleasant company to the journey.

I kindly thank the Arctic Research Center for providing funding and housing throughout the full course of my PhD. I would also like to thank the Natural History Museum in Aarhus, who harbored me during the sorting and identification process and provided materials, room and enjoyable company.

I owe very special gratitude to my son Daq Lindhardt Reisner who has endured my absence, short holidays and parental mood swings with a smile and an amazing attitude. This could not have gone as smoothly without the help of my mom Karin Bursche Leth and the Busk- Veje family, who all have provided my son with a second home, in my absence. Finally, I would like to thank my partner Lars Nygaard Nielsen for bearing with me in the final stages and showing me unconditional encouragement and loving support.

Rikke Reisner Hansen August 2016

SUMMARY

With this thesis I provide new insight into how climate change can be expected to affect different aspects of future arthropod diversity (i.e. composition, diversity and life history traits), through the deciphering of past and contemporary distribution patterns. I identify important drivers of local and regional variation in community and population differences of Arctic terrestrial arthropods. The focus is primarily on Greenland, yet the results are discussed in a wider geographical context. We show that Arctic arthropod assemblages varies over very short distances and that patterns in composition and diversity are dictated by local to landscape scale habitat features (Paper I and II). Large geographical features, such as fjords, glaciers and mountain ranges fragments species pools and creates distinct site specific arthropod communities (Paper I, II and III). These findings are in sharp contrast to the conception of the Arctic as a species poor and homogenous environment. While heterogeneity of the habitat sampled and the overall size of the study area are variables shown to be positively correlated with species diversity, less attention has been given to the temporal accumulation of species within and between years. We demonstrate a non-uniform seasonal development of Arctic arthropod community composition throughout the season along with peaking richness patterns during parts of the season, where conditions are optimal. These patterns are repeated across multiple Arctic sites, spanning a large climatic gradient, but with significant inter-annual and habitat variation (Paper IV). Changes in important life history traits (i.e. body size and phenology) are features that will eventually cause species to alter their occurrences, and ultimately whole community compositions. Species specific investigations revealed that body size variation as a response to climate change, may be masked by local habitat heterogeneity and potentially depend on trophic level of the study organism (Paper V and VI). Inter-annual body size variations of a high Arctic crab spider differ significantly between the habitats, as well as, between males and females (Paper V). Body size of high arctic butterfly is significantly decreasing as a response to increasing temperatures over nearly 20 years (Paper VI). Several studies have demonstrated decoupling of ecological events, such as flight time in butterflies and the peak abundance of floral resources. One of our studies (Paper VII), investigating phenological responses to recent warming for two Arctic butterfly species, show a trend towards earlier flight time for one of the species (Boloria Chariclea) and not the other (Colias hecla). The study furthermore demonstrates contrasting species specific relations with abiotic variables and confirms the need for species specific studies of phenological responses.

RESUMÉ

Med denne afhandling, bidrager jeg med viden om hvordan klimaændringer kan forventes at påvirke forskellige aspekter af fremtidig diversitet af leddyr (i.e. artssammensætning, diversitet og livshistorietræk), via en udredning af fortidige og nutidige udbredelsesmønstre. Jeg identificerer de vigtige drivkræfter bag lokal og regional variation i samfunds- og populationsmæssige forskelle i landlevende Arktiske leddyr. Fokus er primært på Grønland men resultaterne diskuteres i en bredere geografisk kontekst. Vi viser at leddyrs samfund varierer over korte afstande og at mønstre i sammensætning og diversitet er dikteret af forskelle i habitater på lokal til landskabs skala (Artikel I og II). Store geografiske barrierer, såsom fjorde, gletsjere og bjergkæder hjælper til at fragmentere artspuljer og skabe lokalt specifikke artssammensætninger (Artikel I, II og III). Disse resultater står i skarp kontrast til den gængse opfattelse af Arktis som et artsfattigt og homogent terræn. Heterogeniteten af et område, samt størrelsen af arealet som indsamles på, har vist sig at være positivt korreleret med artsdiversitet, mindre opmærksomhed er imidlertid blevet tildelt sammenhængen mellem tidsmæssig akkumulering af arter indenfor og årerne imellem. Vi viser at artsammensætninger af Arktiske leddyr ikke udvikler sig ens gennem sæsonen, men at der er perioder hvor artsrigdom topper og hvor der sker få ændringer i artssammensætning fra en uge til næste. Disse mønstre gentager sig over flere lokaliteter i Arktis, som tilsammen spænder over store klimatiske gradienter (Artikel IV). Ændringer i vigtige livshistorietræk (e.g. kropsstørrelser of fænologi) gør at arter ændrer forekomster og vil i sidste ende ændre artssammensætninger for hele leddyrssamfund. Artsspecifikke undersøgelser viser at variation i kropsstørrelser muligvis maskeres af forskelle i habitater og kan afhænge af trofisk niveau (Artikel V og VI). År til år variation i kropsstørrelser hos en høj-Arktisk krabbe edderkop er signifikant forskellige imellem habitater såvel som mellem hanner og hunner (Artikel V). Størrelsen af en Arktisk sommerfugl er blevet signifikant mindre gennem de seneste næsten 20 år, som et respons på stigende temperature (Paper VI). En del studier har påvist en afkobling mellem økologiske events, såsom flyvetid af sommerfugle og optimal tilgængelighed af deres føderessourcer. Et af vores studier (Artikel VII), som undersøger fænologiske responser til seneste tids opvarmning hos to arter af Arktiske sommerfugle, demonstrerer tendens til tidligere flyvetid hos den ene art (Boloria chariclea) og ikke den anden (Colias hecla). Studiet viser derudover nogle kontrasterende artsspecifikke sammenhæng med abiotiske variable og understreger behovet for artsspecifikke studier af fænologi.

SYNOPSIS

Introduction

Documenting the spatial and temporal distribution of species is essential to grasp the consequences a changing climate has for biodiversity (Heller & Zavaleta 2009). Arthropods as well as plants are central for many food chains and make up the vast majority of global and Arctic biodiversity (Hodkinson 2013). With arthropods being sensitive to abiotic conditions (Danks 2004) and while they provide an array of ecosystem services, a deeper knowledge of their ecology, diversity and distribution is crucial to understand how to conserve Arctic biodiversity as a whole. However, basic knowledge regarding all aspects of Arctic arthropods is still scant and lacking far behind that of plants and vertebrates.

The Arctic is often referred to as a homogenous and biodiversity low region (Hodkinson 2013), yet it is home to more than 2200 species of insects and spiders (Danks 1981), approximately 800 of which live in Greenland (Böcher et al. 2015). Greenland covers an area of 2.166.086 km2, with an area of 410.449 km2 free of ice, and is divided into high- low- and sub- Arctic regions (Greenland Statbank 2015). The North American Arctic spans an area nearly four times the size of the ice-free Greenland and houses, in comparison, approximately 1650 terrestrial arthropod species (Danks 1991). Within Greenland, the inland icecap, fjord and topographical features provide large barriers for species to disperse (Normand et al. 2013; Hansen et al. 2016a). A large difference in species composition and diversity is therefore found between the latitudinal and longitudinal gradients. These gradients and the large variation in e.g. topography and environmental variables, renders Greenland a very interesting geographic focal point for studying distribution patterns and the effects of climate change.

Global warming is unequivocal and during the 20th century the globally averaged combined land and ocean surface temperature has increased by approximately 0.85. Most of this over the course of the past couple of decades (Ipcc 2014b). Climate change across the globe is however, not uniform due to polar amplification and feedback mechanisms (figure 1). Polar amplification refers to the phenomenon that warming or cooling trends are strongest over high latitudes, particularly over the Arctic. This is due to poleward atmospheric heat transfer, oceanic oscillation dynamics (Manabe 1969) and a number of positive feedback mechanisms, including diminishing snow and sea ice cover (Screen & Simmonds 2010), as well as permafrost melt (Kellogg 1983). Melting of the permafrost forms part of a feedback mechanism, contributing to

the release of large quantities of methane encapsulated in the frozen soils as massive carbon stocks. Permafrost dynamics are not directly being addressed in this thesis, but play an indirect role in the altering of tundra habitats (Woo & Young 2006; Myers-Smith et al. 2011; Perreault et al. 2015). Due to the complex and heterogeneous patterns of the climate system, there is vast geographic variation even within the Arctic. Predicting climate change patterns at a scale relevant to organisms with narrow range distributions, like arthropods, therefore requires detailed levels of ground-truthing. I intend to show how the balance between detailed investigations and broader scope approaches can supplement each other and add to pan Arctic generalizations about climate change effects on arthropod community composition.

Figure 1: Change in average surface temperature (a) and change in average precipitation (b) based on multi-model mean projections for 2081–2100 relative to 1986–2005 under two different scenarios. Least case (left) and worst case (right) (Ipcc 2014c). Study sites and approaches

Inferring contemporary patterns in species composition onto future prediction of distributional changes, can partly be achieved through time for space substitution (Pickett 1989). The large variation in environmental variables within Greenland makes for an ideal region for these inferences. Even though the results of the studies in this thesis all pertains to Greenland as the geographic focus of the research, some of the mechanisms responsible for the patterns that I present in this thesis, can be generalized to the Arctic as a whole. For instance, the Greenlandic ice cap with its glacier arms forms a dispersal barrier for species with limited dispersal capacity (i.e. arthropods and plants) (Normand et al. 2013; Hansen et al. 2016a) and affects distributional

patterns. However, general dispersal barriers are not unique for Greenland as large parts of the Arctic is comprised of large water bodies, glaciers, mountain ranges, all limiting dispersal and area of suitable habitats for species (CAFF 2013). Hydrology and vegetation changes, as well as, increasing temperatures are phenomena occurring across the Arctic, yet, the direction and magnitude is not uniform (Woo & Young 2006; Ipcc 2014a; Myers-Smith et al. 2015; Wrona et al. 2016). The Godthåbsfjord area, South-west Greenland covers a large variety in climatic and topographic gradients and serves as a model area to study responses to these variations. It stretches from the coast to the inland ice sheet and branches out, covering a large heterogenous area (Figure 2). The mouth of the fjord is characterized by a coastal climate with relatively high precipitation and narrow annual temperature range. As we move further in the fjord, towards a more continental climate, the gap between winter and summer temperatures widens and humidity decreases. Apart from the more direct influences of climate variability within the fjord system, the fjord also represents a large gradient in habitat variability. This generates a large variety in living conditions for a large suite of organisms. Topographic variation occurs both along the fjord systems, but also within each location and serves as proxy temperature measures.

Figure 2: The two main study sites in Greenland. The Zackenberg valley has been depicted with a white circle

The second focus area is at the Zackenberg research station in North-east Greenland. In a global change context, Zackenberg research station is situated ideally at the border between high- and low Arctic (Figure 2). This region is predicted to undergo the largest climatic changes worldwide (ACIA 2004) and is unaffected by direct human influence. The Zackenberg Basic program is part of the most extensive ecosystem monitoring program in the Arctic; the Greenland Ecosystem Monitoring program (GEM) (Albon et al. 2014). The geographical locations are Nuuk (SW Greenland) and Zackenberg (NE Greenland). Arthropods have been collected as contribution to GEM in Nuuk, and Zackenberg, since 2007 and 1996 respectively. The Zackenberg Basic Monitoring Program has been in operation since 1996 with standardized collection of arthropods via a gridded pitfall trapping regime. The overall aim of the biological monitoring program (Biobasis) is to establish long and simultaneous data series on all trophic levels in a low Arctic ecosystem. The biological program includes a large number of parameters related to terrestrial plants, arthropods, birds and mammal dynamics, and lake ecology. During the past 20 years where Zackenberg basic has been operating, the temperature in the region has increased significantly (Bowden et al. 2015).

Arctic arthropod community composition

Global change is altering the landscape, making new land available for farming, mining and other extractive industries. These activities are predicted to affect large land masses (CAFF 2013), hereby increasing the possibility that rare species will go extinct. In order to foresee how the change in land-use will affect the inhabiting arthropod species, we first need to identify areas of conservation concern and understand how the abiotic environment shapes arthropod species composition. Species distribution patterns vary across multiple spatial scales (Willis & Whittaker 2002), but with four overall amplitudes relevant to arthropods. I) Global scale: Species diversity generally decreases toward the poles (Gaston 2000), with the Arctic as a region relatively low in biodiversity. II) Regional scale: Within the Arctic itself, large climatic gradients divide the Arctic into high, low and sub-Arctic regions, which coupled with large scale dispersal barriers (i.e. inland ice cap), create a variety of living conditions. III) Local scale: Large habitat variability and dispersal barriers, creates variations in species composition. A handful of studies have now shown pronounced compositional changes at regional to local scales within groups of arthropods (Bowden & Buddle 2010; Rich et al. 2013; Sikes et al. 2013). IV) Micro scale. Variation in habitats, operating at relatively minute scales (a few meters), can furthermore cause species to cluster, increasing the potential of species extinction during periods of disturbance.

One major obstacle in identifying how the environment will change future patterns in Arctic arthropod species distributions is limited data on contemporary arthropod species distribution.

In paper I we addressed some of the paucities in Arctic arthropod literature, investigating basic distributional patterns. We examined the degree of spatial variation in species diversity and assemblage structure among five habitat types at two sites of similar abiotic conditions and plant species composition. We discovered distinct differences in arthropod composition and diversity between the two sites and habitats, even though compositions were largely driven by the same, spatially homogenous abiotic variables (soil moisture and temperature). The study demonstrated large turnover, even over much smaller spatial scales than expected and thus highlighted the importance of the heterogeneity of Arctic landscapes on local to landscape scale distributional patterns of Arctic arthropods.

One of ecology’s most general rules states that the size of an area sampled is positively correlated with the number of species detected (Connor & McCoy 1979; Lomolino 2000). This is referred to as the species-area relationship. This rule holds true in the Arctic, where a recent study demonstrated how short-term sampling of multiple habitats, compared to long term sampling in fewer habitats, resulted in drastic increases in the number of species detected (Sikes et al. 2013). The temporal pendant of species-area relationship: species-time relationship, states how observing an area over temporal increments accumulates species (Preston 1960). However, studies also show that there are seasonal windows, where conditions are optimal, resulting in a peak in species numbers (White & Gilchrist 2007; Xu et al. 2015). This suggests that sampling efforts, aimed at capturing the total species pool, could be prioritized to match these windows of optimal conditions. To date, these temporal patterns have received very little, if any, attention in Arctic arthropod research.

In paper IV, we used the Zackenberg family and species level arthropod data, coupled with family level data from three other Arctic sites spanning a large climatic gradient, to explore the seasonal development of Arctic arthropod communities across multiple sites, habitats and taxa. We were able to show how the seasonal patterns of Arctic arthropod composition and richness developed synchronously across a large Arctic gradient at both the species and family level. There was a tendency towards earlier peak in richness at the sub-Arctic sites, but across all sites one week of carefully planned sampling, represented a large proportion of the total season richness. There was, however, large inter-annual variation for the sites with fewer yearly replications, as compared to Zackenberg with 18 years of data. Results also showed large

variation between the habitats. This led us to the conclusion that short-term sampling over multiple habitats and years can detect significant proportions of the full arthropod community that a site holds through a full season.

Potential climate change effects on arthropod communities

The warming of the Arctic region holds dramatic consequences for terrestrial biodiversity (Callaghan et al. 2005; Post et al. 2009) and carries both direct and indirect consequences for Arctic arthropods. Some of these effects are summarized in figure 3. With arthropods being ectothermic, they are directly influenced by changing temperatures (figure 3, pathway 1). This makes them particularly susceptible to the extremes caused by temperature inclinations, which is thought to impact species composition and diversity heavily (Danks 2004). The forecasted change in precipitation, combined with higher temperatures, glacier melt, and permafrost thaw, alters soil moisture content (Rawlins et al. 2010; Perreault et al. 2015) causing dramatic shifts in plant species compositions (Myers-Smith et al. 2011; Elmendorf et al. 2012; Wrona et al. 2016), as well as thermocast erosion, again influencing the hydrological regime (Perreault et al. 2015) (figure 3, pathway 2, 3, 4 and 5). All of which leads increased plant productivity and indirectly effects arthropod composition and diversity (figure 3, pathway 6) (Hodkinson 2013). The hydrological response, however, is not unidirectional - while some areas are becoming wetter and others are drying out (Zona et al. 2009), highly dependent upon permafrost dynamics (Woo & Young 2006). Altered vegetation structure creates a change in surface temperatures, with shading in the summer and snow accumulation in the winter (Blok et al. 2010), impacting the soil dwelling micro arthropods that are important detritivores and food sources for larger arthropods (figure 3, pathway 7 and 8). Changes in surface covers will also undoubtedly alter both micro and macro habitats as structures in the tundra habitats are rearranged (Lawrence & Swenson 2011). At subsurface level, for instance, rhizosphere processes and mycorrhizal associations are influenced by shifts in vegetation (Lindahl & Tunlid 2015) (figure 3, pathway 9). The current observed shrub expansions, tree-line advances and species shifts is reducing local surface albedo (figure 3, pathway 10), exacerbating regional warming through a change in the reflecting snow and ice cover (Wenxin et al. 2013). The transition from tundra to shrubland promotes various additional feedback mechanisms, caused by increased evapotranspiration, resulting in further water loss (figure 3, pathway 11). The idiosyncrasies of these dynamics across the Arctic (Myers-Smith et al. 2015) necessitates multiple site comparisons of habitat and

species responses to climate change and a deeper understanding of the interplay between biotic and abiotic responses.

Figure 3: Conceptual diagram of the some of the direct and indirect effects of climate change on arthropod community composition. The direct effect of warming is indicated by a dashed line Arthropod responses to climate change

There are three ways arthropod species can respond to the direct and indirect effects of a changing climate. Either through range shifts, defined as the change in geographic distributions over time (Lenoir & Svenning 2013), phenological advancement of life history events (Høye & Forchhammer 2008; Høye et al. 2014), and finally through direct alteration of body sizes in organisms adapting to the abrupt changes in physical conditions (Gardner et al. 2011). The three types of species responses to climate change, affect species at the individual and population level, eventually resulting in altered species occurrences, which ultimately sums up to changes in whole arthropod communities. The first four papers of this thesis addresses the potential climate change effects on arthropod communities, while the last three papers examine the species specific effects on life history traits, such as body size and phenology.

- Arthropod species compositional changes to habitat and climatic gradients

The alteration of habitats due to shrub expansion into open tundra and changing wetland hydrology, are likely to affect habitat availability for many organisms through changes in species’ distributions, diversity and composition. Terrestrial arthropods in particular, are associated with specific habitat types and likely respond strongly to habitat changes in the Arctic (Bowden & Buddle 2010; Rich et al. 2013; Sikes et al. 2013; Sweet et al. 2014; Hansen et al. 2016b). Paper II used data collected in the 2013 field campaign to the Godthåbsfjord area, were we sampled transitions in shrub dominance and soil moisture between three different habitats (fen, dwarf shrub heath, and tall shrub tundra) at three different sites. We provided evidence that arthropod diversity and composition vary with soil moisture and vegetation height over very small spatial scales (10–20 m), and discussed the implications of changing tundra habitats to the associated biodiversity. We also identified a handful of species significantly associated with the three different habitats and found that most of the species in the study displayed some form of habitat affinity, matching previous reports (Böcher et al. 2015). Papers I and II both disputed the homogeneity of Arctic arthropod composition and firmly documented the variation in arthropod communities over very small spatial scales.

Range shifts in species-changing occurrences (during their pursuit of optimal conditions) affects whole species compositions, and is a well-documented response to climate change (Lenoir & Svenning 2015). Range shifts include changes in latitude and elevation, and are therefore directly linked to increasing temperatures, as species track their thermal niches. This implies species movements limited to northward or upward shift. However, species composition is highly affected by other factors acting out at the horizontal, and to some extent, local level, such as land use changes (Eskildsen 2015), topographic variation (Rovzar et al. 2016) and soil moisture variation (Le Roux et al. 2013). These gradients have been shown to vary greatly, even at the meter-scale (Suvanto et al. 2014), consequently broadening the need for further inclusion of gradients in range shift studies.

Studies of local climatic gradients can provide us with valuable information on distributional shifts (Lenoir & Svenning 2015), as variations in microclimate can be controlled by topography. Thus, species responses to local topographic variation may help in predicting how compositions are likely to change in the future. Coastal mountainous regions are strongly influenced by continentality (Hodkinson 2005) due to significant differences in thermal capacity between ocean and land. Such differences likely contribute to the structuring of arthropod assemblages. In paper III, we explored the large gradient in continentality represented by the

Godthåbsfjord fjord system and tested whether species diversity and composition changed in response to continentality and micro-climatic conditions. We furthermore tested whether the two groups of arthropods used in the study (beetles and spiders), responded in synchrony to continentality and microclimate. We found that species composition changed significantly along the fjord and that diversity increased with continentality. Spiders and beetles did not, however, show a similar response to the two gradients, as spider diversity responded more to changes in local topography and beetles to continentality. Difference in dispersal capacity between the two groups is likely a contributing factor to this observed pattern, highlighting the need to include functional traits in future studies. A resulting conclusion of this study was that adding details, in terms of multiple environmental gradients and taxonomic groups, increases our understanding of how species assemblages may diversify as a response to climate change.

- Changes to life history traits

Body size is a particularly important trait to the life history of species and has been deemed the third universal response to climate change, alongside shifts in phenology and range (Gardner et al. 2011). In terrestrial arthropods, body size is primarily influenced by food and temperature (Chown & Gaston 2010) and changes in body size affect both fecundity and mass (Bowden & Buddle 2012), as well as overall fitness (Jakob et al. 1996). Snowmelt has proven to be a particularly important predictor of phenology and body size variation (Høye et al. 2009; Høye et al. 2013) and is occurring significantly earlier in the Arctic each year. The timing of snowmelt is modulated by local habitat (Stow et al. 2004), with later snowmelt in low-lying habitats with more or less pronounced shrubby vegetation.

Paper number V explores the effect local habitat has on body size changes. Using individuals of a high Arctic crab spider, collected through the GEM program, we examined how reproductive traits (e.g. body size) and sex ratios have changed over the years with increasing temperatures and earlier snowmelt. The individuals were collected in mesic and arid dwarf shrub heath, and the distinctness of the habitats presented the opportunity to study exactly how the habitat modifies the impact of snowmelt on body size change. Body size variation was significantly related to the timing of snowmelt, and differed significantly between the sexes and between habitats, with the spiders in the mesic habitat showing a stronger response to later snowmelt. This study illustrates how local variation in habitats may conceal climate change effects, highlighting the need to include multiple habitats to the study of species-specific responses to climate change.

Two hypotheses describe how warmer temperatures drive body size changes in ectothermic species. First, the metabolic rates of arthropods increase with warmer temperatures (Gillooly et al. 2001; Gardner et al. 2011). A mismatch may then arise between the increase in metabolic rate and the resources an organism needs to accrue in order to compensate for energy losses related to increased metabolic costs, thus resulting in smaller body size. In contrast, the longer season associated with rising temperatures allows organisms to obtain more resources during the growth season, resulting in larger body sizes (Høye et al. 2009). These two responses may therefor depend upon trophic level and act in concert (Sheridan & Bickford 2011).

In paper number VI, we studied the response of body size to increasing temperatures in two species of butterflies: Colias hecla and Boloria chariclea also collected at Zackenberg as part of the GEM program. We found that wing length decreased in both species in response to warmer summer temperatures, likely caused by increasing metabolism. A previous study on wolf spiders, collected from the same traps, contrasted these findings, as wolf spiders were increasing in body size in response to earlier snowmelt (Høye et al. 2009). Though the study organisms did not respond to the same abiotic variables, the variables are linked to increasing season length. Hence, the two studies corroborate the theory that the contrasting responses depend on trophic level, and thereby energy level of food-resources. If the diet of an organism is energy rich, a longer active season is beneficial. However, for the opposite scenario, a longer active season is costly, and reduces important aspects of fitness related traits, such as body size.

Phenology is a key indicator of species responses to climate change, particularly in polar regions (Parmesan 2006). In recent years, a growing number of studies have documented shift in phenology in a wide array of taxa (Roy & Sparks 2000; Kutz et al. 2005; Høye & Forchhammer 2008; Høye et al. 2014). The changes reported encompass earlier adult emergence and expanded flight season of insects, earlier breeding activity and arrival of migrant birds, as well as earlier flowering of plants (Walther et al. 2002). Set astride the direct influences on the life history of individual species, climate changes also carry the potential to disrupt both inter- and intra- specific interactions (Høye et al. 2013). Decoupling of ecological events, such as mismatches between resources and consumer activity, are some of the more severe examples and pose imminent threats to the persistence of overall ecosystem functioning (Walther et al. 2002; Post et al. 2009).

In paper VII, we demonstrated a trend towards earlier flight season in the most abundant butterfly species (Boloria chariclea) in the Zackenberg valley. The timing of onset, peak and end

of the flight season in B. chariclea were closely related to snowmelt, July temperatures, and dynamics between snowmelt and increased temperature. The duration of the butterfly flight season was significantly positively correlated to the temporal overlap with floral resources, and shows how the aforementioned decoupling of ecological events is already manifesting.

Future recommendations and perspectives

Results presented in this thesis show that Arctic arthropod species respond to very local changes in habitat parameters in terms of compositional changes as well as life history traits (e.g. body size and phenology). Understanding how species respond to local changes in the environment can help us to recognize how species distribute over larger spatial gradients. In order to achieve this, however, we first need to test the generality of mechanisms driving the local patterns. Many studies already point to idiosyncratic site responses, potentially caused by uneven temporal coverage, unmeasured local variation, or low statistical power (Myers-Smith et al. 2011; Oberbauer et al. 2013). We therefore need to assess whether this is an artifact of unstandardized sampling protocols or true site specific characteristics. Hence, spatial upscaling of long-term ecosystem-level monitoring programs, such as Zackenberg and Nuuk basic, can help avoid the caveats of unstandardized studies.

These ecosystem monitoring programs are costly, and to fill all the gaps and represent all compartments of Arctic ecosystems is a daunting task. Additional short-term studies may work as a supplement to current data compilation. Through spatial upscaling of standardized sampling designs of gradient studies mimicking future ecological changes, we can distinguish idiosyncrasies from general mechanisms. With a synchronous seasonal development and aggregation of arthropod species, we are able to upscale compositional studies to cover larger spatial extents with so-called “snapshot” gradient studies. This kind of sampling will provide large-scale insights to the spatial variation in species assemblages and the driving mechanisms behind, as well as standardize sampling procedures, for more reliable pan-Arctic inferences. The Circumpolar Biodiversity Monitoring Program (CBMP) is an international network of scientists, government agencies, Indigenous organizations and conservation groups working together to harmonize and integrate efforts to monitor the Arctic's living resources (CAFF 2013). This “network of networks” facilitates the interplay between full scale monitoring programs and snapshot studies and is a valuable platform in future studies

Most of the papers presented in this thesis hint at responses acting out at a highly detailed and local level. Responses that may go undetected if investigated at more coarse spatial resolutions. Hence the devil is in the details, but detailed investigations of general mechanisms are applicable in wider contexts. The results of this thesis show that the drastic increase in temperatures during the last couple of decades, and the following rearranging of structural components in the Arctic tundra, holds dire implications for arthropod diversity. With consequences encompassing decoupling of important life history events, changes to reproduction and overall fitness with resulting changes in species compositions and diversity reported over multiple taxa. Arthropods are globally the most diverse group of organisms, even in the Arctic. Hence, community- and species-level responses to climate change will reveal overall threats to other aspects of biodiversity. Moreover, arthropod responses are acute due to their ectothermic lifestyle, mobility, and short generation time, rendering them the “canary in the coal mine” for climate change implications at multiple trophic levels. It is clear that arthropod studies are increasingly important and should, in my objective opinion, be granted detailed consideration in future research of climate change effects on Arctic terrestrial environments.

References

ACIA. 2004. Impacts of a Warming Arctic: Arctic Climate Impact Assessment. ACIA Overview report. Cambridge. Albon S, Thiede J, and Holmén K. 2014. Greenland Ecosystem Monitoring. GEM Evaluation. Final Report. Copenhagen: Danish Energy Agency. Blok D, Heijmans MMPD, Schaepman-Strub G, Kononov AV, Maximov TC, and Berendse F. 2010. Shrub expansion may reduce summer permafrost thaw in Siberian tundra. Global Change Biology 16:1296-1305 DOI: 10.1111/j.1365-2486.2009.02110.x. Böcher J, Kristensen NP, Pape T, and Vilhelmsen L. 2015. The Greenland Entomofauna: An Identification Manual of Insects, Spiders and Their Allies: Brill. Bowden JJ, and Buddle CM. 2010. Determinants of ground-dwelling spider assemblages at a regional scale in the Yukon Territory, Canada. Ecoscience 17:287-297 DOI: 10.2980/17- 3-3308. Bowden JJ, and Buddle CM. 2012. Life history of tundra-dwelling wolf spiders (Araneae: Lycosidae) from the Yukon Territory, Canada. Canadian Journal of Zoology-Revue Canadienne De Zoologie 90:714-721 DOI: 10.1139/Z2012-038. Bowden JJ, Eskildsen A, Hansen RR, Olsen K, Kurle CM, and Høye TT. 2015. High-Arctic butterflies become smaller with rising temperatures. Biology Letters 11 DOI: 10.1098/rsbl.2015.0574. CAFF. 2013. Arctic Biodiversity Assessment. Status and trends in Arctic biodiversity. Akureyri. Callaghan T, Björn LO, Chapin FS, Chernov Y, Christensen TR, Huntley B, Ims R, Johansson M, R.D J, Jonasson SE, Matveyeva N, Oechel W, Panikov N, Shaver G, Elster J,

Henttonen H, Jónsdóttir IS, Laine K, Schaphoff S, Stich S, Taulavuori E, Taulavuori K, and Zöckler C. 2005. ACIA 2005. Arctic Climate Impact Assessment. Cambridge: cambridge University Press. Chown SL, and Gaston KJ. 2010. Body size variation in insects: a macroecological perspective. Biological Reviews 85:139-169 DOI: 10.1111/j.1469-185X.2009.00097.x. Connor EF, and McCoy ED. 1979. The Statistics and Biology of the Species-Area Relationship. The American Naturalist 113:791-833 Danks HV. 1981. Arctic arthropods: a review of systematics and ecology with particular reference to the North American fauna: Entomological Society of Canada. Danks HV. 1991. Winter Habitats and Ecological Adaptations for Winter Survival. In: Lee RE, and Denlinger DL, eds. Insects at Low Temperature. Boston, MA: Springer US, 231-259. Danks HV. 2004. Seasonal adaptations in Arctic insects. Integrative and Comparative Biology 44:85-94 DOI: 10.1093/icb/44.2.85. Elmendorf SC, Henry GHR, Hollister RD, Bjork RG, Bjorkman AD, Callaghan TV, Collier LS, Cooper EJ, Cornelissen JHC, Day TA, Fosaa AM, Gould WA, Gretarsdottir J, Harte J, Hermanutz L, Hik DS, Hofgaard A, Jarrad F, Jonsdottir IS, Keuper F, Klanderud K, Klein JA, Koh S, Kudo G, Lang SI, Loewen V, May JL, Mercado J, Michelsen A, Molau U, Myers-Smith IH, Oberbauer SF, Pieper S, Post E, Rixen C, Robinson CH, Schmidt NM, Shaver GR, Stenstrom A, Tolvanen A, Totland O, Troxler T, Wahren CH, Webber PJ, Welker JM, and Wookey PA. 2012. Global assessment of experimental climate warming on tundra vegetation: heterogeneity over space and time. Ecology Letters 15:164-175 DOI: 10.1111/j.1461-0248.2011.01716.x. Eskildsen A. 2015. Long-term Effects of Global Change on the Distribution, Species Richness and Life History of Butterflies: PhD Thesis. Aarhus University. Gardner JL, Peters A, Kearney MR, Joseph L, and Heinsohn R. 2011. Declining body size: a third universal response to warming? Trends in Ecology & Evolution 26:285-291 DOI: http://dx.doi.org/10.1016/j.tree.2011.03.005. Gaston KJ. 2000. Global patterns in biodiversity. Nature 405:220-227 DOI: 10.1038/35012228. Gillooly JF, Brown JH, West GB, Savage VM, and Charnov EL. 2001. Effects of Size and Temperature on Metabolic Rate. Science 293:2248-2251 DOI: 10.1126/science.1061967. Greenland Statbank. 2015. Greenland in figures. Nuuk: Statistics Greenland. Hansen RR, Hansen OLP, Bowden JJ, Normand S, Bay C, Sørensen JG, and Høye TT. 2016a. High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap. Polar Biology:1-10 DOI: 10.1007/s00300-016-1893-2. Hansen RR, Hansen OLP, Bowden JJ, Treier UA, Normand S, and Høye T. 2016b. Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities. Peerj 4:e2224 DOI: 10.7717/peerj.2224. Heller NE, and Zavaleta ES. 2009. Biodiversity management in the face of climate change: A review of 22 years of recommendations. Biological Conservation 142:14-32 DOI: http://dx.doi.org/10.1016/j.biocon.2008.10.006. Hodkinson ID. 2005. Terrestrial insects along elevation gradients: species and community responses to altitude. Biological Reviews 80:489-513 DOI: 10.1017/s1464793105006767. Hodkinson ID. 2013. Terrestrial and Freshwater Invertebrates. In: Meltofte H, ed. Arctic Biodiverisity AssessmentStatus and trends in Arctic biodiversity. Akureyri, 195-219. Høye TT, Eskildsen A, Hansen RR, Bowden JJ, Schmidt NM, and Kissling WD. 2014. Phenology of high-arctic butterflies and their floral resources: Species-specific responses to climate change. Current Zoology 60:243-251 Høye TT, and Forchhammer MC. 2008. Phenology of High-Arctic Arthropods: Effects of Climate on Spatial, Seasonal, and Inter-Annual Variation. In: Hans Meltofte

TRCBEMCF, and Morten R, eds. Advances in Ecological Research: Academic Press, 299-324. Høye TT, Hammel JU, Fuchs T, and Toft S. 2009. Climate change and sexual size dimorphism in an Arctic spider. Biology Letters 5:542-544 DOI: 10.1098/rsbl.2009.0169. Høye TT, Post E, Schmidt NM, Trojelsgaard K, and Forchhammer MC. 2013. Shorter flowering seasons and declining abundance of flower visitors in a warmer Arctic. Nature Climate Change 3:759-763 DOI: 10.1038/Nclimate1909. Ipcc. 2014a. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Barros, V.R., C.B. Field, D.J. Dokken, M.D. Mastrandrea, K.J. Mach, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken, P.R. Mastrandrea, and L.L. White (eds.)]. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. Ipcc. 2014b. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Summaries, Frequently Asked Questions, and Cross-Chapter Boxes. A Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Field, C.B., V.R. Barros, D.J. Dokken, K.J. Mach, M.D. Mastrandrea, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken,P.R. Mastrandrea, and L.L. White (eds.)].190 Ipcc. 2014c. Summary for Policymakers. In: Field CB, Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE, Chatterjee M, Ebi KL, Estrada YO, Genova RC, Girma B, Kissel ES, Levy AN, MacCracken S, Mastrandrea PR, and White LL, eds. Climate Change 2014: Impacts, Adaptation, and Vulnerability Part A: Global and Sectoral Aspects Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom, and New York, NY, USA: Cambridge University Press, 1-32. Jakob EM, Marshall SD, and Uetz GW. 1996. Estimating fitness: A comparison of body condition indices. Oikos 77:61-67 DOI: Doi 10.2307/3545585. Kellogg WW. 1983. Feedback mechanisms in the climate system affecting future levels of carbon dioxide. Journal of Geophysical Research: Oceans 88:1263-1269 DOI: 10.1029/JC088iC02p01263. Kutz SJ, Hoberg EP, Polley L, and Jenkins EJ. 2005. Global warming is changing the dynamics of Arctic host–parasite systems. Proceedings of the Royal Society B: Biological Sciences 272:2571-2576 DOI: 10.1098/rspb.2005.3285. Lawrence DM, and Swenson SC. 2011. Permafrost response to increasing Arctic shrub abundance depends on the relative influence of shrubs on local soil cooling versus large- scale climate warming. Environmental Research Letters 6:045504 Le Roux PC, Aalto J, and Luoto M. 2013. Soil moisture's underestimated role in climate change impact modelling in low-energy systems. Global Change Biology 19 DOI: 10.1111/Gcb.12286. Lenoir J, and Svenning J-C. 2013. Latitudinal and Elevational Range Shifts under Contemporary Climate Change A2 - Levin, Simon A. Encyclopedia of Biodiversity (Second Edition). Waltham: Academic Press, 599-611. Lenoir J, and Svenning JC. 2015. Climate-related range shifts – a global multidimensional synthesis and new research directions. Ecography 38:15-28 DOI: 10.1111/ecog.00967. Lindahl BD, and Tunlid A. 2015. Ectomycorrhizal fungi – potential organic matter decomposers, yet not saprotrophs. New Phytologist 205:1443-1447 DOI: 10.1111/nph.13201.

Lomolino MV. 2000. Ecology’s most general, yet protean 1 pattern: the species-area relationship. Journal of Biogeography 27:17-26 DOI: 10.1046/j.1365-2699.2000.00377.x. Manabe s. 1969. Climate and the ocean circulation. Monthly Weather Review 97:775-805 DOI: doi:10.1175/1520-0493(1969)097<0775:CATOC>2.3.CO;2. Myers-Smith IH, Elmendorf SC, Beck PSA, Wilmking M, Hallinger M, Blok D, Tape KD, Rayback SA, Macias-Fauria M, Forbes BC, Speed JDM, Boulanger-Lapointe N, Rixen C, Levesque E, Schmidt NM, Baittinger C, Trant AJ, Hermanutz L, Collier LS, Dawes MA, Lantz TC, Weijers S, Jorgensen RH, Buchwal A, Buras A, Naito AT, Ravolainen V, Schaepman-Strub G, Wheeler JA, Wipf S, Guay KC, Hik DS, and Vellend M. 2015. Climate sensitivity of shrub growth across the tundra biome. Nature Climate Change 5:887-+ DOI: 10.1038/nclimate2697. Myers-Smith IH, Forbes BC, Wilmking M, Hallinger M, Lantz T, Blok D, Tape KD, Macias- Fauria M, Sass-Klaassen U, Levesque E, Boudreau S, Ropars P, Hermanutz L, Trant A, Collier LS, Weijers S, Rozema J, Rayback SA, Schmidt NM, Schaepman-Strub G, Wipf S, Rixen C, Menard CB, Venn S, Goetz S, Andreu-Hayles L, Elmendorf S, Ravolainen V, Welker J, Grogan P, Epstein HE, and Hik DS. 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environmental Research Letters 6 DOI: 10.1088/1748-9326/6/4/045509. Normand S, Randin C, Ohlemuller R, Bay C, Hoye TT, Kjaer ED, Korner C, Lischke H, Maiorano L, Paulsen J, Pearman PB, Psomas A, Treier UA, Zimmermann NE, and Svenning JC. 2013. A greener Greenland? Climatic potential and long-term constraints on future expansions of trees and shrubs. Philosophical Transactions of the Royal Society B-Biological Sciences 368 DOI: 10.1098/Rstb.2012.0479. Oberbauer SF, Elmendorf SC, Troxler TG, Hollister RD, Rocha AV, Bret-Harte MS, Dawes MA, Fosaa AM, Henry GHR, Hoye TT, Jarrad FC, Jonsdottir IS, Klanderud K, Klein JA, Molau U, Rixen C, Schmidt NM, Shaver GR, Slider RT, Totland O, Wahren CH, and Welker JM. 2013. Phenological response of tundra plants to background climate variation tested using the International Tundra Experiment. Philosophical Transactions of the Royal Society B-Biological Sciences 368 DOI: 10.1098/Rstb.2012.0481. Parmesan C. 2006. Ecological and Evolutionary Responses to Recent Climate Change. Annual Review of Ecology, Evolution, and Systematics 37:637-669 DOI: 10.1146/annurev.ecolsys.37.091305.110100. Perreault N, Lévesque E, Fortier D, and Lamarque LJ. 2015. Thermo-erosion gullies boost the transition from wet to mesic vegetation. Biogeosciences Discuss 12:12191-12228 DOI: 10.5194/bgd-12-12191-2015. Pickett STA. 1989. Space-for-Time Substitution as an Alternative to Long-Term Studies. In: Likens GE, ed. Long-Term Studies in Ecology: Approaches and Alternatives. New York, NY: Springer New York, 110-135. Post E, Forchhammer MC, Bret-Harte MS, Callaghan TV, Christensen TR, Elberling B, Fox AD, Gilg O, Hik DS, Hoye TT, Ims RA, Jeppesen E, Klein DR, Madsen J, McGuire AD, Rysgaard S, Schindler DE, Stirling I, Tamstorf MP, Tyler NJC, van der Wal R, Welker J, Wookey PA, Schmidt NM, and Aastrup P. 2009. Ecological Dynamics Across the Arctic Associated with Recent Climate Change. Science 325:1355-1358 DOI: 10.1126/science.1173113. Preston FW. 1960. Time and Space and the Variation of Species. Ecology 41:612-627 DOI: 10.2307/1931793. Rawlins MA, Steele M, Holland MM, Adam JC, Cherry JE, Francis JA, Groisman PY, Hinzman LD, Huntington TG, Kane DL, Kimball JS, Kwok R, Lammers RB, Lee CM, Lettenmaier DP, McDonald KC, Podest E, Pundsack JW, Rudels B, Serreze MC, Shiklomanov A,

Skagseth Ø, Troy TJ, Vörösmarty CJ, Wensnahan M, Wood EF, Woodgate R, Yang D, Zhang K, and Zhang T. 2010. Analysis of the Arctic System for Freshwater Cycle Intensification: Observations and Expectations. Journal of Climate 23:5715-5737 DOI: 10.1175/2010JCLI3421.1. Rich ME, Gough L, and Boelman NT. 2013. Arctic arthropod assemblages in habitats of differing shrub dominance. Ecography 36:994-1003 DOI: 10.1111/j.1600- 0587.2012.00078.x. Rovzar C, Gillespie TW, and Kawelo K. 2016. Landscape to site variations in species distribution models for endangered plants. Forest Ecology and Management 369:20-28 DOI: http://dx.doi.org/10.1016/j.foreco.2016.03.030. Roy DB, and Sparks TH. 2000. Phenology of British butterflies and climate change. Global Change Biology 6:407-416 DOI: 10.1046/j.1365-2486.2000.00322.x. Screen JA, and Simmonds I. 2010. The central role of diminishing sea ice in recent Arctic temperature amplification. Nature 464:1334-1337 DOI: http://www.nature.com/nature/journal/v464/n7293/suppinfo/nature09051_S1.html. Sheridan JA, and Bickford D. 2011. Shrinking body size as an ecological response to climate change. Nature Clim Change 1:401-406 DOI: http://www.nature.com/nclimate/journal/v1/n8/abs/nclimate1259.html#supplementary- information. Sikes DS, Draney ML, and Fleshman B. 2013. Unexpectedly high among-habitat spider (Araneae) faunal diversity from the Arctic Long-Term Experimental Research (LTER) field station at Toolik Lake, Alaska, United States of America. Canadian Entomologist 145:219-226 DOI: 10.4039/tce.2013.5. Stow DA, Hope A, McGuire D, Verbyla D, Gamon J, Huemmrich F, Houston S, Racine C, Sturm M, Tape K, Hinzman L, Yoshikawa K, Tweedie C, Noyle B, Silapaswan C, Douglas D, Griffith B, Jia G, Epstein H, Walker D, Daeschner S, Petersen A, Zhou L, and Myneni R. 2004. Remote sensing of vegetation and land-cover change in Arctic Tundra Ecosystems. Remote Sensing of Environment 89:281-308 DOI: http://dx.doi.org/10.1016/j.rse.2003.10.018. Suvanto S, Le Roux PC, and Luoto M. 2014. Arctic-alpine vegetation biomass is driven by fine- scale abiotic heterogeneity. Geografiska Annaler Series a-Physical Geography 96:549- 560 DOI: 10.1111/Geoa.12050. Sweet SK, Asmus A, Rich ME, Wingfield J, Gough L, and Boelman NT. 2014. NDVI as a predictor of canopy arthropod biomass in the Alaskan arctic tundra. Ecological Applications 25:779-790 DOI: 10.1890/14-0632.1. Walther GR, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC, Fromentin JM, Hoegh- Guldberg O, and Bairlein F. 2002. Ecological responses to recent climate change. Nature 416:389-395 DOI: 10.1038/416389a. Wenxin Z, Paul AM, Benjamin S, Rita W, Torben K, and Ralf D. 2013. Tundra shrubification and tree-line advance amplify arctic climate warming: results from an individual-based dynamic vegetation model. Environmental Research Letters 8:034023 White EP, and Gilchrist MA. 2007. Effects of population-level aggregation, autocorrelation, and interspecific association on the species-time relationship in two desert communities. Evolutionary Ecology Research 9:1329-1347 Willis KJ, and Whittaker RJ. 2002. Ecology - Species diversity - Scale matters. Science 295:1245-1248 DOI: 10.1126/science.1067335. Woo M-k, and Young KL. 2006. High Arctic wetlands: Their occurrence, hydrological characteristics and sustainability. Journal of Hydrology 320:432-450 DOI: http://dx.doi.org/10.1016/j.jhydrol.2005.07.025.

Wrona FJ, Johansson M, Culp JM, Jenkins A, Mård J, Myers-Smith IH, Prowse TD, Vincent WF, and Wookey PA. 2016. Transitions in Arctic ecosystems: Ecological implications of a changing hydrological regime. Journal of Geophysical Research: Biogeosciences 121:650-674 DOI: 10.1002/2015JG003133. Xu H, Yong J, and Xu G. 2015. Sampling frequency of ciliated protozoan microfauna for seasonal distribution research in marine ecosystems. Marine Pollution Bulletin 101:653- 659 DOI: http://dx.doi.org/10.1016/j.marpolbul.2015.10.034. Zona D, Oechel WC, Kochendorfer J, U KTP, Salyuk AN, Olivas PC, Oberbauer SF, and Lipson DA. 2009. Methane fluxes during the initiation of a large-scale water table manipulation experiment in the Alaskan Arctic tundra. Global Biogeochemical Cycles 23 DOI: 10.1029/2009gb003487.

Polar Biol DOI 10.1007/s00300-016-1893-2

ORIGINAL PAPER

High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap

1 1 1 Rikke Reisner Hansen • Oskar Liset Pryds Hansen • Joseph James Bowden • 2 3 2 1,4,5 Signe Normand • Christian Bay • Jesper Givskov Sørensen • Toke Thomas Høye

Received: 17 August 2015 / Revised: 11 January 2016 / Accepted: 11 January 2016 Ó Springer-Verlag Berlin Heidelberg 2016

Abstract Arthropods form a major part of the terrestrial regression. Species richness and the species pool differed species diversity in the Arctic, and are particularly sensitive between sites, with the latter indicating high species turn- to temporal changes in the abiotic environment. It is assumed over. Local-scale assemblage patterns were related to soil that most Arctic arthropods are habitat generalists and that moisture and temperature. We conclude that Arctic arthro- their diversity patterns exhibit low spatial variation. The pod species assemblages vary substantially over short dis- empirical basis for this assumption, however, is weak. We tances due to local soil characteristics, while regional examine the degree of spatial variation in species diversity variation in the species pool is likely influenced by geo- and assemblage structure among five habitat types at two graphic barriers, i.e., inland ice sheet, glaciers, mountains sites of similar abiotic conditions and plant species compo- and large water bodies. In order to predict future changes to sition in southwest Greenland, using standardized field col- Arctic arthropod diversity, further efforts are needed to lection methods for spiders, beetles and butterflies. We disentangle contemporary drivers of diversity at multiple employed non-metric multidimensional scaling, species spatial scales. richness estimation, community dissimilarity and indicator species analysis to test for local (within site)- and regional Keywords Araneae Á Biodiversity Á Climate change Á (between site)-scale differences in arthropod communities. Coleoptera Á Habitat heterogeneity Á Lepidoptera To identify specific drivers of local arthropod assemblages, we used a combination of ordination techniques and linear Introduction

Electronic supplementary material The online version of this Climate change is causing rapid alterations to Arctic habi- article (doi:10.1007/s00300-016-1893-2) contains supplementary material, which is available to authorized users. tats (Myers-Smith et al. 2011; Elmendorf et al. 2012; Sch- midt et al. 2012), and for many taxonomic groups like & Rikke Reisner Hansen arthropods, we are largely unable to identify biodiversity [email protected] hotspots and prioritize conservation efforts in a changing

1 Arctic environment (CAFF 2013). The climatic changes are Arctic Research Centre, Aarhus University, Ny Munkegade 114, Bldg. 1540, 8000 Aarhus C, Denmark furthermore causing glaciers and the Greenland ice cap to retreat and expose new land, which is already stimulating 2 Department of Bioscience, Aarhus, Aarhus University, Ny Munkegade 116, Bldg. 1540, 8000 Aarhus C, Denmark industrial interests and activities. Such activities will affect large areas, and to identify areas of conservation concern, 3 Department of Bioscience, Roskilde, Aarhus University, Frederiksborgvej 399, 4000 Roskilde, Denmark we have to understand how specific aspects of the abiotic environment shapes Arctic biodiversity (CAFF 2013). 4 Aarhus Institute of Advanced Studies, Aarhus University, Høegh-Guldbergs Gade 6B, Bldg. 1630, 8000 Aarhus C, Species assemblages vary across spatial scales (Willis Denmark and Whittaker 2002) with species diversity generally 5 Department of Bioscience, Kalø, Aarhus University, decreasing toward the poles (Gaston 2000). This also Grena˚vej 14, 8410 Rønde, Denmark applies within the Arctic where large climatic gradients 123 Polar Biol divide the Arctic into high, low and sub-Arctic regions, understand the drivers of changes in arthropod communi- creating a variety of living conditions. While much of the ties. A large body of the literature supports variation in Arctic may be perceived as a collection of homogeneous vegetation structure as a strong driver of arthropod com- landscapes inhabited by generalist species, and rather munities at regional to landscape scales (Halaj et al. 2000; depauperate in diversity, some studies have shown pro- Jimenez-Valverde and Lobo 2007; Bowden and Buddle nounced compositional changes at regional to local scales 2010; Rich et al. 2013). Vegetation structure affects mul- within groups of arthropods (Bowden and Buddle 2010; tiple aspects of microclimatic conditions (e.g., temperature Rich et al. 2013). If we incorrectly assume little or no and moisture regimes) to which arthropods are highly species turnover at small scales, however, we risk the sensitive (Rypstra 1986; Rushton and Eyre 1992; Bonte extinction of local and rare species in connection with et al. 2002). Yet, only a few studies have studied the industrial activities and climate change. Species distribu- specific environmental variables structuring Arctic arthro- tions can furthermore be clustered within microhabitats at pod communities (Bowden and Buddle 2010; Rich et al. very fine spatial scales, increasing the probability of sig- 2013). nificant species loss during periods of disturbance (Sander The goal of this study was to investigate how the and Wardell-Johnson 2011). In the Arctic, arthropods are composition of Arctic arthropod assemblages changes at a of particular interest as this group comprises a large part of local and regional scale by examining species richness, the Arctic biodiversity (Høye and Sikes 2013). Collec- composition and species turnover within and between two tively, the number of Arctic arthropod species greatly sites of similar vegetation composition and abiotic condi- exceeds that of all other non-microbial eukaryotic species tions. Specifically, we were interested in answering two groups combined, including plants and vertebrates (Hod- questions: (1) Do species assemblages and turnover rates kinson 2013). To provide adequate, reliable predictions for change at local and regional scales in low Arctic Green- conservation and survey planning, we need to understand land? (2) Which environmental parameters predict local how arthropods are distributed and to identify the envi- arthropod assemblages in areas of low structural com- ronmental factors which are driving variation in arthropod plexity? We chose three arthropod taxa representing dif- assemblages. ferent functional groups as our focal study organisms: Turnover patterns can be affected by a multitude of butterflies (pollinators that are herbivorous in all life factors. Peninsula effects can serve as a filter to species stages), spiders (generalist predators and strongly influ- dispersal creating a change in species pools over small enced by structural elements for web building, foraging distances (Simpson 1964). The Arctic in general and etc.) and beetles (including both herbivores, fungivores, Greenland in particular is dominated by fjords, glaciers, detritivores and predators). large water bodies and the inland ice cap that can create important dispersal barriers. Species composition and diversity varies with habitat heterogeneity such that a very Materials and methods heterogeneous habitat has more suitable living conditions for more species with a larger range of functional traits Study area and data (Tews et al. 2004; Kostylev et al. 2005; Brind’Amour et al. 2011). Small-scale topographic and abiotic changes can The study was carried out in West Greenland at the remote further create distinct microhabitats supporting certain area of Isua, situated approximately 100 km northeast of the species pools (Suvanto et al. 2014). For arthropods, this is capital Nuuk. The location was selected with the additional particularly relevant as they represent a great variety of life aim to obtain baseline biodiversity data due to plans of history, behavior and morphological and physiological mining the region’s large iron deposits. The two study sites adaptations (Wise 1993; Danks 2004). A recent study on selected, referred to as site A (65°270N, 50°220W) and site B spider diversity at Toolik Lake, Alaska, documented high (65°220N, 50°460W), respectively, are situated approxi- species turnover just by sampling different habitats at a mately 32 km apart. Site A is located close (*2 km) to the local scale (Sikes et al. 2013), highlighting the importance ice edge at an approximate elevation of 600 m.a.s.l, and site of investigating patterns of arthropod diversity at a higher B is further away (*34 km) at an approximate elevation of spatial resolution. 700 m.a.s.l (Fig. 1). Both sites had similar dwarf shrub cover Habitat heterogeneity and complexity is often referred to and comparable plant communities. The vegetation of the as the variation in vegetation structure (Imhoff et al. 1997; two study sites could be divided into the following five Venn and Kotze 2014). However, vegetation structure habitat types: fen, snow-bed, herb slope, exposed ridge and integrates a larger suite of variables, such as solar radiation dry heath. The dominant plant species (occurring in more and soil characteristics. By separating these factors con- than 30 out of the 80 plots selected) found at each site were founded within vegetation structure, we may better Vaccinium uliginosum, Empetrum nigrum, Carex bigelowii 123 Polar Biol

Fig. 1 Map of the study area in southwest Greenland. Site A (white)at65°270N, 50°220W and site B (black)at65°220N, 50°460W are situated Site A approximately 32 km apart. The figure includes an inset map of the entire Greenland with the capital of Greenland (Nuuk) depicted with a diamond-shaped dot, and the study area indicated with a black dot

Site B Ice sheet

Nuuk

10 km and Salix herbacea. Additional information on plant species volumetric water content) and soil temperature (with one is given in Online Resource 1. In each habitat category at decimal accuracy) was measured consistently in five places each site, a 40 9 40 m plot group, comprising 16 different within a 1 m radius from the center of each plot (center, 10 9 10 m plots was established. In total, 160 plots (two north, south, east and west) using a soil moisture sensor and a sites* five habitat types * 16 plots in each plot group) were standard digital cooking thermometer. All spiders, beetles established. and butterflies were identified to the species level based on Arthropods were collected using yellow pitfall traps genital and morphological characters using a WildÒ M5A ([ = 10 cm). Two pitfall traps separated by half a meter stereo microscope. Spiders were identified using the avail- were dug into the ground with the rim of the pitfall flush to the able literature through The World Spider Catalog (World soil surface in each plot (320 pitfall traps in total). The color Spider Catalog 2016) as well as Spiders of North America of the pitfalls was chosen to catch flying as well as surface- (Paquin and Dupe´rre´ 2003). Beetles were identified using active arthropods (Høye et al. 2014). At site B, the traps were both Scandinavian and North American literature (Lindroth active between July 28 and August 1, 2013; at site A the traps 1985, 1986;Bo¨cher 1988) and consulting the collection at were active from August 4 to August 8, 2013, resulting in a the Natural History Museum, Aarhus. All specimens were total of 3200 pitfall trap days. Because the sites were very identified by the corresponding author. Specimens are pre- remote and access to each site was only possible with char- served in 75 % ethanol at the Natural History Museum, tered helicopter, we opted to sample intensely across a high Aarhus. The dataset with habitat information, sampling number of plots to optimize our collection of most arthro- periods and coordinates is available in Online Resource pods in the relatively short time frame. Vegetation cover of 2 and through Global Biodoversity Information Facility shrubs, herbs, graminoids and bare ground was estimated in a (GBIF) (http://dx.doi.org/doi:10.15468/hblhkl). Not all circle with a 5 m radius from the center of each plot as per- juveniles or larvae could be assigned to species, so only adult cent cover and were classified in six categories as not present specimens were included in the data analysis. or in equal-sized percentage ranges of: 1–20 %, 21–40 %, 41–60 %, 61–80 % and 81–100 %. Vegetation height was Data analysis recorded as prevailing height in 50 % of the circle and classified in four groups (0–5 cm, 6–10 cm, 11–15 cm and We ran species richness estimations and ordinations on 16–20 cm). Presence of plant species were registered in the both plant and arthropod community matrices to compare 5-m circle. Slope was measured in vertical meters between patterns of diversity and composition for both groups. the highest and lowest point of the plot. Aspect was recorded Specifically, we wanted to examine whether a potentially using a Garmin S30 and classified to nearest cardinal lower arthropod diversity closer to the ice edge could be an direction (north, south, east, and west). Soil moisture (% effect of the plant communities being at an earlier

123 Polar Biol successional stage. All statistical analyses were conducted and plant species were better represented by amalgamating in R, version 3.1.0 (R Development Core Team 2015). the 16 plots of 10 9 10 m into four clusters of four neighboring plots of 20 9 20 m. Two arthropod plots Species diversity contained no specimens belonging to the three investigated groups, and these plots were omitted from all ordination Species diversity was rarefied and extrapolated for inves- analyses. To further counteract a potential effect of tigation across sites and habitats based on Hill numbers undersampling, we reran the ordinations on arthropod (q = 0; species richness, q = 1; Shannon diversity, q = 2; abundance data without singletons (species represented by Simpson diversity) and standardized by sample coverage one individual). Site and habitat effects in the composition (Chao and Jost 2012; Chao et al. 2014) using the iNEXT of the arthropod assemblages were analyzed with analysis package (Hsieh et al. 2014). Prior to extrapolation, both of dissimilarities, using the function ‘adonis’ in the ‘vegan’ samples were approximating 100 % coverage, so they were package. Analysis of dissimilarities is analogous to dis- both compared at 100 % coverage, minimizing extrapola- tance-based redundancy analysis and is recommended as a tion. We tested whether the results were significantly dif- robust method for permutational multivariate analysis of ferent across sites with a Student’s t test. A Venn diagram variance with community ecology data (McArdle and was created to visually interpret the difference in species Anderson 2001). pools between the two sites. To investigate how turnover rates differed between sites, we used the function ‘be- Environmental correlation with composition tadisper’ in the ‘vegan’ package (Oksanen et al. 2015), followed by ANOVA. This test assesses the variability in The importance of local (within site) predictor variables for average distances from the group centroid among individ- structuring of arthropods at each site was determined by ual sampling units and is recommended as a method to vectors fitted to ordinations. To match the structure of the measure beta diversity (Anderson et al. 2006). For this amalgamated arthropod matrix, all values of predictor analysis, we employed the Chao dissimilarity index which variables were averaged over the four plots and standard- is robust even with under-sampled data, containing many ized to a mean of zero and a standard deviation of one to rare species (Chao et al. 2005). We then used ANOVA to enable a meaningful interpretation of the results. To allow test the difference between the sampling units. This was for a more objective interpretation of the ordination results followed by a partitioning of beta diversity into nestedness and as a basis for choosing variables in the later model and turnover with the use of the package ‘betapart’ selection, the continuous abiotic variables were overlaid on (Baselga and Orme 2012). This function calculates three the NMDS plot as vectors, using the ‘envfit’ function in the measures of beta diversity: overall beta diversity (bSOR), ‘vegan’ package (Oksanen et al. 2015). The direction and spatial turnover rates (bSIM) and nestedness (bNES). The length of each vector indicates the direction of the gradient latter two define how much of the overall beta diversity is and the strength of the correlation, respectively. Only the attributed to pure spatial turnover rates and how much is significant variables were included in further analyses. All due to nestedness based on the Sørensen index (Baselga tested variables can be found in Online Resource 3. We 2010). We compared partitioning of beta diversity between chose not to include plant species composition as a pre- the two sites and within sites among habitats, to assess how dictor as we were interested in testing the specific drivers the actual species turnover differs at the various levels. of arthropod distributions and plant species composition may confound such factors as vegetation structure, soil Species composition moisture and soil temperature. To assure that the difference in arthropod composition Non-metric multidimensional scaling (NMDS) (Legendre and species diversity at the two sites was not caused by and Legendre 1998) was conducted using the function differences in environmental conditions, we tested for metaMDS (K = 2, iterations = 100) based on the Bray– difference in means of the significant measured variables Curtis dissimilarity index. Ordinations were performed with t tests. To account for covariation and colinearity using the ‘vegan’ package in R (Oksanen et al. 2015) which between variables not accounted for by the ‘envfit’ analy- uses multiple random start configurations to find a sis, we employed a generalized linear model (GLM) to fit stable global solution for the ordination and transforms the scores of the two ordination axes. This technique has data to reduce the influence of highly abundant species. been applied previously to spatially predict vegetation Because of the relatively short sampling periods, we were gradients along a climatic gradient (Muenchow et al. concerned about undersampling. Therefore, we investi- 2013). Using this technique, we tested for an interaction gated several combinations of the data when running between habitat and the significant variables from the NMDS. This led us to the conclusion that both arthropods ‘envfit’ analysis, and an interaction was only included in 123 Polar Biol the final model if significant. We explored different pos- Plant species assembly at site B was a subset of species sibilities for nonlinear responses using both correlation represented at site A, i.e., strict nestedness (Online analysis and a generalized additive model and found no Resource 6), but for arthropods, the species pool at each indication of nonlinearity between any of the predictor site consisted partly of species restricted to either site A or variables and the extracted ordination axes. Model B as well as shared species (Fig. 3b). There was a near- parameters were chosen based on the significance of vari- significant difference in turnover rates between the sites ables from the vector fitting described above. Backward (F1,36 = 2.98, p = 0.09), and a significant difference selection was then employed until the simplest GLM between habitats within site A (F4,13 = 4.00, p = 0.03) but explaining the largest portion of the variance was obtained. not at site B (F4,15 = 1.52, p = 0.24), indicating higher Finally, we ran a species indicator analysis to assess the between habitat variation at site A. For plant species, there strength and statistical significance of the relationship was no significant difference in turnover rates between or between species occurrence/abundance and the specific within site. Pure spatial turnover between sites was equally habitats. For this purpose, we used the function ‘multipatt’ influential as nestedness on overall beta diversity in the R package ‘indicspecies’ (Ca´ceres and Legendre (bSOR = 0.37, bSIM = 0.19 and bNES = 0.18). Pure spatial 2009). turnover, however, was high between the habitats within sites, whereas nestedness contributed little to turnover (site

A: bSOR = 0.72, bSIM = 0.66 and bNES = 0.06; site B: Results bSOR = 0.62, bSIM = 0.55 and bNES = 0.08).

A total of 868 adult individuals, constituting 36 species and Species composition 14 families were identified within the three preselected orders: Lepidoptera (178 individuals, two families, two Visual inspection of the NMDS ordination (runs = 100, species), Coleoptera (66 individuals, six families, seven stress = 0.18) revealed segregation at the site level as species) and Araneae (628 individuals, six families, 27 arthropod composition in similar habitats differed between species). Among these was a spider species from the family sites (Fig. 4a). For plants (runs = 4, stress = 0.19), the Linyphiidae that was previously unknown from Greenland habitats at the different sites resembled each other more than [Pelecopsis mengei (Simon, 1884), 7 individuals] and a the arthropods (Online Resource 7). Results of the analysis weevil with only a few previous records from Greenland of dissimilarities (Adonis) showed that arthropod commu-

[Rutidosoma globulus (Herbst, 1795), 22 individuals] nities differed significantly between sites (F1,37 = 8,54, (Bo¨cher 2015; Marusik 2015). This latter finding was p \ 0.0001) and between the habitats at the different sites restricted to two habitats (herb slope and snow-bed) at site (F4,33 = 5.05, p \ 0.0001). Composition of arthropod A (Table 1). communities varied by site, and the analysis of predictor There was no significant difference in the means of the variables were performed at site level to identify the specific environmental predictor variables between the two sites drivers of the variation in within-site arthropod assemblages. except for percent herb cover (Online Resource 3). Soil The plant communities also differed significantly between moisture content was highly variable within sites and sites (F1,39 = 18.30, p \ 0.0001) and between the habitats vegetation height was very similar between sites (Online within site (F4,35 = 5.05, p \ 0.0001). Resource 3). Environmental correlation with composition Species diversity At site A, four significant variables were selected by vector Species richness (q = 0) increased significantly from site fitting: soil moisture, soil temperature, slope and aspect. At A to site B (t =-6.3, p \ 0.0001), but Shannon (q = 1) site B, the variables soil moisture, soil temperature, bare and Simpson (q = 2) diversity did not differ between sites ground and vegetation height were significant (Online (Fig. 2). This was exactly opposite for extrapolated plant Resource 3, Fig. 4b, c). These variables formed the basis species richness as both species richness and Shannon and for the model selection. Only soil moisture and soil tem- Simpsons diversity were significantly higher at site A perature significantly predicted species composition at both (Online Resource 4). Habitat-level species richness for sites, and at site B, vegetation height was also included in arthropods was also significantly higher at site B for herb the final model (Table 2). There were no significant slope, ridge, heath and snow-bed (p \ 0.05), while fen interactions between the variables, and interaction terms habitats did not differ significantly (Fig. 3a). Estimated were not included in the final model. plant species richness at habitat level was higher than or Species indicator analyses selected eight species as similar at site A compared to site B (Online Resource 5). significant for site A (p \ 0.05). Two species were 123 Polar Biol

Table 1 Overview of the abundance of identified arthropods and the habitats they were collected from Site A B Order Family Species Fen Ridge Herb Heath Snow Fen Ridge Herb Heath Snow

Araneae Dictynidae Dictyna major (Menge, 1869) 1 22 2 2 1 Araneae Dictynidae Emblyna borealis (O.P. Cambridge, 1877) 1 Araneae Hahniidae Hahnia glacialis (Sørensen, 1898) 4 14 Araneae Linyphiidae Agyneta nigriceps (Simon, 1884) 1 1 2 Araneae Linyphiidae Agyneta subtilis 1 Araneae Linyphiidae Ceratinella ornatula (Crosby & Bishop, 1 1925) Araneae Linyphiidae Collinsia holmgreni (Thorell, 1871) 2 1 9 Araneae Linyphiidae Collinsia spetsbergensis (Thorell, 1871) 3 Araneae Linyphiidae Erigone arctica (White, 1852) 22 3 1 31 Araneae Linyphiidae Erigone tirolensis (L. Koch, 1872) 2 Araneae Linyphiidae Erigone whymperi (O.P. Cambridge, 1877) 1 1 Araneae Linyphiidae Islandiana princeps (Brændega˚rd, 1932) 3 1 Araneae Linyphiidae Mecynargus borealis (Jackson, 1930) 4 1 1 1 2 Araneae Linyphiidae Mecynargus morulus (O.P. Cambridge, 1873) Araneae Linyphiidae Oreoneta frigida (Thorell, 1872) 2 2 Araneae Linyphiidae Pelecopsis mengei (Simon, 1884) 7 Araneae Linyphiidae Sciastes extremus (Holm, 1967) 1 1 1 Araneae Linyphiidae Semljicola obtusus (Emerton, 1914) 1 1 Araneae Linyphiidae Typhocrestus pygmaeus (Sorensen, 1898) 1 Araneae Linyphiidae Walckenaeria clavicornis (Emerton, 1882) 1 2 4 3 5 1 4 23 Araneae Lycosidae Arctosa insignita (Thorell, 1872) 64 7 11 6 9 92 3 15 36 82 Araneae Lycosidae Pardosa furcifera (Thorell, 1875) 7 2 1 2 Araneae Lycosidae Pardosa glacialis (Thorell, 1872) 1 2 1 2 Araneae Lycosidae Pardosa groenlandica (Thorell, 1872) 1 1 28 2 22 3 9 1 Araneae Lycosidae Pardosa hyperborea (Thorell, 1872) 5 Araneae Philodromidae Thanatus arcticus (Thorell, 1872) 6 1 3 1 Araneae Theridiidae Enoplognatha intrepida (Sorensen, 1898) 2 Coleoptera Byrrhidae Byrrhus fasciatus (Forster, 1771) 1 2 3 5 1 1 Coleoptera Carabidae Bembidion grapii (Gyllenhal, 1827) 8 Coleoptera Carabidae Patrobus septentrionis (Dejean, 1821) 1 Coleoptera Coccinellidae Coccinella transversoguttata (Falderman, 252 1835) Coleoptera Curculionidae Rutidosoma globulus (Herbst, 1795) 6 3 7 13 Coleoptera Dytiscidae Hydroporus morio (Aube´, 1838) 6 1 Coleoptera Staphylinidae Mycetoporus nigrans (Ma¨klin, 1853) 3 1 Lepidoptera Nymphalidae Boloria chariclea (Schneider, 1794) 10 1 1 18 Lepidoptera Pieridae Colias hecla (Lefe`bvre, 1836) 1 1 26 2 13 94 The table is sorted by order, then family and then species significantly associated with fens at site A: the water beetle 1795), was found in both snow-bed and herb slope habitats. Hydroporus morio (Aube´, 1838) and the wolf spider Par- For site B, eight species were also selected as significant dosa furcifera (Thorell, 1875). Two species associated with indicator species (p \ 0.05). The comb-tailed spider Hah- heath habitats: the wolf spider Pardosa groenlandica nia glacialis (Sørensen, 1898) and the wolf spider Pardosa (Thorell, 1872) and the ladybug Coccinella transver- hyperborea (Thorell, 1872) were associated with heath soguttata (Falderman, 1835) and three species were asso- habitats, R. globulus (Herbst, 1795) and the sheet web ciated with ridges: the mesh weaver Dictyna major spider P. mengei (Simon, 1884) were associated with herb (Menge, 1869), the philodromid spider Thanatus arcticus slopes, the sheet web spider Erigone arctica was found in (Thorell, 1872) and the ground beetle Bembidion grapii fen and snow-bed habitats, C. transversoguttata and P. (Gyllenhal, 1827). A rare weevil, R. globulus (Herbst, groenlandica were found in heath and ridge habitats and

123 Polar Biol

50 a ** fen ridge herb slope 40 snow bed heath

30

20

10 Effective species numbers NMDS2

0 Species richness Shannon diversity Simpson diversity

Fig. 2 Diversity profiles, characterized by effective number of species (± Standard Error) for Hill number order q = 0 (species richness), q = 1 (Shannon diversity) and q = 2 (Simpson diversity), at site A (white) and site B (black) extrapolated to a common base coverage of 100 %. Significance (p \ 0.05) of differences between NMDS1 sites is indicated with asterisk bc

Slope Soil temperature Soil temperature Bare ground a Aspect

60 NMDS2 NMDS2 * Soil moisture 50 Soil moisture Vegetation height 40 NMDS1 NMDS1 * * 30 Fig. 4 a Arthropod NMDS plots for site A (white) and B (black). * Each dot represents the species composition in a specific plot sampled 20 in one of five habitat types (fen, snow-bed, herb slope, exposed ridge and dry heath). Arthropod NMDS plot for site A (b) and site B 10 (c) with the significant predictors from the envfit analysis overlaid as vectors. The direction and length of each vector indicates the 0 direction of the gradient and the strength (p \ 0.05) of the correlation, Fen Ridge Herb Snow Heath respectively

b the wolf spider Arctosa insignita (Thorell, 1872) was associated with all habitats except wind-exposed ridges.

Discussion

We found clear regional-scale differences in the species richness and composition of arthropod communities at two remote sites in low Arctic Greenland with the site closest to the ice edge containing the least species. This difference was evident in most of the five habitat types sampled at the two sites and fits poorly with the common perception of the Fig. 3 a Species richness (? Standard Error) for Hill number q = 0 Arctic as a homogenous biome (CAFF 2013). These find- within habitats extrapolated to a common base coverage of 1. Site A ings correspond well with the findings of Sikes et al. and site B are represented by white and black colors, respectively. (2013), where sampling habitats adjacent to two long-term Significance (p \ 0.05) of differences between sites for each habitat experimental research sites (LTER) yielded 24 spider type is indicated with asterisk. b Venn diagrams based on percentages of species unique to site A (white), species unique to site B (black) species that were not found at LTER. Two of these were and species shared among site A and site B (gray) new state records (Sikes et al. 2013). Plant species richness

123 Polar Biol

Table 2 Summary of the non-metric multidimensional scaling the more common species (q = 1 and 2), the coverage- (NMDS) analysis based difference in diversity disappeared. Wolf spiders are Model Parameter Estimate p well adapted to open tundra habitats and are good dis- persers, so when a group (i.e., wolf spiders) is highly Site A, MDS1 Intercept 0.83 ± 0.15 \0.0001 abundant and well adapted to the prevailing habitat type, F1,16 = 33.74 Soil temperature -0.06 ± 0.01 \0.0001 they are likely to be more successful in establishing new 2 RAdj = 0.66 populations than less abundant groups of ballooning spider p \ 0.0001 species that require specific habitat features to establish Site A, MDS2 Intercept 0.62 ± 0.27 \0.0001 (Bonte et al. 2003). Wolf spiders then occur in high F2,15 = 65.54 Soil moisture -0.0053 ± 0.0017 0.009 numbers at both sites and thus dominate the diversity 2 RAdj = 0.30 Soil temperature -0.034 ± 0.016 0.05 measures. Butterflies and wolf spiders are both active dis- p = 0.027 persers, yet the butterflies are mostly restricted to site B. Site B, MDS1 Intercept -0.0013 ± 0.35 \0.0001 The active flight of butterflies allow for more controlled F3,16 = 21.17 Soil moisture 0.0058 ± 0.0022 0.02 habitat selection compared to the random ballooning of 2 RAdj = 0.76 Soil temperature 0.094 ± 0.022 \0.0001 wolf spiders. p \ 0.0001 Vegetation height 0.19 ± 0.061 0.006 At the local scale, i.e., between habitats within sites, we Site B, MDS2 Intercept 0.37 ± 0.12 0.005 found significant species turnover as well as significant

F2,28 = 11.73 Vegetation height -0.23 ± 0.067 0.003 indicator species for each habitat. This indicates that a 2 RAdj = 0.36 number of arthropod species exhibit strong habitat fidelity p = 0.003 driven by different microclimatic conditions (i.e. soil The table presents results of the linear regression models of the moisture and temperature) specific to each habitat. Our extracted NMDS ordination axes 1 and 2 from each site in separate results show that soil moisture and soil temperature are rows. Parameters are chosen based on an environmental fitting important factors in determining arthropod species patterns analysis at the local scale. Plant species composition has previously been shown to influence local arthropod species composi- tion more than soil characteristics (Schaffers et al. 2008). displayed the opposite pattern and suggests that the dif- Plant species composition is comprised of several factors, ference in arthropod communities was not due to an earlier such as soil properties, vegetation structure and solar succession stage at the site closest to the ice sheet. The radiation, and might, as effect, mask specific predictor partitioning of beta diversity suggests that some of the variables. The species found in this study are comprised of turnover between the sites is due to a subset of species at groups with a high amount of generalist predatory species. site A, but there is also some degree of pure spatial turn- Among the beetles, only a few are herbivores, and these are over, also reflected by the Venn diagram. This serves as an believed to be polyphagous (Bo¨cher 2015). The food indication that the arthropod community at site A is not preferences of the two butterfly species are unknown in solely a result of source–sink dynamics between the two Greenland, but in other regions they feed on multiple plant sites, but also different species pools. A distinct difference species in the larval stage (Karlsholt et al. 2015). We thus between the two sites is the effective land area that forms have no reason to believe that any herbivore–plant inter- the basis for species recruitment to each site. The inland ice actions cause a loss of descriptive power in the models. sheet and a large glacier reduce the vegetated area in the Furthermore, it is generally believed that plant–arthropod surroundings of site A. A peninsula effect at site A gen- associations in the Arctic are expressions of the plant- erated by the reduced immigration potential provides a specific abiotic microenvironment and not arthropod–plant likely explanation to our observed differences in turnover relationships (Coulson et al. 2003). and diversity (Simpson 1964). In other words, the sur- Shifts in precipitation regimes are inherent components of roundings at the two sites have similar abiotic conditions, climate change (Zona et al. 2009; Umbanhowar et al. 2013), but very different proportions of suitable habitat. The rel- and the resultant spatial variation in soil moisture is likely to ative abundance of species in the samples affects diversity be an underestimated factor in the Arctic (Le Roux et al. measures, particularly when the common species are 2013). Plant species compositions may not reflect sudden or habitat generalists (Chao et al. 2005). Wolf spiders are very local changes in hydrology, and studies have demon- highly abundant and make up more than 50 % of the strated how the rate of plant compositional responses to individuals in the samples. When adding equal weight to climate change depend heavily on habitat type and latitude common and uncommon species in the samples (q = 0), and responds faster at the northern latitudes (Daniels et al. we found a significantly higher species richness at site B 2011; Schmidt et al. 2012). Arctic arthropods are sensitive to than at site A, but when adding more and more weight to rapid changes in the abiotic environment at the population 123 Polar Biol

(Bowden et al. 2013, 2015; Høye et al. 2014) and community Bonte D, Baert L, Maelfait JP (2002) Spider assemblage structure and level (Høye and Forchhammer 2008a, b). Hence, changes in stability in a heterogeneous coastal dune system (Belgium). J Arachnol 30:331–343. doi:10.1890/0012-9658(2001)082[0290: the composition of Arctic ecological communities in FMMTCD]2.0.CO;2 response to climate change may be more rapidly detected in Bonte D, Vandenbroecke N, Lens L, Maelfait J-P (2003) Low arthropods than in plants. propensity for aerial dispersal in specialist spiders from The scale of observation is important for identifying fragmented landscapes. Proc R Soc B 270:1601–1607. doi:10. 2307/3592291 areas of high species diversity (Willis and Whittaker 2002), Bowden JJ, Buddle CM (2010) Determinants of ground-dwelling and the ability to detect rare species is highly dependent on spider assemblages at a regional scale in the Yukon Territory, the spatial resolution and extent of the survey (Crawley and Canada. Ecoscience 17:287–297. doi:10.2980/17-3-3308 Harral 2001; Lennon et al. 2001). In this study, we have Bowden JJ, Høye TT, Buddle CM (2013) Fecundity and sexual size dimorphism of wolf spiders (Araneae: lycosidae) along an shown that species pools of Arctic arthropods can change elevational gradient in the Arctic. Polar Biol 36:831–836. doi:10. over relatively short distances even in areas of low species 1007/s00300-013-1308-6 diversity and that local conditions can influence turnover Bowden JJ, Hansen RR, Olsen K, Høye TT (2015) Habitat-specific rates. Future studies investigate the detailed habitat effects of climate change on a low-mobility Arctic spider species. Polar Biol 38:559–568. doi:10.1007/s00300-014-1622-7 requirements of individual arthropod species, e.g., by Brind’Amour A, Boisclair D, Dray S, Legendre P (2011) Relation- determining indicator species for different habitat types. A ships between species feeding traits and environmental condi- quantification of key functional traits like dispersal ability tions in fish communities: a three-matrix approach. Ecol Appl will further aid efforts to identify areas of particular con- 21:363–377. doi:10.1890/09-2178.1 Ca´ceres MD, Legendre P (2009) Associations between species and servation concern and will facilitate a more accurate pre- groups of sites: indices and statistical inference. Ecology diction of species vulnerability to resource development 90:3566–3574. doi:10.1890/08-1823.1 and climate change. In this study, we found two species CAFF (2013) Arctic biodiversity assessment. Status and trends in with none or very few previous records from Greenland Arctic biodiversity. Akureyri: Conservation of Arctic Flora and Fauna among the 36 species identified. This highlights the limited Chao A, Jost L (2012) Coverage-based rarefaction and extrapolation: existing knowledge on the occurrence of individual species standardizing samples by completeness rather than size. Ecology of arthropods in Greenland. Our study documents that 93:2533–2547. doi:10.1890/11-1952.1 diversity patterns of arctic arthropods are more complex Chao A, Chazdon RL, Colwell RK, Shen TJ (2005) A new statistical approach for assessing similarity of species composition with than previously thought with considerable local and incidence and abundance data. Ecol Lett 8:148–159. doi:10. regional spatial variation. 1111/j.1461-0248.2004.00707.x Chao A, Gotelli NJ, Hsieh TC, Sander EL, Ma KH, Colwell RK, Acknowledgments We gratefully acknowledge logistical support Ellison AM (2014) Rarefaction and extrapolation with Hill from the Arctic Research Centre (ARC), Aarhus University. This numbers: a framework for sampling and estimation in species work is a contribution to the Arctic Science Partnership (ASP) asp- diversity studies. Ecol Monogr 84:45–67. doi:10.1890/13-0133.1 net.org. We would also like to thank the Natural History Museum, Coulson SJ, Hodkinson ID, Webb NR (2003) Microscale distribution Aarhus for use of laboratory and equipment during the identification patterns in high Arctic soil microarthropod communities: the process. TTH was supported by a Jens Christian Skou fellowship at influence of plant species within the vegetation mosaic. Ecog- the Aarhus Institute of Advanced Studies, and SN was supported by raphy 26:801–809. doi:10.2307/3683865 the Villum foundation’s Young Investigator Programme Crawley MJ, Harral JE (2001) Scale dependence in plant biodiversity. (VKR023456). Science 291:864–868. doi:10.1126/science.291.5505.864 Daniels FJA, de Molenaar JG, Chytry M, Tichy L (2011) Vegetation change in Southeast Greenland? Tasiilaq revisited after 40 years. Appl Veg Sci 14:230–241. doi:10.1111/j.1654-109X.2010.01107.x Danks HV (2004) Seasonal adaptations in Arctic insects. Integr Comp References Biol 44:85–94. doi:10.1093/icb/44.2.85 Elmendorf SC et al (2012) Plot-scale evidence of tundra vegetation Anderson MJ, Ellingsen KE, McArdle BH (2006) Multivariate change and links to recent summer warming. Nat Clim Change dispersion as a measure of beta diversity. Ecol Lett 9:683–693. 2:453–457. doi:10.1038/Nclimate1465 doi:10.1111/j.1461-0248.2006.00926.x Gaston KJ (2000) Global patterns in biodiversity. Nature Baselga A (2010) Partitioning the turnover and nestedness compo- 405:220–227. doi:10.1038/35012228 nents of beta diversity. Global Ecol Biogeogr 19:134–143. Halaj J, Ross DW, Moldenke AR (2000) Importance of habitat doi:10.1111/j.1466-8238.2009.00490.x structure to the arthropod food-web in Douglas-fir canopies. Baselga A, Orme CDL (2012) betapart: an R package for the study of Oikos 90:139–152. doi:10.1034/j.1600-0706.2000.900114.x beta diversity. Methods Ecol Evol 3:808–812. doi:10.1111/j. Hodkinson ID (2013) Terrestrial and freshwater invertebrates. In: 2041-210X.2012.00224.x Meltofte H (ed) Arctic biodiversity assessment. Status and trends Bo¨cher J (1988) The coleoptera of Greenland. Meddelelser om in Arctic biodiversity. Conservation of Arctic Flora and Fauna, Grønland, 26th edn. Bioscience, Copenhagen Akureyri, pp 195–219 Bo¨cher J (2015) Coleoptera. In: Bo¨cher J, Kristensen NP, Wilhelmsen Høye TT, Forchhammer MC (2008a) The influence of weather L (eds) The Greenland entomofauna. An identification manual of conditions on the activity of high-arctic arthropods inferred from insects, spiders and their allies, vol 44., Fauna Entomologica long-term observations. BMC Ecol 8:8. doi:10.1186/1472-6785- ScandinavicaBrill, Leiden, pp 259–292 8-8

123 Polar Biol

Høye TT, Forchhammer MC (2008b) Phenology of high-arctic Wagner H (2015) vegan: Community Ecology Package. R arthropods: effects of climate on spatial, seasonal, and inter- package version 2.2-1. http://CRAN.R-project.org/package= annual variation. Adv Ecol Res 40:299–324. doi:10.1016/S0065- vegan 2504(07)00013-X Paquin P, Dupe´rre´ N (2003) Guide d’identification des araigne´es de Høye TT, Sikes DS (2013) Arctic entomology in the 21st century. Que´bec. Association des entomologistes amateurs du Que´bec, Can Entomol 145:125–130. doi:10.4039/Tce.2013.14 Varennes, Que´be Høye TT, Eskildsen A, Hansen RR, Bowden JJ, Schmidt NM, R Development Core Team (2015) R: a language and environment for Kissling WD (2014) Phenology of high-arctic butterflies and statistical computing. R Foundation for Statistical Computing, their floral resources: species-specific responses to climate Vienna. ISBN 3-900051-07-0 change. Curr Zool 60:243–251 Rich ME, Gough L, Boelman NT (2013) Arctic arthropod assem- Hsieh TC, Ma KH, Chao A (2014) iNEXT: an R package for blages in habitats of differing shrub dominance. Ecography interpolation and extrapolation in measuring species diversity. 36:994–1003. doi:10.1111/j.1600-0587.2012.00078.x Unpublished manuscript Rushton SP, Eyre MD (1992) Grassland spider habitats in North-east Imhoff ML, Sisk TD, Milne A, Morgan G, Orr T (1997) Remotely England. J Biogeogr 19:99–108. doi:10.2307/2845623 sensed indicators of habitat heterogeneity: use of synthetic Rypstra AL (1986) Web spiders in temperate and tropical forests: aperture radar in mapping vegetation structure and bird habitat. relative abundance and environmental correlates. Am Midl Nat Remote Sens Environ 60:217–227. doi:10.1016/S0034- 115:42–51. doi:10.2307/2425835 4257(96)00116-2 Sander J, Wardell-Johnson G (2011) Fine-scale patterns of species Jimenez-Valverde A, Lobo JM (2007) Determinants of local spider and phylogenetic turnover in a global biodiversity hotspot: (Araneidae and Thomisidae) species richness on a regional scale: implications for climate change vulnerability. J Veg Sci climate and altitude vs. habitat structure. Ecol Entomol 22:766–780. doi:10.1111/j.1654-1103.2011.01293.x 32:113–122. doi:10.1111/j.1365-2311.2006.00848.x Schaffers AP, Raemakers IP, Sykora KV, Ter Braak CJF (2008) Karlsholt O, Kristensen NP, Simonsen TJ, Matti A (2015) Lepi- Arthropod assemblages are best predicted by plant species doptera. In: Bo¨cher J, Kristensen NP, Wilhelmsen L (eds) The composition. Ecology 89:782–794. doi:10.1890/07-0361.1 Greenland entomofauna. An identification manual of insects, Schmidt NM, Kristensen DK, Michelsen A, Bay C (2012) High Arctic spiders and their allies, vol 44., Fauna Entomologica Scandi- plant community responses to a decade of ambient warming. navicaBrill, Leiden, pp 324–326 Biodiversity 13:191–199. doi:10.1080/14888386.2012.712093 Kostylev VE, Erlandsson J, Ming MY, Williams GA (2005) The Sikes DS, Draney ML, Fleshman B (2013) Unexpectedly high relative importance of habitat complexity and surface area in among-habitat spider (Araneae) faunal diversity from the Arctic assessing biodiversity: fractal application on rocky shores. Ecol Long-Term Experimental Research (LTER) field station at Complex 2:272–286. doi:10.1016/j.ecocom.2005.04.002 Toolik Lake, Alaska, United States of America. Can Entomol Le Roux PC, Aalto J, Luoto M (2013) Soil moisture’s underestimated 145:219–226. doi:10.4039/tce.2013.5 role in climate change impact modelling in low-energy systems. Simpson GG (1964) Species density of North American recent Global Change Biol. doi:10.1111/Gcb.12286 mammals. Syst Zool 13:57–73. doi:10.2307/2411825 Legendre P, Legendre L (1998) Numerical ecology. Elsevier, Suvanto S, Le Roux PC, Luoto M (2014) Arctic-alpine vegetation Amsterdam biomass is driven by fine-scale abiotic heterogeneity. Geogra- Lennon JJ, Koleff P, Greenwood JJD, Gaston KJ (2001) The fiska Annaler Series a-Physical Geography 96:549–560. doi:10. geographical structure of British bird distributions: diversity, 1111/Geoa.12050 spatial turnover and scale. J Anim Ecol 70:966–979. doi:10. Tews J, Brose U, Grimm V, Tielbo¨rger K, Wichmann MC, Schwager 1046/j.0021-8790.2001.00563.x M, Jeltsch F (2004) Animal species diversity driven by habitat Lindroth CH (1985) The Carabidae (Coleoptera) of Fennoscandia and heterogeneity/diversity: the importance of keystone structures. Denmark, vol 1. Brill, Leiden J Biogeogr 31:79–92. doi:10.1046/j.0305-0270.2003.00994.x Lindroth CH (1986) The Carabidae (Coleoptera) of Fennoscandia and Umbanhowar C et al (2013) Contrasting changes in surface waters Denmark, vol 2. Brill, Leiden and barrens over the past 60 years for a subarctic forest-tundra Marusik YM (2015) Araneae. In: Bo¨cher J, Kristensen NP, Wilhelm- site in northern Manitoba based on remote sensing imagery. Can sen L (eds) The Greenland entomofauna. An identification J Earth Sci 50:967–977. doi:10.1139/cjes-2012-0162 manual of insects, spiders and their allies, vol 44., Fauna Venn S, Kotze DJ (2014) Benign neglect enhances urban habitat Entomologica ScandinavicaBrill, Leiden, pp 666–703 heterogeneity: responses of vegetation and carabid beetles McArdle BH, Anderson MJ (2001) Fitting multivariate models to (Coleoptera: carabidae) to the cessation of mowing of park community data: a comment on distance-based redundancy lawns. Eur J Entomol 111:703–714. doi:10.14411/eje.2014.089 analysis. Ecology 82:290–297. doi:10.1890/0012-9658(2001) Willis KJ, Whittaker RJ (2002) Ecology—species diversity—scale 082[0290:Fmmtcd]2.0.Co;2 matters. Science 295:1245–1248. doi:10.1126/science.1067335 Muenchow J, Feilhauer H, Brauning A, Rodriguez EF, Bayer F, Wise DH (1993) Spiders in ecological webs, 16.5th edn. Cambridge Rodriguez RA, von Wehrden H (2013) Coupling ordination University Press, Cambridge. doi:10.1017/CBO9780511623431 techniques and GAM to spatially predict vegetation assemblages World Spider Catalog (2016) Natural History Museum Bern. http:// along a climatic gradient in an ENSO-affected region of wsc.nmbe.ch. Accessed 1 Feb 2014 extremely high climate variability. J Veg Sci 24:1154–1166. Zona D et al (2009) Methane fluxes during the initiation of a large- doi:10.1111/Jvs.12038 scale water table manipulation experiment in the Alaskan Arctic Myers-Smith IH et al (2011) Shrub expansion in tundra ecosystems: tundra. Global Biogeochem Cy. doi:10.1029/2009gb003487 dynamics, impacts and research priorities. Environ Res Lett. doi:10.1088/1748-9326/6/4/045509 Oksanen J, Guillaume Blanchet F, Kindt R, Legendre P, Minchin PR, O’Hara RB, Simpson GL, Solymos P, Henry M, Stevens H,

123 Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities

Rikke Reisner Hansen1,2, Oskar Liset Pryds Hansen1,2, Joseph J. Bowden1, Urs A. Treier1,3,4, Signe Normand1,3,4 and Toke Høye1,2,5 1 Arctic Research Centre, Department of Bioscience, Aarhus University, Aarhus C, Denmark 2 Department of Bioscience, Kalø, Aarhus University, Rønde, Denmark 3 Ecoinformatics & Biodiversity, Department of Bioscience, Aarhus University, Aarhus, Denmark 4 Swiss Federal Institute for Forest, Snow and Landscape Research, Birmensdorf, Switzerland 5 Aarhus Institute of Advanced Studies, Aarhus University, Aarhus, Denmark

ABSTRACT The Arctic is warming at twice the rate of the rest of the world. This impacts Arctic species both directly, through increased temperatures, and indirectly, through structural changes in their habitats. Species are expected to exhibit idiosyncratic responses to structural change, which calls for detailed investigations at the species and community level. Here, we investigate how arthropod assemblages of spiders and beetles respond to variation in habitat structure at small spatial scales. We sampled transitions in shrub dominance and soil moisture between three different habitats (fen, dwarf shrub heath, and tall shrub tundra) at three different sites along a fjord gradient in southwest Greenland, using yellow pitfall cups. We identified 2,547 individuals belonging to 47 species. We used species richness estimation, indicator species analysis and latent variable modeling to examine differences in arthropod community structure in response to habitat variation at local (within site) and regional scales (between sites). We estimated species responses to the environment by fitting species-specific generalized linear models with Submitted 5May2016 environmental covariates. Species assemblages were segregated at the habitat and site Accepted 15 June 2016 level. Each habitat hosted significant indicator species, and species richness and Published 14 July 2016 diversity were significantly lower in fen habitats. Assemblage patterns were Corresponding author significantly linked to changes in soil moisture and vegetation height, as well as Rikke Reisner Hansen, geographic location. We show that meter-scale variation among habitats affects [email protected] arthropod community structure, supporting the notion that the Arctic tundra is a Academic editor heterogeneous environment. To gain sufficient insight into temporal biodiversity Dezene Huber change, we require studies of species distributions detailing species habitat Additional Information and preferences. Declarations can be found on page 13 DOI 10.7717/peerj.2224 Subjects Biodiversity, Biogeography, Conservation Biology, Ecology, Entomology Copyright Keywords Coleoptera, Araneae, Environmental gradients, Biodiversity, Habitat suitability 2016 Hansen et al. Distributed under INTRODUCTION Creative Commons CC-BY 4.0 Understanding the factors that structure ecological communities on continental, regional and local scales provides the basis for understanding how global changes might affect

How to cite this article Hansen et al. (2016), Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities. PeerJ 4:e2224; DOI 10.7717/peerj.2224 species composition and biodiversity (Vellend et al., 2013; Dornelas et al., 2014). Climate change is happening at an accelerated pace in the Arctic (Callaghan et al., 2004; IPCC, 2014), and altered moisture regimes and shrub expansion are two of the most prominent habitat-altering phenomena caused by these changes (Rouse et al., 1997; Tape, Sturm & Racine, 2006; Myers-Smith et al., 2011; Elmendorf et al., 2012). Shrub expansion and altered moisture regimes represent considerable consequences of climate change to the Arctic tundra; altering unique habitats such as open heath, wetlands and grasslands (ACIA, 2004). Firstly, warming in the Arctic has led to accelerated plant growth, particular for woody plants, causing a shift towards greater shrub cover, and a northward migration of the tree line (Callaghan, Tweedie & Webber, 2011), increased biomass (Epstein et al., 2012), and changes in plant species composition (Walker et al., 2012). These trends are expected to continue during future climate change (Normand et al., 2013; Pearson et al., 2013). Secondly, a changing Arctic climate with changes in precipitation, glacial melt, and permafrost degradation may alter the spatial extent of wetlands (Avis, Weaver & Meissner, 2011). In areas with continuous permafrost, top soils become wetter due to the impermeable strata that prevent infiltration and percolation (Woo & Young, 2006). Some areas with discontinuous permafrost, however, become drier, due to increased net evapotranspiration and increased drainage due to permafrost thaw (Zona et al., 2009; Perreault et al., 2015). The long term persistence of Arctic wetlands is debated, but climate change projections and field studies indicate deterioration and ultimate destruction of Arctic wetlands (Woo & Young, 2006). These habitat changes, both shrubification and wetland deterioration, will trigger several feedback loops within the climate system (Chapin et al., 2005) and may have profound effects on ecosystems (Post et al., 2009). In order to understand how these habitat changes, affect Arctic biodiversity, we need to adequately describe how Arctic species composition responds to environmental changes. The alteration of habitats, due to e.g., shrub expansion into open tundra and changing wetland hydrology, are likely to affect habitat availability for many organisms, through changes in species’ distributions, diversity, and composition. Terrestrial arthropods (e.g. insects and spiders) in particular, are associated with specific habitat types and likely respond strongly to habitat changes in the Arctic (Bowden & Buddle, 2010; Rich, Gough & Boelman, 2013). Arthropods have long been recognized as valuable indicators of changing environments because of their relatively short lifecycles and their physiology being directly driven by the external environment (ectothermic). Studies of the impacts of habitat changes upon Arctic arthropod communities are, however, only beginning to emerge (Bowden & Buddle, 2010; Rich, Gough & Boelman, 2013; Sikes, Draney & Fleshman, 2013; Sweet et al., 2014; Hansen et al., 2016). In spite of the common conception of the Arctic as a species-poor and relatively homogenous environment, studies have shown that arthropod assemblages vary substantially over short distances (Hansen et al., 2016), with species responding to local and regional climatic gradients (O. L. P. Hanser et al., 2013, unpublished data). Arthropod communities are expected to change in response to the direct effects of increasing temperatures and prolonged growing seasons (Høye et al., 2013; Høye et al., 2014), but also indirectly through changes in soil moisture and vegetation

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 2/18 structure (Bowden & Buddle, 2010; Hansen et al., 2016), changes to snowmelt dynamics (Høye et al., 2009; Bowden et al., 2015b), and shrub expansion (Rich, Gough & Boelman, 2013). Several studies indicate direct effects of temperature change on arthropods (Post et al., 2009; Høye et al., 2013; Bowden et al., 2015a), but we do not yet fully comprehend the distribution of, or habitat requirements for, the majority of Arctic arthropod species. Arctic and alpine tundra areas are vast, and the knowledge of geographical variation associated with recent environmental and ecosystem change is limited. In this study, we explore the influence of moisture regime and habitat structure on the composition and diversity of Arctic arthropod communities, and investigate the site specific effects of the drivers of change. We propose the following hypothesis: Arctic arthropod assemblages and diversity vary with soil moisture and vegetation height at very small spatial scales (10–20 m). Specifically, we compare beetle and spider communities sampled in different habitats (fen, dwarf shrub heath, and tall shrub tundra) at three sites along a large scale gradient. We expect to find distinct arthropod communities in each habitat, and that abundances of groups like wolf spiders, and other active hunters, will be lower in the tall shrub tundra compared to open habitats. METHODS Study area and sampling design Arthropods were sampled with uncovered pitfall traps from the 29th of June to the 23rd of July 2013 at three sites (1, 2, and 3) along the Godthaabsfjord in West Greenland (Fig. 1). Site 1 was situated at the mouth of the fjord and thus characterized by a coastal climate with relatively high precipitation, narrow annual temperature range, and topographic variation (app. 0–300 m.a.s.l.). The shrub community at site 1 was dominated by dwarf shrubs and a very sparse cover of tall shrub species such as Salix glauca (Lange (family: Salicidae)). Site 2 was low lying and flat, and characterized by a mosaic of low shrub vegetation (< 50 cm), dominated by S. glauca, mixed with Betula nana (Lange (family: Betulacae)), Vaccinium uliginosum (L., (family: Ericacae)), Rhododendron groenlandicum (Oeder (family: Ericacae)), and Empetrum nigrum (Lange (family: Ericacae)). Site 3 was characterized by a continental climate and pronounced topographic variation (app. 0–600 m.a.s.l.) with well-defined tall shrub patches dominated by high growth of S. glauca and Alnus crispa (Aiton (family: Betulacae)) (> 50 cm). These patches were mainly located at south facing slopes below 100 m.a.s.l. All dwarf shrub species at site 2 were also present at site 3. Moisture transitions (fen-heath) were sampled at sites 1 and 2, while transitions in vegetation height and cover of tall shrubs (heath-shrub) were sampled at sites 2 and 3. Four fen-heath plots were established, two at site 1 and two at site 2. Each fen-heath plot consisted of two sub-plots placed 10 m apart. Each sub-plot was situated exactly 5 m from a distinct fen-heath transition zone (Fig. 2). Twelve heath-shrub plots were established at site 2 and site 3 (six at each site). Each heath-shrub plot consisted of two sub-plots 20 m apart; one located at the center of a patch of tall shrubs and one in the adjacent open dwarf shrub heath. Each sub-plot was delineated by a circle with a 5 m radius. At the center of each sub-plot, two yellow pitfall traps (9 cm diameter) were placed 50 cm

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 3/18 A

Site 3

Site 2

Site 1 B

Nuuk

0 50 Km

Figure 1 Map of the study area. (A) The Godtha˚bsfjord area, South-West Greenland (6411′N, 5144′W), showing the three study sites (1, 2 and 3) depicted with a circle and the capital Nuuk depicted with a diamond. (B) Greenland with the study area framed in a square (Loecher & Ropkins, 2015). Mapdata: © Google 2016.

apart (Fig. 2). The traps were dug down such that the rim was flush with the surface and filled one third with a soap water solution. There was no overflow due to rainwater accumulation during sampling. The color of the pitfalls was chosen to catch flying as well as surface-active arthropods (Høye et al., 2014). Pitfall traps were emptied twice, once halfway through and once at the end of the sampling period. Samples were stored separately. The following structural and environmental parameters were measured in each sub-plot: (i) percent cover of shrubs, herbs, graminoids and bare ground in six categories: 0, 1–20, 21–40, 41–60, 61–80, and 81–100%, (ii) height (to the nearest 5 cm) of the vegetation height with the highest coverage in the sub-plot, (iii) presence of plant species, (iv) slope in vertical meters between the highest and lowest point of the sub-plot, (v) aspect, recorded using a handheld GPS and classified to nearest cardinal direction (North, South, East, and West), (vi) pH, measured directly with a soil pH measurement kit, model HI 99121, (vii) soil type at 15 cm depth was recorded as humus or sand.

Specimens and data All spiders and beetles were sorted from the samples and the adult specimens were identified (by RRH) to species based on morphological characters using a Wild M5A stereo microscope. Not all juveniles could be assigned to species, so only adult specimens were included in the analysis. Spiders were identified using the available literature through The World Spider Catalog (2016) and Spiders of North America (Paquin & Dupe´rre´, 2003).

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 4/18 A Fen B Shrub

10 m 20 m

5m radius 5m radius

Heath Heath

Figure 2 Sampling design. Conceptual figure of the sampling design showing fen transects in the (A) and shrub transects in the (B).

Beetles were identified using both Scandinavian and North American literature (Lindroth, 1985; Lindroth, 1986; Bo¨cher, 1988) and by consulting the collection at the Natural History Museum Aarhus, Denmark. Specimens are preserved in 75% ethanol at the Natural History Museum Aarhus. The dataset is available through the Global Biodiversity Information Facility (http://doi.org/10.15468/li6jkm).

Data analysis The mean and standard error were calculated for significant environmental variables across all habitats at each site. We ran a correlation analysis of all potential variables, based on Pearson’s correlation coefficient, and tested whether the uncorrelated variables differed significantly between sites and habitats with a MANOVA. To counteract effects of potential-under sampling, all analyses were carried out excluding singletons. All analyses were carried out in R version 3.2.2.

Species diversity Species diversity was rarefied and extrapolated for investigation across habitats based on Hill numbers (q = 0; species richness, q = 1; Shannon diversity, q = 2; Simpson diversity) and standardized by sample coverage (Chao & Jost, 2012; Chao et al., 2014) using the iNEXT R-package (TC Hsieh, KH Ma & A Chao, 2014. unpublished data). We extrapolated to double the reference sample of the habitat with the smallest sample coverage (shrub).

Samples were compared at base-coverage, estimated as a minimum of Ca and Cb, where

Ca is maximum coverage at reference sample size and Cb is minimum coverage at two times reference sample size. iNEXT computes bootstrap confidence bands around the sampling curves, facilitating the comparisons of diversity across multiple assemblages. We then visually assessed if diversity measures differed significantly between habitats. We ran a species indicator analysis to assess the strength and statistical significance of the relationship between species abundances and the specific habitats. We used the function ‘multipatt’ in the R package ‘indicspecies’ (De Ca´ceres, Legendre & Moretti, 2010). This analysis provides a specificity value ‘A’(0–1), which indicates the probability of a certain species occurring in a certain habitat as well as a sensitivity value ‘B’(0–1), which

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 5/18 indicates how many of the plots belonging to a certain habitat the target species is located in. Significance (p < 0.05) is assessed based on the A and B values (De Ca´ceres & Legendre, 2009). In addition to significance testing, we opted to describe habitat preferences more broadly by assigning all species with an A value for a given habitat larger than 0.8 and a B value larger than 0.1 to that specific habitat. In this way, the importance of the sensitivity value is down weighted.

Species composition Traditional methods to visually investigate how arthropod species composition varies between habitats, such as non-metric multidimensional scaling (NMDS), have been shown to confound trends in location with changes in dispersion, leading to potentially misleading results (Warton, Wright & Wang, 2012). To avoid these issues while still enabling visualization, we employed latent variable modelling through the R package ‘boral’ (Hui, 2016). Latent variable modelling is a Bayesian model-based approach that explains community composition through a set of underlying latent variables to account for residual correlation, for example due to biotic interaction. This method offers the possibility to adjust the distribution family to e.g., negative binomial distribution which better accounts for over-dispersion in count data. Thus, it accounts for the increasing mean-variance relationship without confounding location with dispersion (Hui et al., 2015). Three “types” of models may be fitted: (1) With covariates and no latent variables, boral fits independent response General Linear Models (GLMs) such that the columns of y are assumed to be independent; (2) With no covariates, boral fits a pure latent variable model (Rabe-Hesketh, Skrondal & Pickles, 2004) to perform model-based unconstrained ordination (Hui et al., 2015); (3) With covariates and latent variables, boral fits correlated response GLMs, with latent variables. Sub-plots placed in dwarf shrub heath could potentially differ depending on the transition examined (fen-heath or shrub-heath). Therefore, we created latent variable plots for both plants and arthropods to visually assess if the heath sub-plots in the fen- heath and shrub-heath plots groups were distinguishable. In the latent variable plot for plant species composition, heath sub-plots were not segregated (Fig. S1) and all heath plots were hereafter treated as one category. We modelled arthropod species’ distributions with two latent variables to enable visualization comparable to a two dimensional NMDS. From the latent variable model, we extracted the posterior median values of the latent variables which we used as coordinates on ordination axes to represent species composition at plot level (Hui et al., 2015). We then tested the difference in local species composition based on these coordinates between the paired samples (fen-heath or shrub-heath) for each transect using paired t-tests. To test the significance of and, interactions between, the environmental variables and site, we used a multivariate extension of GLMs using the function ‘manyglm’ in the package ‘mvabund’ (Wang et al., 2012). This recently developed method offers the possibility to model distributions based on count data by assuming a negative binomial distribution. Vegetation height and graminoid cover have higher resolutions compared to

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 6/18 the classifications ‘fen’ and ‘shrub’ as these are measured on a continuous scale. We used vegetation height as a proxy for shrub treatment effects and cover of graminoids as a proxy for soil moisture. The gradients in these variables are representative of the moisture transition of fen-heath plot groups and the shrub dominance transition of the shrub- heath plot groups (Fig. S2). We tested for main effects of all the un-correlated variables, selected by the Pearson correlation analyses, and for an interaction between the variables and site. The model assumptions of mean-variance and log-linearity were examined with residual vs. fit plots and a normal quantile plot, and no transformations were needed.

RESULTS A total of 2,547 individuals, constituting 45 species and 13 families were identified within the two orders: Araneae (2,223 individuals, seven families, 37 species) and Coleoptera (324 individuals, 6 families, 8 species). We found a species of sheet web spider (Wabasso cacuminatus (Millidge, 1984)) not previously known from Greenland, represented by one individual. One species (Pelecopsis mengei, (Simon, 1884)), represented in our samples by three individuals, remained unknown from Greenland until recently (Marusik, 2015; Hansen et al., 2016)(Table 1). Extrapolated species richness (q = 0) did not differ significantly between habitats due to overlapping confidence intervals but there was a trend towards higher species richness in heath sub-plots, lower in shrub sub-plots, and lowest in fen sub-plots (Fig. 3). The same pattern was observed for Shannon diversity (q = 1) as well as for Simpson diversity (q = 2), however both these indices differed significantly between habitats, with the highest diversity in the shrub sub-plots, intermediate in the heath sub-plots, and lowest in the fen sub-plots (Fig. 3). The three species significantly (p < 0.05) associated with fen habitats were all sheet web spiders. Erigone whymperi (O.P. Cambridge, 1877), Mecynargus paetulus (O.P. Cambridge, 1875), and Wabasso quaestio (Chamberlin, 1948). Just one species, the ladybird Coccinella transversoguttata (Faldermann, 1835), was significantly associated with heath habitats. Shrub habitats housed six significantly associated species: the comb-footed spider Ohlertidion lundbecki (Sørensen, 1894), and five species of sheet web spiders: Dismodicus decemoculatus (Emerton, 1852), Improphantes complicates (Emerton, 1882), Pocadicnemis americana (Millidge, 1984), Semljicola obtusus (Emerton, 1914), Sisicus apertus (Holm, 1939) (Table 1). The latent variable plots showed that the plant species composition of the shrub sub- plots overlapped with the composition of the heath plots (Fig. S1), but vegetation height was significantly different (Table 2). The plant species composition of the fen plots was different from both the heath and shrub sub-plots (Fig. S1). Arthropod species composition was segregated both at site and habitat level, but the distribution of sub-plots in the latent variable arthropod plot indicated interaction between site and treatment (Fig. 4). Vegetation height in the shrub sub-plots at site 2 was significantly lower than that at

site 3 (F1,21 = 13.46, p = 0.001), and the overall cover of graminoids was significantly lower

at the fen sub-plots at site 1, compared to the fen sub-plots at site 2 (F1,29 = 0.21,

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 7/18 Table 1 Arthropod species. List of arthropod species sampled and their abundance in three habitats; fen, dwarf shrub heath, and tall shrub tundra at three sites along the Nuuk fiord in Western Greenland. The last column shows the results of a species indicator analysis (for details see main text). Species were assigned to one of the three habitats if A (specificity value) > 0.8 and B (sensitivity value) > 0.1. The table is sorted by order, family, and species, respectively. Order Family Species Abundance Habitat

Fen Heath Shrub Araneae Dictynidae Dictyna major (Menge, 1869) 1 No classification Gnaphosidae Haplodrassus signifer (C.L. Koch, 1839) 1 No classification Hahniidae Hahnia glacialis (Sørensen, 1898) 1 7 1 No classification Linyphiidae Agyneta jacksoni (Simon, 1884) 3 8 1 No classification Agyneta nigripes (Brændega˚rd, 1937) 2 3 Fen and heath Bathyphantes simillimus (L. Koch, 1879) 1 No classification Dismodicus decemoculatus (Emerton, 1852) 1 2 10 Shrub* Erigone arctica (White, 1852) 6 Fen Erigone psycrophila (Thorell, 1871) 1 No classification Erigone whymperi (O.P. Cambridge, 1877) 8 Fen* Hilaira herniosa (Thorell, 1875) 1 No classification Hybauchenidium gibbosum (Sørensen, 1898) 5 3 Heath and shrub Hypsosinga groenlandica (Simon, 1889) 2 2 4 Heath and shrub Improphantes complicatus (Emerton, 1882) 2 8 Shrub* Lepthyphantes turbatrix (O.P. Cambridge, 1877) 1 No classification Mecynargus borealis (Jackson, 1930) 4 Heath Mecynargus morulus (O.P. Cambridge, 1873) 2 1 Heath and shrub Mecynargus paetulus (O.P. Cambridge, 1875) 33 Fen* Oreonetides vaginatus (Thorell, 1872) 1 No classification Pelecopsis mengei (Simon, 1884) 2 1 Heath and shrub Pocadicnemis americana (Millidge, 1976) 6 18 Shrub* Sciastes extremus (Holm, 1967) 1 No classification Scotinotylus sacer (Crosby, 1929) 5 Shrub Semljicola obtusus (Emerton, 1914) 3 6 15 Shrub* Sisicus apertus (Holm, 1939) 1 3 Shrub* Tiso aestivus (L. Koch, 1872) 1 31 1 Heath Wabasso cacuminatus (Millidge, 1984) 1 No classification Wabasso quaestio (Chamberlin, 1948) 12 Fen* Walckenaeria karpinskii (O.P. Cambridge, 1873) 6 21 Fen and heath* Thomisidae Xysticus durus (Sørensen, 1898) 17 Heath Lycosidae Arctosa insignita (Thorell, 1872) 17 29 2 Fen and heath* Pardosa furcifera (Thorell, 1875) 524 552 257 No classification Pardosa groenlandica (Thorell, 1872) 17 23 8 No classification Pardosa hyperborea (Thorell, 1872) 6 347 140 Heath and shrub* Philodromidae Thanatus arcticus (Thorell, 1872) 2 10 Fen and heath Theridiidae Robertus fuscus (Emerton, 1894) 1 No classification Ohlertidion lundbecki (Sørensen, 1898) 2 Shrub

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 8/18 Table 1 (continued). Order Family Species Abundance Habitat

Fen Heath Shrub Coleoptera Byrrhidae Byrrhus fasciatus (Forster, 1771) 1 11 Heath Carabidae Patrobus septentrionis (Dejean, 1821) 50 17 23 Fen and shrub* Coccinellidae Coccinella transversoguttata (Falderman, 1835) 51 2 Heath* Cryptophagidae Caenoscelis ferruginea (Sahlberg, 1820) 38 2 Heath and shrub Curculionidae Otiorynchus arcticus (O. Fabricius, 1780) 1 20 1 Heath Otiorynchus nodosus (Mu¨ller, 1764) 18 66 19 No classification Staphylinidae Mycetoporus nigrans (Ma¨klin, 1853) 2 No classification Quedius fellmanni (Zetterstedt, 1838) 2 No classification Note: *Indicates Significance, p < 0.05.

Shannon diversity Simpson diversity Species richness

4.0 7 30 3.5 6 Habitat 3.0 Fen 25 Heath 5 Shrub Estimates 2.5

4 20 2.0

3

Figure 3 Diversity profiles. Species richness, Shannon diversity and Simpson diversity coloured by habitat. Error bars represent 95 percent confidence intervals.

Table 2 Environmental variables. Mean (± S.E) of the environmental variables included in GLM’s and latent variable models, showing the difference between sites and treatments. Graminoid cover was measured in six categories: 0, 1–20, 21–40, 41–60, 61–80, and 81–100%. Vegetation height was measured (classified to the nearest 5 cm) of the vegetation height with the highest coverage in the sub-plot. Site Habitat Vegetation height Graminoid pH (height classes) (% cover) Site 1 Heath 2.6 (0.2) 15 (5) 5.8 (0.1) Fen 2.5 (0.2) 55 (6.3) 5.5 (0.2) Site 2 Heath 2.4 (0.2) 18.6 (3.7) 6.4 (0.1) Fen 2.3 (0.3) 75 (6.3) 6.5 (0.04) Shrub 7.5 (1.2) 10.3 (3.5) 6.2 (0.3) Site 3 Heath 3.2 (0.4) 12.7 (11.4) 6.2 (0.2) Shrub 28.5 (4.1) 4 (1.9) 6.5 (0.04)

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 9/18 Shrub LV2 Heath

Fen Site 1 Site 2 Site 3 LV1

Figure 4 Latent variable plot for arthropod species composition. Species distribution plot of the best fitted latent variable model showing the mean of the latent variable with a negative binomial dis- tribution. Ellipses represent 95 percent confidence intervals around the centroids of each habitat.

Table 3 Table of deviance. Results of the multivariate generalized linear model, including all variables tested, along with residual degrees of freedom, degrees of freedom, deviance and p-value. Parameter Residual degrees Degrees of Deviance p-value of freedom freedom Intercept 55 Vegetation height (height class) 54 1 117.9 0.001 Graminoid cover (% cover) 53 1 93.2 0.001 pH 52 1 43.0 0.389 Site 50 2 296.2 0.001 Vegetation height: site 48 2 35.0 0.639 Graminoid cover: site 46 2 103.8 0.003 pH: site 44 2 40.8 0.568

p = 0.049). The pH levels were nearly significantly different between the shrub sub-plots

from site 2 to site 3 (F1,21 = 4.17, p = 0.054) and highly significantly different between

the fen sub-plots from site 1 to site 2 (F1,29 = 66.01, p < 0.0001). pH did not differ between

fen and heath (F1,29 = 0.69, p = 0.41) or shrub and heath (F1,21 = 0.50, p = 0.49). Cover of graminoids was significantly lower for heath sub-plots compared to fen sub-plots

(F1,29 = 74.54, p < 0.0001). Vegetation height differed significantly between shrub and

heath habitats (F1,21 = 26.04, p < 0.0001), with lower vegetation height in the heath sub-plots compared to shrub sub-plots (Table 2; Fig. S2). Arthropod species composition differed significantly due to different moisture regimes and different height classes. pH levels were not a significant driver of arthropod communities, nor was there a significant interaction between site and the levels of pH.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 10/18 Table 4 T-test of local transitions. Paired t-test of the local transitions in soil moisture and shrub dominance. LV1 and LV2 represent the first and second coordinate of the latent variable. Model Residual degrees Estimates tp of freedom Fen transect site 1 LV1 7 -0.86 -5.32 0.001 Fen transect site 1 LV2 7 -0.43 -4.78 0.002 Fen transect site 2 LV1 7 -1.70 -0.26 0.13 Fen transect site 2 LV2 7 -0.37 -3.21 0.02 Shrub transect site 2 LV1 5 -0.72 -3.90 0.01 Shrub transect site 2 LV2 5 -0.35 -3.10 0.03 Shrub transect site 3 LV1 5 -1.16 -5.50 0.003 Shrub transect site 3 LV2 5 -0.62 -3.28 0.02

There was a significant interaction between cover of graminoid species and site, but no significant interaction was detected between height class and site (Table 3). Arthropod species composition differed significantly between the local fen-heath transitions, but for site 2 only; one latent variable axis differed significantly between fen-heath transitions. The local shrub-heath transects differed significantly for both axes and both sites (Table 4).

DISCUSSION Although Arctic tundra is often perceived as a relative homogenous biome, it consists of a wide range of habitat types due to strong environmental transitions occurring over short spatial scales. In this study, we have demonstrated clear effects of vegetation height and soil moisture on diversity and composition of spiders and beetles in low Arctic Greenland. This effect is evident across distances of 10–20 m. Fen, heath, and shrub vegetation hosted distinct arthropod communities differing in both composition and diversity. While previous studies have emphasized the importance of vegetation structure as predictors of Arctic arthropod communities (Bowden & Buddle, 2010; Rich, Gough & Boelman, 2013; Sweet et al., 2014), it has not been demonstrated that such effects are visible at the scale of meters. Existing literature generally agrees with the habitat classifications we have assigned the species in this study. According to existing descriptions of habitat preferences, the wetland species we find in this study are found strictly in wet open habitats, whereas both shrub and heathland species mostly have a more general distribution (Leech, University of Alberta & Department of Entomology, 1966; Bo¨cher, 2015; Marusik, 2015), indicating a higher degree of habitat specialization in the fens. The sheet web spider, Erigone arctica, was significantly linked to wet fen habitats in an alpine study site in West Greenland (Hansen et al., 2016), and in this study, E. arctica was also linked to fen plots, further suggesting habitat specialization. We found the lowest diversity in the fens, which are spatially limited, compared to much more wide spread heathland habitats. Both tall shrub tundra and dwarf shrub heath are comprised of different habitats with open patches, moist areas, and varying vegetation structure. Such variation in habitat structure likely leads to higher diversity compared to the fen habitats, which are rather homogenous.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 11/18 This particular study area is characterized as low Arctic with discontinuous permafrost unaffected by glacial meltwater. Models predict that this area will experience increased evapotranspiration and precipitation (Rawlins et al., 2010). Increased drainage due to permafrost melt coupled with evapotranspiration is likely to lead to wetland deterioration. Shrubification has been forecasted to be most pronounced at the boundary between high and low Arctic where permafrost is melting and in areas where soil moisture is greatest (Myers-Smith et al., 2015). In the Godtha˚bsfjord, it is therefore likely that shrub expansion will be most notable in the fens and snow-beds. With shrubification (Myers-Smith et al., 2011; Elmendorf et al., 2012), as well as, increased land use such as forestry and agriculture (ACIA, 2004), wetland habitats are at risk (CAFF, 2013). Our results suggest that wetland deterioration and shrubification will strongly affect arthropod communities, and may compromise the living conditions of individual specialized species. We found an interaction between site and graminoid cover, suggesting that the fens differ between sites. Wetlands with coastal proximity are known to be impacted by salt influx from the sea (Woo & Young, 2006). This is supported by the slightly higher pH at site 1 compared to site 2, but does not explain differences in arthropod composition in the fens between the coastal (site 1) and intermediate site (site 2), as pH was not significant in the multivariate GLM. Even though plant species composition showed clear segregation of wet and dry plots, conditions may be drier at the intermediate site than at the coastal site, where summer precipitation is higher. Plant species composition reflects an integration of seasonal variation in soil moisture conditions (Daniels et al., 2011) such that they may not reflect sudden soil moisture changes. The variation in moisture regime only partially explained arthropod species composition at the intermediate site and supports the idea of drier conditions at the intermediate site affecting arthropod species composition differently. We expected the effect of vegetation height to be less pronounced at the intermediate site due to the patchy structure of the shrubs and overall lower vegetation; yet, we did not find an interaction between site and treatment. We studied mostly mobile predatory species. The few herbivores like the weevils: Otiorynchus arcticus (O. Fabricius, 1780) and Otiorynchus nodosus (Mu¨ller, 1764) are mostly found in open heath plots with low vegetation. It is conceivable that even a small change in vegetation height has an effect on the surface active predatory species, because vegetation height may also affect the composition and abundance of prey items. The web builders, like sheet web spiders, require some amount of vegetation structure to form webs, but even low shrubs provide structure and shelter. Rich, Gough & Boelman (2013) found that overall arthropod abundance and species richness increased in shrub plots in Arctic Alaska, but suggested that for groups like wolf spiders and other active hunters, full shrub encroachment of open habitats could be detrimental. Our results show that abundances of these groups are lower in shrub sub-plots and support this notion.

CONCLUSION We have established a baseline of species occurrence in relation to transition in soil moisture and shrub dominance which will facilitate future assessment of changes in Arctic

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 12/18 arthropod communities, where these transitions in habitat structure are likely to change. The variation in community composition at the scale of meters was surprising and suggests drastic changes in arthropod species composition given continued shrubification and wetland deterioration. We found that the strength of the environmental predictor variables varied among sites. Understanding the sources of such site variation is an important topic for future studies. Two important steps are needed to further the knowledge of arthropod responses to changing habitats. Primarily, we need information on species occurrence across multiple taxa and multiple environmental gradients. Habitat preferences of species are needed to determine the effects that climate change will have in Arctic ecosystems. Spiders and butterflies have proven useful for detection of rapid environmental change due to climate change (Høye et al., 2014; Bowden et al., 2015a; Bowden et al., 2015b) and may serve as important indicator taxa in future studies. Secondly, we need further studies of spatial variability and change in environmental gradients like soil moisture.

ACKNOWLEDGEMENTS This work is a contribution to the Arctic Science Partnership (ASP: http://asp-net.org). We would also like to thank the Natural History Museum Aarhus for use of laboratory and equipment during the identification process.

ADDITIONAL INFORMATION AND DECLARATIONS

Funding Toke Thomas Høye was supported by a Jens Christian Skou fellowship at the Aarhus Institute of Advanced Studies and Signe Normand was supported by the Villum foundation’s Young Investigator Programme (VKR023456). Logistical support was provided by the Arctic Research Centre (ARC), Aarhus University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Grant Disclosures The following grant information was disclosed by the authors: Jens Christian Skou Fellowship at the Aarhus Institute of Advanced Studies. Villum foundation’s Young Investigator Programme: VKR023456. Arctic Research Centre (ARC), Aarhus University.

Competing Interests The authors declare there are no competing interests.

Author Contributions  Rikke Reisner Hansen conceived and designed the experiments, performed the experiments, analyzed the data, wrote the paper, prepared figures and/or tables, reviewed drafts of the paper.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 13/18  Oskar Liset Pryds Hansen performed the experiments, contributed reagents/materials/ analysis tools, reviewed drafts of the paper, contributed substantially to the design of the experiments, contributed to data interpretation and approved the final version.  Joseph J. Bowden contributed reagents/materials/analysis tools, reviewed drafts of the paper, contributed substantially to data interpretation and approved the final version.  Urs A. Treier reviewed drafts of the paper, contributed to the fieldwork, data interpretation and approved the final version.  Signe Normand reviewed drafts of the paper, contributed to the fieldwork, data interpretation and approved the final version.  Toke Høye conceived and designed the experiments, reviewed drafts of the paper, contributed substantially to the fieldwork, data interpretation and approved the final version.

Data Deposition The following information was supplied regarding data availability: Global Biodiversity Information Facility: http://www.gbif.org/dataset/59d25ddf-355d-47d7-b8a1-c1e2819014c7 DOI: 10.15468/li6jkm

Supplemental Information Supplemental information for this article can be found online at http://dx.doi.org/ 10.7717/peerj.2224#supplemental-information.

REFERENCES ACIA. 2004. Impacts of a Warming Arctic: Arctic Climate Impact Assessment. ACIA Overview report. Cambridge: Cambridge University Press, 140. Avis CA, Weaver AJ, Meissner KJ. 2011. Reduction in areal extent of high-latitude wetlands in response to permafrost thaw. Nature Geoscience 4(7):444–448 DOI 10.1038/ngeo1160. Bo¨cher J. 1988. The Coleoptera of Greenland. Copenhagen: Bioscience. Bo¨cher J. 2015. Coleoptera. In: Bo¨cher J, Kristensen NP, Wilhelmsen L, eds. The Greenland Entomofauna: An Identification Manual of Insects, Spiders and Their Allies. Leiden: Brill, 259–292. Bowden JJ, Buddle CM. 2010. Determinants of ground-dwelling spider assemblages at a regional scale in the Yukon Territory, Canada. Ecoscience 17(3):287–297 DOI 10.2980/17-3-3308. Bowden JJ, Eskildsen A, Hansen RR, Olsen K, Kurle CM, Høye TT. 2015a. High-Arctic butterflies become smaller with rising temperatures. Biology Letters 11(10):20150574 DOI 10.1098/rsbl.2015.0574. Bowden JJ, Hansen RR, Olsen K, Høye TT. 2015b. Habitat-specific effects of climate change on a low-mobility Arctic spider species. Polar Biology 38(4):559–568 DOI 10.1007/s00300-014-1622-7. CAFF. 2013. Arctic Biodiversity Assessment: Status and Trends in Arctic Biodiversity. Akureyri: Conservation of Arctic Flora and Fauna. Callaghan TV, Johansson M, Heal OW, Saelthun NR, Barkved LJ, Bayfield N, Brandt O, Brooker R, Christiansen HH, Forchhammer M, Høye TT, Humlum O, Ja¨rvinen A, Jonasson C,

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 14/18 Kohler J, Magnusson B, Meltofte H, Mortensen L, Neuvonen S, Pearce I, Rasch M, Turner L, Hasholt B, Huhta E, Leskinen E, Nielsen N, Siikama¨ki P. 2004. Environmental changes in the North Atlantic Region: SCANNET as a collaborative approach for documenting, understanding and predicting changes. Ambio 13:39–50. Callaghan TV, Tweedie CE, Webber PJ. 2011. Multi-decadal changes in tundra environments and ecosystems: the International Polar Year-Back to the Future Project (IPY-BTF). Ambio 40(6):555–557 DOI 10.1007/s13280-011-0162-4. Chao A, Gotelli NJ, Hsieh TC, Sander EL, Ma KH, Colwell RK, Ellison AM. 2014. Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecological Monographs 84(1):45–67 DOI 10.1890/13-0133.1. Chao A, Jost L. 2012. Coverage-based rarefaction and extrapolation: standardizing samples by completeness rather than size. Ecology 93(12):2533–2547 DOI 10.1890/11-1952.1. Chapin FS, Sturm M, Serreze MC, McFadden JP, Key JR, Lloyd AH, McGuire AD, Rupp TS, Lynch AH, Schimel JP, Beringer J, Chapman WL, Epstein HE, Euskirchen ES, Hinzman LD, Jia G, Ping CL, Tape KD, Thompson CDC, Walker DA, Welker JM. 2005. Role of land-surface changes in arctic summer warming. Science 310(5748):657–660 DOI 10.1126/science.1117368. Daniels FJA, de Molenaar JG, Chytry M, Tichy L. 2011. Vegetation change in Southeast Greenland? Tasiilaq revisited after 40 years. Applied Vegetation Science 14(2):230–241 DOI 10.1111/j.1654-109X.2010.01107.x. De Ca´ceres M, Legendre P. 2009. Associations between species and groups of sites: indices and statistical inference. Ecology 90:3566–3574 DOI 10.1890/08-1823.1. De Ca´ceres M, Legendre P, Moretti M. 2010. Improving indicator species analysis by combining groups of sites. Oikos 119(10):1674–1684 DOI 10.1111/j.1600-0706.2010.18334.x. Dornelas M, Gotelli NJ, McGill B, Shimadzu H, Moyes F, Sievers C, Magurran AE. 2014. Assemblage time series reveal biodiversity change but not systematic loss. Science 344(6181):296–299 DOI 10.1126/science.1248484. Elmendorf SC, Henry GHR, Hollister RD, Bjork RG, Boulanger-Lapointe N, Cooper EJ, Cornelissen JHC, Day TA, Dorrepaal E, Elumeeva TG, Gill M, Gould WA, Harte J, Hik DS, Hofgaard A, Johnson DR, Johnstone JF, Jonsdottir IS, Jorgenson JC, Klanderud K, Klein JA, Koh S, Kudo G, Lara M, Levesque E, Magnusson B, May JL, Mercado-Diaz JA, Michelsen A, Molau U, Myers-Smith IH, Oberbauer SF, Onipchenko VG, Rixen C, Schmidt NM, Shaver GR, Spasojevic MJ, Porhallsdottir PE, Tolvanen A, Troxler T, Tweedie CE, Villareal S, Wahren CH, Walker X, Webber PJ, Welker JM, Wipf S. 2012. Plot-scale evidence of tundra vegetation change and links to recent summer warming. Nature Climate Change 2(6):453–457 DOI 10.1038/Nclimate1465. Epstein HE, Raynolds MK, Walker DA, Bhatt US, Tucker CJ, Pinzon JE. 2012. Dynamics of aboveground phytomass of the circumpolar Arctic tundra during the past three decades. Environmental Research Letters 7(1):015506 DOI 10.1088/1748-9326/7/1/015506. Hansen RR, Hansen OLP, Bowden JJ, Normand S, Bay C, Sørensen JG, Høye TT. 2016. High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap. Polar Biology 1–10 DOI 10.1007/s00300-016-1893-2. Høye TT, Eskildsen A, Hansen RR, Bowden JJ, Schmidt NM, Kissling WD. 2014. Phenology of high-arctic butterflies and their floral resources: species-specific responses to climate change. Current Zoology 60(2):243–251 DOI 10.1093/czoolo/60.2.243. Høye TT, Hammel JU, Fuchs T, Toft S. 2009. Climate change and sexual size dimorphism in an Arctic spider. Biology Letters 5(4):542–544 DOI 10.1098/rsbl.2009.0169.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 15/18 Høye TT, Post E, Schmidt NM, Trøjelsgaard K, Forchhammer MC. 2013. Shorter flowering seasons and declining abundance of flower visitors in a warmer Arctic. Nature Climate Change 3(8):759–763 DOI 10.1038/Nclimate1909. Hui FKC. 2016. Boral–Bayesian ordination and regression analysis of multivariate abundance data in R. Methods in Ecology and Evolution 7(6):744–750 DOI 10.1111/2041-210X.12514. Hui FKC, Taskinen S, Pledger S, Foster SD, Warton DI. 2015. Model-based approaches to unconstrained ordination. Methods in Ecology and Evolution 6(4):399–411 DOI 10.1111/2041-210X.12236. IPCC. 2014. Summary for policymakers. In: Field CB, Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE, Chatterjee M, Ebi KL, Estrada YO, Genova RC, Girma B, Kissel ES, Levy AN, MacCracken S, Mastrandrea PR, White LL, eds. Climate Change 2014: Impacts, Adaptation, and Vulnerability Part A: Global and Sectoral Aspects Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge and New York: Cambridge University Press, 1–32. Leech RE, University of Alberta, Department of Entomology. 1966. The spiders (Araneida) of Hazen Camp 81@ 49′N, 71@ 18′W. Edmonton: Department of Entomology, University of Alberta. Lindroth CH. 1985. The Carabidae (Coleoptera) of Fennoscandia and Denmark. Leiden: Brill. Lindroth CH. 1986. The Carabidae (Coleoptera) of Fennoscandia and Denmark. Leiden: Brill. Loecher M, Ropkins K. 2015. RgoogleMaps and loa: unleashing R graphics power on map tiles. 2015 63(4):18 DOI 10.18637/jss.v063.i04. Marusik YM. 2015. Araneae. In: Bo¨cher J, Kristensen NP, Wilhelmsen L, eds. The Greenland Entomofauna: An Identification Manual of Insects, Spiders and Their Allies. Leiden: Brill, 666–703. Myers-Smith IH, Elmendorf SC, Beck PSA, Wilmking M, Hallinger M, Blok D, Tape KD, Rayback SA, Macias-Fauria M, Forbes BC, Speed JDM, Boulanger-Lapointe N, Rixen C, Levesque E, Schmidt NM, Baittinger C, Trant AJ, Hermanutz L, Collier LS, Dawes MA, Lantz TC, Weijers S, Jorgensen RH, Buchwal A, Buras A, Naito AT, Ravolainen V, Schaepman-Strub G, Wheeler JA, Wipf S, Guay KC, Hik DS, Vellend M. 2015. Climate sensitivity of shrub growth across the tundra biome. Nature Climate Change 5(9):887–891 DOI 10.1038/nclimate2697. Myers-Smith IH, Forbes BC, Wilmking M, Hallinger M, Lantz T, Blok D, Tape KD, Macias-Fauria M, Sass-Klaassen U, Levesque E, Boudreau S, Ropars P, Hermanutz L, Trant A, Collier LS, Weijers S, Rozema J, Rayback SA, Schmidt NM, Schaepman-Strub G, Wipf S, Rixen C, Menard CB, Venn S, Goetz S, Andreu-Hayles L, Elmendorf S, Ravolainen V, Welker J, Grogan P, Epstein HE, Hik DS. 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environmental Research Letters 6(4):045509 DOI 10.1088/1748-9326/6/4/045509. Normand S, Randin C, Ohlemuller R, Bay C, Høye TT, Kjaer ED, Korner C, Lischke H, Maiorano L, Paulsen J, Pearman PB, Psomas A, Treier UA, Zimmermann NE, Svenning JC. 2013. A greener Greenland? Climatic potential and long-term constraints on future expansions of trees and shrubs. Philosophical Transactions of the Royal Society B: Biological Sciences 368(1624):20120479 DOI 10.1098/Rstb.2012.0479. Paquin P, Dupe´rre´ N. 2003. Guide D’identification Des Araigne´es De Que´bec. Varennes: Association des entomologistes amateurs du Que´bec. Pearson RG, Phillips SJ, Loranty MM, Beck PSA, Damoulas T, Knight SJ, Goetz SJ. 2013. Shifts in Arctic vegetation and associated feedbacks under climate change. Nature Climate Change 3(7):673–677 DOI 10.1038/nclimate1858.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 16/18 Perreault N, Le´vesque E, Fortier D, Lamarque LJ. 2015. Thermo-erosion gullies boost the transition from wet to mesic vegetation. Biogeosciences Discuss 12(15):12191–12228 DOI 10.5194/bgd-12-12191-2015. Post E, Forchhammer MC, Bret-Harte MS, Callaghan TV, Christensen TR, Elberling B, Fox AD, Gilg O, Hik DS, Hoye TT, Ims RA, Jeppesen E, Klein DR, Madsen J, McGuire AD, Rysgaard S, Schindler DE, Stirling I, Tamstorf MP, Tyler NJC, van der Wal R, Welker J, Wookey PA, Schmidt NM, Aastrup P. 2009. Ecological dynamics across the Arctic associated with recent climate change. Science 325(5946):1355–1358 DOI 10.1126/science.1173113. Rabe-Hesketh S, Skrondal A, Pickles A. 2004. Generalized multilevel structural equation modeling. Psychometrika 69(2):167–190 DOI 10.1007/bf02295939. Rawlins MA, Steele M, Holland MM, Adam JC, Cherry JE, Francis JA, Groisman PY, Hinzman LD, Huntington TG, Kane DL, Kimball JS, Kwok R, Lammers RB, Lee CM, Lettenmaier DP, McDonald KC, Podest E, Pundsack JW, Rudels B, Serreze MC, Shiklomanov A, Skagseth Ø, Troy TJ, Vo¨ro¨smarty CJ, Wensnahan M, Wood EF, Woodgate R, Yang D, Zhang K, Zhang T. 2010. Analysis of the Arctic System for freshwater cycle intensification: observations and expectations. Journal of Climate 23(21):5715–5737 DOI 10.1175/2010JCLI3421.1. Rich ME, Gough L, Boelman NT. 2013. Arctic arthropod assemblages in habitats of differing shrub dominance. Ecography 36(9):994–1003 DOI 10.1111/j.1600-0587.2012.00078.x. Rouse WR, Douglas MSV, Hecky RE, Hershey AE, Kling GW, Lesack L, Marsh P, McDonald M, Nicholson BJ, Roulet NT, Smol JP. 1997. Effects of climate change on the freshwaters of Arctic and Subarctic North America. Hydrological Processes 11(8):873–9021 DOI 10.1002/(SICI)1099-1085(19970630)11:8<873::AID-HYP510>3.0.CO;2-6. Sikes DS, Draney ML, Fleshman B. 2013. Unexpectedly high among-habitat spider (Araneae) faunal diversity from the Arctic Long-Term Experimental Research (LTER) field station at Toolik Lake, Alaska, United States of America. Canadian Entomologist 145(2):219–226 DOI 10.4039/tce.2013.5. Sweet SK, Asmus A, Rich ME, Wingfield J, Gough L, Boelman NT. 2014. NDVI as a predictor of canopy arthropod biomass in the Alaskan arctic tundra. Ecological Applications 25(3):779–790 DOI 10.1890/14-0632.1. Tape K, Sturm M, Racine C. 2006. The evidence for shrub expansion in Northern Alaska and the Pan-Arctic. Global Change Biology 12(4):686–702 DOI 10.1111/j.1365-2486.2006.01128.x. Vellend M, Baeten L, Myers-Smith IH, Elmendorf SC, Beause´jour R, Brown CD, De Frenne P, Verheyen K, Wipf S. 2013. Global meta-analysis reveals no net change in local-scale plant biodiversity over time. Proceedings of the National Academy of Sciences of the United States of America 110(48):19456–19459 DOI 10.1073/pnas.1312779110. Walker DA, Epstein HE, Raynolds MK, Kuss P, Kopecky MA, Frost GV, Danie¨ls FJA, Leibman MO, Moskalenko NG, Matyshak GV, Khitun OV, Khomutov AV, Forbes BC, Bhatt US, Kade AN, Vonlanthen CM, Tichy´ L. 2012. Environment, vegetation and greenness (NDVI) along the North America and Eurasia Arctic transects. Environmental Research Letters 7(1):015504 DOI 10.1088/1748-9326/7/1/015504. Wang Y, Naumann U, Wright ST, Warton DI. 2012. mvabund–an R package for model-based analysis of multivariate abundance data. Methods in Ecology and Evolution 3(3):471–474 DOI 10.1111/j.2041-210X.2012.00190.x. Warton DI, Wright ST, Wang Y. 2012. Distance-based multivariate analyses confound location and dispersion effects. Methods in Ecology and Evolution 3(1):89–101 DOI 10.1111/j.2041-210X.2011.00127.x.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 17/18 Woo M-K, Young KL. 2006. High Arctic wetlands: their occurrence, hydrological characteristics and sustainability. Journal of Hydrology 320(3–4):432–450 DOI 10.1016/j.jhydrol.2005.07.025. World Spider Catalog. 2016. Available at http://wsc.nmbe.ch (accessed 1 February 2014 ). Zona D, Oechel WC, Kochendorfer J, Paw UKT, Salyuk AN, Olivas PC, Oberbauer SF, Lipson DA. 2009. Methane fluxes during the initiation of a large-scale water table manipulation experiment in the Alaskan Arctic tundra. Global Biogeochemical Cycles 23(2):GB2013 DOI 10.1029/2009gb003487.

Hansen et al. (2016), PeerJ, DOI 10.7717/peerj.2224 18/18 Title Diversity and composition changes along a continentality gradient in the Arctic

Authors

1. Oskar L. P. Hansen1,2,3 a. [email protected] b. Corresponding author 2. Rikke R. Hansen1,2 . [email protected] 3. Joseph J. Bowden1,3 . [email protected] 4. Signe Normand4 . [email protected] 5. Urs Treier1,4 . [email protected] 6. Toke T. Høye1,2,5 . [email protected]

1 Department of Bioscience - Arctic Research Centre, Aarhus University, Ny Munkegade 114, bldg. 1540, DK-8000 Aarhus C

2 Department of Bioscience - Biodiversity and Conservation, Aarhus University, Grenåvej 14, Kalø, DK-8410 Rønde

3 Natural History Museum Aarhus, Wilhelm Meyers Allé 210, Universitetsparken, DK-8000 Aarhus C

4 Department of Bioscience - Ecoinformatics and Biodiversity, Aarhus University, Ny Munkegade 116, bldg. 1540, DK-800 Aarhus C

5 Aarhus Institute of Advanced Studies, Aarhus University, Høegh-Guldbergs Gade 6B, bldg. 1630, DK-8000 Aarhus C

Abstract

Background: Studies using climatic gradients can be valuable in order to predict range shifts triggered by climate change. These studies and their impact on species abundances, however, are unevenly distributed among the ecoregions of the world and are generally scarce in the Arctic. We tested whether species diversity and composition changed in response to change in continentality and micro-climatic gradients in the Arctic. We also tested the similarity of diversity and composition patterns along the gradient among the two dominant groups of ground dwelling arthropods, spiders (Araneae) and beetles (Coleoptera). The study was conducted at the Godthaabsfjord (64° 10'N 51° 45'W), west Greenland during the summer 2013. Three sites were selected along the fjord to represent a continentality gradient. At each site three micro climates, defined as north-facing, south-facing and horizontal plots, were selected at which pitfall trapping methods were used to collect ground-dwelling arthropods. In this study we will first quantify the compositional and diversity changes with increasing continentality and with increasing solar radiation. For each of the gradients we will assess if spider (Araneae) and beetle (Coleoptera) community composition are varying in the same way along these climatic gradients.

Results: In total, 53 species (43 spider species, 10 beetle species) of 107 species known in Greenland were identified from 4,181 individuals. Species diversity increased towards continental sites, while species composition changed significantly along the gradient. Along the micro climatic gradient species diversity increased towards warmer habitats for beetles, while spider diversity peaked at horizontal plots. Nor were the best explanatory variables the same for both taxa, as beetles were best correlated to continentality, while changes in spider communities were best explained by local topography.

Conclusions: The diversity and community composition does change along increasing continentality and solar radiation. However, the community composition for the dominant wolf spiders, were markedly different from other groups. Also, species diversity profiles showed contrasting patterns along the two gradients. This study emphasizes the need to include multiple gradients and multiple taxa to understand the complexity of local and regional responses to climate.

Keywords

Continentality, Micro climate, Biodiversity, Arctic, Climate, Araneae, Coleoptera, Indicator taxa

Background

Climatic gradients are important to understand in order to predict range shifts triggered by climate change (Lenoir & Svenning 2015). Still, changes in composition and diversity of species along climatic gradients (e.g., elevation) are not well documented in Arctic arthropods and plants (But see, Bowden & Buddle, 2010; Bruun et al., 2006). Studies on contemporary range shifts have typically focused on uni-variate range shifts, such as latitude or altitude. These clines may represent narrow gradients that do not consider east- to west-ward range shifts. The effects of local gradients have not been considered along with regional structuring of species and communities (Lennoir & Svenning 2015).

Climatic gradients structure community composition of species across a large range of spatial scales, from micro climates varying within a few meters, depending on topography, a few hundred altitudinal meters on a mountain side to continental distances of hundreds or thousands kilometres (Hodkinson 2005; Whittaker 1960). Therefore, climatic gradients are important to include in a broader framework for understanding the distribution of species. In coastal mountainous areas, continentality is a strong gradient due to the large differences in thermal capacity between the ocean and land. It is, however, an infrequently considered gradient that changes in temperature lapse rate along elevation (Hodkinson 2005). It could, therefore, also strongly affect arthropod species composition and diversity.

Continentality is a pronounced climatic gradient in Northern ecosystems (e.g. Hein et al. 2014a; Hein, et al. 2014b), which can be described by three physical properties: diurnal temperature range, proportion of heat shared by interface substances and the depth or height to which heat is transported (Driscoll & Fong 1992). Continental areas are thus more susceptible to temperature fluctuations than coastal areas, due to their physical properties. Apart from the physical properties, Löffler & Finch (2005), proposed a set of additional covariates along the continentality gradient, which includes increased solar radiation, less precipitation, higher maximum temperature, less soil moisture, and lower wind velocity in continental areas. These geophysical properties do not only affect arthropods directly, but could also cause indirect effects

on arthropods through plant growth or other secondary biotic interactions. With increased primary production and more stable weather, a continental site could be expected to hold greater species diversity than coastal sites.

Snow dynamics (i.e. snow cover and snowmelt timing) and shrub cover (Myers-Smith et al. 2011; Löffler & Finch 2005) are both linked to temperature and wind, and therefore change with elevation and continentality. Snowmelt timing determines the onset of the summer season for Arctic arthropods and plants (Høye et al. 2007), and is therefore an important driver of plant and arthropod community composition (Kreyling et al. 2012; Dollery et al. 2006).

Studies of climatic gradients and their impact on communities of species are unevenly distributed among ecoregions and highly desirable for the Arctic in order to assess current patterns, and predict future changes in range shifts for populations and communities (Lenoir & Svenning 2015; Meltofte 2013). Greenland, in particular, poses a huge potential for knowledge gain in climate related range shifts through the exploration of climatic gradients, as historical sampling is scarce or non-existent (Jensen & Christensen 2003). Furthermore, the diversity of arthropod communities in the Arctic have traditionally been seen as more simple than they are (Meltofte 2013), yet arthropod species compositions have been shown to be heterogeneous and vary in diversity and composition over short distances (Hansen et al 2016).

Differences in arthropod communities between coastal and continental areas around Greenland have been registered based on scattered, non-systematic sampling and anecdotal evidence (Jensen & Christensen 2003). While this has increased our knowledges on the distributions of a few species, data on arthropod community composition is lacking for most regions in Greenland (e.g. Böcher et al 2015; Høye et al. 2013; Meltofte 2013).

A single taxon may not act as a suitable surrogate measure for the diversity and compositional responses of all other taxa along climatic gradients, as individual taxa may respond differently (Axmacher et al. 2011). This is especially true at small spatial scales, where the variability in the environment does not correspond to the same variability in species richness in all taxa (Ricketts et al. 2002; Giorgini et al. 2015). At larger scales, however, a single taxon may be a reasonable predictor of another (Mikusinski et al. 2001; Giorgini et al. 2015). It is however not clear if all taxa will vary in similar ways along climatic gradients or if some taxa respond differently than others.

This study addresses the following questions 1) is species diversity increasing and does species composition change with increasing continentality and with increasing solar radiation (from north to south facing slopes)? 2) are spiders (Araneae) and beetle (Coleoptera) community composition varying in the same way along these climatic gradients?

Results

In total, 4,181 adult individuals of spiders and beetles were collected and identified from 96 plots, comprising 43 spider species and 10 beetle species, table S1. The wolf spiders (Lycosidae) with 3,114 individuals and four species were the most numerous and widely distributed family, leaving 527 and 540 individuals, belonging to the beetles (Coleoptera) and other spiders (Araneae excl. Lycosidae), respectively (table S1). Wolf spiders were present in all 96 plots, while other spiders and beetles were present in 91 and 88 plots, respectively. Hence the analysis of spiders was done separately with and without wolf spiders.

Species diversity

Species richness increased significantly from the coastal site, 1, to the intermediate and continental sites, 2 and 3, for all groups combined (figure 1a). Shannon diversity increased significantly between site 1 and 3, but when up weighing abundant species (Simpson diversity) this increase along the continentality gradient disappeared. For beetles, the greatest increase in diversity occurred between the intermediate site, 2, and the continental site, 3, closest to the head of the fjord. For spiders, the increase primarily occurred closer to the mouth of the fjord, between site 1 and 2 (figure 1a).

North-facing plots were significantly less species rich than horizontal plots along the local climatic gradients (figure 1b), while no difference was found for Simpson and Shannon diversity. For beetles no differences were detected between any of the micro-climates. The horizontal plots yielded a significantly higher species richness of spiders than both the south- and north-facing plots. For spiders, excluding the wolf spiders, the confidence interval of north-facing plots spanned the confidence intervals of the two other groups and was therefore excluded to optimize the reference coverage for the remaining groups and thus increase estimate precision.

Composition

The three sites separated in ordination space (figure 2) along the continentality gradient for all combinations of taxonomic groups. It is also clear that the micro-climates tended to vary along the same trajectory as continentality, such that communities from cold microclimates at the continental site are similar to the communities from warmer microclimates at the coastal sites. The fitted environmental variables revealed that slope and continentality separated the communities in a perpendicular direction from each other consistently for all taxonomic subsets. Species composition of spiders, beetles, and both groups combined, differed significantly between sites and between micro-climates (table 1). We found a significant interaction term between site and microclimate for all taxonomic subsets except for the beetles (table 1). Some of the abundant species differed significantly between micro-climates and/or sites (table S2). All significant species showed increased activity-abundance from the coastal site 1 to the continental site 3 and from North- to South-facing plots (table S3). The structuring of sites was best explained by continentality and slope (table 2). The interpretation is highly sensitive to the selection of taxonomic group; however wolf spiders dictated the overall picture of community structure. For beetles and spiders, excluding wolf spiders, continentality alone was the best explanatory variable. Topography did not add explanatory power when continentality was included (table 2). Point-biserial correlation coefficients were calculated between species abundances and sites, 19 of 53 species were significantly correlated with one or two sites in the fjord (table S4). The highest number of indicator species were found at the most continental site with 11 of the species associated with that climate. Pardosa groenlandica, Erigone whymperi, and Mecynargus paetulus were significant indicators for the most oceanic site. For the most continental site 11 species were significant indicators, among those Pardosa hyperborea, Coccinella transversoguttata, Pocadicnemis americana and Caenoscelis ferruginea had a point- biserial correlation above 0.450. Pocadicnemis americana, who was almost exclusively found on south-facing plots, was in combination with Pardosa groenlandica and P. hyperborea highly correlated with the warm south facing plots (point-biserial correlation 0.360, pval = 0.010). Interestingly Pardosa hyperborea in combination with Mycetoporus nigrans were suggested as the best indicators for north facing plots (point-biserial correlation 0.376, pval = 0.005), however P. hyperborea was not significant alone contrary to M. nigrans (point-biserial correlation 0.422, pval = 0.002).

Discussion

Species diversity increased significantly along the continentality gradient and along the local climate gradient. Species composition also varied significantly along the two gradients. The changes were, however, not consistent among taxonomic groups, nor were the variables predicted to have the most explanatory power for species composition the same among the taxonomic groups.

Even though the general pattern, when assessing the extreme endpoints of the gradients were similar, beetles and spiders did not respond similarly along the gradients. While beetle diversity decreased with temperature, i.e. highest at south-facing plots and lowest at north-facing plots, spider diversity was estimated to be most diverse at horizontal plots. This difference could be caused by scale, topographic context, other covariables or a combination.

Species diversity

While spiders and beetles both had significantly higher species diversity at the continental site relative to the coastal site, there was a discrepancy at the intermediate site, where beetle diversity did not increase, contrary to the spiders (figure 1a). This discrepancy is not visible in the model combining all taxonomic groups, indicating that patterns of diversity of specific taxonomic or functional groups can be masked by more diverse groups. Further the differences between spiders and beetles could be better understood in an analysis of species diversity by functional traits.

Composition

We found continentality to be a strong gradient for structuring of arthropod community composition. The main driver of the other spiders and beetles lies within the continentality gradient, which clearly separates the sites.

Taxonomic differences

If the taxonomic groups varied similarly along the climatic gradients, it would be possible to predict both abundances and species richness based on information about a selected indicator group. This has been done in several studies covering plants, birds, butterflies and beetles (Axmacher et al. 2011; Giorgini et al. 2015; Ricketts et al. 2002; Mikusinski et al. 2001); the conclusions, however, have been inconsistent. Ricketts et al. (2002), for example, suggest that

differences in habitat preferences between arthropod orders makes it difficult to assume indicator taxa at small scales. Also, Giorgini et al. (2015) find that species richness is more reliably estimated at coarser scales. While the overall patterns of species diversity and community composition between the two groups at the extremes in this study correspond, idiosyncrasies exists along both the continentality and micro-climatic gradients that may not be apparent in simple correlations between taxonomic groups. Thus the scale in this study is too fine to imply consistent changes among taxonomic groups, as scale matters in the ability to predict diversity of one group from another (Giorgini et al. 2015).

Topographic context

The sourrounding topography may also interact with the local topography. The topography on larger scale creates different living conditions for otherwise equal micro-climates. e.g. some south facing plots on the continental site, situated on a mountain-side, are not only differenct from other sites, but also from other south-facing plots at the same site, suggesting that topographic context could enhance or decrease the effect of a slope at a small scale. When the contiguous area becomes smaller the very local ability to avoid extinctions and to be a subject for immigration decreases, creating an area in which virtually all populations would be sink populations and only maintained by immigration from adjacent habitats (Hanski 1998; Eriksson et al. 2002). This could also create a more mixed community that is not entirely defined by the micro-climate, but subject to an edge effect between habitats (Ries & Sisk 2004).

Covariables

The interaction term between the two gradients for arthropod communities in this study may be partly explained by micro-climate- and continentality- specific variation in snowmelt timing. On larger scales snowmelt is driven by radiative fluxes and energy advection, which varies greatly (Mioduszewski et al. 2014; Mioduszewski et al. 2015). Prevailing wind and wind speed on the other hand controls snow accumulation at smaller scales (Pomeroy et al. 1997). At smaller scales, snowmelt timing is also highly variable in space. Snow accumulation in itself may not have any short term impact on arthropod survival or community composition (Convey et al. 2014; Legault & Weis 2013). Slope is an important predictor of wolf spider composition and is somewhat orthogonal to continentality (figure 2). While slope could be a proxy for soil moisture, as meso-topography is well correlated with soil moisture (Le Roux et al. 2013) it is not, however,

possible to disentangle the ridges and depressions in this study. Horizontal plots could thus include both the highest and lowest soil moisture. Other methods using digital elevation models have shown not to produce good predictions of soil moisture (Le Roux et al. 2013).

These results indicate that future studies, investigating broad scale variation in species diversity and composition, should include continentality, as well as, local topography. The results also show that the diversity of taxonomic groups respond at different stages to climatic gradients, though overall trends are similar. In future research, we recommend that a broad variety of taxa should be included, whenever possible to understand such idiosyncratic responses.

Conclusions

Both the continentality and micro-climate affected the species diversity and species composition. 2) The two arthropod orders, beetles and spiders, did not respond completely synchronously to either of the two gradients, albeit, the overall pattern was the same.

This study emphasizes the need for future studies to include multiple gradients and multiple taxa to better understand the complexity of climatic responses in what has been thought of as simple systems. This study also shows how two gradients can interact and even though having different scales and different covariables. This may also suggest temperature to be the dominant variable for some, but not all species.

Methods

This study was conducted at Godthaabsfjorden (64° 10'N 51° 45'W), west Greenland (figure 3) between June 27 and July 20, 2013. Three sites were selected along the fjord from the mouth to the head. The resulting continentality gradient stretched approximately 60 km in total with 30 km between each site. Microclimates at each site were categorized as North facing, South facing or horizontal so as to represent a local micro-climatic gradient of solar radiation defined by slope and aspect. Horizontal plots did not have a slope greater than 11°, while north and south were greater than 11° and within 70° from geographic north/south. From the 186 plots sampled, 96 plots with equal, concurrent sampling effort (figure S1) and situated below 60 m.a.s.l. were chosen within the aforementioned climate gradient criteria. The plots (table 3) were distributed slightly unevenly among the gradient categories with 30, 50 and 16 on site 1, 2 and 3 respectively and of which 37, 11 and 48 categorized as south, north and horizontal plots. Each

plot was sampled using yellow pitfall traps (H: 81 mm, ⌀: 100 mm) for 16 to 20 (mean 17.4) days. Samples from plots extending more than 20 days and plots with a total sampling duration less than 14 days were excluded, in order to keep the sampling effort at a comparable level (figure S1). At each plot, two pitfall traps were set at a distance of about 2 meters from each other.

Species Identification

Spiders were identified using literature available through The World Spider Catalog (World Spider Catalog 2014). Beetles were identified using various Scandinavian and North American literature (i.e., Böcher 1988; Lindroth et al. 1985; Lindroth et al. 1986; Palm 1966)⁠ and through consultation of the collection at the Natural History Museum Aarhus, Denmark. All individuals were identified to the lowest possible taxonomic level, however only adult individuals identified to species level were used in the analysis. Specimens are preserved in 75% vol. ethanol at the Natural History Museum, Aarhus.

Data analysis

Species Diversity Analysis

Data analysis was conducted using R version 3.2.3 (2015-12-10) (R Core Team 2015). Diversity for each micro-climate at each site was rarefied and extrapolated, using abundance data with the iNEXT R-package (Hsieh et al. 2014; Chao et al. 2014). Initially all groups (each local micro- climate for each site, n = 9, and local micro-climates pooled together, n = 3) were compared, but due to low sample-sizes (table 3) and to reduce the within group variation, it was not feasible to compare other than the best replicated set of groups for the two gradients (i.e. site 2 for micro- climates and south facing plots for continentality).

Hill numbers, which represent species richness, shannon diversity, weighting species by relative frequency and simpson diversity, weighting dominant species (Chao et al. 2014), were used to analyze species diversity. The units of the Hill numbers are effective number of species that is the number of equally abundant species equivalent to a diversity measure (Chao et al. 2014). Maximum extrapolation ratio, r, (i.e the ratio of extrapolated sample size and reference sample size) was set to r = 2, as recommended by Chao et al. (2014). Each model was compared at a common base coverage that was set to Cbase = min{Ca, Cb}, where Ca = max{C(n1), C(n2), ...,

C(nk)} and Cb = min{C(rn1), C(rn2), ..., C(rnk)}, where n1, n2, ..., nk is the reference sample sizes for the groups. In this way, at least one group was not extrapolated and thus optimizing the precision and contrasts of the estimates. If a group confidence interval at Cbase spanned the two other groups, within a given order, this group was removed and the analysis rerun, to increase contrast among the remaining groups.

Composition analysis

To assess species composition along the two gradients and among taxonomic groups, four non- metric multidimensional scaling (NMDS) ordinations were run using the vegan package version 2.3-5 (Oksanen et al. 2015). For each analysis a maximum of 5000 random starts were chosen to find the global minimum stress solution. Spearman rank correlation between community matrix and the gradients were used to check that Bray-Curtis distance was the most appropriate dissimilarity index to be used in the distance matrix. Monte Carlo simulations with randomized data (n = 25) of ten runs each at maximum 20 random starts and visual inspection of scree plots were used to choose the optimal number of dimensions and insure that stress levels by the data were not higher than what could be generated from randomized data. Ellipsoids describing the standard error of the average of scores for each site were added to assess the differences among sites in ordination space. Continuous environmental variables were fitted to each plot as vectors to describe the ordination space and included slope, topography, continentality, species richness and abundance. Slope was assessed as the field-measure: number of vertical meters per 10 horizontal meters. Topography was calculated as a continuous index to quantify the combined information of slope and aspect (equation 1 and 2). For plots with aspect towards north the index would generate a negative value, and for plots with aspect towards south a positive value dependent on the steepness of the slope. Horizontal plots received values near zero.

To test for differences in communities between sites and habitats a model approach was used (Warton et al. 2014). This was done for two reasons. First, more common and widely used PERMANOVA and related distance-metric tests are not well-suited, being either too liberal or too conservative in their conclusions for unbalanced data sets and requires similar variability in all species for non-confounding interpretation of dispersion and displacement (Warton et al. 2012; Anderson & Walsh 2013). The null hypothesis stating no difference between sites and habitats was tested using the manyglm function in the R-package mvabund (Wang et al. 2012).

This function fitted a generalized linear model for high dimensional data with a negative binomial error distribution. Analysis of deviance was assessed using likelihood-ratio-test and species specific contributions were assessed using p-values adjusted for multiple testing. The model assumptions of mean-variance and log-linearity were controlled through residual vs. fit plots. No patterns were found to indicate violations of the model assumptions.

Point-biserial correlation coefficients (i.e Pearson correlation between species abundance and a binary vector for site membership) were used to identify indicator species for each site or combination of sites following De Cáceres & Legendre (2009) and De Cáceres et al. (2010). Spearman rank correlations between community dissimilarity matrices and environmental variables (Clarke & Ainsworth 1993) were used to assess the set of environmental variables that best explained the data. Full documentation of the analysis can be found in the appendix.

Declarations

Authors' contributions

OLPH, RRH and TTH conceived the idea and planned the field work. OLPH, RRH, TTH, SN and UT carried out the fieldwork. OLPH, RRH and JJB identified the species. OLPH carried out the analyses and wrote the first draft. Everyone contributed to the interpretation of results and preparation of the final version of the paper.

Acknowledgements

We would like to acknowledge the Natural History Museum, Aarhus for laboratory space and curation of specimens.

Funding

Availability of data and material

References Anderson, M.J. & Walsh, D.C.I., 2013. PERMANOVA, ANOSIM, and the Mantel test in the face of heterogeneous dispersions: What null hypothesis are you testing? Ecological Monographs, 83(4), pp.557–574.

Axmacher, J.C. et al., 2011. Spatial α-diversity patterns of diverse insect taxa in Northern China: Lessons for biodiversity conservation. Biological Conservation, 144(9), pp.2362–2368. Available at: http://dx.doi.org/10.1016/j.biocon.2011.06.016. Böcher, J., 1988. The Coleoptera of Greenland. Meddelelser om Grønland, 26, p.100. Böcher, J. et al., 2015. The Greenland Entomofauna: An Identification Manual of Insects, Spiders and Their Allies, Brill Academic Publishers. Available at: https://books.google.dk/books?id=JODHoQEACAAJ. Bowden, J.J. & Buddle, C.M., 2010. Spider Assemblages across Elevational and Latitudinal Gradients in the Yukon Territory, Canada. Arctic, 63(3), pp.261–272. Bruun, H.H. et al., 2006. Effects of altitude and topography on species richness of vascular plants , bryophytes and lichens in alpine communities. Journal of Vegetation Science, 17, pp.37–46. De Cáceres, M. & Legendre, P., 2009. Associations between species and groups of sites : indices and statistical inference. Ecology, 90(12), pp.3566–3574. De Cáceres, M., Legendre, P. & Moretti, M., 2010. Improving indicator species analysis by combining groups of sites. Oikos, 119(10), pp.1674– 1684. Chao, A. et al., 2014. Rarefaction and extrapolation with Hill numbers: A framework for sampling and estimation in species diversity studies. Ecological Monographs, 84(1), pp.45–67. Clarke, K.R. & Ainsworth, M., 1993. A method of linking multivariate community structure to environmental variables. Marine Ecology Progress Series, 92(3), pp.205–219. Convey, P. et al., 2014. Survival of rapidly fluctuating natural low winter temperatures by High Arctic soil invertebrates. Journal of Thermal Biology, pp.1–7. Available at: http://dx.doi.org/10.1016/j.jtherbio.2014.07.009. Dollery, R., Hodkinson, I.D. & Jónsdóttir, I.S., 2006. Impact of warming and timing of snow melt on soil microarthropod assemblages associated with Dryas-dominated plant communities on Svalbard. Ecography, 29(1), pp.111–119. Driscoll, D.M. & Fong, J.M.Y., 1992. Continentality: A basic climatic parameter re-examined. Int. J. Climatol., 12, pp.185–192. Available at: http://dx.doi.org/10.1002/joc.3370120207.

Eriksson, O., Cousins, S. a O. & Bruun, H.H., 2002. Land-use history and fragmentation of traditionally managed grasslands in Scandinavia. Journal of Vegetation Science, 13(5), pp.743–748. Available at: ://BIOSIS:PREV200300186878. Giorgini, D. et al., 2015. Woody species diversity as predictor of vascular plant species diversity in forest ecosystems. Forest Ecology and Management, 345, pp.50–55. Available at: http://dx.doi.org/10.1016/j.foreco.2015.02.016. Hansen, R.R. et al., 2016. High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap. Polar Biology. Available at: http://link.springer.com/10.1007/s00300-016- 1893-2. Hanski, I., 1998. Metapopulation dynamics. Nature, 396(November), pp.41– 49. Available at: http://noss.cos.ucf.edu/papers/Hanski 1998.pdf [Accessed June 4, 2012]. Hein, N., Pape, R., et al., 2014b. Alpine activity patterns of Mitopus morio (Fabricius, 1779) are induced by variations in temperature and humidity at different scales in central Norway. Journal of Mountain Science, 11, pp.644–655. Available at: http://link.springer.com/article/10.1007/s11629-013-2913-0. Hein, N., Feilhauer, H., et al., 2014a. Snow cover determines the ecology and biogeography of spiders (Araneae) in alpine tundra ecosystems. Erdkunde, 68(3), pp.157–172. Available at: http://www.erdkunde.uni- bonn.de/archive/2014/snow-cover-determines-the-ecology-and- biogeography-of-spiders-araneae-in-alpine-tundra-ecosystems. Hodkinson, I.D., 2005. Terrestrial insects along elevation gradients: species and community responses to altitude. Biological reviews of the Cambridge Philosophical Society, 80, pp.489–513. Høye, T.T. et al., 2007. Rapid advancement of spring in the High Arctic. Current Biology, 17, pp.449–451. Høye, T.T. et al., 2013. Shorter flowering seasons and declining abundance of flower visitors in a warmer Arctic. Nature Climate Change, 3(8), pp.759–763. Available at: http://www.nature.com/doifinder/10.1038/nclimate1909. Hsieh, T.C., Ma, K.H. & Chao., A., 2014. iNEXT: iNterpolation and EXTrapolation for species diversity. Available at: http://chao.stat.nthu.edu.tw/blog/software-download. Jensen, D.B. & Christensen, K.D., 2003. The Biodiversity of Greenland – a country study, Available at:

http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:The+B iodiversity+of+Greenland+?+a+country+study#0 [Accessed January 23, 2014]. Kreyling, J., Haei, M. & Laudon, H., 2012. Absence of snow cover reduces understory plant cover and alters plant community composition in boreal forests. Oecologia, 168(2), pp.577–587. Legault, G. & Weis, A.E., 2013. The impact of snow accumulation on a heath spider community in a sub-Arctic landscape. Polar Biology, 36(6), pp.885–894. Lenoir, J. & Svenning, J.C., 2015. Climate-related range shifts - a global multidimensional synthesis and new research directions. Ecography, 38(1), pp.15–28. Lindroth, C.H. et al., 1985. The Carabidae (Coleoptera) of Fennoscandia and Denmark, Volume 1, Leiden: Brill/Scandinavian Science Press Ltd. Available at: http://www.google.no/books?id=R-cUAAAAIAAJ. Lindroth, C.H. et al., 1986. The Carabidae (Coleoptera) of Fennoscandia and Denmark, Volume 2, Leiden: Brill/Scandinavian Science Press Ltd. Available at: http://books.google.no/books?id=prsbYAAACAAJ. Löffler, J. & Finch, O.-D., 2005. Spatio-temporal gradients between high mountain ecosystems of central Norway. Arctic, Antarctic, and Alpine Research, 37(4), pp.499–513. Meltofte, H. ed., 2013. Arctic Biodiversity Assessment. Status and trends in Arctic biodiversity, Akureyri: Conservation of Arctic Flora and Fauna (CAFF). Available at: http://www.researchgate.net/publication/256486980_Arctic_Biodiversity _Assessment_Status_and_trends_in_Arctic_biodiversity/file/e0b495230d d7cb512c.pdf [Accessed September 2, 2014]. Mikusinski, G., Gromadzki, M. & Chylarecki, P., 2001. Woodpeckers as indicators of forest bird diversity. Conservation Biology, 15(1), pp.208– 217. Mioduszewski, J.R. et al., 2015. Controls on Spatial and Temporal Variability in Northern Hemisphere Terrestrial Snow Melt Timing, 1979–2012. Journal of Climate, 28(6), pp.2136–2153. Available at: http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-14-00558.1. Mioduszewski, J.R. et al., 2014. Journal of Geophysical Research : Atmospheres Attribution of snowmelt onset in Northern Canada. , pp.9638–9653.

Myers-Smith, I.H. et al., 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environmental Research Letters, 6, p.045509. Oksanen, A.J. et al., 2015. vegan: Community Ecology Package. Available at: http://cran.r-project.org/package=vegan. Palm, T., 1966. Svensk insektfauna 9, Skalbaggar, Coleoptera Kortvingar: Fam. Staphylinidae., Stockholm: Entomologiska föreningen i Stockholm. Pomeroy, J.W., Marsh, P. & D.M. Gray, 1997. Application of a distributed blowing snow model to the Arctic. Hydrological Processes, 11(11), pp.1451–1464. Available at: http://dx.doi.org/10.1002/(SICI)1099- 1085(199709)11:11<1451::AID-HYP449>3.0.CO;2-Q. R Core Team, 2015. R: A Language and Environment for Statistical Computing. Available at: http://www.r-project.org/. Ricketts, T.H., Daily, G.C. & Ehrlich, P.R., 2002. Does butterfly diversity predict moth diversity? Testing a popular indicator taxon at local scales. Biological Conservation, 103(3), pp.361–370. Ries, L. & Sisk, T.D., 2004. A predictive model of edge effects. Ecology, 85(11), pp.2917–2926. Le Roux, P.C., Aalto, J. & Luoto, M., 2013. Soil moisture’s underestimated role in climate change impact modelling in low-energy systems. Global Change Biology, 19(10), pp.2965–2975. Wang, Y. et al., 2012. Mvabund- an R package for model-based analysis of multivariate abundance data. Methods in Ecology and Evolution, 3(3), pp.471–474. Warton, D.I. et al., 2014. Model-based thinking for community ecology. Plant Ecology. Warton, D.I., Wright, S.T. & Wang, Y., 2012. Distance-based multivariate analyses confound location and dispersion effects. Methods in Ecology and Evolution, 3(1), pp.89–101. Whittaker, R.H., 1960. Vegetation of the Siskiyou Mountains, Oregon and California. Ecological Monographs, 30(3), pp.279–338. Available at: http://www.esajournals.org/doi/abs/10.2307/1943563\nhttp://www.esa journals.org/action/showCitFormats?doi=10.2307/1943563. World Spider Catalog, 2014. World Spider Catalog. World Spider Catalog. Available at: http://wsc.nmbe.ch [Accessed December 1, 2014]. Figures

Figure 1

Estimated effective number of species (± 95% confidence interval) calculated for species richness, shannon diversity, and simpson diversity, along two gradients: (a) Continentality for south-facing plots and (b) local temperature gradient (north, horizontal and south facing plots) at the intermediate site, 2. (a) Continentality: For spiders and beetles the species richness increases along the gradient, but the increase happens in different parts of the fjord. Spider species diversity increases between the coastal and intermediate sites, 1 and 2, while the increase for beetles happens between the intermediate and continental sites, 2 and 3. (b) The local micro-climatic gradient suggest that spiders have the highest species richness on horizontal plots.

1 Figure 2 2 Two dimensional non-metric multidimensional scaling (NMDS) ordination plots, 3 based on Bray-Curtis distances. For each ordination a subset is shown for each site. 4 Centroids (± SE) for sites and covariates are printed in the rightmost panel of each 5 row. (a) represent all species combined together, while (b) represent subsets of 6 taxonomic groups. Wolf spiders heavily dominate the overall structure (top and 7 third row), while the remaining spiders and beetles are structured along the 8 continentality gradient. Stress spiders and beetles: 14, beetles: 18, wolf spiders: 9 12, spiders excl wolf spiders: 14

10

11

12 Tables

13 Table 3 14 Number of plots that fulfilled the selection criteria for each category.

North Horizontal South

Site 1, Coastal 2 15 13

Site 2, Intermediate 6 31 13

Site 3, Continental 3 2 11

15

16 Table 1 17 Analysis of deviance tables. Testing the hypothesis that the community between 18 sites and habitats is not different (based on glm for each species, see table S2). 19 The test was run for all data combined and for subsets of taxonomic groups. All but 20 the interaction term for beetles were significant at α = 0.05.

Res.Df Df.diff Dev Pr(>Dev)

Spiders and Beetles

(Intercept) 95

site 93 2 408.128 0.001

micro climate 91 2 275.343 0.001

site:micro climate 87 4 156.992 0.001

Beetles

(Intercept) 87

site 85 2 146.752 0.001

micro climate 83 2 64.594 0.002

site:micro climate 79 4 23.934 0.229

Spiders

(Intercept) 95

site 93 2 254.544 0.001

micro climate 91 2 213.367 0.001

site:micro climate 87 4 134.112 0.001

Wolf spiders

(Intercept) 95

site 93 2 68.789 0.001

micro climate 91 2 57.797 0.001

site:micro climate 87 4 60.099 0.001

Spiders excl. wolf spiders

(Intercept) 90

site 88 2 184.903 0.001

micro climate 86 2 156.411 0.001

site:micro climate 82 4 75.452 0.001 21

22 Table 2 23 Spearman rank correlations between community dissimilarities and the best 24 combination of environmental variables. Based on bioenv in vegan R-package.

Variables size correlation

Spiders and Beetles

topography 1 0.354

continentality slope 2 0.444

continentality topography slope 3 0.443

Beetles

continentality 1 0.310

continentality topography 2 0.286

continentality topography slope 3 0.246

Wolf spiders

topography 1 0.352

continentalityslope 2 0.421

continentality topography slope 3 0.426

Spiders excl. wolfspiders

continentality 1 0.103

continentality slope 2 0.104

continentality topography slope 3 0.089 25

26 Equations

27 Eq 1

28

29 Eq 2

30

31 Supplementary material

32 Figure S1 33 Samples used in the analysis. Dots denotes start and endpoints for samples.

34 35 Table S1 36 Occurrences of species and number of individuals in the nine different combinations of site (Coastal, 37 Intermediate and Continental) and local climatic gradient (horizontal, north or south).

Coastal Intermediate Continental

Horizont Nort Sout Horizont Nort Sout Horizont Nort Sout Tota al h h al h h al h h l

Beetles

Ground beetles

Bembidion 0 0 0 0 0 0 0 0 1 1 grapii

Patrobus 39 0 3 32 0 6 3 1 3 87 septentrionis

Ladybirdbeetl es

Coccinella 0 1 1 13 4 12 3 7 44 85 transversogutta ta

Nephus 0 0 0 0 0 0 0 0 2 2 redtenbacheri

Weevils

Dorytomus 0 0 0 1 0 0 0 0 0 1 imbecillus

Otiorhynchus 3 0 13 0 0 0 5 0 21 42 arcticus

Otiorhynchus 10 1 12 100 21 18 3 8 11 184 nodosus

Other beetles

Byrrhus 2 0 3 8 0 11 0 2 5 31 fasciatus

Caenoscelis 0 0 0 0 0 0 3 0 79 82 ferruginea

Mycetoporus 0 1 0 0 1 1 0 7 2 12 nigrans

Spiders

Wolf spiders

Arctosa 6 0 10 45 3 10 0 0 0 74 insignita

Pardosa 438 0 33 802 36 230 29 40 8 161 furcifera 6

Pardosa 25 7 21 5 2 24 1 5 12 102 groenlandica

Pardosa 6 0 39 64 104 291 20 56 742 132 hyperborea 2

Theridiidae

Enoplognatha 0 0 0 0 0 0 0 0 1 1 intrepida

Ohlertidion 0 0 0 0 0 0 0 0 1 1 lundbecki

Robertus fuscus 0 0 0 1 0 0 0 0 0 1

Other spiders

Dictyna major 0 0 0 0 0 1 0 0 0 1

Emblyna 0 0 0 1 0 1 0 0 0 2 borealis

Hahnia glacialis 1 0 10 0 0 17 1 12 16 57

Haplodrassus 0 0 3 1 0 0 0 1 0 5 signifer

Hypsosinga 0 0 0 30 0 3 0 0 11 44 groenlandica

Thanatus 2 0 2 7 1 7 0 0 16 35 arcticus

Xysticus durus 0 0 2 6 1 7 0 0 11 27

Linyphiidae

Agyneta 0 0 0 2 0 0 0 0 0 2 allosubtilis

Agyneta 4 0 6 14 0 2 0 1 4 31 jacksoni

Agyneta 0 0 0 1 0 0 0 0 0 1 nigripes

Agyneta subtilis 0 0 0 4 0 0 0 0 0 4

Bathyphantes 0 0 0 1 0 0 0 0 0 1 eumenis

Dismodicus 0 0 0 11 0 1 0 11 2 25 decemoculatus

Erigone arctica 6 0 0 0 0 0 0 0 0 6

Erigone 1 0 0 0 0 0 0 0 0 1 psychrophila

Erigone 7 0 0 1 0 0 0 0 0 8 whymperi

Hilaira herniosa 0 0 0 0 1 0 0 0 0 1

Hybauchenidiu 0 0 0 0 0 0 1 2 4 7 m gibbosum

Improphantes 0 0 2 2 1 2 0 1 1 9

complicatus

Islandiana 0 0 0 0 1 0 0 0 0 1 princeps

Lepthyphantes 0 0 0 1 0 0 0 0 1 2 turbatrix

Mecynargus 2 0 1 4 0 0 0 0 0 7 borealis

Mecynargus 0 0 4 4 1 3 0 0 0 12 morulus

Mecynargus 31 0 0 4 0 0 0 1 0 36 paetulus

Oreonetides 0 0 0 1 0 1 0 0 0 2 vaginatus

Pelecopsis 0 0 0 3 0 2 0 0 0 5 mengei

Pocadicnemis 0 0 1 1 0 12 0 0 41 55 americana

Sciastes 0 0 0 1 0 0 0 0 0 1 extremus

Semljicola 4 3 4 19 0 11 0 0 0 41 obtusus

Sisicus apertus 0 0 0 3 0 1 0 0 1 5

Tiso aestivus 0 0 0 1 6 6 0 0 11 24

Wabasso 0 0 0 0 0 0 0 4 0 4 cacuminatus

Wabasso 10 0 0 5 0 0 0 0 0 15 quaestio

Walckenaeria 0 0 0 0 0 0 0 0 1 1 castanea

Walckenaeria 0 3 0 2 0 0 0 0 0 5 clavicornis

Walckenaeria 8 2 9 26 2 5 0 0 2 54 karpinskii

Total 605 18 179 1227 185 685 69 159 105 418 4 1 38

39 Table S2 40 Analysis of deviance table for each species. Pr(>Dev) and deviance in brackets is reported

41 for each species. 42 Species Stat p-val

Site 1

Pardosa groenlandica 0.298 0.039

Erigone whymperi 0.258 0.046

Mecynargus paetulus 0.238 0.039

Site 2

Otiorhynchus nodosus 0.382 0.003

Arctosa insignita 0.363 0.001

Site 3

Pardosa hyperborea 0.662 0.001

Coccinella transversoguttata 0.507 0.001

Pocadicnemis americana 0.467 0.001

Caenoscelis ferruginea 0.461 0.001

Hybauchenidium gibbosum 0.413 0.001

Hahnia glacialis 0.389 0.001

Thanatus arcticus 0.362 0.003

Otiorhynchus arcticus 0.346 0.004

Mycetoporus nigrans 0.337 0.001

Tiso aestivus 0.282 0.025

Dismodicus decemoculatus 0.270 0.027

Site 1 & 2

Pardosa furcifera 0.350 0.005

Walckenaeria karpinskii 0.273 0.039

Site 2 & 3

Hypsosinga groenlandica 0.264 0.043 43

44 Table S3 45 Univariate abundance estimates from the manyglm. 46

(Inter sit sit habita habitatH site2:hab site3:hab site2:habita site3:habita cept) e2 e3 tNorth orisontal itatNorth itatNorth tHorisontal tHorisontal

Gen.Var 0 0 0 0 0 0 0 0 0

Beetles

Ground beetles

Bembidio - 0 12 0 0 0 -12.42 0 -12.42 n grapii 14.82 .4 2

Patrobus -1.47 0. 0. -13.35 2.42 -0.69 13.55 -1.62 -0.72 septentri 69 17 onis

Weevils

Otiorhyn 0 - 0. -14.82 -1.61 14.82 -0.65 1.61 1.88 chus 14 65 arcticus .8 2

Otiorhyn -0.08 0. 0. -0.61 -0.33 1.54 1.59 1.17 0.73 chus 41 08 nodosus

Dorytom - 0 0 0 0 0 0 11.38 0 us 14.82 imbecillu s

Ladybir dbeetles

Coccinell -2.56 2. 3. 1.87 -12.25 -2.2 -2.41 11.46 11.27 a 48 95 transvers oguttata

Nephus - 0 13 0 0 0 -13.11 0 -13.11 redtenba 14.82 .1 cheri 1

Other beetles

Byrrhus -1.47 1. 0. -13.35 -0.55 -1.3 13.73 -0.64 -13.48 fasciatus 3 68

Caenosce - 0 16 0 0 0 -16.79 0 -1.57 lis 14.82 .7 ferrugine 9 a

Mycetopo - 12 13 14.12 0 -13.35 -11.57 -12.25 -13.11 rus 14.82 .2 .1 nigrans 5 1

Spiders

Other spiders

Emblyna - 12 0 0 0 -12.25 0 -0.87 0 borealis 14.82 .2 5

Dictyna - 12 0 0 0 -12.25 0 -12.25 0 major 14.82 .2 5

Hahnia -0.26 0. 0. -14.55 -2.45 -0.53 15.56 -12.64 1.38 glacialis 53 64

Haplodra -1.47 - - -13.35 -13.35 13.35 27.07 24.73 13.35 ssus 13 13 signifer .3 .3 5 5

Hypsosin - 13 14 0 0 -13.35 -14.82 1.43 -14.82 ga 14.82 .3 .8 groenlan 5 2 dica

Thanatus -1.87 1. 2. -12.94 -0.14 11.77 -2.25 -0.73 -15.05 arcticus 25 25

Xysticus -1.87 1. 1. -12.94 -12.94 11.77 -1.87 11.92 -1.87 durus 25 87

Wolf spiders

Arctosa -0.26 0 - -14.55 -0.65 14.12 14.55 1.29 0.65 insignita 14 .5 5

Pardosa 0.93 1. - -15.75 2.44 14.67 18.66 -2.06 0.55 furcifera 94 1. 25

Pardosa 0.48 0. - 0.77 0.03 -2.48 -0.35 -2.47 -0.81 groenlan 13 0. dica 39

Pardosa 1.1 2. 3. -15.91 -2.01 15.66 14.63 -0.37 0.11 hyperbor 01 11 ea

Theridii dae

Enoplogn - 0 12 0 0 0 -12.42 0 -12.42 atha 14.82 .4 intrepida 2

Robertus - 0 0 0 0 0 0 11.38 0 fuscus 14.82

Ohlertidi - 0 12 0 0 0 -12.42 0 -12.42 on 14.82 .4 lundbecki 2

Linyphii dae

Agyneta - 0 0 0 0 0 0 12.07 0 allosubtili 14.82 s

Agyneta -0.77 - - -14.04 -0.55 1.1 13.96 1.63 -13.26 jacksoni 1. 0. 1 24

Agyneta - 0 0 0 0 0 0 11.38 0 nigripes 14.82

Agyneta - 0 0 0 0 0 0 12.77 0 subtilis 14.82

Bathypha - 0 0 0 0 0 0 11.38 0 ntes 14.82 eumenis

Dismodic - 12 13 0 0 -12.25 3 1.53 -13.11 us 14.82 .2 .1 decemoc 5 1 ulatus

Erigone - 0 0 0 13.9 0 0 -13.9 -13.9 arctica 14.82

Erigone - 0 0 0 12.11 0 0 -12.11 -12.11 psychrop 14.82 hila

Erigone - 0 0 0 14.05 0 0 -2.67 -14.05 whymper 14.82 i

Hilaira - 0 0 0 0 13.02 0 0 0 herniosa 14.82

Hybauch - 0 13 0 0 0 0.61 0 0.32 enidium 14.82 .8 gibbosu m

Improph -1.87 0 - -12.94 -12.94 13.02 14.24 12.07 0.53 antes 0.

complicat 53 us

Islandian - 0 0 0 0 13.02 0 0 0 a 14.82 princeps

Lepthyph - 0 12 0 0 0 -12.42 11.38 -12.42 antes 14.82 .4 turbatrix 2

Mecynarg -2.56 - - -12.25 0.55 12.25 12.25 12.22 -0.55 us 12 12 borealis .2 .2 5 5

Mecynarg -1.18 - - -13.64 -13.64 13.31 13.64 13.06 13.64 us 0. 13 morulus 29 .6 4

Mecynarg - 0 0 0 15.54 0 13.72 -2.77 -15.54 us 14.82 paetulus

Oreoneti - 12 0 0 0 -12.25 0 -0.87 0 des 14.82 .2 vaginatu 5 s

Pelecopsi - 12 0 0 0 -12.94 0 -0.46 0 s mengei 14.82 .9 4

Pocadicn -2.56 2. 3. -12.25 -12.25 -2.48 -3.88 8.9 -3.88 emis 48 88 american a

Sciastes - 0 0 0 0 0 0 11.38 0 extremus 14.82

Semljicol -1.18 1. - 1.58 -0.14 -16.23 -1.58 -0.18 0.14 a 01 13 obtusus .6 4

Sisicus - 12 12 0 0 -12.25 -12.42 0.23 -12.42 apertus 14.82 .2 .4 5 2

Tiso - 14 14 0 0 0.77 -14.82 -2.66 -14.82 aestivus 14.82 .0 .8 4 2

Wabasso - 0 0 0 0 0 15.1 0 0 cacumina 14.82 tus

Wabasso - 0 0 0 14.41 0 0 -1.42 -14.41 quaestio 14.82

Walckena - 0 12 0 0 0 -12.42 0 -12.42 eria 14.82 .4 castanea 2

Walckena - 0 0 15.22 0 -15.22 -15.22 12.07 0 eria 14.82 clavicorni s

Walckena -0.37 - - 0.37 -0.26 -0.51 -13.48 1.04 -12.85 eria 0. 1. karpinskii 59 34 47

48 Table S4 49 Point-biserial correlations between species abundance and stations (Association function: r.g.). 50 Selected number of species: 19.

Species Stat p-val

Site 1

Pardosa groenlandica 0.298 0.039

Erigone whymperi 0.258 0.046

Mecynargus paetulus 0.238 0.039

Site 2

Otiorhynchus nodosus 0.382 0.003

Arctosa insignita 0.363 0.001

Site 3

Pardosa hyperborea 0.662 0.001

Coccinella transversoguttata 0.507 0.001

Pocadicnemis americana 0.467 0.001

Caenoscelis ferruginea 0.461 0.001

Hybauchenidium gibbosum 0.413 0.001

Hahnia glacialis 0.389 0.001

Thanatus arcticus 0.362 0.003

Otiorhynchus arcticus 0.346 0.004

Mycetoporus nigrans 0.337 0.001

Tiso aestivus 0.282 0.025

Dismodicus decemoculatus 0.27 0.027

Site 1 & 2

Pardosa furcifera 0.35 0.005

Walckenaeria karpinskii 0.273 0.039

Site 2 & 3

Hypsosinga groenlandica 0.264 0.043 51

52 Table S5 53 54 familyname genusnam speciesname authorname wdescriptio wingr wdevelopm flying e n ef ent

Carabidae Nebria rufescens (Ström, Fully 2 Full no 1768) developed, but never observed flying (cf Lindroth 1945a and Lindroth et al 1974). In Scandinavi an specimens the flight muscles may be reduced (cf Gislén & Brinck, 1950). cf Larson and Gigja (1959) mentioned a few icelandic individuals with strikingly better developed wings than in most speciemens

.

Carabidae Patrobus septentrionis Dejan, 1821 Hind wings 2 Full yes always fully developed, and spontaneou s flight has been observed in northern Norway (cf Lindroth 1945a)

Carabidae Bembidion grapii Gyllenhal, Polymorphi 2 Full/reduce yes/n 1827 c hind d o wings. Four types from fully developed to quite rudimentar y, scale like (cf Lindroth 1945a, 1949, 1957, 1958, 1963a, 1979; Lindroth et al 1973). High spatial variation in Greenland cf Lindroth (1959) and table 3.

Carabidae Trichocellu cognatus (Gyllenhal, Hind wings 2 Full yes s 1827) always well developed and at least in europe flight is frequent (cf Lindroth 1945a, Den Boer et al 1980)

Dytiscidae Hyrdoporu morio Aubé, 1838 Hind wings 2 Full yes s always fully developed and able to

fly.

Dytiscidae Colymbete dolabratus (Paykull, Hind wings 2 Full yes s 1798) constantly full and functional, and the species has frequently been observed flying

Gyrinidae Gyrinus opacus Sahlberg, The hinds 2 Full yes 1819 wings are always fully developed and functional

Hydrophilida Helophorus brevipalpis Bedel, 1881 The hind 2 Full yes e wings are always fully developed, and the species is an active flyer (cf Balfour- Browne, 1958), frequently found in wind drift material on seashores in Denmark (cf. M Hansen 1986)

Staphylinida Omalium excavatum Stephens, The hind 2 Full yes e 1834 wings are fully developed and no doubt functional (cf. Lindroth et al 1973)

Staphylinida Xylodromu concinnus (Marsham, No 2 NA NA e s 1802) information

Staphylinida Euaestethu laeviusculus Mannerheim No 1 NA NA e s , 1844 information

Staphylinida Lathrobium fulvipenne Gravenhorst In Iceland 2 Full/reduce yes/n e , 1806 (and d o Greenland) the hind wings are rudimentar y, but a macroptero us form is known from the European Continent (cf. Lindroth et al 1973)

Staphylinida Quedius mesomelinus (Marsham, The hind 2 reduced no e 1802) wings are slightly reduced and according to Lindroth et al (1973) possibly nonfunctio nal.

Staphylinida Quedius fellmanni (Zetterstedt Wings 2 reduced no e , 1838) reduced and not fit for flight (cf Smetana 1964)

Staphylinida Mycetopor nigrans Mäklin, No 1 NA NA e us 1853 information

Staphylinida Micralymm marinum (Ström, The hind 2 reduced no e a 1783) wings are always entirely reduced

Staphylinida Micralymm brevilingue Schiödte, Hind wings 2 reduced no e a 1845 always entirely reduced

Staphylinida Gnypeta cavicollis Sahlberg, The hind 2 Full NA e 1880 wings are fully developed.

Staphylinida Atheta groenlandica Mahler, The hind 2 Full yes

e 1988 wings are always fully developed and no functional doubt.

Staphylinida Atheta islandica (Kraatz, The hind 2 Full yes e 1856) wings are full and functional (Lindroth et al 1973)

Staphylinida Atheta hyperborea Brudin, Hind wings 2 Full NA e 1940 fully developed

Staphylinida Atheta vestita (Gravenhor The hind 2 Full yes e st, 1806) wings are fully developed and probably functional (cf Lindroth et al 1973)

Byrrhidae Simplocari metallica (Sturm, The hind 2 Full NA a 1807) wings are always fully developed.

Byrrhidae Simplocari elongata Sahlberg, The hind 2 Full NA a 1903 wings are fully developed.

Byrrhidae Arctobyrrh subcanus (LeConte, At least in 2 reduced no us 1878) greenland the hind wings are so short that flight is impossible

Byrrhidae Byrrhus fasciatus Forster, The hind 2 Full yes 1771 wings are fully developed and the species regularly flies in sunny wearher (cf Lindroth

1931 and Lindroth et al 1973).

Cryptophagi Caenosceli ferruginea (Sahlberg, Wings fully 2 Full NA dae s 1820) developed

Coccinellidae Nephus redtenbacheri (Mulsant, Wings fully 2 Full yes 1846) developed and functional

Coccinellidae Coccinella transversogutt Falderman, The hind 2 Full yes ata 1835 wings are always fully developed and the species are quite often seen flying short distances, especially when disturbed

Latridiidae Latridius minutus (Linnaeus, Hind wings 2 Full yes 1767) are fully developed and functional

Latridiidae Latridius transversus (Olivier, No 1 NA NA 1790) information

Latridiidae Corticaria rubripes Mannerheim Hind wings 2 Full NA , 1844 are always fully developed

Chrysomelid Phratora polaris (Sparre No 1 NA NA ae Scheider, information 1886)

Curculionida Otiorhynch arcticus (O. The hind 2 rudimentar no e us Fabricius, wings are y 1780) always rudimentar y

Curculionida Otiorhynch nodosus (Müller, Always 2 Wingless no e us 1764) wingless

Curculionida Hypera diversipunctat (Schrank, In 2 reduced no e a 1798) greenland wings highly reduced

Curculionida Dorytomus imbecillus Faust, 1882 No 2 NA NA e information

Curculionida Rutidosom globulus (Herbst, The 2 rudimentar no e a 1795) hindwings y are rudimentar y

Hydrophilida Cercyon obsoletus (Gyllenhal, No 2 NA NA e 1808) information

Staphylinida Euphaleru sorbi (Gyllenhal, No 2 NA NA e m 1810) information

Staphylinida Ocalea cf. picata (Stephens, No 2 NA NA e 1832) information

Staphylinida Othius angustus Stephens, No 2 NA NA e 1833 information

Staphylinida Philonthus cf. cepholates (Gravenhor Macroptero 2 Full NA e st, 1802) us.

Staphylinida Philonthus politus (Linnaeus, Macroptero 2 Full yes e 1758) us and often flying (cf Lindroth et al 1973)

Buprestidae Melanophil acuminata (De Geer, No 2 NA NA a 1774) information

Dermestidae Anthrenus museorum (Linnaeus, No 2 NA NA 1758) information

Dermestidae Attagenus pellio (Linnaeus, No 2 NA NA 1758) information

Dermestidae Dermestes lardarius (Linnaeus, No 2 NA NA 1758) information

Dermestidae Reesa vespulae (Milliron, No 2 NA NA 1939) information

Lyctidae Lyctus brunneus (Stephens, No 2 NA NA 1830) information

Anobiidae Ernobius mollis (Linnaeus, No 2 NA NA 1758) information

Anobiidae Anobium punctatum (De Geer, No 2 NA NA 1774) information

Ptinidae Ptinus fur (Linnaeus, No 2 NA NA 1758) information

Ptinidae Ptinus tectus Boieldieu, No 2 NA NA

1856 information

Ptinidae Ptinus raptor Sturm, No 2 NA NA 1837 information

Trogossitida Tenebroide mauritanicus (Linnaeus, No 2 NA NA e s 1758) information

Malachiidae Malachius aeneus (Linnaeus, No 2 NA NA 1758) information

Cucujidae Oryzaephil surinamensis (Linnaeus, No 2 NA NA us 1758) information

Cryptophagi Cryptopha acutangulus Gyllenhal, No 2 NA NA dae gus 1827 information

Cryptophagi Cryptopha lapponicus Gyllenhal, No 2 NA NA dae gus 1827 information

Latridiidae Corticaria serrata (Paykull, No 2 NA NA 1798) information

Latridiidae Dienerella filum (Aubé, No 2 NA NA 1850) information

Latridiidae Thes bergrothi (Reitter, No 2 NA NA 1880) information

Tenebrionida Tribolium confusum Jacquelin du No 1 NA NA e Val, 1863 information

Tenebrionida Tribolium destructor Uyttenboog No 2 NA NA e art, 1934 information

Tenebrionida Tenebrio obscurus Fabricius, No 2 NA NA e 1792 information

Cerambycida Callidium violaceum (Linnaeus, No 2 NA NA e 1758) information

Cerambycida Gracilia minuta (Fabricius, No 2 NA NA e 1781) information

Cerambycida Molorchus minor (Linnaeus, No 2 NA NA e 1758) information

Cerambycida Phymatode testaceus (Linnaeus, No 2 NA NA e s 1758) information

Cerambycida Pogonoche fasciculatus (De Geer, No 2 NA NA e rus 1775) information

Cerambycida Tetropium castaneum (Linnaeus, No 2 NA NA e 1758) information

Cerambycida Xylotrechu colonus (Fabricius, No 2 NA NA e s 1775) information

Curculionida Otiorhynch sulcatus (Fabricius, No 1 NA NA e us 1775) information

Scolytidae Pityogenes chalcographus (Linnaeus, No 2 NA NA 1761) information 55

Synchronous seasonal dynamics in composition, richness and turnover of arthropod assemblages across the Arctic

Rikke Reisner Hansena*, Ashley Asmusb, Joseph James Bowdena, Sarah Lobodac, Amy M. Ilerd, Paul J. CaraDonnad, Oskar Liset Pryds Hansena,f, Signe Normanda,g, Maarten J.J.E. Loonene, Niels Martin Schmidta,h, Peter Aastrupa,h, Toke Thomas Høyea,i

a. Arctic Research Centre, Department of Bioscience, Aarhus University, Aarhus, Denmark b. Department of Biology, University of Texas, Arlington, USA c. Mc. Gill University, Montréal, Canada d. Aarhus Institute of Advanced Studies, Aarhus University, Aarhus, Denmark e. Arctic Centre, University of Groningen, The Netherlands f. Natural history museum of Aarhus, Aarhus, Denmark g. Ecoinformatics & Biodiversity, Department of Bioscience, Aarhus University, Aarhus, Denmark h. Arctic Ecosystem Ecology, Department of Bioscience, Roskilde, Denmark i. Department of Bioscience, Kalø, Aarhus University, Rønde, Denmark

* Corresponding author: Rikke Reisner Hansen, E-mail: [email protected]

Abstract

Tracking changes in community composition in response to environmental change, requires knowledge on species distributions in space and time. For Arctic Arthropods this knowledge is accumulating slowly due to the prohibitive costs of field work and species identification. Accordingly, it is important to determine to which extent the timing and duration of sampling affects detectability of species. With a solid base in a monitoring scheme incorporating nearly 20 years of consecutive weekly arthropod sampling in North-Greenland, we studied how sampling time and duration of sampling, affected the distribution patterns of Arctic arthropods at family and species level resolutions. Through multivariate models, turnover measures, and species richness, we assessed the seasonal window where arthropod diversity was stable and compared the results to similar data from three other sites in the Arctic, spanning a large geographic and climatic. To further document how this seasonal window could be expected to deviate in response to climate change, we used environmental fitting analysis to correlate findings with environmental variables. Arthropod compositions displayed a similar multivariate pattern across sites, habitats taxonomic resolution and years. However, compositions differed significantly between habitats as well as sampling dates, and there was an interaction between sampling dates and habitats for most sites. Increasing temperatures and advancement in timing of snowmelt extended the seasonal window where richness peaked, and this also varied slightly among sites. The recommendations made here are a useful tool for accruing large scale data on distributional patterns.

Introduction

Global biodiversity loss is occurring at an alarming rate, due to global change and increasing human influence (Butchart et al. 2010). Monitoring of biodiversity, as well as, gaining a full comprehension of the spatial and temporal distribution of species are among the top priorities in ecology (Heller & Zavaleta 2009). Recent accelerating climate change, which alters habitats and compromises species living conditions, increases relevance further (Elmendorf et al. 2012). In spite of increasing focus, sufficient information about the distribution of species richness in large regions, such as the Arctic, remains sparse for many biological groups (CAFF 2013; Hodkinson 2013). Many ecological studies seek to track changes in species diversity and composition throughout the season and over multiple years (Høye & Forchhammer 2008; Spitzer & Jaroš

2009; Bowden et al. 2015), in order to adequately describe the changes in communities brought on by environmental disturbances, such as climate change (Magurran et al. 2010). Although sampling all taxa at multiple sites through a full season of activity is preferable, it may be a daunting task, due to logistical constraints and limited resources available. This predicament, however, raises an interesting question. What fraction of the community is ignored by not covering full seasons?

Numerous studies on species-area relationships have shown strong positive relationships between species richness and turnover and the area or the heterogeneity of the habitat sampled (Connor & McCoy 1979; Lomolino 2000; Hansen et al. 2016a; Hansen et al. 2016b; Vandvik et al. 2016). The temporal pendant of species area relationship, the species–time relationship (Adler & Lauenroth 2003; White & Gilchrist 2007; McGlinn & Palmer 2009), describes the similarly positive relationship between species richness and the length of time a site is studied (Preston 1960), but has not yet been studied as intensively. Temporal species accumulation operates at multiple time scales. Firstly species accumulates intraannually as seasons and conditions shifts. Secondly, variability in weather patterns between years accumulates species over decades or even centuries. Thirdly, there is the evolutionary change in extinction and speciation rates, operating over centuries and millennia (Preston 1960; Magurran et al. 2010; Boggs 2016). Therefore the seasonal development of species is expected to change across years. Moreover, changes over time in a community may occur as the observed accumulation of richness as more species are added throughout the time frame, but also from turnover events where one species replaces another, making both community composition and richness important measures (Yen et al. 2016). Most past and contemporary studies of species-time relationships refer to the interannual variation of species aggregating over the course of multiple years.

A few studies have indicated that lowering the temporal extent and resolution of sampling may have limited impact on species compositions and diversity (Nally et al. 2004; Xu et al. 2015). It is, however, unknown to what spatial and taxonomic extent the generality of these patterns can be inferred. Furthermore, when combining spatial and temporal patterns, the influence of increasing the spatial extent exceeds that of temporal increments on species richness (Erös & Schmera 2010). One way of capturing the seasonal window where local colonization and extinction patterns are low, is to look for statistical regularity and aggregation in the

observed compositional patterns (White & Gilchrist 2007). Considering this, the rate of species accumulation might not be constant over the span of a season, and specific subsets in time may capture significant proportions of the community.

Multiple studies of Arctic distributional patterns have disputed the common conception of the Arctic as a rather homogenous landscape and that heterogeneity between habitats cause great diversification (Bowden & Buddle 2010; Normand et al. 2013; Sikes et al. 2013; Ernst et al. 2016; Hansen et al. 2016a), yet little focus has been given to the temporal variation in arthropod compositions across the active season in Arctic regions. While species in the Arctic are under strong seasonal impacts, the region offers a powerful model system for studying changes over time. Hence temporal studies of Arctic communities can prove useful for overcoming the challenges brought on by sampling large or multiple areas. Moreover, the Arctic region spans large climatic gradients, rendering results of temporal studies applicable at the global scale.

Invertebrates have long been recognized as good indicators of changing communities due to their small body sizes, relatively short generation time and ectothermic lifestyle. Terrestrial arthropods furthermore play an important role in the Arctic food webs as they represent more than one third of all terrestrial organisms, including plants and fungi (CAFF 2013). Since timing of snowmelt and increasing summer temperatures have been identified as important drivers of the active season length of Arctic arthropods (Høye & Forchhammer 2008), they are likely to be important factors in controlling seasonal dynamics of Arctic arthropod communities.

In this study, we evaluate the effect of sampling time and duration on arthropod diversity, turnover and composition. More specifically, we study the seasonal development across multiple sites, taxa and habitats. By including both family- and species level data, we verify whether the patterns are replicated at lower taxonomic resolutions. We investigate the difference in species composition and diversity between habitats and discuss this in relation to temporal variation in species composition and species richness. We propose an optimal timeframe where communities are stable and furthermore suggest a tradeoff between sampling more habitats and shorter timeframes for answering questions regarding compositional changes. An optimal sampling window (a) maximizes community diversity; (b) is characteristic of whole-season community composition; and (c) is robust to effects of inter-annual weather variability and long-term climate

change. The goal of this study was to determine the “optimal” sampling window (date and duration of sampling) for arthropod communities across a wide array of terrestrial arctic locations and habitat types. We divided this goal into two objectives:

Objective 1: We describe seasonal dynamics in community composition and compare indices of arthropod community composition, stability and species richness for sampling windows that vary in date and duration across multiple Arctic sites and habitat types;

Objective 2: We evaluate the effects of inter-annual weather variability and long-term climate trends on community assembly and community indices for each sampling window within sites and habitat types.

Methods

Data

We compiled datasets entailing three different regions within the Arctic (Svalbard, Greenland and Alaska). The Greenland datasets are comprised of two datasets collected as part of Greenland ecosystem Monitoring (GEM). One consists of 18 consecutive years (1996 – 2014) (samples from 2010 were lost in transit from Greenland) of weekly pitfall trap samples from Zackenberg valley, North-east Greenland (74º28’ N, 20º34’ W). The Nuuk dataset, consists of weekly pitfall trap samples (2008 – 2010) in Kobbefjord, South-west Greenland (64°07’N / 51°21'W). Both datasets have been identified to family level and furthermore butterflies, spiders and muscid flies have been identified to the species level for Zackenberg. The Alaskan dataset, sampled at Toolik lake at (68°38' N, 149°35' W) holds two consecutive years (2010 and 2011) of weekly pitfall trap sampling, and has been identified to family level (Rich et al. 2013). From Svalbard (78°55' N, 11°55' W), pitfall sampling occurred from 2009 to 2011 and has been identified to the lowest possible taxonomic level (Table 1). For each dataset, we selected the weeks where all habitats had been equally sampled across all years and standardized to counts per trap per week. The week numbers were calculated from day of year.

Analysis of site characteristics

We examined the distribution of the four regions in relation to climatic gradients within the Arctic using Principal Component Analysis (PCA). Data for five climatic variables were extracted from the worldclim data set (Hijmans et al. 2005): annual mean temperature, mean temperature of the warmest quarter, precipitation of the warmest quarter, precipitation of the coldest quarter, and minimum temperature of the coldest month. PCA was conducted with climate data within the sub- to high Arctic and the most common climate (75-percentile) within Greenland, subarctic, low arctic, and high arctic was delineated. The four regions were plotted according to their climatic conditions with the multidimensional climate space.

Arthropod community composition and seasonal assembly

To visually identify the time frame of the season where arthropod species composition was most stable, we first modeled each dataset through latent variable modelling. Latent variable modelling is a Bayesian model-based approach that explains community composition through a set of underlying latent variables to account for residual correlation, for example due to biotic interaction. This method offers the possibility to adjust the distribution family to account for over-dispersion in count data via negative binomial distribution. Thus, it also accounts for the increasing mean-variance relationship without confounding location with dispersion (Hui et al. 2015). Three “types" of models may be fitted: 1) With covariates and no latent variables, boral fits independent response GLMs such that the columns of y are assumed to be independent; 2) With no covariates, boral fits a pure latent variable model (Rabe-Hesketh et al. 2004) to perform model-based unconstrained ordination (Hui et al. 2015); 3) With covariates and latent variables, boral fits correlated response GLMs, with latent variables.

We created latent variable models at site levels for each dataset with two latent variables. to visualize how the arthropod communities were distributed. This method is comparable to a two dimensional non metric multidimensional scaling (NMDS) plot. From the latent variable model, we extracted the posterior median values of the latent variables which we used as coordinates on ordination axes to represent family level composition at plot level (Hui et al. 2015). We drew convex ellipses around posterior median values belonging to each habitat for each dataset based on 95 percent confidence limits of the yearly averages. For this purpose, we used the function ‘ordiellipse’ in the r package ‘Vegan’ (Oksanen et al. 2016). We repeated this analysis at species level for the Zackenberg dataset.

We then tested whether the seasonal development in arthropod composition differed significantly between habitats through a multivariate extension of General Linear Models (GLMs), using the function ‘manyglm’ in the package ‘mvabund’ (Wang et al. 2012). This recently developed method offers the possibility to model distributions based on count data by assuming a negative binomial distribution. We tested for main effects of week and habitats and for an interaction between the two terms. We then repeated the latent variable modeling at habitat level for all datasets with a significant interaction term, and calculated convex ellipsoids around posterior median values belonging to each week based on the standard error of the yearly averages. We extracted the centroid of each ellipsoid, which we plotted on top of the latent variable plot. At habitat level, we tested if the seasonal development differed significantly between years, by testing for an interaction between time of season (week) and year.

Sampling window – based arthropod species richness and turnover

Multivariate methods for measuring beta diversity have been shown robust compared to classical methods. This is due to the dependency upon gamma diversity in classical methods, such that when sampling effort increases, so does beta diversity along with gamma diversity (Bennett & Gilbert 2016). We used the centroids extracted from the latent variable models for each habitat from the Zackenberg family level dataset. While the distance to the group centroid is a common way of measuring spatial beta diversity (Anderson et al. 2006), it describes the yearly variations in beta diversity within the weeks and not the unidirectional, temporal drift in multivariate space from one week to the next. We used the distance between week to week centroids as a measure of temporal turnover, and used the average area of the two ellipsoids as error bars to represent the interannual variation. We calculated species richness for each week and each habitat across all years. To identify the timeframe and length of timeframe in which the highest number of species were represented, we calculated species richness for each combination of timing and duration; four week sampling at nine different points, three week sampling at ten different points, two week sampling at eleven different points of the season and one week sampling at 12 different points. We explored how what proportion of total season species richness was captures for each of the sampling strategies.

Climatic variability and the effect on seasonal development

We used the Zackenberg family level dataset to analyze the sensitivity of diversity to climatic variables. This dataset is the only one, where environmental variables were measured every year. Timing of snowmelt has been shown to occur significantly earlier, as well as average may-august temperatures to become significantly warmer (Bowden et al. 2015). Timing of snowmelt has furthermore proven a significant predictor of season length for arthropods (Høye et al. 2014). To identify if the different seasonal development between years was climate related, we ran a multivariate correlation analysis with the function ‘envfit’ in the ‘vegan’ package. We used the posterior median values of the latent variables as response variables and timing of snowmelt, average may-august temperature, year and week as predictor variables. We examined how the changing climate has affected seasonal development of arthropod assemblages by dividing the Zackenberg family level dataset into years of early and late snowmelt, as well as, warm and cold years. Early snowmelt years were categorized as years where the average day of year for snowmelt lay below the average day of year for all years. Similarly, warm years where categorized as years where the average summer (may august) temperature lay above the average across years. Following these divisions, we calculated species richness for each category.

Results

Analysis of site characteristics

The investigated sites were distributed across the climatic space of the PCA. Two sites were distributed within the high Arctic climate (Ny Aalesund and Zackenberg) and two sites in the low Arctic (Nuuk and Toolik) (Fig. 1). The first two axis of the PCA explained 87% of the variation and was mainly correlated with precipitation of the warmest quarter (1. Axis) and temperature of the warmest quarter (2. axis). Toolik have warmer summers then the other sites and Nuuk, Ny Aalesund and Zackenberg is distributed along a gradient from wet to dryer conditions.

Arthropod community composition and seasonal assembly

Arthropod compositions differed significantly between all of the investigated habitats at each site and there was a significant effect of week (Table 2). The moisture gradient was a driving component across sites. At Zackenberg, there was a clear distinction between arthropod communities, in the wet habitats and arthropod communities in the dry and mesic habitats. For

Nuuk habitats, the biggest distinction was found between the shrub covered habitat and the wet, with the dry habitat as an intermediate (Fig. 2). However, the interaction between seasonal development and habitat was not significant for Toolik and Ny Aalesund, even though arthropod compositions differed significantly between both habitats and week (Table 2).

The weekly development through the season, displayed a characteristic u-shaped pattern for all the habitats at all the sites. The arthropod compositions were more similar towards the beginning and the end of the season. There was a change in multivariate space from week to week throughout the season, yet, the communities seemed to stabilize during mid-season with less distance between the weekly centroids (Fig. 3).

Sampling window – based on arthropod species richness and turnover

At Zackenberg, the weeks of significantly (non-overlapping error bands) highest richness where between week 29 until week 33 for the dry habitat. The mesic habitat varied from week to week and richness never reached a plateau. In the wet habitat, richness increased until week 29 and stabilized before dropping in week 32. Turnover patterns were opposite to richness and stabilized at the lowest point between week 29 and 30, until increasing again between week 32 and 33. Turnover in the mesic habitat did not reach a stable low point until between week 32 and 33. In Nuuk, there was inter annual variability both in species richness and turnover, but species richness in the shrubs and the dry habitat peaked from week 29 to 32. The wet habitat had significantly higher species richness in week 32 and it was overall lower than in the shrub and dry habitat. Turnover did not differ significantly in any of the habitats between any of the weeks for Nuuk. Species richness at Toolik was significantly highest in week 25 to 30 and declined drastically in week 31. Turnover was lowest between weeks 26 to 28. In Svalbard, species richness was significantly highest in week 29 with a maximum of 13 taxa present, and turnover did not change significantly between any of the weeks (Fig. 4).

Analysis of sampling strategy in Zackenberg mirrored the weeks of highest richness and turnover and revealed that sampling either week in a one-week sampling strategy between weeks 28-32, yielded 69-72 percent of the taxa present when sampling a full season. Four weeks of sampling that time frame yielded 79-81 percent of the full season taxa detected. The optimal timespan for one week sampling strategy in Nuuk was starting week 28 and 29 which detected

55-57 percent of the full season taxonomic richness, whereas four weeks of sampling with the same starting weeks presented 64-69 percent. In Toolik, the optimal starting week was between week 26 and 28 for one week sampling duration, which detected 59-61 percent of the taxa present throughout a full season and four week in with the same starting weeks yielded 75-77 percent. In Ny-Aalesund, there was no significant difference between sampling duration (Fig. 5).

The effect of climatic variability on seasonal development

Temperatures had a significant effect on arthropod compositions in the dry and mesic habitat at Zackenberg and year was highly correlated with temperature. In the wet habitat, there was no significant effect of snowmelt, temperature and year (Fig. 6). The average day of year for snowmelt across all years was day 152. For six years (2004, 2005, 2009, 2010, 2011 and 2013) the average snowmelt date lay below this threshold, and these were categorized as early snowmelt years. Average may-august temperature was 2.23 °C and the years 1996 – 2002 and 2014 had an average may-august temperature below this threshold. Taxonomic richness was higher and peaked earlier for the warm and early snowmelt years (Fig. 7).

Discussion

Across sites, habitats, and taxonomic levels, arthropod compositions displayed similar trends in seasonal development with less seasonal drift during the peak season weeks (i.e. weeks with low turnover). In present study one week of carefully planned sampling at Zackenberg research station, represented up to 72 percent of richness detected through a full season of sampling.

Zackenberg monitoring program presents us with a unique chance of studying temporal trends as it has been operating for 20 years and has collected data on arthropods with weekly intervals throughout the active season. A long term monitoring program at ecosystem level is quite unique and can answer questions entailing interaction dynamics through trophic mismatch and cascades (Mortensen et al. 2014). Additional short term sampling schemes with high spatial resolution across multiple sites could help fill gaps in our knowledge of species distributions and add information to the model predictions of climate change effects on biodiversity. This could be partly accomplished through citizen science schemes as well as cross disciplinary cooperation among researches, where full seasonal commitments may not be an option, but simultaneous

multiple year and site collections are. With a strong standardized sampling design available, people without scientific training can carry out fieldwork and will be capable of generating data, which will help biologists answer questions regarding species distributions. Utilizing local communities is a valuable way of accruing large amounts of data otherwise unattainable. It is important to note that we are not trying to limit the temporal extent, nor resolution of long term monitoring schemes. These recommendations are solely for studies of distributional patterns and inferences on life history traits, species interactions and other species specific responses, require longer timeframes to answer. We are still dependent on long term sampling strategies spanning full seasons with higher detailed resolution levels. However, these findings have merits in addition to the high resolution monitoring schemes.

Climate changes are altering habitats (Myers-Smith et al. 2011; Elmendorf et al. 2012) and expanding the active season for arthropods (Høye & Forchhammer 2008), affecting arthropod compositions across the Arctic (Bowden & Buddle 2010; Hansen et al. 2016b). Time and duration of peak richness may therefor change with a changing climate. Our results show that increasing temperatures does not cause shifts in timing of peak richness, but increases species richness and extends the time frame, where species richness is significantly highest. Sampling strategies have previously been studied in other ecosystems and for other groups of organisms with the intent of describing effect of lower sampling extent and intensity on diversity patterns. For instance temporal aggregation of species had a substantial effect on the species-time relationship of rodent populations in Arizona (White & Gilchrist 2007). A study of ciliated protozoa in the Yellow sea, north China, a sampling regime with one third the amount of original sampling, recovered > 75% information of the total seasonal variability and > 90 percent of protozoan ciliate species present (Xu et al. 2015). Compared to our findings of little effect of changing climates on timing of optimal sampling window, as well as, the similar u-shaped pattern across a large climatic gradient, super generalizations to other ecosystems and regions is possible.

The study indicates that short term sampling procedures comes with some considerations. The peak in species richness varies between sites with earlier seasonal peaks at the lower latitudes. As a consequence, sampling one week requires careful consideration of the geographic location of the study. Interaction between seasonal development and habitat coupled with

different temporal development in species richness underlines not only a need to sample multiple habitats, but also a detailed consideration towards habitat characteristics, such as soil moisture and vegetation structures. For instance, wet and mesic habitats display the lowest and latest peaks in species richness. These results mirror a previous study from the Godthåbsfjord area in Nuuk, where diversity and species richness where overall lowest in the fens (Hansen et al. 2016b). The year to year variation in turnover for some sites (Nuuk and Ny Aalesund), which is less pronounced for species richness, indicates that abundances vary substantially between years. However, multiple inter annual short term samplings will help counterbalance the year-to-year variability. Shorter sampling windows are more sensitive to stochastic events, such as less optimal weather and it is therefore always desirable to sample the longest term possible.

This study is to our knowledge the first to answer questions regarding sampling time and duration of Arctic arthropods, and may prove valuable in the planning of logistically challenging field work (Post & Høye 2013). In conclusion, should compromises be made between spatial and temporal sampling resolution and extent, we believe that an increase of spatial, at the cost of temporal intra annual extent and resolution, is preferable. In combination with the long term monitoring programs, these recommendations will go a long way in mapping species distributions, as well as, responses to climate change.

Acknowledgements

We gratefully acknowledge the tremendous efforts put forward in connection with data collection. All staff included in the GEM programs, as well as people involved in data collection at the two field sites Toolik and Ny-Aalesund. SN was supported by the Villum foundation’s Young Investigator Programme (VKR023456).

Adler PB, and Lauenroth WK. 2003. The power of time: spatiotemporal scaling of species diversity. Ecology Letters 6:749-756 DOI: 10.1046/j.1461-0248.2003.00497.x. Anderson MJ, Ellingsen KE, and McArdle BH. 2006. Multivariate dispersion as a measure of beta diversity. Ecology Letters 9:683-693 DOI: 10.1111/j.1461-0248.2006.00926.x. Bennett JR, and Gilbert B. 2016. Contrasting beta diversity among regions: how do classical and multivariate approaches compare? Global Ecology and Biogeography 25:368-377 DOI: 10.1111/geb.12413. Boggs CL. 2016. The fingerprints of global climate change on insect populations. Current Opinion in Insect Science 17:69-73 DOI: http://dx.doi.org/10.1016/j.cois.2016.07.004. Bowden JJ, and Buddle CM. 2010. Determinants of ground-dwelling spider assemblages at a regional scale in the Yukon Territory, Canada. Ecoscience 17:287-297 DOI: 10.2980/17-3-3308.

Bowden JJ, Eskildsen A, Hansen RR, Olsen K, Kurle CM, and Høye TT. 2015. High-Arctic butterflies become smaller with rising temperatures. Biology Letters 11 DOI: 10.1098/rsbl.2015.0574. Butchart SHM, Walpole M, Collen B, van Strien A, Scharlemann JPW, Almond REA, Baillie JEM, Bomhard B, Brown C, Bruno J, Carpenter KE, Carr GM, Chanson J, Chenery AM, Csirke J, Davidson NC, Dentener F, Foster M, Galli A, Galloway JN, Genovesi P, Gregory RD, Hockings M, Kapos V, Lamarque J-F, Leverington F, Loh J, McGeoch MA, McRae L, Minasyan A, Morcillo MH, Oldfield TEE, Pauly D, Quader S, Revenga C, Sauer JR, Skolnik B, Spear D, Stanwell-Smith D, Stuart SN, Symes A, Tierney M, Tyrrell TD, Vié J-C, and Watson R. 2010. Global Biodiversity: Indicators of Recent Declines. Science 328:1164-1168 DOI: 10.1126/science.1187512. CAFF. 2013. Arctic Biodiversity Assessment. Status and trends in Arctic biodiversity. Akureyri. Connor EF, and McCoy ED. 1979. The Statistics and Biology of the Species-Area Relationship. The American Naturalist 113:791-833 Elmendorf SC, Henry GHR, Hollister RD, Bjork RG, Boulanger-Lapointe N, Cooper EJ, Cornelissen JHC, Day TA, Dorrepaal E, Elumeeva TG, Gill M, Gould WA, Harte J, Hik DS, Hofgaard A, Johnson DR, Johnstone JF, Jonsdottir IS, Jorgenson JC, Klanderud K, Klein JA, Koh S, Kudo G, Lara M, Levesque E, Magnusson B, May JL, Mercado-Diaz JA, Michelsen A, Molau U, Myers-Smith IH, Oberbauer SF, Onipchenko VG, Rixen C, Schmidt NM, Shaver GR, Spasojevic MJ, Porhallsdottir PE, Tolvanen A, Troxler T, Tweedie CE, Villareal S, Wahren CH, Walker X, Webber PJ, Welker JM, and Wipf S. 2012. Plot-scale evidence of tundra vegetation change and links to recent summer warming. Nature Climate Change 2:453-457 DOI: 10.1038/Nclimate1465. Ernst CM, Loboda S, and Buddle CM. 2016. Capturing northern biodiversity: diversity of arctic, subarctic and north boreal beetles and spiders are affected by trap type and habitat. Insect Conservation and Diversity 9:63-73 DOI: 10.1111/icad.12143. Erös T, and Schmera D. 2010. Spatio-temporal scaling of biodiversity and the species–time relationship in a stream fish assemblage. Freshwater Biology 55:2391-2400 DOI: 10.1111/j.1365- 2427.2010.02438.x. Hansen RR, Hansen OLP, Bowden JJ, Normand S, Bay C, Sørensen JG, and Høye TT. 2016a. High spatial variation in terrestrial arthropod species diversity and composition near the Greenland ice cap. Polar Biology:1-10 DOI: 10.1007/s00300-016-1893-2. Hansen RR, Hansen OLP, Bowden JJ, Treier UA, Normand S, and Høye T. 2016b. Meter scale variation in shrub dominance and soil moisture structure Arctic arthropod communities. Peerj 4:e2224 DOI: 10.7717/peerj.2224. Heller NE, and Zavaleta ES. 2009. Biodiversity management in the face of climate change: A review of 22 years of recommendations. Biological Conservation 142:14-32 DOI: http://dx.doi.org/10.1016/j.biocon.2008.10.006. Hijmans RJ, Cameron SE, Parra JL, Jones PG, and Jarvis A. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965-1978 DOI: 10.1002/joc.1276. Hodkinson ID. 2013. Terrestrial and Freshwater Invertebrates. In: Meltofte H, ed. Arctic Biodiverisity AssessmentStatus and trends in Arctic biodiversity. Akureyri, 195-219. Høye TT, Eskildsen A, Hansen RR, Bowden JJ, Schmidt NM, and Kissling WD. 2014. Phenology of high- arctic butterflies and their floral resources: Species-specific responses to climate change. Current Zoology 60:243-251 Høye TT, and Forchhammer MC. 2008. Phenology of High-Arctic Arthropods: Effects of Climate on Spatial, Seasonal, and Inter-Annual Variation. In: Hans Meltofte TRCBEMCF, and Morten R, eds. Advances in Ecological Research: Academic Press, 299-324.

Hui FKC, Taskinen S, Pledger S, Foster SD, and Warton DI. 2015. Model-based approaches to unconstrained ordination. Methods in Ecology and Evolution 6:399-411 DOI: 10.1111/2041- 210X.12236. Lomolino MV. 2000. Ecology’s most general, yet protean 1 pattern: the species-area relationship. Journal of Biogeography 27:17-26 DOI: 10.1046/j.1365- 2699.2000.00377.x. Magurran AE, Baillie SR, Buckland ST, Dick JM, Elston DA, Scott EM, Smith RI, Somerfield PJ, and Watt AD. 2010. Long-term datasets in biodiversity research and monitoring: assessing change in ecological communities through time. Trends in Ecology & Evolution 25:574-582 DOI: http://dx.doi.org/10.1016/j.tree.2010.06.016. McGlinn DJ, and Palmer MW. 2009. Modeling the sampling effect in the species–time–area relationship. Ecology 90:836-846 DOI: 10.1890/08-0377.1. Mortensen LO, Jeppesen E, Schmidt NM, Christoffersen KS, Tamstorf MP, and Forchhammer MC. 2014. Temporal trends and variability in a high-arctic ecosystem in Greenland: multidimensional analyses of limnic and terrestrial ecosystems. Polar Biology 37:1073-1082 DOI: 10.1007/s00300- 014-1501-2. Myers-Smith IH, Forbes BC, Wilmking M, Hallinger M, Lantz T, Blok D, Tape KD, Macias-Fauria M, Sass- Klaassen U, Levesque E, Boudreau S, Ropars P, Hermanutz L, Trant A, Collier LS, Weijers S, Rozema J, Rayback SA, Schmidt NM, Schaepman-Strub G, Wipf S, Rixen C, Menard CB, Venn S, Goetz S, Andreu-Hayles L, Elmendorf S, Ravolainen V, Welker J, Grogan P, Epstein HE, and Hik DS. 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environmental Research Letters 6 DOI: 10.1088/1748-9326/6/4/045509. Nally RM, Fleishman E, and Murphy DD. 2004. Influence of Temporal Scale of Sampling on Detection of Relationships between Invasive Plants and the Diversity Patterns of Plants and Butterflies. Conservation Biology 18:1525-1532 DOI: 10.1111/j.1523-1739.2004.00399.x. Normand S, Randin C, Ohlemuller R, Bay C, Hoye TT, Kjaer ED, Korner C, Lischke H, Maiorano L, Paulsen J, Pearman PB, Psomas A, Treier UA, Zimmermann NE, and Svenning JC. 2013. A greener Greenland? Climatic potential and long-term constraints on future expansions of trees and shrubs. Philosophical Transactions of the Royal Society B-Biological Sciences 368 DOI: 10.1098/Rstb.2012.0479. Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, Simpson GL, Solymos P, Stevens MHH, and Wagner H. 2016. vegan: community ecology package. 2.0-10 ed. Post E, and Høye TT. 2013. Advancing the long view of ecological change in tundra systems. Philosophical Transactions of the Royal Society B-Biological Sciences 368 DOI: 10.1098/Rstb.2012.0477. Preston FW. 1960. Time and Space and the Variation of Species. Ecology 41:612-627 DOI: 10.2307/1931793. Rabe-Hesketh S, Skrondal A, and Pickles A. 2004. Generalized multilevel structural equation modeling. Psychometrika 69:167-190 DOI: 10.1007/bf02295939. Rich ME, Gough L, and Boelman NT. 2013. Arctic arthropod assemblages in habitats of differing shrub dominance. Ecography 36:994-1003 DOI: 10.1111/j.1600-0587.2012.00078.x. Sikes DS, Draney ML, and Fleshman B. 2013. Unexpectedly high among-habitat spider (Araneae) faunal diversity from the Arctic Long-Term Experimental Research (LTER) field station at Toolik Lake, Alaska, United States of America. Canadian Entomologist 145:219-226 DOI: 10.4039/tce.2013.5. Spitzer K, and Jaroš J. 2009. Long-term monitoring of moth populations (Lepidoptera) associated with a natural wetland forest: synthesis after 25 years. Terrestrial Arthropod Reviews 1:155-163

Vandvik V, Klanderud K, Meineri E, Måren IE, and Töpper J. 2016. Seed banks are biodiversity reservoirs: species–area relationships above versus below ground. Oikos 125:218-228 DOI: 10.1111/oik.02022. Wang Y, Naumann U, Wright ST, and Warton DI. 2012. mvabund– an R package for model-based analysis of multivariate abundance data. Methods in Ecology and Evolution 3:471-474 DOI: 10.1111/j.2041- 210X.2012.00190.x. White EP, and Gilchrist MA. 2007. Effects of population-level aggregation, autocorrelation, and interspecific association on the species-time relationship in two desert communities. Evolutionary Ecology Research 9:1329-1347 Xu H, Yong J, and Xu G. 2015. Sampling frequency of ciliated protozoan microfauna for seasonal distribution research in marine ecosystems. Marine Pollution Bulletin 101:653-659 DOI: http://dx.doi.org/10.1016/j.marpolbul.2015.10.034. Yen JDL, Thomson JR, Keith JM, Paganin DM, and Nally RM. 2016. How do different aspects of biodiversity change through time? A case study on an Australian bird community. Ecography:n/a-n/a DOI: 10.1111/ecog.02306.

1 Table 1

2 Overview of sites: The asterisk indicates that the dataset is used for supportive analyses and the 3 results can be found in supplement material

Dataset Position Years Selected Sampling Taxonomic Habitats sampling intensity resolution period (weeks) Zackenberg 74º28’ N, 1996 – 2014 24:35 Weekly Family level Dry,mesic (Greenland) 20º34’ W and wet Nuuk 64°07’N / 2008 – 2014 25:39 Weekly Family level Dry, wet (Greenland) 51°21'W and shrub Ny 78°55' N, 2009 – 2011 27:33 Two-day Family, Mesic and Aalesund 11°55' W interval super family dry (Svalbard) and order Toolik lake 68°38' N, 2010 – 2011 21:30 Weekly Family level Open and 149°35' W shrub Zackenberg* 74º28’ N, 1996 – 2014 24:35 Weekly Species level Dry, mesic 20º34’ W (butterflies, and wet spiders and muscid flies)

4

5

6

7

8

9

10

11

12

13

14 Table 2

15 Table of deviance. Table showing results of the multivariate GLM, testing the difference 16 between habitat and weeks, as well as the interaction between them at family and species level. 17 The Zackenberg family level interaction between year and week at habitat level is also shown.

Dataset Variable Degrees of Deviance P freedom Zackenberg Week 1 753 <0.001 Habitat 2 2848.1 <0.001 Week:Habitat 2 554.7 <0.001 Nuuk Week 1 493.2 <0.001 Habitat 2 761.7 <0.001 Week:Habitat 2 186.1 <0.001 Ny Aalesund Week 1 133.6 <0.001 Habitat 1 130.8 <0.001 Week:Habitat 1 23.9 0.190 Toolik lake Week 1 508.1 0.010 Habitat 1 244.8 <0.001 Week:Habitat 1 91.6 0.265 Zackenberg Week 1 511.2 <0.001 species level Habitat 2 2194.7 <0.001 Week:Habitat 2 97.5 <0.001 Zackenberg wet Year 1 198 <0.001 Week 1 382.2 <0.001 Year:Week 1 65.9 0.020 Zackenberg mesic Year 1 233.5 <0.001 Week 1 640.9 <0.001 Year:Week 1 93.2 0.010 Zackenberg dry Year 1 406 <0.001 Week 1 321.3 <0.001 Year:Week 1 130.3 0.010

18

19

20

21

22

23

24 Figure 1

25 Climate space. Plot of the principal component analyses with the following variables 26 downloaded from http://www.worldclim.org/bioclim: annual mean temperature, minimum 27 temperature of the coldest month, mean temperature of the warmest quarter, precipitation of the 28 warmest quarter and precipitation of the coldest quarter. Climate data is plotted with grey colors 29 and the sites in blue. Dashed lines delineate the 75-percentile of the climatic conditions within 30 high-, low- and sub-Arctic, respectively.

31

32

33

34 Figure 2

35 Site level latent variable plots. Latent variable plots displaying how arthropod communities are 36 distributed along two latent variables. Ellipsoids show the 95 percent confidence bands around 37 habitats.

38

39

40

41

42 Figure 3

43 Seasonal variation in arthropod community composition. Latent variable plots showing the latent 44 variable models divided onto habitats. The division was only made if there was a significant 45 interaction between habitat and week. The points below are centroids of the standard error of the 46 weekly centroids.

47

48

49

50

51 Figure 4

52 Richness and turnover plots. Family level richness and turnover plots showing species richness, 53 as well as, distance between centroids for the weekly average of across years. Both indices are 54 shown with standard errors of interannual variation.

55

56

57

58

59

60

61

62

63

64

65

66 Figure 5

67 Proportion of total species richness. Line graph showing how much of the total species richness 68 is detected with four different sampling durations (one week, two week, three week and four 69 week). The week number on the x-axis shows starting week and the y-axis shows the proportion 70 of the total species richness detected within a year. Error bars are standard error of the 71 interannual variation.

72

73

74

75

76

77

78

79 Figure 6

80 Correlation with environmental variables. Biplots showing only significant (p<0.05) variables 81 from the environmental fitting analysis

82

83

84

85

86

87

88

89

90

91

92

93

94

95 Figure 7

96 A) Climatic variability and effect on species richness. Seasonal development of species richness 97 divided into years with late (blue colors) versus early (red colors) snowmelt and cold (blue 98 colors) versus warm (red colors) years. Error bars represents standard error of the weekly 99 average across years.

100

101

102

103 Supplements

104 Fig S1: Species level latent variable models for Zackenberg data. The top figure shows all 105 habitats and the three figures below shows the seasonal weekly development divided onto 106 habitats

107

Polar Biol (2015) 38:559–568 DOI 10.1007/s00300-014-1622-7

ORIGINAL PAPER

Habitat-specific effects of climate change on a low-mobility Arctic spider species

Joseph J. Bowden • Rikke R. Hansen • Kent Olsen • Toke T. Høye

Received: 10 June 2014 / Revised: 8 October 2014 / Accepted: 12 November 2014 / Published online: 2 December 2014 Springer-Verlag Berlin Heidelberg 2014

Abstract Terrestrial ecosystems are heterogeneous hab- morphology. This provided a unique opportunity to mea- itat mosaics of varying vegetation types that are differen- sure recruitment potential because we could assume that all tially affected by climate change. Arctic plant juvenile crab spiders belonged to that species. We deter- communities, for example, are changing faster in moist mined sex, stage, and size of all specimens (n = 2,115). habitats than in dry habitats and abiotic changes like Body size variation was significantly related to the timing snowmelt vary locally among habitats. Such differences of snowmelt and differed significantly between the sexes between habitats may influence the effects of climate and habitats with the spiders in the mesic habitat showing a changes on animals and this could be especially true in stronger temporal response to later snowmelt. Juvenile/ low-mobility species. Suitable model systems to test this female ratios also differed significantly between habitats; idea, however, are rare. We examined how proxies of as did the overall abundance of individuals. We found reproductive success (body size, juvenile/female ratios) significant main effects of snowmelt and habitat on sex and sex ratios have changed in low-mobility crab spiders ratio with the proportion of females decreasing signifi- collected systematically over a 17-year period (1996–2012) cantly in response to later snowmelt only in the mesic sites. from two distinct habitats (mesic and arid dwarf shrub Effects of climate change may be masked by habitat dif- heath) at Zackenberg in northeast Greenland. We identified ferences and have implications for future range changes of all adults in the collection to confirm that they represented species at both small and large spatial extents. We rec- just one species (Xysticus deichmanni Sørensen) based on ommend that local habitat differences are included in analyses of species responses to climate change.

Electronic supplementary material The online version of this Keywords High Arctic Arthropod Zackenberg article (doi:10.1007/s00300-014-1622-7) contains supplementary Thomisidae Mesic heath Dry heath material, which is available to authorized users.

J. J. Bowden (&) R. R. Hansen T. T. Høye Arctic Research Centre, Aarhus University, C.F. Møllers Alle´ 8, Introduction Bldg. 1110, 8000 A˚ rhus C, Denmark e-mail: [email protected] Terrestrial ecosystems are heterogeneous mosaics of eco- K. Olsen regions at a global scale (Olson et al. 2001) and vegetation Natural History Museum Aarhus, Wilhelm Meyers Alle´ 210, types at regional scales (Walker et al. 2005). Similarly, ˚ 8000 Arhus C, Denmark there is great heterogeneity in regional (Ru¨hlund et al. T. T. Høye 2013) and local changes to temperature and soil moisture Aarhus Institute of Advanced Studies, Aarhus University, across regions like the Arctic tundra. Specifically, some Høegh-Guldbergs Gade 6B, Bldg. 1632, 8000 A˚ rhus C, Denmark areas are becoming wetter, due to permafrost melt and some dryer, due to increased net evapotranspiration (Zona T. T. Høye Department of Bioscience, Kalø, Aarhus University, Grena˚vej et al. 2009). Recent climate change has been occurring at 14, 8410 Rønde, Denmark an accelerated rate in the Arctic with substantial ecological 123 560 Polar Biol (2015) 38:559–568 repercussions (ACIA 2005; Post et al. 2009; Høye et al. remain on the ground below or adjacent to vegetation 2013); however, local habitat may modulate these changes, (Leech 1996) due to their slow-moving sit-and-wait life- for example, snowmelt occurs later and vegetation is typ- style. Crab spiders, like other arthropods with low mobility ically higher in wet low-lying areas (Stow et al. 2004). (e.g., Thomas 2000; Thomas et al. 2001; Warren et al. Studies and syntheses of Arctic plant communities have 2001), are therefore likely more strongly influenced by revealed that species’ responses to climate change are local scale differences in habitat. Additionally, snowmelt- idiosyncratic (Myers-Smith et al. 2011; Elmendorf et al. induced variation over time or between habitats in the 2012), but generally these changes (species composition Arctic could directly (e.g., sex-specific responses to climate and/or local abundance) have been more pronounced in wet change/increased juvenile mortality) or indirectly (e.g., tundra habitats (Elmendorf et al. 2012; Schmidt et al. alterations to female body size) influence sex ratios, 2012a). Local effects to individuals are highly dependent fecundity and/or subsequent recruitment. Since females upon small-scale variation in temperature and moisture command significantly greater reproductive requirements, regime, and climate-induced shifts can thus have signifi- they frequently display greater responses in body size to cant impacts upon populations and communities over small gradients in resource availability (e.g., Høye et al. 2009; spatial scales (Kudo and Hirao 2006; Høye et al. 2007a, Høye and Hammel 2010). 2013). The biological effects of climate change occur Terrestrial arthropods comprise a substantial portion of in situ and act upon traits that dictate whether individuals the Arctic’s terrestrial biomass and species diversity (Cal- in a population will be able to adapt to a changing envi- laghan et al. 2004). Spiders serve as a central component to ronment or disperse to another more suitable habitat Arctic terrestrial food webs (Hodkinson and Coulson 2004; (Bourne et al. 2014). Roslin et al. 2013), serving as food for many birds and Body size is a particularly important trait to the life small mammals, but also as top predators of mesofauna and history (e.g., reproductive success), physiology (e.g., ther- are well represented in diversity and abundance on the mal limits) and ecology (e.g., resource acquisition) of Arctic tundra (e.g., Bowden and Buddle 2010). Long-term individuals in a population (Chown and Gaston 2010). In data on Arctic arthropods is scarce; Zackenberg, Green- terrestrial arthropods, body size is primarily influenced by land, however, represents a long-term monitoring program food availability and temperature (Chown and Gaston that has been collecting terrestrial arthropods and envi- 2010) and is often positively correlated to fecundity, mass ronmental data for nearly 20 years in the high Arctic. Two (e.g., Bowden and Buddle 2012) and overall fitness (Jakob of the habitat types being monitored are mesic and arid et al. 1996). In the Arctic, however, snowmelt has proven heath in which the mesic habitat undergoes substantially to be a particularly important predictor of phenology and later snowmelt on average (Høye and Forchhammer 2008). body size variation (e.g., Høye et al. 2009, 2013). Like Previous literature suggests that our focal species, Xysticus insects (Stillwell et al. 2010), female spiders are frequently deichmanni Sørensen, can be found more frequently in arid known to exhibit higher phenotypic plasticity in body size tundra habitats (Leech 1996; Wyant et al. 2011). Our than males in response to environmental gradients (Uhl objective was to determine how habitat-specific snowmelt et al. 2004). has influenced inter-annual variation in body size, recruit- Previous studies have shown that variation in body size ment and sex ratio in a species with low mobility (X. de- is a good indicator of sensitivity to variation in growth ichmanni). We predicted that body size would decrease season length in wolf spiders (Lycosidae) (Høye et al. significantly with later snowmelt due to the shorter season 2009; Høye and Hammel 2010; Bowden et al. 2013). It is, with which to obtain resources. We also predicted that however, still unclear whether cohort-specific changes in there would be a significant difference between the two body size are reflected in more direct measures of fitness habitats with smaller body size in the mesic habitat due to and whether similar responses are found in other taxa and the substantially shorter snow-free period, resulting in a in different habitats. The crab spiders (Thomisidae) are significant snowmelt–habitat interaction. Furthermore, closely related to wolf spiders, but represent an ecologi- because of their relatively higher costs of reproduction, we cally distinct group of spiders, which comprises species of predicted that females would exhibit a stronger response to non-web building, sit-and-wait predators. While it is clear later snowmelt than males. We also predicted that there that the timing of snowmelt significantly affects body size would be significant main effects of snowmelt and habitat, variation in the Arctic wolf spider [Pardosa glacialis as well as, a significant interaction between snowmelt and (Thorell)] (Høye et al. 2009), the crab spiders are less habitat on recruitment (indexed by proportion of juveniles mobile, meaning that as juveniles they possess a low pro- to adult females) and sex ratios (indexed by proportion of pensity for dispersal (Morse 1992) and as adults that they females to males).

123 Polar Biol (2015) 38:559–568 561

Materials and methods per the Monitoring Program (see Høye and Forchhammer 2008); representing each two mesic and two arid sites, Sampling and specimens respectively, over a 17-year period from 1996 to 2012 (2,010 samples were unfortunately not available). Over this Arthropods were collected weekly during the summer period, average summer (June–August) temperature was months (June, July and August) each year and sorted to the 4.63 ± 0.22 C(±SE) as recorded 2 m above the surface family level as part of the Zackenberg Basic Monitoring by a centrally located meteorological station. The mesic Program (Schmidt et al. 2012b) at Zackenberg; located in and arid sites are dominated by similar vegetation cover, northeast Greenland (74280N, 20340W, Fig. 1). Speci- largely made up of Dryas octopetala, Salix arctica, lichen mens are preserved in 70 % ethanol and currently housed and grasses with Cassiope tetragona in the mesic sites and at the Natural History Museum in Aarhus, Denmark. We lie approximately 500 m apart from one another (Høye and identified all specimens of Thomisidae to the species level Forchhammer 2008). The most distinct difference between and confirmed that X. deichmanni is the only crab spider the two habitat types is the moisture regime (mesic versus species in the collection. It is distinguishable based on arid) and the related timing of snowmelt. Specifically, morphology from other such species in the genus (Dondale snowmelt occurs earlier and at a more consistent date, and Redner 1978). Y. Marusik had also previously identi- inter-annually, in the two arid sites. Similar mesic sites in fied numerous specimens from the collection from 2002 the area average 34.85 ± 1.90 % (±SE) water by volume and confirmed that all individuals examined were X. de- while similar dry sites average 18.85 ± 0.95 % water by ichmanni. The family is represented by just two known volume at a depth of 5 cm (Sigsgaard et al. 2014). The species in Greenland, Xysticus durus (Sørensen) in the Zackenberg Basic Monitoring Program has been in oper- south and X. deichmanni in the northeast (Marusik et al. ation since 1996 with standardized collection of arthropods 2006). X. deichmanni is a ground-dwelling species dis- via a gridded pitfall trapping regime (Schmidt et al. 2012b). tributed broadly across the Arctic from Alaska across Each plot consists of a 10 m 9 20 m grid utilizing 8 Canada and into northern Greenland (Leech 1996). We (1996–2006) or 4 (2007–present) pitfall traps. Pitfalls at only used individuals collected from plots 3, 4, 5 and 7 as Zackenberg consist of 10-cm-diameter yellow cups and are

Fig. 1 Map of study region with inset orthophoto indicating the two (3, 4) mesic and two (5, 7) arid sites sampled at Zackenberg (74280N, 20340W) in northeast Greenland between 1996 and 2012

123 562 Polar Biol (2015) 38:559–568 open during the activity season when plots are snow free sex ratios and juvenile/female ratios (also determined to fit (mid-June) until snow begins to fall again (early Septem- Gaussian error distributions) differed through time and ber). To obtain body size, we measured the carapace width between the two habitats in response to snowmelt using of all specimens using a Wild M5A stereo microscope linear models. We expressed juvenile/female ratio [juve- with an ocular micrometer. niles/(juveniles ? females)] and sex ratio [females/(fema- les ? males)] as proportions (Wilson and Hardy 2002); Statistical analysis and data providing us with a standardized index that allowed us to compare potential changes in recruitment (Clark et al. Although temperature is important to the growth and sur- 2004) and sex ratio with snowmelt and between the two vival of arthropods, timing of snowmelt is a particularly habitat types. We also tested how the total abundance good predictor of body size in wolf spiders (Høye et al. (measured as activity density) differed between the habitats 2009) and phenology of terrestrial arthropods at Zacken- and with snowmelt. We recognize that the pitfall collection berg (Høye et al. 2007b; Høye and Forchhammer 2008). method is biased toward mature males searching for mates We used habitat-specific snowmelt date, calculated as the and our index of sex ratio underestimates the actual pro- day of year (DOY) from January 1st, when \50 % of the portion of females in the populations; similarly, we opted plot about the respective traps was covered with snow. to represent recruitment by the proportion of juveniles to While snowmelt between the two habitats is highly corre- females due to the large number of males collected. We lated (r = 0.81, Pearson), it occurs much later in the mesic used all specimens available from the four sites for our habitat, which results in higher soil moisture. Furthermore, analyses, but due to the influence that sample size may we included a habitat effect as the inter-annual variation is have on proportions in some years with relatively low over twice that of the arid sites (Fig. 2). We also included sample size, we conducted tests at higher restrictions (at sex as a factor due to the influence of sexual size dimor- least 10, 15, 20 individuals owing to the response) and our phism in the species (Dondale and Redner 1978). results proved to be robust. We included interactions in the We used general linear models weighted by sampling initial models, but only retained them if they were signif- error in R (R Development Core Team 2013) to determine icant and all responses were log-transformed to improve how annual mean adult body size of X. deichmanni chan- linearity and confirmed to meet assumptions of normality ged over 17 years. Weights were based on the reciprocal of and homoscedasticity. standard deviation so as to up-weight the years with the smallest errors; singletons and doubletons were removed from the analyses (one female from 1999 and two females Results from 2008). The terms snowmelt and habitat were included in every full model. We also ran within-habitat models to We identified a total of 2,115 (1,259 adult) individuals test for habitat-specific effects of snowmelt. We tested how belonging to the species X. deichmanni for our analyses (Table S1). Overall, 2 times more males and 1.7 times more females were collected in the arid habitat than the 190 mesic habitat.

180 Body size and abundance

Year vs Arid.Snow.melt 170 Year vs Mesic.Snow.melt Body size (Fig. 3a) varied greatly over the sampling Year vs ChirArid.1 Year vs ChirMesic.1 period with snowmelt and sex both contributing signifi- 160 cantly to the model describing body size variation (Table 1). There was no significant effect of inter-annual 150

Snowmelt (DOY) snowmelt in the arid habitat, but the individuals from the mesic habitat were negatively influenced by later snow- 140 melt (Table 1). We also noted no significant (p = 0.123)

130 effect of snowmelt, controlling for sex, on body size using 1996 1998 2000 2002 2004 2006 2008 2010 2012 a global averaged snowmelt with individuals averaged Year 2 over both habitats (F2,29 = 117.8, RAdj = 0.88). Females were significantly larger than males, body size decreased Fig. 2 Difference in inter-annual snowmelt date between mesic and with later snowmelt dates and mesic sites yielded sig- arid habitats from 1996 to 2012 at Zackenberg, Greenland. Snowmelt date, measured as day of year (DOY) from the 1st day of each annum nificantly larger individuals than their arid counterparts for arid (solid line) and mesic (dashed line) habitats (Table 1). Females and males were 0.072 ± 0.019 123 Polar Biol (2015) 38:559–568 563

a 2.6 Table 1 Results of weighted linear regression models (weighted by 1/standard deviation/sex/habitat/year) and parameters of body size 2.5 variation in adult Xysticus deichmanni from Zackenberg, Greenland Model Parameter Estimate p 2.4 Full Intercept 1.0 ± 4.5E-2 \0.001

2.3 F3,59 = 113.6 Sex (male) -9.0E-2 ± 5.0E-3 \0.001 2 RAdj = 0.85 Snowmelt -1.0E-3 ± 3.0E-4 0.002 2.2 p B 0.001 Habitat (mesic) 3.0E-2 ± 8.0E-3 \0.001

Carapace Width (mm) Arid Intercept 9.1E-1 ± 5.7E-2 \0.001

2.1 F2,28 = 150.5 Sex (male) -8.5E-2 ± 5.0E-2 \0.001 2 RAdj = 0.91 Snowmelt -4.0E-4 ± 4.0E-4 0.32 2.0 p B 0.001 1996 1998 2000 2002 2004 2006 2008 2010 2012 Mesic Intercept 1.1 ± 7.5E-2 \0.001 Year F2,29 = 69.22 Sex (male) -9.4E-2 ± 8.0E-3 \0.001 1.0 2 b RAdj = 0.81 Snowmelt -1.0E-3 ± 5.0E-4 0.01 p B 0.001 0.8 Male Intercept 9.1E-1 ± 5.0E-2 \0.001

F2,29 = 4.88 Habitat (mesic) 2.5E-2 ± 8.0E-3 0.005 R2 = 0.20 Snowmelt -9.0E-4 ± 3.0E-4 0.009 0.6 Adj p = 0.01 Female Intercept 1.0 ± 7.8E-2 \0.001 0.4 F2,29 = 3.68 Habitat (mesic) 3.5E-2 ± 1.3E-2 0.01 2 RAdj = 0.15 Snowmelt -1.0E-3 ± 5.0E-4 0.06 0.2 p = 0.04 Juveniles/(Juveniles + Females) Specimens were collected from four sites (two arid, two mesic) and 0.0 totaling 1,259 individuals collected over a 17-year period 1996 1998 2000 2002 2004 2006 2008 2010 2012 Year Recruitment index Fig. 3 Inter-annual variation in body size and juvenile/female ratios between mesic and arid habitats at Zackenberg, Greenland. a Body Juvenile/female ratios varied greatly over the sampling size of male (open) and female (closed) Xysticus deichmanni from period (Fig. 3b) and were significantly lower in the mesic 1996 to 2012 collected in arid (triangles) and mesic (circles) heath sites in Zackenberg, Greenland. Body size was measured as carapace habitat (Table 2; Fig. 4d). Juvenile/female ratios did not width in millimeters to the nearest hundredth of a millimeter. Error vary significantly with snowmelt within the mesic habitat, bars represent standard errors from each yearly mean. b Variation in but did decrease significantly with later snowmelt within the juvenile/female ratios, measured as proportion of juveniles in the arid habitat (Table 2). Significantly more (F1,30 = 29.1, population, of Xysticus deichmanni from 1996 to 2012 collected from arid (solid line) and mesic (dashed line) habitats in Zackenberg, p B 0.001) juveniles were captured from the arid habitat Greenland. No juveniles were collected in 1998 at the mesic sites, and than the mesic habitat (1.23 ± 0.18 and 0.27 ± 0.05 indi- no samples were available for 2010 viduals per trap day per annum ± SE, respectively). No juveniles were collected during the year 1998 at the mesic sites. (Fig. 4a) and 0.034 ± 0.009 mm (Fig. 4b) larger (±SE) in the mesic sites, respectively, and differed significantly Sex ratios between habitats. Surprisingly, while females exhibited higher variation intra-annually and between habitats, only There were significant effects of snowmelt and habitat on males responded significantly to inter-annual variation in female/male ratios overall revealing a higher proportion of snowmelt date (Table 1). The total abundance of X. de- females in the mesic site. Although there was an overall higher ichmanni was significantly (p = 0.005) lower in the mesic proportion of females/males in the mesic habitat, the proportion 2 sites (F2,29 = 13.7, RAdj = 0.45; Fig. 4c), despite a sig- of females decreased significantly in years of later snowmelt nificantly later average snowmelt (F1,30 = 38.06, (Table 3;Fig.5). Sex ratio did not change significantly with 2 RAdj = 0.54, p \ 0.001). Total abundance did not vary snowmelt in the arid habitat, but there was no significant significantly with inter-annual snowmelt (p = 0.77). interaction between snowmelt and habitat (p =0.07).

123 564 Polar Biol (2015) 38:559–568

ab2.5 2.5 p = 0.01 p = 0.005

2.4 2.4

2.3 2.3

2.2 2.2

2.1 2.1 Male carapace width (mm)

Female carapace width (mm) 2.0 2.0 Mesic Arid Mesic Arid cd 0.12 1.0 p = < 0.001 p = 0.04 0.10 0.8

0.08 0.6 0.06 0.4 0.04

0.2 0.02 Abundance (per trap day) 0.00 0.0 Mesic Arid Juveniles/(Juveniles + Females) Mesic Arid

Fig. 4 Bar graphs showing significant differences in body size, individuals per trap day per annum) of all individuals and d juve- juvenile/female ratios and abundance between mesic and arid habitats nile/female ratios (proportion of juveniles in the population) between at Zackenberg, Greenland. a female and b male body size (measured the two habitats. Error bars represent standard errors. p values are as carapace width in millimeters) c abundance (averaged to based on full models including snowmelt and habitat

Table 2 Results of full and habitat-specific linear regression models Table 3 Results of full and habitat-specific linear regression models with parameters of juvenile/female ratios in Xysticus deichmanni with parameters of sex ratios in Xysticus deichmanni from Zacken- from Zackenberg, Greenland berg, Greenland Model Parameter Estimate p Model Parameter Estimate p

Full Intercept -8.8E-2 ± 1.2 0.49 Full Intercept 8.7E-1 ± 5.7E-1 0.008

F2,28 = 3.52 Habitat (mesic) -4.3E-1 ± 2E-1 0.04 F2,29 = 3.35 Habitat (mesic) 1.3E-1 ± 5.0E-2 0.02 2 2 RAdj = 0.14 Snowmelt 4.0E-3 ± 8.0E-3 0.65 RAdj = 0.13 Snowmelt -4.3E-3 ± 2.0E-3 0.05 p = 0.04 p = 0.05 Arid Intercept 1.6 ± 7.4E–1 0.05 Arid Intercept 5.3E-2 ± 4.5E-1 0.91

F1,14 = 6.40 Snowmelt -1.3E-2 ± 5.0E-3 0.02 F1,14 = 0.17 Snowmelt 1.0E-3 ± 3.0E-3 0.69 2 2 RAdj = 0.26 RAdj B 0.01 p = 0.02 p = 0.69 Mesic Intercept -2.6 ± 2.3 0.29 Mesic Intercept 1.4 ± 4.5E-1 0.007

F1,13 = 0.69 Snowmelt 1.2E-2 ± 1.4E-2 0.42 F1,14 = 6.43 Snowmelt -6.8E-3 ± 2.7E-3 0.025 2 2 RAdj B 0.01 RAdj = 0.26 p = 0.42 p = 0.025 Specimens were collected from four sites (two arid, two mesic) col- Specimens were collected from four sites (two arid, two mesic) col- lected over a 17-year period lected over a 17-year period

Discussion Høye et al. 2013). In these systems, snowmelt effectively Local, habitat-specific variation in species’ phenological opens the activity/growth season for many species of ani- responses to climate change is particularly important in mals and plants. The snow-free period determines the Arctic and alpine ecosystems (e.g., Kudo and Hirao 2006; amount of resources that can be accumulated during the

123 Polar Biol (2015) 38:559–568 565

0.6 expected, but females did exhibit higher intra-annual var- iation in size as compared to males (Fig. 3a). Size variation 0.5 in males is also affected by resource limitation and thought to be primarily under sexual selection (Stillwell et al. 0.4 2010). Small size in males, therefore, could be advanta- geous despite strong environmental constraints on size 0.3 because early maturation may ensure mating success. Hence, any size variation in individual males is likely 0.2 affected by resource availability and modulated by the balance between the benefits of early maturation versus

Females/(Females + Males) 0.1 large size in the competition for mates.

0.0 The larger body size and smaller juvenile/female ratio in 130 140 150 160 170 180 190 the mesic habitat are somewhat surprising and suggests Snowmelt (DOY) habitat differences are very important in this species. Habitat differences in the species may arise through a Fig. 5 Linear relationships between sex ratio (measured as propor- combination of mechanisms to achieve optimal growth and tion of females in population) and habitat-specific snowmelt date. Annual sex ratios in response to habitat-specific snowmelt in the fitness such as delayed maturation, higher resource avail- mesic (closed circles) and arid (open circles) habitats. Data represent ability in mesic sites and possibly, migration of specific-size proportions of females in the adult population collected using pitfall individuals to preferable sites. A delayed maturation is quite traps between 1996 and 2012 at Zackenberg, Greenland possible given the delay in season onset due to the signifi- cantly later average snowmelt in the mesic habitat. Later snowmelt does, however, mean that maturation occurs growing season. Snow is unevenly distributed across the during a warmer part of the growing season, where landscape and the habitat-specific timing of snowmelt resources may be more abundant (Høye and Forchhammer influences both incident radiation and subsequent soil 2008). Moreover, the variance in snowmelt date is over two moisture levels to create habitats of varying aridity. Here, times greater in the mesic sites than the arid sites. Such we have shown that such variation in the abiotic environ- unpredictable timing of snowmelt compounded with the ment also affects fitness-related traits like body size and significantly shorter snow-free period could lead to signif- juvenile/female ratios, as well as sex ratios in a species icant alterations in life history strategy (e.g., delaying with low mobility. Importantly, we identified effects of maturation to maximize reproductive output or general climate when accounting for habitat variation which was reductions in reproductive fitness). Still, there could be not detectable in models not taking local habitat variation higher resource availability in the mesic sites leading to the into account. significantly larger body size detected in these sites. Indeed, Previously, Høye et al. (2009) have shown that date of mesic habitats likely support a higher abundance of small snowmelt significantly affects average annual body size in flies that require an aquatic or semi-aquatic life stage and the Arctic wolf spider, such that years with a longer snow- may constitute a sizeable portion of the spider’s diet (Leech free period are associated with a larger mean body size. 1996). We have witnessed substantial pulses of some Conceivably, the additional time enables individuals to groups in the mesic habitat. For example, the Chironomidae obtain more resources and achieve a larger adult size. Our are on average higher in abundance in the mesic sites finding is in contrast to Høye et al. (2009) who showed that (0.52 ± 0.17 individuals per trap day per annum) than in the highly mobile wolf spiders seem less affected by hab- the arid sites (0.34 ± 0.07 individuals per trap day per itat-specific snowmelt although they did identify plot dif- annum) although, not significantly. Lastly, it is possible that ferences in males and juveniles. This emphasizes the large individuals preferably migrate to mesic habitats sim- importance of species-specific responses to climate change, ply to avoid desiccation stress. Indeed, species vary greatly such that low-mobility species like X. deichmanni are in their desiccation tolerance and distribute themselves likely more affected by landscape scale habitat variation. accordingly over fine spatial scales (e.g., Devito et al. Similar to Høye et al. (2009), we suggest that, since 2004), but surface area to volume ratio decreases with females are driven to obtain a large body size under increasing size; making larger individuals less susceptible fecundity selection (Stillwell et al. 2010), both the cost to desiccation. It is unlikely that there is a great degree of associated with reproduction and environmental constraints movement between the populations as spiders in the genus impose strong limits upon final body size in the Arctic. We Xysticus are slow-moving ground dwellers that do not show did not, however, detect a significant response in female a high propensity for aerial dispersal (Morse 1992; Larrive´e body size to inter-annual variation in snowmelt as and Buddle 2011) or movement as adults (Leech 1996). 123 566 Polar Biol (2015) 38:559–568

Contrary to our prediction, we found significantly larger (given the collection technique), the collection methods body sizes in the mesic habitat, yet there was a significantly among habitats and over time are standardized. Like many lower abundance of crab spiders at these sites. Similarly, others with Arctic and/or alpine distributions (e.g., Schm- Wyant et al. (2011) found substantial differences in the oller 1970; Høye et al. 2009), X. deichmanni seems to abundance of X. deichmanni between dry and moist tundra require multiple years to mature (J.J.B. personal observa- (89 and five individuals, respectively) over a three year tion), so lagged effects may also play a role in the popula- sampling period in Alaska, adding support to our desig- tion dynamics of these species (e.g., Høye et al. 2009) and nation of sites as being arid or mesic. Furthermore, as early this should be investigated further with an experimental as 1966, Leech observed that X. deichmanni ‘‘…is a approach. Here, we have shown the value of long-term member of the arid Arctic faunal element.’’ This, com- arthropod collections to reveal how habitat can mediate the pounded with evidence from sex ratios and juvenile/female effects of climate change on species and that it may have ratios suggests that mesic sites serve better for resource particularly strong effects upon low-mobility species in the acquisition (potentially with the cost of delayed matura- Arctic. We predicted that body size, recruitment (measured tion/reproduction) and that the arid habitat represents better as proportion of juveniles in the population) and sex ratios sites for mating and reproduction; especially in years of in a low-mobility spider would be significantly affected by late snowmelt. temporal variation in snowmelt, habitat and their interac- While larger size in arthropods in theory provides the tion. We also predicted that females would exhibit a necessary resources for offspring production, trade-offs in stronger response in body size to snowmelt than males. We reproductive effort may prevent the production of many fit found that body size varied significantly with snowmelt and offspring (Bowden and Buddle 2012) that can successfully between the habitats with a stronger effect of snowmelt in be recruited into the population. Similarly, Puzin et al. (2011) the mesic habitat, but females did not show a stronger found higher average egg-sac mass (reproductive output) in response than males. We found that our index of recruit- two species of wolf spider when collected from their ment was significantly higher in the arid tundra and varied respective optimal habitats. They also detected a trade-off significantly with snowmelt in this habitat. Lastly, there between egg size–number in one species, such that more were significant main effects of habitat and snowmelt on smaller progeny were produced in the optimal habitat while sex ratio with the proportion of females in the mesic habitat fewer large progeny were produced in the sub-optimal hab- responding negatively to later snowmelt, but did not find itat. This may help explain the higher juvenile/female ratios any significant interactions. These relatively contemporary detected in the arid tundra habitat where females were sig- effects may be intensified by future habitat-specific climate nificantly smaller. Moreover, the mesic habitat could pose a changes. While increased warming and net solar radiation higher risk of juvenile mortality due to predation, parasitism could allow for more efficient incubation of eggs and or disease, yet, increased juvenile mortality could come incident heat, for example, increased alteration to the tundra about either through juvenile susceptibility to overwintering (e.g., shrubification), related to an increase in soil moisture (Martyniuk and Wise 1985) or simply through increased (Zhang et al. 2013) may exacerbate the effects of habitat- growth rate with high food availability (Higgins and Rankin specific climate change on ground-dwelling terrestrial 2001). Furthermore, Benton and Uetz (1986) have shown arthropods in the Arctic whose physiology is directly con- that survival of juveniles from arid habitats is higher than trolled by the abiotic environment. juveniles from moist habitats in a tropical spider at a broad array of humidity treatments, suggesting that they possess a Acknowledgments Samples from Zackenberg, Northeast Green- stronger tolerance to desiccation stress. Whether differences land, were provided by BioBasis, the Department of Bioscience, Aarhus University, Denmark, and Natural History Museum Aarhus, between habitats are primarily driven by time constraints Denmark. We would also like to acknowledge Climate Basis for related to timing of snowmelt, differences in resource access to environmental data. availability or susceptibility to desiccation stress should be tested experimentally. The significant decrease in female/male ratios may explain why larger females in the mesic sites do not trans- References late into higher abundance of juveniles (and overall indi- viduals) in these sites. Fewer females in the population in ACIA (2005) Arctic climate impact assessment: impacts of a relation to later snowmelts in the mesic sites, due to, for warming Arctic. Cambridge University Press, Cambridge example, sexual mismatch or delayed maturation could Benton MJ, Uetz GW (1986) Variation in life-history characteristics negatively affect population dynamics (Miller-Rushing over a clinal gradient in three populations of a communal orb- weaving spider. Oecologia 68:395–399 et al. 2010) in the mesic habitat. Although our assessment of Bourne EC, Bocedi G, Travis JMJ, Pakeman RJ, Brooker RW, the proportion of females in the population is male biased Schiffers K (2014) Between migration load and evolutionary 123 Polar Biol (2015) 38:559–568 567

rescue: dispersal, adaptation and the response of spatially flower visitors in a warmer Arctic. Nat Clim Change 3:759–763. structured population to environmental change. Proc R Soc B doi:10.1038/nclimate1909 281:20132795. doi:10.1098/rspb.2013.2795 Jakob EM, Marshall SD, Uetz GW (1996) Estimating fitness: a Bowden JJ, Buddle CM (2010) Determinants of ground-dwelling comparison of body condition indices. Oikos 77:61–67 spider assemblages at a regional scale in the Yukon Terrritory, Kudo G, Hirao AS (2006) Habitat-specific responses in the flowering Canada. E´ coscience 17:287–297. doi:10.2980/17-3-3308 phenology and seed set of alpine plants to climate variation: Bowden JJ, Buddle CM (2012) Life history of tundra-dwelling wolf implications for global-change impacts. Popul Ecol 48:49–58. spiders (Araneae: Lycosidae) from the Yukon Territory, Canada. doi:10.1007/s10144-005-0242-z Can J Zool 90:714–721. doi:10.1139/z2012-038 Larrive´e M, Buddle CM (2011) Ballooning propensity of canopy and Bowden JJ, Høye TT, Buddle CM (2013) Fecundity and sexual size understorey spiders in a mature temperate hardwood forest. Ecol dimorphism of wolf spiders (Araneae: Lycosidae) along an Entomol 36:144–151. doi:10.1111/j.1365-2311.2010.01255.x elevational gradient in the Arctic. Polar Biol 36:831–836. doi:10. Leech RE (1996) The spiders (Araneidae) of Hazen Camp 81490 N, 1007/s00300-013-1308-6 71 180 W. Quaest Entomol 2:153–212 Callaghan TV, Bjo¨rn LO, Chernov Y, Chapin T, Christensen TR, Martyniuk J, Wise DH (1985) Stage-biased overwintering survival of Huntley B, Ims RA, Johansson M, Jolly D, Jonasson S, the filmy dome spider (Araneae, Linyphiidae). J Arachnol Matveyeva N, Panikov N, Oechel W, Shaver G, Elster J, 13:321–329 Henttonen H, Laine K, Taulavuori E, Zo¨ckler C (2004) Marusik YM, Bo¨cher J, Koponen S (2006) The collection of Biodiversity, distributions and adaptations of Arctic species in Greenland spiders (Aranei) kept in the Zoological Museum, the context of environmental change. AMBIO 33:404–417. University of Copenhagen. Arthropoda Sel 15:59–80 doi:10.1579/0044-7447-33.7.404 Miller-Rushing AJ, Høye TT, Inouye DW, Post E (2010) The effects Chown SL, Gaston KJ (2010) Body size variation in insects: a of phenological mismatches on demography. Proc R Soc B macroecological perspective. Biol Rev 85:139–169. doi:10.1111/ 365:3177–3186. doi:10.1098/rstb.2010.0148 j.1469-185X.2009.00097.x Morse DH (1992) Dispersal of the spiderlings of Xysticus emertoni Clark JA, Robinson RA, Clark NA, Atkinson PW (2004) Using the (Araneae, Thomisidae), a litter-dwelling crab spider. J Arachnol proportion of juvenile waders in catches to measure recruitment. 20:217–221 Wader Study Group Bull 104:51–55 Myers-Smith IH, Forbes BC, Wilmking M, Hallinger M, Lantz T, DeVito J, Meik JM, Gerson MM, Formanowicz DR Jr (2004) Blok D, Tape KD, Macias-Fauria M, Sass-Klaassen U, Le´vesque Physiological tolerances of three sympatric riparian wolf spiders E, Boudreau S, Ropars P, Hermanutz L, Trant A, Collier LS, (Araneae: Lycosidae) correspond with microhabitat distribu- Weijers S, Rozema J, Rayback SA, Schmidt NM, Schaepman- tions. Can J Zool 82:1119–1125. doi:10.1139/z04-090 Strub G, Wipf S, Rixen C, Me´nard CB, Venn S, Goetz S et al Dondale CJ, Redner JH (1978) the insects and Arachnids of Canada: (2011) Shrub expansion in tundra ecosystems: dynamics, part 5, the crab spiders of Canada and Alaska. Supply and impacts and research priorities. Environ Res Lett 6:045509. Services Canada, Hull doi:10.1088/1748-9326/6/4/045509 Elmendorf SC, Henry GHR, Hollister RD, Bjo¨rk RG, Boulager- Olson DM, Dinerstein E, Wikramanayake ED, Burgess ND, Powell Lapointe N, Cooper EJ, Cornelissen JHC, Day TA, Dorrepaal E, GVN, Underwood EC, D’Amico JA, Itoua I, Strand HE, Morrison Elumeeva TG, Gill M, Gould WA, Harte J, Hik DS, Hofgaard A, JC, Loucks CJ, Allnutt TF, Ricketts TH, Kura Y, Lamoreux JF, Johnson DR, Johnstone JF, Jo´nsdo´ttir IS, Jorgenson JC, Wettengel WW, Hedao P, Kassem KR (2001) Terrestrial ecore- Klanderud K, Klein JA, Koh S, Kudo G, Lara M, Le´vesque E gions of the world: a new map of life on Earth. Bioscience et al (2012) Plot-scale evidence of tundra vegetation change and 51:933–938. doi:10.1641/0006-3568(2001)051[0933:TEOTWA]2. links to recent summer warming. Nat Clim Change 2:453–457. 0.CO;2 doi:10.1038/nclimate1465 Post E, Forchhammer M, Bret-Harte MS, Callaghan TV, Christensen Higgins LE, Rankin MA (2001) Mortality risk of rapid growth in the TR, Elberling B, Fox AD, Gilg O, Hik DS, Høye TT, Ims RA, spider Nephila clavipes. Funct Ecol 15:24–28. doi:10.1046/j. Jeppesen E, Klein DR, Madsen J, McGuire AD, Rysgaard S, 1365-2435.2001.00491.x Schindler DE, Stirling I, Tamstorf MP, Tyler NJC, van der Wal Hodkinson ID, Coulson SJ (2004) Are high Arctic terrestrial food R, Welker J, Wookey PA, Schmidt NM, Aastrup P (2009) chains really that simple? The bear Island food web revisited. Ecological dynamics across the Arctic associated with recent Oikos 106:427–431. doi:10.1111/j.0030-1299.2004.13091.x climate change. Science 325:1355–1358 Høye TT, Forchhammer MC (2008) Phenology of high-arctic Puzin C, Acou A, Bonte D, Pe´tillon J (2011) Comparison of arthropods: effects of climate on spatial, seasonal and inter- reproductive traits between two salt-marsh wolf spiders (Ara- annual variation. Adv Ecol Res 40:299–324. doi:10.1016/S0065- neae, Lycosidae) under different habitat suitability conditions. 2504(07)00013-X Anim Biol 61:127–138. doi:10.1163/157075511X566461 Høye TT, Hammel JU (2010) Climate change and altitudinal variation R Development Core Team (2013) R: a language and environment for in sexual size dimorphism of arctic wolf spiders. Clim Res statistical computing, version 3.0.1. R Foundation for Statistical 41:259–265. doi:10.3354/cr00855 Computing, Vienna Høye TT, Post E, Meltofte H, Schmidt NM, Forchhammer MC Roslin T, Wirta H, Hopkins T, Hardwick B, Va´rkonyi G (2013) (2007a) Rapid advancement of spring in the high Arctic. Curr Indirect interactions in the high Arctic. PLoS One e67367. Biol 17:R449–R451. doi:10.1016/j.cub.2007.04.047 doi:10.1371/journal.pone.0067367 Høye TT, Ellerbjerg SM, Philipp M (2007b) The impact of climate on Ru¨hlund KM, Paterson AM, Keller W, Michelutti N, Smol JP (2013) flowering in the high Arctic the case of Dryas in a hybrid zone. Global warming triggers the loss of a key Arctic refugium. Proc Arct Antarct Alp Res 39:412–421. doi:10.1657/1523-0430(06- R Soc B 280:20131887. doi:10.1098/rspb.2013.1887 018)[HOYE]2.0.CO;2 Schmidt NM, Kristensen DK, Michelsen A, Bay C (2012a) High Høye TT, Hammel JU, Fuchs T, Toft S (2009) Climate change and Arctic plant community responses to a decade of ambient sexual size dimorphism in an Arctic spider. Biol Lett 5:542–544. warming. Biodiversity 13:191–199. doi:10.1080/14888386.2012. doi:10.1098/rsbl.2009.0169 712093 Høye TT, Post E, Schmidt NM, Trøjelsgaard K, Forchhammer MC Schmidt NM, Hansen LH, Hansen J, Berg TB, Meltofte H (2012b) (2013) Shorter flowering seasons and declining abundance of BioBasis manual: conceptual design and sampling procedures of

123 568 Polar Biol (2015) 38:559–568

the biological monitoring programme within Zackenberg Basic. Walker DA, Raynolds MK, Danie¨ls FJA, Einarsson E, Elvebakk A, Aarhus University Press, Roskilde Gould WA, Katenin AE, Kholod SS, Markon CJ, Melnikov ES, Schmoller R (1970) Life histories of alpine tundra Arachnida in Moskalenko NG, Talbot SS, Yurtsev BA, The other members of Colorado. Am Midl Nat 83:119–133 the CAVM Team (2005) The circumpolar Arctic vegetation Sigsgaard C, Mylius MR, Skov K (2014) GeoBasis Manual: map. J Veg Sci 16:267–282. doi:10.1111/j.1654-1103.2005. Guidelines and sampling procedures for the geographical tb02365.x monitoring programme of Zackenberg Basic. Aarhus University Warren MS, Hill JK, Thomas JA, Asher J, Fox R, Huntley B, Roy Press, Roskilde DB, Telfer MG, Jeffcoate S, Harding P, Jeffcoate G, Willis SG, Stillwell RC, Blanckenhorn WU, Teder T, Davidowitz G, Fox CW Greatorex-Davies JN, Moss D, Thomas CD (2001) Rapid (2010) Sex differences in phenotypic plasticity affect variation in responses of British butterflies to opposing forces of climate sexual size dimorphism in insects: from physiology to evolution. and habitat change. Nature 414:65–69. doi:10.1038/35102054 Annu Rev Entomol 55:227–245. doi:10.1146/annurev-ento- Wilson K, Hardy ICW (2002) Statistical analysis of sex ratios: an 112408-085500 introduction. In: Hardy ICW (ed) Sex ratios: concepts and research Stow DA, Hope A, McGuire D, Verbyla D, Gamon J, Huemmrich F, methods. Cambridge University Press, Cambridge, pp 48–92 Houston S, Racine C, Sturm M, Tape K, Hinzmon L, Yoshikawa Wyant KA, Draney ML, Moore JC (2011) Epigeal spider (Araneae) K, Tweedie C, Noyle B, Silapaswan C, Douglas D, Griffith B, Jia communities in moist acidic and dry heath tundra at Toolik G, Epstein H, Walker D, Daeshner S, Petersen A, Zhou L, Lake, Alaska. Arct Antarct Alp Res 43:301–312. doi:10.1657/ Myneri R (2004) Remote sensing of vegetation and land-cover 1938-4246-43.2.301 change in Arctic Tundra Ecosystems. Remote Sens Environ Zhang W, Miller PA, Smith B, Wania R, Koenigk T, Do¨scher R 89:281–308. doi:10.1016/j.rse.2003.10.018 (2013) Tundra shrubification and tree-line advance amplify Thomas CD (2000) Dispersal and extinction in fragmented land- arctic climate warming: results from and individual-based scapes. Biol Sci 267:139–145. doi:10.1098/rspb.2000.0978 dynamic vegetation model. Environ Res Lett 8:034023. doi:10. Thomas CD, Bodsworth EJ, Wilson RJ, Simmons AD, Davies ZG, 1088/1748-9326/8/3/034023 Musche M, Conradt L (2001) Ecological and evolutionary Zona D, Oechel WC, Kochendorfer J, Paw UKT, Salyuk AN, Olivas processes at expanding range margins. 411:577–581. doi:10. PC, Oberbauer SF, Lipson DA (2009) Methane fluxes during the 1038/35079066 initiation of a large-scale water table manipulation in the Uhl G, Schmitt S, Scha¨fer MA, Blanckenhorn WU (2004) Food and Alaskan Arctic tundra. Glob Biogeochem Cycles 23:GB2013. sex-specific strategies in a spider. Evol Ecol Res 6:523–540 doi:10.1029/2009GB003487

123 Downloaded from http://rsbl.royalsocietypublishing.org/ on August 28, 2016

Global change biology High-Arctic butterflies become smaller rsbl.royalsocietypublishing.org with rising temperatures

Joseph J. Bowden1, Anne Eskildsen2,4, Rikke R. Hansen1,4, Kent Olsen2,5, Carolyn M. Kurle6 and Toke T. Høye1,3,4 Research 1Arctic Research Centre, 2Department of Bioscience, and 3Aarhus Institute of Advanced Studies, Cite this article: Bowden JJ, Eskildsen A, Aarhus University, 8000 Aarhus, Denmark 4Department of Bioscience, Kalø, Aarhus University, 8410 Rønde, Denmark Hansen RR, Olsen K, Kurle CM, Høye TT. 2015 5Natural History Museum Aarhus, 8000 Aarhus, Denmark High-Arctic butterflies become smaller with 6Division of Biological Sciences, University of California, San Diego, La Jolla, CA 92093, USA rising temperatures. Biol. Lett. 11: 20150574. http://dx.doi.org/10.1098/rsbl.2015.0574 The response of body size to increasing temperature constitutes a universal response to climate change that could strongly affect terrestrial ectotherms, but the magnitude and direction of such responses remain unknown in most species. The metabolic cost of increased temperature could reduce body size Received: 3 July 2015 but long growing seasons could also increase body size as was recently shown in an Arctic spider species. Here, we present the longest known time Accepted: 9 September 2015 series on body size variation in two High-Arctic butterfly species: Boloria char- iclea and Colias hecla. We measured wing length of nearly 4500 individuals collected annually between 1996 and 2013 from Zackenberg, Greenland and found that wing length significantly decreased at a similar rate in both species Subject Areas: in response to warmer summers. Body size is strongly related to dispersal ecology capacity and fecundity and our results suggest that these Arctic species could face severe challenges in response to ongoing rapid climate change. Keywords: Lepidoptera, insect, terrestrial arthropod 1. Introduction Author for correspondence: Body size change is regarded as a third universal species response to climate change along with shifts in phenology and range [1,2]. Body size is a key trait Joseph J. Bowden related to the life history of individuals with implications for reproductive success e-mail: [email protected] [3,4] and dispersal capacity [5,6]. In ectothermic species, like arthropods, body size is especially influenced by the abiotic environment. Two patterns in ecology summarize the general pattern of ectothermic body size variation in response to temperature change. The adaptation of Bergmann’s rule to ectotherms describes latitudinal or thermal variation in size, such that larger individuals occur at higher latitudes and in colder environments [7]. Similarly, the specific pattern of higher temperatures resulting in smaller adult size of ectotherms, deemed the temperature–size rule, has been commonly found via experiments on numerous taxa [8–10]. While there are several proposed ways that temperature may influ- ence final adult body size (including effects on food limitation and predation rates), the role of metabolism is central [1,9]. Although both Bergmann’s rule and the temperature–size rule predict larger individuals in colder environments, the opposite pattern also occurs [7,9]. Hence, two hypotheses detail how external temperatures may drive size responses to cli- mate change in arthropods in seasonal environments. First, the metabolic rates of arthropods increase with warmer temperatures [1,11]. Therefore, in warmer environments, organisms become smaller if they cannot offset energy losses related to increased metabolic costs during growth. Conversely, rising temperatures in sea- Electronic supplementary material is available sonal environments associated with longer growing seasons may allow arthropods at http://dx.doi.org/10.1098/rsbl.2015.0574 or to obtain more resources and grow larger [2,12]. These two mechanisms may act in concert, moderating the outcomes of each or the effect of a longer growing season via http://rsbl.royalsocietypublishing.org. could depend upon trophic level [2]. Although feeding rates of herbivores may

& 2015 The Author(s) Published by the Royal Society. All rights reserved. Downloaded from http://rsbl.royalsocietypublishing.org/ on August 28, 2016

increase to some peak [13], extended seasons could also mean a) 170( 2 lower plant-food quality, especially during late season [14]. rsbl.royalsocietypublishing.org Temperature and timing of snowmelt define the activity and growing season for numerous Arctic taxa and influence 160 life histories (e.g. [15,16]). Here, we present the longest time series of body size variation within arthropod (i.e. butterflies) species at high latitude. We tested which of the hypotheses 150 was better supported by our data using snowmelt as a proxy for season length and temperature as a proxy for the

metabolism hypothesis and predicted that previous year’s (DOY) snowmelt temperature would best explain variation in body size as it 140 covers the majority of the larval growing period. Lett. Biol.

130

2. Material and methods (b) 11 20150574 : (a) Study area and data 0 Butterfly specimens and climate (snowmelt and temperature) data were collected in northeast Greenland (748280 N, 208340 W) from 1996 to 2013 as part of the Zackenberg Basic Monitoring Pro- –1 (°C) gramme [17]. See the electronic supplementary material (detailed t Material and methods) and references therein for more details on climate data collection and plant community characteristics. –2 Arthropods were collected weekly during the activity season each year in one window trap plot and six pitfall trap plots in an average temp average area less than 1 km2, but the peak flight period for adult butterflies occurs mid–late July depending upon snowmelt and temperature –3 [18]. The butterfly community at Zackenberg has four species but is dominated by the Arctic fritillary, Boloria chariclea Schneider and the northern clouded yellow, Colias hecla Lefe`bvre. We measured (c) all 3629 (1934 males and 1695 females) B. chariclea and 847 (531 4 males and 316 females) C. hecla collected. Both of these species most probably have a 2 year generation time at our site [19] with adult size depending on resources accrued during their larval stages. We determined sex, and measured the length of one fore- 3 wing from thoracic attachment to apex of the wing to the nearest (°C) t –1 0.01 mm21 using a Diesellaw 150 mm digital caliper. 2 (b) Data analysis We used current year’s May–June (temp ) and previous year’s

t temp average May–August (tempt21) temperatures to encompass the larval growing season. Current and previous years’ snowmelt date 1

(snowt and snowt21) were used as proxies for season length. We used linear mixed-models with a Gaussian error distribution to 1995 2000 2005 2010 2015 determine which predictor(s) best explained variation in wing year length using annual averages (years as replicates) for each sex in each species. We composed an apriorilist of candidate models Figure 1. (a) Timing of snowmelt (F1,16 ¼ 8.92, estimate ¼ 21.05, 2 ¼ based upon what we currently know about the development of RAdj 0:32, p ¼ 0.008), (b) May–Junet temperature (F1,16 ¼ 10.87, 2 ¼ these species (electronic supplementary material, table S1) and the estimate ¼ 0.11, RAdj 0:37, p , 0.005) and (c) May–Augustt21 temp- 2 ¼ best model was determined by model selection using the Akaike erature (F1,16 ¼ 22.52, estimate ¼ 0.11, RAdj 0:56, p ¼ 0.001) from information criterion corrected for small sample size (AICc). We 1996–2013 at Zackenberg, Greenland. repeated the analyses using individuals as replicates to test for random effects of day of year (DOY) and plot using log-likelihood ratio tests. We tested for plot variation using Tukey’s honestly sig- correlated, there was no indication of multicollinearity nificant difference (HSD). Differences in body size between plots (electronic supplementary material, table S2). were tested in subsequent models using individuals as replicates Average wing length + s.e. in male (17.96 + 0.02 mm) when there was a significant plot effect. Normality was assessed and female (18.87 + 0.02 mm) B. chariclea differed significantly using q–q-plots. Statistical tests were conducted using the R (F1,3626 ¼ 1145, p , 0.001, analysis of variance (ANOVA)). The environment for statistical computing [20]. wing lengths of male (22.45 + 0.04 mm) and female (23.19 +

0.06 mm) C. hecla also differed (F1,845 ¼ 108.8, p , 0.001, ANOVA). Wing lengths varied greatly in both species over 3. Results time (figure 2a,b), yet were strongly correlated between the Timing of snowmelt is occurring significantly earlier (figure 1a) sexes in B. chariclea (r ¼ 0.88, p , 0.001) and C. hecla (r ¼ 0.80, and average temperatures over the activity period of cater- p , 0.001), and between species (all r . 0.6, p , 0.01). pillars have increased significantly (figure 1b,c) during the Average annual body size decreased significantly in study period. While some of the variables were significantly response to previous year’s temperature (tempt21) for both Downloaded from http://rsbl.royalsocietypublishing.org/ on August 28, 2016

(a) 24 (c) 3 rsbl.royalsocietypublishing.org

23

22 average wing length (mm) average il Lett. Biol. 21 (b) (d)

20 11 20150574 :

19

18 average wing length (mm) average

17 1996 2000 2004 2008 2012 0 1 2 345 year average tempt–1 (°C) Figure 2. Inter-annual variation in average male (open circles) and female (filled circles) wing length over the sampling period for (a) Colias hecla and (b) Boloria chariclea and their responses (c,d, respectively) to average May–Augustt21 temperature. Error bars represent s.e. Data for 2010 are not available. (Online version in colour.)

Table 1. Summary statistics showing the best-fit models as selected by AICc on wing length for each sex of Boloria chariclea and Colias hecla collected between 1996 and 2013 at Zackenberg, Greenland using annual averages. Significance of individual parameters is indicated by asterisks.

2 species sex intercept snowt snowt21 tempt tempt21 p-value RAdj F (d.f.) Boloria chariclea F 20.41** 20.01 n.a. n.a. 20.31** 0.01 0.42 6.46 (13) C 19.38** n.a. n.a. n.a. 20.25** 0.007 0.38 10.16 (14) Colias hecla F 25.68** 20.02 n.a. n.a. 20.29* 0.19 0.06 1.88 (13) C 26.25** 20.02 n.a. n.a. 20.32* 0.05 0.27 3.72 (13) *p , 0.05; **p , 0.01. n.a., not applicable. species (table 1 and figure 2c,d) and was consistently selected in effectively controlled for seasonal variation. These models simi- the top models for all tests (electronic supplementary material, larly included the negative effects of temperature on body size table S3). The models using individuals as replicates also in the top models (electronic supplementary material, table S7). revealed significant effects of current year’s snowmelt on wing length (electronic supplementary material, table S4). DOY and plot significantly improved model fit for B. chariclea, but not for C. hecla (electronic supplementary material, 4. Discussion table S5) and some sites differed significantly from one another We show that body size of males and females in two High- (electronic supplementary material, table S6). We re-analysed Arctic butterfly species fluctuated in synchrony from year to the dataset excluding the plots that showed significant differ- year, strongly supporting the influence of external factors on ences from others, but as temperature remained included in inter-annual size variation in these species. We further show the top models with very similar estimates (electronic sup- that increasing summer temperatures lead to smaller adult plementary material, table S5) data from all plots were body size, thus corroborating earlier short-term experiments retained in the final models (table 1). We tested for inter- suggesting that higher temperatures result in smaller adult annual variation in size for B. chariclea, controlling for collection size [3,10]. Even though longer, warmer seasons may mean a date by re-analysing the data with individuals collected from longer period of time to obtain resources, the seasonal quality the same peak abundance day of each season; in this way, we of resources, combined with the higher cost of obtaining them, Downloaded from http://rsbl.royalsocietypublishing.org/ on August 28, 2016

suggests that the larvae cannot compensate for energy losses. bioclimatic models [23]. While these models include dispersal 4 While some species could be capable of increasing their feeding capacity, in situ adaptation to climate change or plasticity rsbl.royalsocietypublishing.org rate to offset increased metabolic costs with warming, this may enable some populations of a species to persist. Indeed, ability appears to be relatively uncommon and species likely the degree to which phenotypic plasticity and adaptation ulti- possess some locally adapted optimal feeding temperature mately play a role in this system remains to be thoroughly [21]. Indeed, Barrio et al. [22] recently showed that respiration investigated. A recent review by Seebacher et al. [24] suggests rates in the Arctic moth (Gynaephora groenlandica) were signifi- that terrestrial ectotherms in less stable environments are less cantly higher and growth rates significantly lower at lower capable of physiological plasticity in response to climate elevation, adding support to the metabolism hypothesis. change, particularly at high latitudes. Hence, species such as While we believe our study presents the longest time series these found at high latitudes, adapted to cold climates, could available on body size variation in butterflies, the mechanistic suffer from further warming. basis for the observed variation remains to be demonstrated. Lett. Biol. We also cannot rule out that generation time is extended in particularly short growing seasons. Data accessibility. Supporting data can be accessed via the Dryad data We have demonstrated that two butterfly species in the repository (http://dx.doi.org/10.5061/dryad.43gt3). 11 20150574 : High-Arctic responded similarly and negatively to warming Authors’ contributions. A.E. and T.T.H. conceived the idea and J.J.B. temperatures over 18 years. Smaller body size in these Arctic measured specimens, generated data, performed analyses and species could have significant consequences for their popu- prepared the first draft. A.E., T.T.H., C.M.K., R.R.H. and K.O. contributed to data interpretation, article revisions and final approval. lation dynamics by leading to decreased dispersal capacity or Competing interests. We have no competing interests. lower fecundity and fitness. B. chariclea has already demon- Funding. T.T.H. and A.E. acknowledge 15 Juni Fonden, Denmark for strated a significant shift towards earlier and shorter flight support. seasons at Zackenberg [18] and both B. chariclea and C. hecla Acknowledgements. Access to butterfly specimens and climate data were are considered under extremely high climate change risk by kindly provided by the Greenland Ecosystem Monitoring Programme. the Climatic risk atlas of European Butterflies given future Specimens are curated by the Natural History Museum Aarhus.

References

1. Gardner JL, Peters A, Kearney MR, Joseph L, 9. Angilletta Jr MJ. 2009 Thermal adaptation: a design and sampling procedures of the biological Heinsohn R. 2011 Declining body size: at third theoretical and empirical synthesis. Oxford, UK: monitoring programme within Zackenberg Basic. universal response to warming? Trends Ecol. Evol. Oxford University Press. Roskilde: Aarhus University. 26, 285–291. (doi:10.1016/j.tree.2011.03.005) 10. Fischer K, Fielder K. 2002 Reaction norms for age 18. Høye TT, Eskildsen A, Hansen RR, Bowden JJ, 2. Sheridan JA, Bickford D. 2011 Shrinking body size as and size at maturity in response to temperature: a Schmidt NM, Kissling WD. 2014 Phenology of high- an ecological response to climate change. Nat. Clim. test of the compound interest hypothesis. Evol. Ecol. arctic butterflies and their floral resources: species- Change 1, 401–406. (doi:10.1038/nclimate1259) 16, 333–349. (doi:10.1023/A:1020271600025) specific responses to climate change. Curr. Zool. 60, 3. Jones RE, Hart JR, Bull JD. 1982 Temperature, size 11. Gillooly JF, Brown JH, West GB, Savage VM, Charnov 243–251. and egg production in the cabbage butterfly, Pieris EL. 2001 Effects of size and temperature on 19. Eliasson CU, Ryrholm N, Ga¨rdenfors U. 2005 The rapae L. Aust. J. Zool. 30, 223–231. (doi:10.1071/ metabolic rate. Science 293, 2248–2251. (doi:10. Encyclopedia of the Swedish Flora and Fauna, ZO9820223) 1126/science.1061967) Fja¨rilar, Dagfja¨rilar [Swedish]: Hesperiidae— 4. Berger D, Walters R, Gotthard K. 2008 What limits 12. Høye TT, Hammel JU, Fuchs T, Toft S. 2009 Climate Nymphalidae. Uppsala E: Artdatabanken, Sveriges insect fecundity? Body size- and temperature- change and sexual size dimorphism in an Arctic spider. Lantbruksuniversitet. dependent egg maturation and oviposition in a Biol. Lett. 5, 542–544. (doi:10.1098/rsbl.2009.0169) 20. R Core Development Team. 2014 R: a language and butterfly. Funct. Ecol. 22, 523–529. (doi:10.1111/j. 13. Sherman PW, Watt WB. 1973 The thermal ecology environment for statistical computing. Vienna, 1365-2435.2008.01392.x) of some Colias butterfly larvae. J. Comp. Physiol. 83, Austria: R Foundation for Statistical Computing. 5. Jenkins DG et al. 2007 Does size matter for dispersal 25–40. (doi:10.1007/BF00694570) (http://www.R-project.org/) distance? Global Ecol. Biogeogr. 16, 415–425. 14. Awmack CS, Leather SR. 2002 Host plant quality 21. Lemoine NP, Burkepile DE, Parker JD. 2014 Variable (doi:10.1111/j.1466-8238.2007.00312.x) and fecundity in herbivorous insects. Annu. Rev. effects of temperature on insect herbivory. PeerJ 2, 6. Sekar S. 2012 A meta-analysis of the traits affecting Entomol. 47, 817–844. (doi:10.1146/annurev.ento. e376. (doi:10.7717/peerj.376) dispersal ability in butterflies: can wingspan be 47.091201.145300) 22. Barrio IC, Bueno CG, Hik DD. In press. Warming the used as a proxy? J. Anim. Ecol. 81, 174–184. 15. Høye TT, Post E, Meltofte H, Schmidt NM, tundra: reciprocal responses of invertebrate (doi:10.1111/j.1365-2656.2011.01909.x) Forchhammer MC. 2007 Rapid advancement herbivores and plants. Oikos. (doi:10.1111/oik. 7. Blanckenhorn WU, Demont M. 2004 Bergmann and of spring in the High Arctic. Curr. Biol. 17, 02190) converse Bergmann latitudinal clines in arthropods: R449–R451. (doi:10.1016/j.cub.2007.04.047) 23. Settele J et al. 2008 Climatic risk atlas of European two ends of a continuum? Integr. Comp. Biol. 44, 16. Post E, Forchhammer MC. 2008 Climate change butterflies. BioRisk 1, 1–710. (doi:10.3897/biorisk.1) 413–424. (doi:10.1093/icb/44.6.413) reduces reproductive success of an Arctic herbivore 24. Seebacher F, White CR, Franklin CE. 2014 8. Atkinson D. 1994 Temperature and organism through trophic mismatch. Phil. Trans. R. Soc. B Physiological plasticity increases resilience of size—a biological law for ectotherms? Adv. Ecol. 363, 2369–2375. (doi:10.1098/rstb.2007.2207) ectothermic animals to climate change. Nat. Res. 25, 1–58. (doi:10.1016/S0065-2504(08) 17. Schmidt NM, Hansen LH, Hansen J, Berg TB, Clim. Change 5, 61–66. (doi:10.1038/NCLI 60212-3) Meltofte H. 2012 BioBasis manual—conceptual MATE2457) Current Zoology 60 (2): 243–251, 2014

Phenology of high-arctic butterflies and their floral resources: Species-specific responses to climate change

Toke T. HØYE1,2,3*, Anne ESKILDSEN2,4 , Rikke R. HANSEN3, Joseph J. BOWDEN3, Niels M. SCHMIDT3,5, W. Daniel KISSLING6 1 Aarhus Institute of Advanced Studies, Høegh-Guldbergs Gade 6B, bldgs. 1630-1632, DK-8000 Aarhus C, Denmark 2 Department of Bioscience, Kalø, Aarhus University, Grenåvej 14, DK-8410 Rønde, Denmark 3 Arctic Research Centre, Aarhus University, C.F. Møllers Allé 8, bldg. 1110, DK-8000 Aarhus C, Denmark 4 Department of Bioscience, Aarhus, Aarhus University, Ny Munkegade, bldg. 1540, DK-8000 Aarhus C, Denmark 5 Department of Bioscience, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmark 6 Institute for Biodiversity and Ecosystem Dynamics (IBED), University of Amsterdam, P.O. Box 94248, 1090 GE Amsterdam, The Netherlands

Abstract Current global warming is particularly pronounced in the Arctic and arthropods are expected to respond rapidly to these changes. Long-term studies of individual arthropod species from the Arctic are, however, virtually absent. We examined butterfly specimens collected from yellow pitfall traps over 14 years (1996–2009) at Zackenberg in high-arctic, north-east Greenland. Specimens were previously sorted to the family level. We identified them to the species level and examined long-term species-specific phenological responses to recent summer warming. Two species were rare in the samples (Polaris fritillary Bolo- ria polaris and Arctic blue Plebejus glandon) and statistical analyses of phenological responses were therefore restricted to the two most abundant species (Arctic fritillary, B. chariclea and Northern clouded yellow Colias hecla). Our analyses demonstrated a trend towards earlier flight seasons in B. chariclea, but not in C. hecla. The timing of onset, peak and end of the flight season in B. chariclea were closely related to snowmelt, July temperature and their interaction, whereas onset, peak and end of the flight season in C. hecla were only related to timing of snowmelt. The duration of the butterfly flight season was significantly positively related to the temporal overlap with floral resources in both butterfly species. We further demonstrate that yellow pitfall traps are a useful alternative to transect walks for butterfly recording in tundra habitats. More phenological studies of Arctic arthropods should be carried out at the species level and ideally be analysed in context with interacting species to assess how ongoing cli- mate change will affect Arctic biodiversity in the near future [Current Zoology 60 (2): 243–251, 2014]. Keywords Arctic, Arthropod, Flight period, Greenland, Pitfall trap, Zackenberg

Phenology is a key indicator of species responses to species across broad taxonomic scales are responding climate change (Parmesan, 2006) and among terrestrial rapidly to these changes (Høye et al., 2007; Post et al., animals, ectotherms are particularly sensitive, probably 2009). Unfortunately, long-term records of individual because their developmental rates are closely related to arthropod species from the Arctic are virtually absent variation in the abiotic environment (Høye et al., 2007; and the phenological sensitivity to climate change in Thackeray et al., 2010). Butterflies in particular are most arctic arthropod species remains to be estimated frequently used as model organisms in phenological (Bale et al., 2002; Hodkinson et al., 1996; Hodkinson studies, because they respond rapidly to environmental and Bird, 1998). The biological monitoring program at change (Diamond et al., 2011; Illan et al., 2012; Roy Zackenberg, north-east Greenland, represents an excep- and Sparks, 2000; Westwood and Blair, 2010). Addi- tion, where arthropods have been collected weekly dur- tionally, their appeal to the general public has stimula- ing the growing season since 1996 and subsequently ted monitoring programs and has generated high-quality sorted to the family level (Schmidt et al., 2012). data, especially in temperate and boreal regions Very little information exists about the drivers of (Karlsson, In press; Roy and Sparks, 2000). Recent phenological variation at the species level for Arctic global warming, however, is particularly pronounced in arthropods, but family-level studies suggest that timing the Arctic and empirical evidence suggests that Arctic of snowmelt and temperature are important (Danks and

Received Dec. 17, 2013; accepted Mar. 20, 2014.  Corresponding author. E-mail: [email protected]. © 2014 Current Zoology

244 Current Zoology Vol. 60 No. 2

Oliver, 1972; Høye and Forchhammer, 2008a, b; availability of nectar resources can affect adult longev- Strathdee and Bale, 1998). Snow cover acts as a con- ity and population dynamics (Boggs and Inouye, 2012; straint on arthropod growth and development in arctic Cahenzli and Erhardt, 2012). Hence, if adult butterflies and alpine environments (Forrest and Thomson, 2011; emerge in asynchrony with the flowering season or Iler et al., 2013a). Effectively, the timing of snowmelt during a period of low flower availability, their flight opens and closes the activity season for many arthro- season may become shorter with potential detrimental pods in the Arctic. For many species, this time period effects on population dynamics (Nilsson et al., 2008). dictates the amount of resources they can accrue and We have recently demonstrated that the flowering sea- utilize in the current season e.g. for reproduction, dis- son at our study site is shortening and this could affect persal or to carry them over the winter into the follow- resource availability and flight season duration in but- ing season. Temperature during the active season may terfly species (Høye et al., 2013). modulate the effect of snowmelt on arthropod phenol- Here, we examine species-specific phenological re- ogy by affecting their developmental rate (Hodkinson et sponses to recent warming in high-arctic butterflies and al., 1996). In addition, many Arctic arthropod species, their temporal synchrony with floral resources. We use like spiders (Bowden and Buddle, 2012; Høye et al., specimens collected as part of the Zackenberg Basic 2009) and insects (Butler, 1982; Morewood and Ring, monitoring programme (Meltofte and Rasch, 2008) that 1998), have multi-annual life cycles and their phenolo- were previously identified to the family level. We iden- gical responses to climate change may further depend tify these specimens to the species level and establish on overwintering strategy or species-specific develop- species-specific estimates of onset, peak and end of the mental cues. flight season. For two species with sufficient data, we Like their predicted shifts in distribution under cli- ask two specific questions: 1) Are onset, peak, and mate change (Eskildsen et al., 2013; Leroux et al., end of the flight seasons affected by recent rapid 2013), phenological responses to climatic change in warming and changes in timing of snowmelt at Zack- arthropods could be species-specific and have concealed enberg? 2) To what extent is the duration of the flight consequences for trophic interactions. In arctic insects, season related to the timing of the flight season and the the emergence of adults normally coincides with the temporal overlap with the flowering season of relevant peak in food resources, the availability of mates, or plant species. suitable egg-laying habitats (Danks, 2004; Høye and 1 Material and Methods Forchhammer, 2008b; MacLean, 1980). Arctic pollina- tor communities are dominated by Diptera species that 1.1 Study area and data are generalized in their flower visitation patterns Data were collected at Zackenberg, north-east (Elberling and Olesen, 1999; Lundgren and Olesen, Greenland (74°28'N, 20°34'W) as part of the Zacken- 2005). Among arctic flower visitors, however, butter- berg Basic monitoring programme. Although expanding, flies visit only a subset of the available flowering plants the growing season is currently limited to between early (Olesen et al., 2008). Hence, butterflies form a relevant June and early September during which the average air taxonomic group with which to examine species-speci- temperature is around 4.5°C. Throughout the study pe- fic phenological responses to climate change. Even if riod, temperature and snow depth were recorded hourly different butterfly species respond to changes in the by an automated weather station (Hansen et al., 2008). same phenological cue, they can do so at different rates Timing of snowmelt was estimated as the first date (Hegland et al., 2009). Hence, more subtle changes in when less than 10 cm of snow was measured (Hinkler et the phenological response of a pollinator community al., 2008). The vegetation of the study area can be may only be fully resolved with species-level informa- roughly divided into five major plant communities: fen, tion (Iler et al., 2013b). grassland, Salix snow-bed, Cassiope heath, and Dryas Pollinator populations are often limited by the availa- heath (Elberling et al., 2008). bility of floral resources (Potts et al., 2010). Such re- Arthropods were monitored during 14 consecutive source limitation can emerge if the flight season of a years (1996–2009) in one window trap plot and six pit- pollinator shifts relative to the flowering season of its fall trap plots in the vicinity (<2 km) of the weather plant resource causing a decrease in the temporal over- station. Each pitfall trap plot (10 m × 20 m) consisted of lap of pollinators and flowers (Høye et al., 2013; Miller- eight pitfall traps during 1996–2006 and four pitfall Rushing et al., 2010). Moreover, in butterflies, the traps thereafter. The pitfall traps were yellow plastic HØYE TT et al.: Phenology of high-arctic butterflies 245 cups 10 cm in diameter. The colour was chosen to at- mates, we assumed that all Boloria specimens not sub- tract flying insects while also catching surface-active ject to species identification were B. chariclea since B. arthropods. The window trap plot consisted of two traps polaris made up such a small subset of the Boloria sp. each with two chambers. The two traps were placed specimens in the subset subject to species identification. perpendicular to each other. Arthropods caught in the The Lycaenid species P. glandon was even rarer (n = 23 traps were collected weekly during June, July and Au- specimens) than B. polaris in the samples across all gust. The window trap plot and the pitfall traps in arid years (Table 1). The low sample sizes prevented us from heath (one plot), mesic heath (two plots) and in the fen estimating inter-annual variation in the phenology of B. (one plot) were in operation during the entire study pe- polaris and P. glandon. Adults of all four species of but- riod 1996–2009, while one snow-bed plot and one addi- terflies were only observed during July, August and early tional arid heath plot were only operated during the pe- September (Table 1). Traps were occasionally flooded, riods 1996–1998 and 1999– 2009, respectively (see destroyed by arctic foxes (Vulpes lagopus, Linneaus), or Schmidt et al., 2012 for details). trampled by muskoxen (Ovibos moschatus, Zimmer- The entire butterfly community at Zackenberg con- mann), so the capture numbers in each plot were con- sists of four species: Arctic fritillary, Boloria chariclea verted to individuals caught per trap per day for each (Schneider), Polaris fritillary, B. polaris (Boisduval), trapping period, i.e. scaled by the number of active traps Arctic blue, Plebejus glandon (de Prunner), and North- and the number of days each trap was active (see Høye ern clouded yellow, Colias hecla (Lefèbvre). Only one and Forchhammer 2008b for further information). additional species, Small copper, Lycaena phlaeas The Zackenberg Basic monitoring program also in- (Linnaeus) has been observed in Greenland. All speci- cluded weekly observations of the number of buds, mens caught in all years were previously identified to open and senescent flowers from the same locality and the family level (Nympalidae, Lycaenidae and Pieridae) the same time periods within and across years on six as part of Zackenberg Basic monitoring program. We flowering plant species (Cassiope tetragona, Dryas identified the specimens to the species level and calcu- octopetala, Papaver radicatum, Salix arctica, Saxifraga lated species-specific estimates of onset, peak and end oppositifolia and Silene acaulis). These plant species of the flight season. are very common at the study site. The butterfly species The data set included a total of 3,868 specimens re- B. chariclea has been observed on flowers of all six trieved from the pitfall and window trap plots. As part species and C. hecla has only been observed on D. oc- of this study, we revisited a subset of 1,660 Boloria topetala, S. arctica and S. acaulis (Olesen et al., 2008). specimens, distributed evenly across all years. Because For B. chariclea, we used data on all six plant species our subset represents more than half (51%) of all Bolo- and for C. hecla we used only data on the three plant ria specimens in the collection, differences in phenolo- species upon whose flowers the species has been ob- gy between B. chariclea and B. polaris in the subset served. Each plant species was monitored in 3–6 plots would also likely reflect the total sample of specimens. (see Schmidt et al., 2012 for details). Both butterfly A total of 35 specimens (out of 1,660) were identified as species have also been observed on flowers of other B. polaris, equalling just 2.1% of the Boloria specimens. plant species, for which we have no data on flowering For the purpose of making robust phenological esti- phenology. We used the data on flower phenology to

Table 1 Summary of phenological observations for four species of butterflies at Zackenberg, north-east Greenland (Bol- oria chariclea, B. polaris, Colias hecla and Plebejus glandon)

Species First Onset Peak End Last n B. chariclea 168 200±2.0 212±2.3 224±1.7 245 3,235* B. polaris 176 190 217 232 238 35* C. hecla 176 194±2.4 205±2.6 218±2.3 238 575 P. glandon 196 196 204 211 224 23 All dates are given as days after 1st January. The first observation is the first date at which a butterfly of a particular species was found in a trap across all years (1996–2009) and the last observation is the last date across all years. Onset, peak and end indicate the mean day of year ± SE at which 10%, 50% and 90% of the annual sum of individuals were caught, respectively. Asterisks indicate that only 51% of the total collection of Boloria specimens was checked for B. polaris and the remaining specimens were assumed to be B. chariclea. The sample size, n gives the total number of specimens of a particular species identified from the traps. 246 Current Zoology Vol. 60 No. 2 characterize the temporal variability in floral resources type III sums of squares and starting with interaction available for each of the two butterfly species. An inde- terms and evaluating main effects only if interaction pendent study of the pollination network at our site terms were non-significant. The correlation between documented visits by B. chariclea to 15 plant species timing of snowmelt and July temperature as well as and to 5 plant species by C. hecla (Olesen et al., 2008). between peak flight time of B. chariclea and C. hecla 1.2 Data analysis was estimated with Pearson correlation analysis. We Following Høye and Forchhammer (2008b), we de- calculated the temporal extent of overlap between the fined onset, peak and end of the flight season of adult flowering season of relevant plant species (see above) butterflies by dates at which 10%, 50% and 90% of the and the butterfly flight season in days for B. chariclea seasonal capture of a species was reached, respectively. and C. hecla, separately. We tested for trends in the du- The dates for onset, peak and end of the flight season ration of overlap during the study period and whether were estimated by linear interpolation between weekly duration of the flight season was affected by onset of trapping periods. For instance, peak flight season was the flight season, the duration of the overlap with floral estimated by interpolation between the latest weekly resources or their interaction. Finally, for analyses of trapping period in which less than 50% of the individu- temporal trends in peak flight time and overlap with als were caught and the earliest weekly trapping period flower seasons, year was used as a continuous predictor in which more than 50% of the individuals were caught. variable in simple linear regression models. For each weekly trapping period, we pooled the number 2 Results of individuals from all plots to base the phenological estimates on the largest number of individuals and be- The number of specimens was highest for B. chari- cause previous analyses of Boloria specimens demon- clea (n = 3,235) and C. hecla (n = 575) and very low strated limited phenological variation among plots for B. polaris (n = 35) and P. glandon (n = 23). The (Høye and Forchhammer, 2008b). Although low sample earliest and the latest butterfly records were observa- sizes prohibited us from estimating yearly phenology tions of B. chariclea, while the earliest observation of P. metrics for B. polaris and P. glandon, we calculated an glandon was later than all other species and the latest average across years of onset, peak, and end of the observation of this species was also earlier than all other flight season for future comparisons. All dates were species. The average peak flight time was late July for C. expressed as day of year with 1st January equalling day hecla and P. glandon, and about ten days later in B. 1. Flight season duration was estimated as the number chariclea and B. polaris (Table 1). For B. chariclea and of days between onset and end of the flight season. We C. hecla, the timing of peak flight season varied marke- quantified onset and end of the community-wide flow- dly from year to year (Fig. 1), with a significant trend ering season at a landscape scale based on the relevant towards earlier peak flight season in B. chariclea (slope = plant species for each of the two butterfly species fol- -1.22, F1,12 = 6.14, P = 0.029), but not in C. hecla (slope = lowing the approach of Høye et al. (2013). -0.68, F1,12 = 1.12, P = 0.31). In B. chariclea, onset, At the study site, snowmelt typically takes place peak, and end of the flight period showed a statistically during June, while the flight season of butterflies is significant relationship with timing of snowmelt, mean typically in July and early August. We were interested in daily July temperature, and their interaction (Table 2). separating the effect of snowmelt from the effect of In C. hecla, the onset, peak, and end of the flight season temperature after snowmelt on the timing of the butter- was, however, only significantly related to timing of fly flight season. Hence, we used mean daily July tem- snowmelt (Table 2). Timing of snowmelt and July tem- perature as a predictor of timing of the butterfly flight peratures were not significantly correlated (Pearson cor- season in addition to timing of snowmelt. We applied relation: r = -0.38, n = 13, P = 0.18). The flight season of generalized linear models with a Gaussian error distri- both butterfly species was generally later than the flower- bution (McCullagh and Nelder, 1998) with onset, peak, ing season of the plant species that are visited by the end, and duration of the flight season of B. chariclea butterflies and for which we had flower phenology data and C. hecla as response variables. As predictors for (Fig. 2). The duration of the flight season showed a sta- models with onset, peak and end of the flight season, we tistically significant and positive relationship to the du- used the mean daily temperature during July, the timing ration of overlap with floral resources in both B. chari- of snowmelt, and their interaction. Non-significant clea (Fig. 3a) and C. hecla (Fig. 3b), but not with onset terms were removed successively using F-tests based on of the flight period or their interaction in any of the two HØYE TT et al.: Phenology of high-arctic butterflies 247

species (Table 3). The duration of overlap became smaller ficant for B. chariclea (slope = -0.080, F1,12 = 0.13, P = during the study period, but the relationship was not signi- 0.72) or C. hecla (slope = -0.77, F1,12 = 3.56, P = 0.084).

Fig. 1 Peak flight times of two butterfly species A) Boloria chariclea and B) Colias hecla during 1996–2009 at Zackenberg, north-east Greenland. Summary statistics (R2 and P-values) based on simple linear regression analysis are given in each panel Regression lines indicate that slopes are significantly different from zero.

Table 2 Results of generalized linear models of onset, peak and end of the flight season for two butterfly species (Boloria chariclea and Colias hecla) at Zackenberg, north-east Greenland, across 1996–2009 as a function of mean daily temperature during July and timing of snowmelt, as well as their interaction as continuous variables Species Event Intercept Snowmelt Temp Snowmelt:Temp df R2 P B. chariclea Onset -144.2±111.8 2.05±0.66 35.4±14.4 -0.21±0.09 10 0.73 0.0032 Peak -266.1±136.6 2.90±0.80 54.5±17.6 -0.33±0.10 10 0.73 0.0037 End -54.4±103.9 1.71±0.61 30.9±13.4 -0.19±0.08 10 0.69 0.0069 C. hecla Onset 109.7±27.5 0.50±0.16 - - 12 0.44 0.0096 Peak 104.1±26.5 0.60±0.16 - - 12 0.55 0.0025 End 131.3±24.6 0.51±0.15 - - 12 0.51 0.0043 Only models with significant parameters (at α = 0.05) are shown. Parameters ± SE are presented along with residual degrees of freedom, coefficient of determination (R2) and P-value.

Fig. 2 The onset and end of the flight season of A) Boloria chariclea and community-wide flowering seasons for six flower- ing plant species (Cassiope tetragona, Dryas octopetala, Papaver radicatum, Salix arctica, Saxifraga oppositifolia and Silene acaulis) which the butterfly species has been observed visiting and B) Colias hecla and community-wide flowering seasons for three flowering plant species (Dryas octopetala, Salix arctica and Silene acaulis) on which the butterfly species has been observed visiting The lower line in each set is onset and the upper line is end of butterfly flight seasons (hatched lines) and flowering season (full lines). 248 Current Zoology Vol. 60 No. 2

Fig. 3 Duration (end minus onset) in days of the flight season for A) Boloria chariclea and B) Colias hecla in relation to the overlap with the flowering season of six relevant plant species for B. chariclea and three relevant plant species for C. hecla (see text for details) The overlap is negative in some years, because onset and end of seasons are defined by the date of 10% and 90% of butterflies and flowers observed during the season, respectively. Regression lines based on simple linear regression analysis along with R2 and p-values are given in each panel.

Table 3 Results of generalized linear models of flight season duration for two butterfly species (Boloria chariclea and Colias hecla) at Zackenberg, north-east Greenland, from 1996–2009 as a function of the overlap with the flowering season of their respective plant species (upon which they have been observed), the onset of flight season, and their interaction as con- tinuous variables Species Intercept Overlap Onset Overlap:Onset df R2 P B. chariclea 21.41±1.46 0.83±0.33 - - 12 0.35 0.027 C. hecla 17.32±2.52 0.49±0.17 - - 12 0.39 0.016 Only models with significant parameters (at α = 0.05) are shown. Parameters ± SE are presented along with residual degrees of freedom, coefficient of determination (R2) and P-value.

3 Discussion By identifying the individual species of a butterfly assemblage from the monitoring program at Zacken- The literature on species-specific phenological re- berg, we have made progress towards estimating spe- sponses to ambient climatic variability of Arctic ar- cies-specific phenological responses to recent climatic thropods is scant and to our knowledge this is the first variation in the Arctic. The two abundant butterfly study to span more than a decade of observations species indeed differ in their trend towards earlier (Høye and Sikes, 2013; Leung and Reid, 2013). Given flight time, their phenological responses to climatic the species richness of arthropods in marine, freshwa- variability, and in inter-annual variation in degree of ter and terrestrial environments in the Arctic, this illu- temporal overlap with floral resources. This suggests strates an important knowledge gap that limits our that phenological responses measured at coarse taxo- ability to predict the consequences of rapid climate nomic resolutions can mask important variability at the change for Arctic biodiversity (Meltofte 2013; Callag- species level. han et al., 2004b; Post et al., 2009). We have previ- Our results indicate that timing of the flight season in ously demonstrated large family-level variation in es- high-arctic butterfly species is closely related to the timates of timing of emergence over a broad range of timing of snowmelt. While the timing of the flight sea- terrestrial arthropods at Zackenberg (Høye et al., 2007; son in C. hecla is only related to timing of snowmelt, Høye and Forchhammer, 2008b). However, with family- there is an additional effect of July temperature in B. level taxonomic resolution, it is not possible to sepa- chariclea, and this temperature effect interacts with rate the effects of changing species composition be- timing of snowmelt. This suggests that the advancement tween years from the inter-annual variation in phe- of the flight season in B. chariclea with warmer tem- nology of individual species (Callaghan et al., 2004a). peratures is most pronounced with late snowmelt. The HØYE TT et al.: Phenology of high-arctic butterflies 249 lack of relationship to summer temperatures in timing of flowering season is rather conservative. Indeed, in a the flight season for C. hecla may be the reason why C. pollination network study from our site, the same plant hecla is not showing a trend towards earlier flight time species were observed to flower later than the end of during the study period. The summer temperature has flowering in our permanent plots (J.-M. Olesen unpub- increased dramatically during the study period at our lished data). site (Høye et al., 2013). With future warming, the flight Our results demonstrate that for high-arctic study seasons of B. chariclea and C. hecla may shift relative sites it is possible to use pitfall trapping to census but- to each other. Since peak flight date is generally ten terfly species. We used yellow pitfall traps as a method days later in B. chariclea than in C. hecla, their flight to trap both ground-active and flying insects in the same times are likely to become more synchronized. A know- trap (Böcher and Meltofte, 1997). The large sample ledge gap is whether future timing of snowmelt in the sizes for at least two species of butterflies using this region will advance or be delayed, which is currently technique suggests that this approach is useful for re- uncertain (Stendel et al., 2008). mote tundra localities as it does not require trained but- The duration of butterfly flight seasons could be affe- terfly observers who are otherwise needed for conven- cted by the timing and duration of flowering seasons if tional transect counts (Pollard and Yates, 1993). The the lifespan of each individual butterfly is constrained separation of B. chariclea and B. polaris is difficult in by access to floral resources. Adult longevity has been the field and we provide new information on their rela- linked to nectar resources in some butterfly species tive abundance. We found less than 3% of the Boloria (Cahenzli and Erhardt, 2012; Murphy et al., 1983). Our specimens to be B. polaris, suggesting that this species results demonstrate that the duration of the flight season is much rarer at Zackenberg than B. chariclea. This varied by a factor of two over the study period in both could be linked to the distribution of their larval food butterfly species and was related to the duration of plants or because of habitat segregation not picked up overlap with relevant floral resources. Other studies by our sampling design. We consider it unlikely that our have found extending flight seasons, but these studies subsampling procedure, where we identified 51% of the have used observations collected at much larger spatial Boloria specimens, could have accidentally missed extents (Roy and Sparks, 2000; Westwood and Blair, samples with a large proportion of B. polaris specimens. 2010). For instance, a recent study demonstrated ex- The low capture numbers for B. polaris and P. glandon tended flight seasons in northern butterfly communities could potentially be the result of these butterfly species in response to recent warming across Sweden (Karlsson not being attracted to the yellow colour of our pitfall In press). Flight seasons may also be extended when the traps, but independent data from a study of the pollina- flight seasons start earlier. This could happen by more tion network at our study site support the idea that B. pronounced protandry (i.e. males emerging earlier than polaris and P. glandon are rare species in the area (J.-M. females) in years of early snowmelt and warmer tem- Olesen unpublished data). peratures. The strong phenological response to climate varia- Our estimates of the onset and end of the flowering tion that we document for high-arctic butterflies and the season of plant species known to be visited by the studi- clear differences between the two most abundant spe- ed butterfly species is generally earlier than the flight season of the butterflies. Both species of butterflies are cies suggest that species-specific studies of populations known to visit flowers of other plants as well and some of arctic arthropods should be a research priority. Future of these are late flowering species (e.g. Arnica angusti- phenological studies of Arctic arthropods should ideally folia and Bistorta vivipara). Hence, the apparent mis- be analysed in context with interacting species to assess timing of the butterfly flight season with the flowering how ongoing climate change will affect Arctic biodiver- season presented here could be due to the subset of sity. Long-term records will be particularly helpful in plant species for which we have data on flowering phe- attempts to identify the climatic constraints on the fu- nology. The butterfly flight season may also be timed ture persistence of high-arctic butterfly species (Post primarily with particular phenological stages of the and Høye, 2013). plant species used for oviposition, although this would Acknowledgements We thank Zackenberg Basic for access not explain why flight season duration is related to the to the ecosystem data. T.T.H. and A.E. acknowledge 15. Juni overlap with our subset of plant species. We consider it Fonden for funding and W.D.K. acknowledges a University of more likely that our measure of onset and end of the Amsterdam (UvA) starting grant. We appreciate comments on 250 Current Zoology Vol. 60 No. 2

a previous version of the manuscript by Oliver Schweiger and Ecol. Lett. 12: 184–195. two anonymous reviewers. Hinkler J, Hansen BU, Tamstorf MP, Sigsgaard C, Petersen D, 2008. Snow and snow-cover in central northeast Greenland. References Adv. Ecol. Res. 40: 175–195. Hodkinson ID, Coulson SJ, Webb NR, Block W, Strathdee AT et Bale JS, Masters GJ, Hodkinson ID, Awmack C, Bezemer TM et al., 1996. Temperature and the biomass of flying midges (Dip- al., 2002. Herbivory in global climate change research: Direct tera: Chironomidae) in the high Arctic. Oikos 75: 241–248. effects of rising temperatures on insect herbivores. Global. Hodkinson ID, Bird J, 1998. Host-specific insect herbivores as Change Biol. 8: 1–16. sensors of climate change in arctic and alpine environments. Böcher J, Meltofte H, 1997. Comparison of three different types of arthropod traps. In: Meltofte H, Thing H ed. Zackenberg Arct. Alp. Res. 30: 78–83. Ecological Research Operations, 2nd Annual Report, 1996. Høye TT, Post E, Meltofte H, Schmidt NM, Forchhammer MC, Copenhagen, Denmark: Danish Polar Center, Ministry of Re- 2007. Rapid advancement of spring in the High Arctic. Curr. search & Information Technology, 66–67. Biol. 17: R449–R451. Boggs CL, Inouye DW, 2012. A single climate driver has direct Høye TT, Forchhammer MC, 2008a. The influence of weather and indirect effects on insect population dynamics. Ecol. Lett. conditions on the activity of high-arctic arthropods inferred 15: 502–508. from long-term observations. BMC Ecol. 8:8. Bowden JJ, Buddle CM, 2012. Life history of tundra-dwelling Høye TT, Forchhammer MC, 2008b. Phenology of high-arctic wolf spiders (Araneae: Lycosidae) from the Yukon Territory, arthropods: Effects of climate on spatial, seasonal and inter- Canada. Can. J. Zool. 90: 714–721. annual variation. Adv. Ecol. Res. 40: 299–324. Butler MG, 1982. A 7-year life-cycle for 2 Chironomus species in Høye TT, Hammel JU, Fuchs T, Toft S, 2009. Climate change and arctic alaskan tundra ponds (Diptera, Chironomidae). Can. J. sexual size dimorphism in an arctic spider. Biol. Lett. 5: 542– Zool. 60: 58–70. 544. Cahenzli F, Erhardt A, 2012. Nectar sugars enhance fitness in Høye TT, Post E, Schmidt NM, Trøjelsgaard K, Forchhammer male Coenonympha pamphilus butterflies by increasing lon- MC, 2013. Shorter flowering seasons and declining abundance gevity or realized reproduction. Oikos 121: 1417–1423. of flower visitors in a warmer Arctic. Nature Clim. Change 3: Callaghan TV, Bjorn LO, Chernov Y, Chapin T, Christensen TR et 759–763. st al., 2004a. Responses to projected changes in climate and Høye TT, Sikes DS, 2013. Arctic entomology in the 21 century. UV-B at the species level. Ambio 33: 418–435. Can. Entomol. 145: 125–130. Callaghan TV, Bjorn LO, Chernov Y, Chapin T, Christensen TR et Iler AM, Høye TT, Inouye DW, Schmidt NM, 2013a. Nonlinear al., 2004b. Effects on the structure of arctic ecosystems in the flowering responses to climate: Are species approaching their short- and long-term perspectives. Ambio 33: 436–447. limits of phenological change? Phil. Trans. R. Soc. B 368 Danks HV, Oliver DR, 1972. Seasonal emergence of some high 20120489. arctic Chironomidae (Diptera). Can. Entomol. 104: 661–686. Iler AM, Inouye DW, Høye TT, Miller-Rushing AJ, Burkle LA et Danks HV, 2004. Seasonal adaptations in arctic insects. Integr. al., 2013b. Maintenance of temporal synchrony between syr- Comp. Biol. 44: 85–94. phid flies and floral resources despite differential phenological Diamond SE, Frame AM, Martin RA, Buckley LB, 2011. Species' responses to climate. Glob. Change Biol. 19: 2348–2359. traits predict phenological responses to climate change in but- Illan JG, Gutierrez D, Diez SB, Wilson RJ, 2012. Elevational terflies. Ecology 92: 1005–1012. trends in butterfly phenology: Implications for species respon- Elberling B, Tamstorf MP, Michelsen A, Arndal MF, Sigsgaard C ses to climate change. Ecol. Entomol. 37: 134–144. et al., 2008. Soil and plant community-characteristics and dy- Karlsson B, In press. Extended season for northern butterflies. Int. namics at Zackenberg. Adv. Ecol. Res. 40: 223–248. J. Biometeorol.: doi:10.1007/s00484–00013–00649–00488. Elberling H, Olesen JM, 1999. The structure of a high latitude Leroux SJ, Larrivee M, Boucher-Lalonde V, Hurford A, Zuloaga J plant–flower visitor system: The dominance of flies. Ecogra- et al., 2013. Mechanistic models for the spatial spread of spe- phy 22: 314–323. cies under climate change. Ecol. Appl. 23: 815–828. Eskildsen A, le Roux PC, Heikkinen RK, Høye TT, Kissling WD Leung MC, Reid DG, 2013. New species records for butterflies et al., 2013. Testing species distribution models across space (Lepidoptera) on Herschel Island, Yukon, Canada, with notes and time: High latitude butterflies and recent warming. Global on natural history. Can. Entomol. 145: 227–234. Ecol. Biogeogr. 22: 1293–1303. Lundgren R, Olesen JM, 2005. The dense and highly connected Forrest JRK, Thomson JD, 2011. An examination of synchrony world of Greenland's plants and their pollinators. Arct. Antarct. between insect emergence and flowering in Rocky Mountain Alp. Res. 37: 514–520. meadows. Ecol. Monogr. 81: 469–491. MacLean SF, Jr., 1980. The detritus-based trophic system. In: Hansen BU, Sigsgaard C, Rasmussen L, Cappelen J, Hinkler J et (Brown J, Miller PC, Tieszen LL, Bunnell FL ed. An Arctic al., 2008. Present-day climate at Zackenberg. Adv. Ecol. Res. Ecosystem: The Coastal Tundra at Barrow, Alaska. Strouds- 40: 111–149. burg, PA: Dowden, Hutchinson & Ross, Inc., 411–457. Hegland SJ, Nielsen A, Lázaro A, Bjerknes AL, Totland, 2009. McCullagh P, Nelder JA, 1998. Generalized Linear Models. 2nd How does climate warming affect plant-pollinator interactions? edn. Boca Raton: Chapman & Hall/CRC. HØYE TT et al.: Phenology of high-arctic butterflies 251

Meltofte H, 2013. Arctic Biodiversity Assessment: Status and sen TR et al., 2009. Ecological dynamics across the Arctic associ- Trends in Arctic Biodiversity. Akureyri: Conservation of Arc- ated with recent climate change. Science 325: 1355–1358. tic Flora and Fauna. Post E, Høye TT, 2013. Advancing the long view of ecological Meltofte H, Rasch M, 2008. The study area at Zackenberg. Adv. change in tundra systems. Phil. Trans. R. Soc. B 368 20120477. Ecol. Res. 40: 101–110. Potts SG, Biesmeijer JC, Kremen C, Neumann P, Schweiger O et Miller-Rushing AJ, Høye TT, Inouye DW, Post E, 2010. The ef- al., 2010. Global pollinator declines: Trends, impacts and dri- fects of phenological mismatches on demography. Phil. Trans. vers. Trends Ecol. Evol. 25: 345–353. R. Soc. B 365: 3177–3186. Roy DB, Sparks TH, 2000. Phenology of British butterflies and Morewood WD, Ring RA, 1998. Revision of the life history of climate change. Glob. Change Biol. 6: 407–416. the High Arctic moth Gynaephora groenlandica (Wocke) Schmidt NM, Hansen LH, Hansen J, Berg TB, Meltofte H, 2012. (Lepidoptera: Lymantriidae). Can. J. Zool. 76: 1371–1381. BioBasis: conceptual design and sampling procedures of the Murphy DD, Launer AE, Ehrlich PR, 1983. The role of adult biological monitoring programme within Zackenberg basic. feeding in egg-production and population dynamics of the che- Aarhus: Aarhus University, Department of Bioscience. ckerspot butterfly Euphydryas editha. Oecologia 56: 257–263. Stendel M, Christensen JH, Petersen D, 2008. Arctic climate and Nilsson SG, Franzen M, Jonsson E. 2008. Long-term land-use climate change with a focus on Greenland. Adv. Ecol. Res. 40: changes and extinction of specialised butterflies. Insect. Con- 13–43. serv. Diver. 1: 197–207. Strathdee AT, Bale JS, 1998. Life on the edge: Insect ecology in Olesen JM, Bascompte J, Elberling H, Jordano P, 2008. Temporal arctic environments. Annu. Rev. Entomol. 43: 85–106. dynamics in a pollination network. Ecology 89: 1573–1582. Thackeray SJ, Sparks TH, Frederiksen M, Burthe S, Bacon PJ et Parmesan C, 2006. Ecological and evolutionary responses to recent al., 2010. Trophic level asynchrony in rates of phenological climate change. Annu. Rev. Ecol., Evol. Syst. 37: 637–669. change for marine, freshwater and terrestrial environments. Pollard E, Yates TJ, 1993. Monitoring butterflies for ecology and Glob. Change Biol. 16: 3304–3313. conservation: The British butterfly monitoring scheme. Lon- Westwood AR, Blair D, 2010. Effect of regional climate warming don: Chapman & Hall. on the phenology of butterflies in boreal forests in Manitoba, Post E, Forchhammer MC, Bret-Harte MS, Callaghan TV, Christen- Canada. Environ. Entomol. 39: 1122–1133.