Greater shrub dominance alters breeding habitat and food resources for migratory songbirds in Alaskan arctic tundra
Abstract
Climate warming is affecting the Arctic in multiple ways, including via increased dominance of deciduous shrubs. Although many studies have focused on how this vegetation shift is altering nutrient cycling and energy balance, few have explicitly considered effects on tundra fauna, such as the millions of migratory songbirds that breed in northern regions every year. To understand how increasing deciduous shrub dominance may alter breeding songbird habitat, we quantified vegetation and arthropod community characteristics in both graminoid and shrub dominated tundra. We combined measurements of preferred nest site characteristics for Lapland longspurs (Calcarius lapponicus) and Gambel's White-crowned sparrows (Zonotrichia leucophrys gambelii) with modeled predictions for the distribution of plant community types in the Alaskan arctic foothills region for the year 2050. Lapland longspur nests were found in sedge-dominated tussock tundra where shrub height does not exceed 20 cm, whereas White-crowned sparrows nested only under shrubs between 20 cm and 1 m in height, with no preference for shrub species. Shrub canopies had higher canopy-dwelling arthropod availability (i.e. small flies and spiders) but lower ground-dwelling arthropod availability (i.e. large spiders and beetles). Since flies are the birds' preferred prey, increasing shrubs may result in a net enhancement in preferred prey availability. Acknowledging the coarse resolution of existing tundra vegetation models, we predict that by 2050 there will be a northward shift in current White-crowned sparrow habitat range and a 20–60% increase in their preferred habitat extent, while Lapland longspur habitat extent will be equivalently reduced. Our findings can be used to make first approximations of future habitat change for species with similar nesting requirements. However, we contend that as exemplified by this study's findings, existing tundra modeling tools cannot yet simulate the fine-scale habitat characteristics that are critical to accurately predicting future habitat extent for many wildlife species.
Introduction
Since 1960, arctic regions have been warming at a rate two to three times higher than the global average (Anisimov et al., 2007), and the physical and biological responses are proving acute (Callaghan et al., 2004). The main documented responses include increasing vegetation productivity (e.g. Goetz et al., 2007; Epstein et al., 2012), lengthened growing seasons (Zeng et al., 2011), enhanced fire regimes (Jones et al., 2009; Bret-Harte et al., 2013), thawing permafrost, and increasing active layer depth and temperature (Anisimov et al., 2007). In addition, woody deciduous shrubs have become increasingly dominant in many arctic regions, including the North Slope of Alaska, over the past 50 years (Sturm et al., 2001; Tape et al., 2006), and several experimental and modeling studies show that this trend is expected to continue over the next several decades (Pearson et al., 2013; Zhang et al., 2013). As highlighted in a recent review of impacts of ‘shrubification’ (Myers-Smith et al., 2011), multiple studies have focused on how shrub expansion on the tundra alters nutrient cycling and energy balance, however, fewer have explicitly considered the role that shrub expansion may have on tundra fauna (Zöckler, 2005; Henden et al., 2011; Ehrich et al., 2012; Sokolov et al., 2012).
Millions of migratory songbirds migrate to the arctic tundra biome every year (Pielou, 1994) due in part to the abundant summer food resources, long day length, and fewer predators and parasites (in most years) relative to more southern ecosystems (e.g. Richardson et al., 1995; Piersma, 1997; McKinnon et al., 2010). Songbirds are important prey for many tundra predators, and also provide essential ecosystem services – such as seed dispersal and insect control (Sekercioglu, 2006) – not only in the Arctic, but in the more southern and often disparate ecosystems they transit during migration and inhabit over winter months. However, as deciduous shrubs become more dominant in the arctic tundra, songbird breeding habitats are being altered in a variety of ways that have currently unknown consequences for birds. Previous work has shown that deciduous tundra shrubs [willow, birch and alder (Salix, Betula and Alnus spp., respectively)] have greater dominance at the expense of understory plant species such as mosses, graminoids, and evergreens (e.g. Cornelissen et al., 2001), thereby not only shifting plant community composition but also creating a taller and structurally more complex canopy (Walker et al., 2005; Macias-Fauria et al., 2012). Increasing shrub dominance in tundra regions is occurring in three main ways: infilling of existing shrub patches, increases in vertical and lateral growth of existing shrubs, and advancement of shrubline (Myers-Smith et al., 2011). These changes could significantly alter availability and variety of suitable nesting and feeding habitats utilized by different bird species, as well as availability of shelter from inclement weather and predation. Songbirds that breed in northern Alaska every year include both shrub nesting passerine species [e.g. Gambel's white-crowned sparrows (Zonotrichia leucophrys gambelii), American tree sparrows (Spizella arborea), and American robins (Turdus migratorius)] and open tundra nesters [e.g. Lapland longspurs (Calcarius lapponicus) and Smiths longspurs (Calcarius pictus)] (Poole, 2005). This broad distinction in nesting habitat characteristics suggests that as arctic tundra regions become increasingly shrubby, nest habitat quality will increase or decrease depending on individual species preferences (Sokolov et al., 2012; Henden et al., 2013). However, the potential importance of more nuanced habitat characteristics associated with increasing shrub dominance – such as shrub height and species dominance – also require consideration. While a few studies have examined how some tundra nesting species such as Lapland longspurs and White-crowned sparrows discriminate among microtopographical features or the presence of water (Rodrigues, 1994; Wingfield et al., 2004; Boal & Andersen, 2005), most studies have only broadly characterized nest site preferences based on hill slope position and vegetation community type cover (e.g. Blanchard Oakeson, 1954). However, Norment (1993) conducted a study in a forest-tundra ecotone that explicitly showed the importance of forest structural characteristics to nesting sparrows, suggesting that changes to the tundra's physical structure that accompany increasing deciduous shrub dominance require consideration. Further, to our knowledge, no tundra studies have yet considered whether shrub-nesting songbirds exhibit nesting preferences for or against any particular deciduous shrub species (i.e. willow, birch, and alder).
In addition to vegetation height and species composition, shrub-dominated communities likely differ relative to graminoid-dominated tundra in the quality and quantity of nutritional resources – such as berries, inflorescences, green foliage, and arthropods – that are available to songbirds. How increasing shrub dominance will impact food resource availability for tundra breeding songbirds has not been explored. In a previous study (Rich et al., 2013), we found greater arthropod availability in graminoid compared to shrub-dominated tundra communities, but the scope was limited to ground-dwelling arthropods (i.e. spiders and beetles), and excluded the contribution of canopy-dwelling communities (i.e. flies). As arthropods are a high-protein food resource required in abundance by chicks (Visser et al., 1998), landscape-level changes in arthropod availability that may accompany increasing shrub dominance could play an essential role in determining songbird reproductive success in arctic tundra.
The overall goal of this study was to understand and predict how increasing deciduous shrub dominance may alter nesting habitat extent and food availability in northern Alaskan tundra for two migratory songbird species – Gambel's White-crowned sparrows (a largely boreal species) and Lapland longspurs (an arctic specialist). Because these species represent two broad groups of arctic tundra breeding songbirds with contrasting habitat requirements – shrub nesting vs. open tundra nesting, respectively – they serve as indicator species that inform how songbirds with similar nest site preferences may experience shifts in habitat as warming continues. Our specific objectives were to: (i) determine how vegetation height, plant species presence/absence, and plant community type influence nest site selection for Gamble's White-crowned sparrows and Lapland longpsurs; (ii) quantify and characterize food resources (arthropods, berries, and inflorescences) available to nesting songbirds in several shrub and graminoid-dominated communities; (iii) obtain a first approximation of future songbird nesting habitat availability in our study region by combining existing knowledge and our own field-based characterization of nest site vegetation with previously modeled predictions for the distribution of arctic tundra vegetation community types by the year 2050 (Pearson et al., 2013). To our knowledge, this is the first attempt to use a published vegetation model to quantifiably predict changes in nesting habitat of songbirds in the Arctic.
Materials and methods
Study area
The study area includes tundra in the vicinity of the Arctic Long Term Ecological Research (ARC LTER) site at Toolik Field Station in the arctic foothills region of Alaska (68°38′ N, 149°34′ W, elevation 760 m) (Fig. 1). The four research sites include Roche Mountonee (ROMO), Toolik Lake Field Station (TLFS), Imnavait Creek (IMVT), and Sag River-Department of Transportation Camp (SDOT). Sites were chosen in May 2010 to represent the most common shrub tundra types in the northern foothills of the Brooks Range. Each of the four sites includes two 20 000 m2 plots: one graminoid-dominated plot (Open plot) and one woody deciduous shrub-dominated plot (Shrub plot). These two plot types were selected based on a priori information of the general tundra vegetation community types preferred by the two study songbird species – White-crowned sparrows (shrub tundra nesters) and Lapland longspurs (open tundra nesters) – so that our plot vegetation and arthropod food resource availability measured within each plot can be directly related to the nearby habitats in which study nests were found. Within each Open and Shrub plot, two permanent 100-m transects were established, for a total of eight Open and eight Shrub transects. Along each transect, ten 1-m2 permanent vegetation quadrats were established at 10-m intervals for repeated, nondestructive vegetation measurements. Open plots are made up of short tussock tundra vegetation, comprised of sedges (e.g. Eriophorum spp. and Carex spp.), forbs (e.g. Polygonum spp. and Pedicularis spp.), dwarf evergreen shrubs (e.g. Vaccinium vitis-idaea), occasional dwarf deciduous shrubs (e.g. Betula nana, Salix pulchra), mosses, and lichens (see Table S1). Shrub plots are made up of either dwarf or erect-deciduous shrub vegetation, dominated by a mixture of Salix pulchra, Salix glauca, Salix alaxensis, Salix richardsonii, Betula nana, and Vaccinium uliginosum, with a mixture of sedges, evergreens, and other herbaceous perennials (IMVT and TLFS), or, riparian shrub tundra vegetation dominated by taller shrubs, mainly Salix alaxensis and Betula nana (ROMO and SDOT) (see Table S1). The height of the tallest deciduous shrub in each quadrat was measured in each Open and Shrub plot, and the mean of these values was used to characterize the average maximum height of shrubs in each of the four Open and Shrub plots: 16 ± 1 cm/22 ± 2 cm (IMVT Open/Shrub), 23 ± 2 cm/35 ± 4.5 cm (TLFS Open/Shrub), 13 ± 2 cm/86 ± 11 cm (ROMO Open/Shrub), and 28 ± 2 cm/84 ± 20 cm (SDOT Open/Shrub).
Songbird nest habitat characteristics
Lapland longspur and White-crowned sparrow nests were located by searching our study plots, as well as any breeding habitat located within approximately 0.5–1 km of the Dalton Highway between ROMO and SDOT (Fig. 1), during the month of June. Due to the relative ease of finding nests in graminoid-dominated compared to shrub-dominated vegetation, once a minimum number of Lapland longspur nests (~25) were found in graminoid-dominated communities, nest searching efforts switched to shrub-dominated tundra to ensure enough time remained in the incubation period to find a similar sample number of White-crowned sparrow nests. Nest searching was conducted along a full range of vegetation community heights (<20 cm to >2 m tall) that are found in the foothills region of the Brooks Range. Nests were found by flushing incubating birds or following birds to their nests. Twenty-seven Lapland longspur and 20 White-crowned sparrow nests were found in 2012, and 30 Lapland longspur and 24 White-crowned sparrow nests were found in 2013. Nest locations were recorded with a handheld GPS so that they could be easily relocated. Once nestlings had fledged (end of June to beginning of July), each nest was revisited and characterized. In a 1-m2 quadrat immediately surrounding the nest, the most abundant plant growth form (e.g. deciduous shrub, evergreen shrub, graminoid, and moss) was estimated by percent aerial plant cover. Within the same 1-m2 area, mean vegetation height was determined by taking the average of five height measurements made at the four corners and near the center. In addition, the specific plant species under/in which the nest was placed was recorded.
Arthropod biomass and abundance
We employed two arthropod sampling approaches that are biased separately towards ground- and canopy-dwelling arthropods. Ground-dwelling arthropod abundance was measured weekly via pitfall traps from June 6 to July 19, 2012 (Norment, 1987) with traps (7.5 cm diameter cups) placed so that the top of each cup was flush with the ground/vegetation surface. The traps were placed at 10-m intervals between the vegetation quadrats along each of the sixteen permanent transects and filled with ~2 cm of 50/50 ethanol–water mix. Trap contents were collected after a 48-h sampling period and transferred to vials to be counted and identified in the laboratory. Twenty open and shrub samples were collected per plot with a total of forty per site. The samples were dried for a minimum of 48 h at 40 °C, after which arthropods were weighed to determine dry biomass. For the analyses presented here, the samples found in each trap across the entire sampling period were aggregated.
Canopy-dwelling arthropod biomass was also measured weekly via sweep netting using a standard insect net (Norment, 1987), from June 6 to July 19, 2012. One 100-m sweep net transect was established in each of the eight 20 000-m2 plots. Using an iron bar as the center point, the direction of the transect was randomly determined at each sample time using a compass bearing generated from a random numbers table to avoid repeated sampling of the same area. Each sweep set involved ten horizontal passes of the net (at 10-m intervals) along the ground vegetation and up into the shrubs to about 2 m where necessary. The ten-sweep set was duplicated on each side of the transect line (out to approximately 5 m on each side), so that each sampling area was approximately 10 m2. The contents of the samples from each sweep set were transferred to plastic bags with 1-cm2 pieces of Shell pest strips to kill arthropods, and then transferred to glass vials to be counted and identified in the laboratory. The samples were dried for a minimum of 48 h at 40 °C, after which arthropods were weighed for dry biomass. Similar to the pitfall samples, weekly sweep net samples collected along the same transect were aggregated across the seasonal sampling period for analysis.
Berry and inflorescence abundance
Berry and inflorescence abundances were determined in late July of 2012, by counting the number of current year's berries and catkins in the front 20% (0.2 m2) of each vegetation quadrat along each of the sixteen permanent transects. The berry producing species along the transects include two evergreen shrub species (Vaccinium vitis-idaea, and Empetrum nigrum), two dwarf deciduous shrubs (Vaccinium uligonosum, and Arctuos alpina), and one forb species (Rubus chamaemorus). The inflorescence producing species along the transects include five erect-deciduous shrubs (Salix pulchra, Salix glauca, Salix alaxensis, Salix richardsonii, and Betula nana).
Songbird habitat mapping and predictions
Pearson et al.'s (2013) ecological niche model estimates present and future (year 2050) pan-arctic tundra vegetation types as defined by the dominant vegetation physiognomy used in the Circumpolar Arctic Vegetation Map (CAVM) (Walker et al., 2005) and two tree cover classes from the Global Land Cover 2000 database (GLC2000). For this study, we used the output from a subset of two of the possible scenarios modeled; details on all modeling approaches and bioclimatic, substrate, tree dispersal into tundra, and climate variables used can be found in Pearson et al. (2013). Briefly, for 2050, we took advantage of the range of scenarios that Pearson et al. (2013) generated using the Random Forests (RF) model, including modeled vegetation distribution assuming both restricted tree dispersal (5 km) and unrestricted colonization of trees (equilibrium dispersal) from current latitudinal treeline. We focus here on predictions of the Hadley Centre coupled atmosphere-ocean general circulation model (HadCM3) under the A2a emissions scenario. We chose to utilize these specific combinations because they were those selected by Pearson et al. (2013) as representative of the various scenarios they developed, and also because they represent outcomes that are most consistent with observations of atmospheric carbon dioxide (CO2) concentrations (Le Quéré et al., 2014).
The Pearson et al. (2013) maps were limited spatially to include only the arctic foothills region of Alaska where our fieldwork was conducted. White-crowned sparrow habitat – characterized by a significant deciduous shrub component – was defined by present day and future vegetation modeled as Erect dwarf shrub (CAVM vegetation type S1), Low-shrub tundra (CAVM vegetation type S2), and Tree cover mosaic (GLC2000 vegetation class T1). Although we did not conduct field work in T1 vegetation because it is not currently present in our study region, we include it as suitable for White-crowned sparrow nesting because previous work has shown this species nests under shrubs within forested areas of the sub-Arctic (Norment, 1993). Dominance of Lapland longspur habitat was defined as locations modeled as Tussock-sedge, dwarf-shrub, moss tundra (CAVM vegetation type G4), and nontussock-sedge, dwarf-shrub, moss tundra (CAVM vegetation class G3) in which dwarf woody shrubs are present but not abundant. Although we did not conduct field work in G3 vegetation because it is not currently present in our study region, we include it as suitable for Lapland longspur nesting because previous work has shown this species nests in nontussock forming sedge tundra where it is common (Rodrigues, 1994).
Statistical analysis
Data reflecting nesting habitat were analyzed separately for both bird species using chi-squared analysis to test the null hypothesis that vegetation height or plant growth form did not affect nesting location. Arthropod data (collected in study plots) were analyzed separately for number of arthropods caught and biomass as a one-way anova with vegetation type as the main effect and site as a blocking factor. The canopy-dwelling and ground-dwelling arthropods were analyzed separately to reflect the different sampling design (see methods above). Berries and inflorescences (counted in study plots) were analyzed separately as a one-way anova with vegetation type as the main effect after summing the number of each category at the transect level. Statistical analyses were conducted with sas (Version 9.3, SAS Institute, Cary, NC, USA) with the exception of arthropod data, which were analyzed with R (R version 3.0.2, R Development Core Team).
Results
Nest habitat characteristics
Lapland longspurs and White-crowned sparrows breeding in the vicinity of Toolik Lake Field Station established nests in habitat that significantly differed in vegetation height (Χ2 = 79.8, df = 5, P < 0.0001) (Fig. 2a) and plant growth form composition (Χ2 = 89.6, df = 2, P < 0.0001) (Fig. 2b). Lapland longspurs almost exclusively selected short vegetation (<20 cm tall), with nearly all (>95%) of those nests located in moss and/or graminoid-dominated communities (Fig. 2a and b). The majority (80%) of these were placed directly in the sides of tussocks formed by the sedge Eriophorum vaginatum (cottongrass), with the remaining 20% of nests placed either amongst other vascular species, (10%, including Carex bigelowii) or directly in moss mats (10%) (Fig. 3a). In contrast, White-crowned sparrow nests were situated almost exclusively at the base of deciduous shrubs (Fig. 2b) that were 20 cm but less than 100-cm tall (Fig. 2a). Equal numbers of White-crowned sparrow nests were placed directly under Betula nana, Salix pulchra, Salix glauca, and Salix alaxensis (Fig. 3b), the tall shrub species that are most common in our study area. Both bird species were therefore targeting particular microsites within these communities (see Table S1 for growth form composition in shrub and open tundra at each site for comparison).
Food availability
Arthropod biomass and abundance
Canopy-dwelling arthropod biomass and abundance were significantly greater in Shrub plots compared with Open plots (F1,75 = 62.9 and 141.1, respectively, both P < 0.0001) (Fig. 4a and b). In contrast, ground-dwelling arthropod biomass was significantly greater in Open compared with Shrub plots (F1,155 = 11.5, P = 0.0009), while abundance did not differ between vegetation types (P > 0.05, Fig. 4c and d).
Berries and inflorescence abundance
There were no significant differences in either berry or inflorescence abundance in Shrub relative to Open plots (P > 0.05). Berry abundance across the four sites averaged 308 m−2 (±164 SEM) in Open and 380 m−2 (±174 SEM) in Shrub plots, while inflorescence abundance averaged 847 m−2 (±347 SEM) in Open and 677 m−2 (±373 SEM) in Shrub plots.
Future songbird habitat availability
We found that relative to the present day (Fig. 5a), White-crowned sparrow habitat extent is predicted to increase by 22% under the 5 km (Fig. 5b) and by 58% under the unlimited seed dispersal (Fig. 5c) scenarios. Lapland longspur habitat is predicted to be reduced between 18% under the 5-km dispersal scenario (Fig. 5b) and 56% under the unlimited dispersal scenario (Fig. 5c).
Discussion
Vegetation height plays a key role in nesting habitat selection
Our findings strongly suggest that increasing shrub dominance in northern Alaska will enhance White-crowned sparrow habitat quality and extent and diminish that of Lapland longspurs, but also reveal more nuanced preferences for nest site characteristics, particularly for White-crowned sparrows. Our 2 years of nest site data suggest that White-crowned sparrows prefer to nest under shrubs of moderate height but not under shrubs taller than 1 m. However, due to the inherent difficulty in locating nests under large compared to smaller shrubs, we may have slightly underestimated the use of shrubs taller than 1 m as nesting habitat by White-crowned sparrows. As shrubs become more abundant and expand their range, taller, denser shrubs may actually constrain habitat availability for this species. Although it is possible that arctic White-crowned sparrows breeding north of the Brooks Range may adapt by nesting under taller shrubs more than they do at present, this seems unlikely to occur as even populations breeding in forested ecosystems do not nest directly in or under trees, instead preferring the shelter of moderate height shrubs (Norment, 1993; Chilton et al., 1995). White-crowned sparrows do utilize tall shrubs as perches for singing, territory defense and predator lookouts, suggesting that even though tall shrubs may be unsuitable for nesting under, they likely serve an important secondary role in nest site selection (J.C. Wingfield, personal observation). In our study, we documented that White-crowned sparrow nests were only found on the edges of dense shrub thickets (where tall shrubs formed a near-complete canopy; also see Chilton et al., 1995; Norment, 1993), and rarely, if ever, located within them, suggesting that documented increases in shrub density by existing shrubs (Myers-Smith et al., 2011; Tape et al., 2012) may actually reduce White-crowned sparrow nesting habitat (J.C. Wingfield, personal observation).
In contrast to the strong influence of structural characteristics on nest site selection, we found that White-crowned sparrows show no preference for nesting among the four most common tall birch and willow species present in the arctic foothills region of Alaska. This provides new insight into current and future White-crowned sparrow habitat availability given there is significant heterogeneity in the dominant deciduous shrub species among various locations in northern Alaska (Sturm et al., 2001; Tape et al., 2006, 2010). Few studies have explored the relative rates at which different species of deciduous shrubs become more dominant as shrubification of tundra regions continues, but there is some evidence of variability in growth rates among species as a result of both direct and indirect effects of warming (Bret-Harte et al., 2002). Although alder is much less common in our study region than birch or willow, White-crowned sparrows do nest in alder communities in central Alaska (Blanchard Oakeson & Erickson, 1949). Our findings suggest that nesting habitat availability for White-crowned sparrows may be unrelated to variation in which species of shrub is most dominant. Instead, White-crowned sparrow nesting habitat will be largely determined by changes in shrub height and their distribution across the landscape relative to other plant growth forms.
In sharp contrast to White-crowned sparrows, Lapland longspurs nest exclusively in graminoid-dominated communities where deciduous shrubs are sparsely distributed and do not exceed 20 cm in height. Further, nesting Lapland longspurs exhibit very strong plant species specificity, placing their nests primarily into the sides of tussocks that are created by a single sedge species, E. vaginatum. This species specificity is presumably associated with the biophysical structure of the tussock, which likely offers protection from predators, overheating caused by direct solar radiation, and insulation from low temperatures and inclement weather (Williamson, 1968; Boal & Andersen, 2005). By building on what is already known regarding Lapland longspur nest site preferences (Hunt et al., 1995; Hussell & Montgomerie, 2002), our findings provide confirmation that preferred nesting habitat of Lapland longspurs in the arctic foothills region of Alaska is likely to decline as deciduous shrub dominance increases at the expense of graminoid and moss species in low arctic, wet tundra types (e.g. Cornelissen et al., 2001; Wahren et al., 2005; Walker et al., 2006; Elmendorf et al., 2012).
Food resources differ between graminoid- and shrub-dominated communities
In addition to vegetation compositional and structural characteristics that determine the availability and location of songbird nesting habitat, food resource availability for both adults and young is of critical importance to their reproductive success (Lack, 1968; Visser et al., 1998). Our quantitative measurements of food resource availability show that canopy-dwelling arthropod biomass (i.e. flies and web-building spiders) was higher in shrub tundra relative to open tundra; however, ground-dwelling arthropod biomass (i.e. ground spiders and beetles) was lower in shrub tundra likely due to the greater number of large spiders and beetles active in more open habitat (Rich et al., 2013).
Although White-crowned sparrows and Lapland longspurs have highly specialized and distinct nesting requirements, like most passerines they are opportunistic feeders whose summer diet includes plant-based resources and arthropods (Custer & Pitelka, 1977; Seastedt & Maclean, 1979; Chilton et al., 1995; Hussell & Montgomerie, 2002). The diet of arctic breeding populations of both species is determined by seasonal and geographic variation in availability of these food sources rather than food selection in northern Alaska, Canada, and in Greenland (Chilton et al., 1995; Hussell & Montgomerie, 2002). Stomach content and observational data have both shown that during the nesting period – when arthropod availability is high – adult diet is comprised primarily of arthropods and plant matter (mainly seeds, but also leaf material), while nestlings are fed arthropods almost exclusively (Seastedt & Maclean, 1979; Chilton et al., 1995; Hussell & Montgomerie, 2002). Arthropod consumption by both songbird species is dominated by crane flies (Tipula carinifrons), sawflies (Tenthredinidae), midges (Chironomidae), and muscoid flies (Muscidae), but also includes beetles (Coleoptera), spiders (Araneida), caterpillars (Lepidoptera), true bugs (Hemiptera), and various other insects (Poole, 2005). Analysis of our own limited collection of fecal samples from adult and nestling of both species confirms this diet (unpublished data). Our own observations confirm those in older literature that suggest arctic breeding Lapland longspurs forage primarily in short vegetation and remain on the ground, picking insects and spiders off low lying tundra plants or snatching them from the air (Drury, 1961; Williamson, 1968; Watson, 1957), while White-crowned sparrows have a much broader foraging niche that also includes picking from taller shrub canopies (Morton, 1967; Hussell & Montgomerie, 2002; J.C. Wingfield, personal observation). Further, during the nesting period, both species limit the spatial extent of their foraging to be closer to their nests such that Lapland longspurs foraged primary in nearby open tundra, while White-crowned sparrows forage mainly in nearby shrub dominated tundra (Blanchard Oakeson & Erickson, 1949; Chilton et al., 1995, J.C. Wingfield, personal observation). Due to these differences in foraging habitat, future changes in arthropod availability are likely to affect these two species differently.
We hypothesize that arthropod community composition will likely shift towards more abundant small flies and spiders, and fewer large spiders and beetles as shrubs become more dominant. This prediction conservatively assumes that arthropod niches will remain similar as climate continues to warm, although arthropod populations may respond in somewhat unpredictable ways. Given that both White-crowned sparrows and Lapland longspurs consume primarily flies during the nesting period, their preferred prey availability may be enhanced. This supports the hypothesis by Ims & Henden (2012) that willow thickets harbor an abundant and rich community of arthropods that may benefit all songbird species, although the longspurs will only benefit if they expand their foraging habitat during nesting to include taller, shrub dominated communities. Currently, longspurs mostly do not venture into shrub habitats to forage so such a behavioral change seems unlikely. It is also important to consider that temporal changes in availability and quality of this critical, high quality food source may have more of an impact on songbird body condition and reproductive success (Durant et al., 2007; Both et al., 2009; Bolduc et al., 2013) than the spatial differences explored here, as the timing of arthropod emergence is predicted to become mismatched with the timing of songbird nesting in the Arctic (Tulp & Schekkerman, 2008; Both et al., 2009).
Greater shrub dominance will have contrasting effects on future habitat extent
Using the results presented above with the vegetation model, we predict a general reduction in Lapland longspur and concurrent expansion in White-crowned sparrow nest habitat availability as tundra vegetation shifts from graminoid dominated to erect dwarf- and low-shrub tundra, or tree cover mosaic (Pearson et al., 2013). Our findings support Sokolov et al.'s (2012) expectation that shrub specialist songbirds in Eurasian arctic tundra will become more abundant as tundra shrubs become increasingly dominant, largely due to boreal species expanding northward into tundra at the expense of arctic tundra specialists. Similarly, our model predictions at least partially support findings by Henden et al. (2013) that suggest that the abundance of shrub nesting songbird species will increase while open nesting songbirds will be unaffected. Our predictions are both complementary to these studies and novel. While previous research surveyed the abundance of songbirds in relation to vegetation community characteristics, we focused explicitly on nest site selection to allow us to understand a crucial factor in reproductive success. In addition, this study was conducted in northern Alaska, which complements previous studies focused on Eurasian arctic tundra, thereby contributing towards a pan-Arctic perspective on changing songbird habitat.
While White-crowned sparrows are likely to benefit from increasing shrub dominance in the Arctic, unless Lapland longspurs are able to adapt to the ongoing compositional and structural changes in large portions of their current nesting habitat, a decline in their breeding populations is a likely outcome and may be further amplified by ongoing changes in their wintering habitats (Newton, 2004; Holmes, 2007). Further, unlike more southern species, Lapland longspurs have very limited opportunity to disperse northward as the tundra gives way only to the polar marine environment. Alterations of nesting habitat will determine future reproductive success and population viability in the long term, although changes to these species' wintering grounds can also affect their populations (Norris et al., 2004).
This study focuses on Lapland longspurs and White-crowned sparrows, but our first approximations of future spatial extent of both habitat types are relevant to habitat change for species with similar nesting requirements. For example, American tree sparrows and American robins are obligate shrub nesters, while Smith's longspurs nest in graminoid dominated tundra communities (Naugler, 2014; Hunt et al., 1995; Sallabanks & James, 1999). However, given our field-based findings suggest nuanced preferences for specific nest site characteristics of White-crowned sparrows (e.g. moderate height shrubs), accurate species level predictions require closer examination of species-specific nest site preferences and more refined modeling tools than currently exist.
Current vegetation model constraints limit habitat predictions
Relative to other ecosystems, there are very few predictive studies of the spatial distribution of vegetated habitat for tundra wildlife (Pearce et al., 2012; Hope et al., 2013; Gustine et al., 2014; Stralberg et al., 2014). This is despite overwhelming evidence from existing literature that tundra vegetation communities are changing significantly (ACIA, 2004), and that such changes affect tundra wildlife by altering food and shelter availability (Klein et al., 2005; Post & Forchhammer, 2008; Joly et al., 2009; Willerslev et al., 2014). We suspect that this gap exists largely because, unlike forested habitats where gap dynamics models are widely available and mature (as reviewed by Bugmann, 2001; and Busing & Daniel, 2004), existing tundra modeling tools cannot yet simulate the fine-scale habitat characteristics that are critical to many wildlife species. For example, willows, but not dwarf birch nor alder, are the strongly preferred forage species for caribou, moose, ptarmigan, and other wildlife species (Viereck & Little, 2007; den Herder et al., 2008; Tape et al., 2010), but current models do not resolve species dominance in deciduous shrub communities. In addition, occurrence of communities dominated by woody shrubs is spatially heterogeneous and dependent on the fine-scale topography of the landscape. Similarly, while tundra breeding caribou preferentially forage on willow shrubs during summer months, they tend to avoid tall, dense shrub canopies that are attractive to both moose and wolves (Briand et al., 2009). These habitat preferences highlight the need for habitat models to include high spatial resolution predictions of the physical structure of tundra vegetation.
As one example, Pearson et al.'s (2013) ecological niche modeling of tundra vegetation communities (i.e. CAVM classes) was spatially explicit at 4.5 km (~20 km2) resolution, and had broad spatial coverage (i.e. the pan-Arctic domain). Despite being the most advanced predictions of future tundra vegetation communities to date, Pearson et al.'s (2013) effort has two major limitations that make it a first approximation of the potential extent and spatial pattern of future habitat change in northern Alaska. First, the fine-scale spatial heterogeneity of vegetation types in the arctic foothills region is not captured at 4.5 km spatial resolution. This limitation in spatial scale of the modeling was determined by the baseline CAVM and GLC2000 reference maps used in the predictions. Second, the broad characterization of vegetation communities by the CAVM vegetation classes did not provide information on the more nuanced vegetation community characteristics that are often critical to habitat quality and species preferences. Thus, the predictions made by Pearson et al. (2013) enable only a coarse estimation of the future of the contrasting habitat types used by nesting White-crown sparrows vs. Lapland longspurs. Not only does the spatial resolution mask the fine-scale patchy nature of the vegetation communities important to the birds but also physical properties of the vegetation are unaccounted for. For example, the presence and size of shrub thickets cannot be considered because shrub density and height is not known (although can be roughly approximated by the CAVM classes with height and structure definitions). For example, the CAVM G4 vegetation class includes dwarf shrubs up to 40-cm tall, whereas our field observations of nest site height (see 3) indicate Lapland longspurs nest exclusively in tussock tundra of less than 20-cm height while White-crowned sparrows nest in shrub tundra greater than 20-cm tall. This mismatch between model resolution and the scale of habitat characteristics important to songbirds allows for somewhat limited predictions of future habitat extents used by shrub or graminoid nesting songbirds, and this limitation likely extends to other wildlife species.
Refining habitat predictions for habitat specialists would require tundra vegetation models (i) be based on fine spatial resolution (<50 m2) and highly resolved vegetation community characteristics (i.e. species level composition and biomass, vertical canopy structure, and height); (ii) include complex interactions between the many factors controlling tundra plant growth through dynamic vegetation modeling; and (iii) be applicable at regional to biome level spatial scales. It is unlikely that these criteria will be met near-term, even though vegetation models continue to improve based on a growing number of studies across the arctic tundra biome that are enhancing the mechanistic understanding of what controls vegetation change. Nevertheless, there may be ways to develop phenomenological relationships to down-scale the relatively coarse predictions of these process-based models to predict habitat abundance.
To refine these down-scaling tools for accurate future habitat prediction over large spatial scales, a backdrop of high spatial and vegetation type resolution data on key tundra vegetation characteristics will be critical. Recent and upcoming advances in remote sensing may in part facilitate this effort. For example, a growing suite of satellite-based sensors currently quantifies tundra vegetation cover, community type, and productivity at fine spatial scales (<1 m2) (i.e. WorldView, QuickBird, GeoEye, IKONOS). Further, there is mounting evidence from studies of more southern ecosystems that air and space-borne hyper-spectral sensors (e.g. Hyperion, AVIRIS, EnMap) can be used to quantify detailed vegetation community attributes such as species composition (Asner et al., 2008), species diversity (Oldeland et al., 2010), and forage quality (Skidmore et al., 2010). Targeted assessments based on field surveys of vegetation characteristics will continue to be essential to develop robust and representative empirical relationships so that such imagery can be used to extend field measurements over large areas via field-calibrated remote sensing observations (Turner et al., 2003). Additionally, airborne LiDAR imagery shows great promise, as in forested ecosystems, that will provide detailed information on the 3-D physical structure of tundra ecosystems (e.g. height and vertical complexity) (Vierling et al., 2013; H. Greaves, L.A. Vierling, J.U.H. Eitel, T.S. Magney, N.T. Boelman, C. Prager and K.L. Griffin, in review), and thus spatial coverage of these data is likely to greatly increase in the region over the next decade.
The tundra biome modeling community should keep the need for fine-scale habitat prediction in mind as they continue to develop models for the broader community to explore downscaling strategies used in predictions of habitat abundance from the general characteristics of vegetation classifications. We note the upcoming NASA-led multi-year Arctic Boreal Vulnerability Experiment (ABoVE) (http://above.nasa.gov) prioritizes the use of geospatial models of current and predicted vegetation and wildlife population dynamics at landscape-to-regional scales by linking models of vegetation dynamics, hydrology, permafrost, and disturbance.
Our findings, integrated across the arctic foothills region of Alaska, indicate that: (i) Lapland longspurs and White-crowned sparrows select distinct and nonoverlapping nest sites, with Lapland longspurs nesting only in E. vaginatum-dominated tussock tundra where shrub height does not exceed 20 cm, and White-crowned sparrows nesting primarily under moderate height shrubs ranging from 20 cm to 1 m in height, regardless of shrub species; (ii) increasing shrub dominance in northern Alaska may be accompanied by a reduction in total arthropod biomass, but since flies are the primary food source for both nesting Lapland longspurs and White-crowned sparrows and their nestlings, and they are more abundant in shrubs, the availability of their preferred food source may be enhanced; (iii) by 2050, we predict the expansion of shrub-dominated communities and tree cover mosaic will result in a ~20–60% increase in White-crowned sparrow nesting habitat extent, and a decline of ~20–60% in Lapland longspur nesting habitat, suggesting that habitat for songbird species with similar nesting preferences will change in a similar manner; (iv) there is a mismatch between the resolution of existing vegetation models and the scale of habitat characteristics that are important to songbirds which enables only first approximations of their future habitat extents, and this limitation likely extends to other wildlife. Our study thus complements and broadens current understanding of how changing tundra vegetation cover will impact arctic breeding birds since the majority of research in this field has focused on migratory waterbirds whose arctic breeding habitat extent is predicted to shrink by 25–85% over the current century (Zöckler, 1998; Zöckler & Lysenko, 2000; Tomkovich et al., 2002).
Acknowledgements
We thank Shae Bowman, Kathryn Daly, Adam Formica, Jessica Gersony, Kathleen Hunt, Michaela McGuigan, Simone Meddle, Lisa Quach, and Jake Schas for field assistance. We thank Ed Rastetter and Howie Epstein for helpful discussions regarding current tundra vegetation model constraints. We thank Toolik Field Station (Institute of Arctic Biology, University of Alaska Fairbanks) and CH2M HILL for providing support and logistics. This project has been funded by a collaborative NSF grant from the Office of Polar Programs (ARC 0908444 to N. Boelman, ARC 0908602 to L. Gough, and ARC 0909133 to J. Wingfield), and another to S. Goetz (ARC 0732954).
