Arthropod prey and diets of woodland migrants are similar between natural riparian woodlands and anthropogenic woodlots in the northern prairie region

,


INTRODUCTION
Migration is an energetically expensive period in the annual cycle and woodland migrants typically cannot carry enough energy stores to complete migration in a single flight between breeding and wintering grounds. Migrants thus need to replenish fuel stores at stopover habitats along the migratory route to support subsequent flights (Hussel 1969, Blem 1980, Wang and Moore 1997. Stopover habitat quality, especially the availability of energetic resources, is a crucial factor influencing the population dynamics of migrants (Moore and Wang 1991, Newton 2006, Schmaljohann and Eikenaar 2017. Failure to find suitable stopover habitats may lead to lower fat deposition rates, and, thereby, lengthen the stopover duration and delay arrival on the breeding or wintering grounds (Francis and Cooke 1986, Moore and Wang 1991, Schmaljohann and Eikenaar 2017. Thus, stopover habitat quality affects the refueling performance of migrants, which, in turn, impacts migration speed and timing (Piersma 1987, Pierce and McWilliams 2004, Moore et al. 2017. Migratory timing influences nesting success and survival rates and can thereby impact migratory bird population trends (Newton 2006, Emmenegger et al. 2014, Janiszewski et al. 2014. Consequently, identification of high-quality stopover habitat along the migratory route is important for conservation of woodland migrant birds (Donovan et al. 2002, Mehlman et al. 2005, Newton 2006, Faaborg et al. 2010).
However, stopover habitat quality can be difficult to assess (Donovan et al. 2002, Melhman et al. 2005, Berlanga et al. 2010. Fattening rates derived from plasma metabolite profiling can serve as an integrated measure of refueling success of migrant birds at stopover sites, and by extension, habitat quality at stopover sites along the migratory route Guglielmo 2010, Smith et al. 2015). Cues for habitat selection in migrant birds, however, may not always involve food abundance that would be expected to result in high refueling rates. For example, food abundance appeared to the be the most important cue for stopover habitat selection for frugivorous migrants, but insectivorous migrants appeared to use vegetative cues as the primary cue for stopover site selection (Wolfe et al. 2014). Habitat size, structure, vegetative diversity, thermal microclimates, competition intensities, predation pressures, and food resources may synergistically influence stopover habitat quality and affect prey choices, stopover behaviors, and physiological capacities of woodland migrants (Olson and Grubb 2007, Olson et al. 2010, Liu and Swanson 2014a. Indeed, Cerasale and Guglielmo (2010) found that ecological (stopover behavior and food resources) and physiological (fattening rates and stress physiology) measures are not always in accord with respect to stopover habitat quality, because habitats with higher arthropod biomass and insectivorous bird densities produced lower fattening rates for some insectivorous bird species. High bird density at a stopover site can suppress the arthropod prey base (Moore and Wang 1991) and fattening rates (Kelly et al. 2002, Cerasale andGuglielmo 2010) for migrants. Competitor and predator densities may also influence migrant behavior (Cimprich et al. 2005, Smith and Hatch 2008, Carlisle et al. 2009, Hope et al. 2014) and fattening rates (Lennox et al. 2016) at stopover sites, thus contributing to overall stopover habitat quality. Finally, stopover sites perform other functions for migrants than just refueling, including resting, waiting, recovering, sleeping, information collection, and social interaction (Linscott and Senner 2021), so fattening rates alone may not provide a completely accurate picture of intrinsic stopover habitat quality. Consequently, assessments of habitat quality using both ecological and physiological metrics are required to fully evaluate the factors and their interactions affecting habitat quality for woodland migrants.
Directly measuring habitat attributes, such as food resources, that influence stopover site suitability is a basic approach to assess habitat quality (Johnson 2007). Suitable stopover habitats must provide sufficient food resources of the right types as well as sufficient access to food for migrants (Martin andFinch 1995, Buler et al. 2007). Most migrant landbirds are primarily insectivorous, but frugivorous and granivorous migrants may expand diets to consume invertebrates to help meet the energetic requirements for migration and to enhance lipid and protein intake (Rappole 1995, Wolfe 2009, Pageau et al. 2020. Thus, documenting arthropod prey abundance and correlating arthropod abundance with migrant energetic condition are necessary to evaluate suitability of stopover habitats.
In the northern prairie region, woodland habitats were historically concentrated along river corridors, such as the Missouri River (Castonguay 1982, van Bruggen 1996. Since Euro-American settlement, these natural riparian corridor woodlands (hereafter, corridors) have been greatly reduced and degraded (Johnson 1992, Hesse 1996, Dixon et al. 2012), but these losses may potentially be mitigated to some extent by the appearance of anthropogenic woodlands (hereafter, woodlots) on the landscape. Both woodland types provide stopover habitat for woodland birds during spring and fall migrations and migrants occupy both corridor and woodlot habitats in this region during stopover in roughly similar numbers (Martin 1980, Swanson et al. 2003.
The Missouri River riparian corridor (the focus of this study) contains greater contiguous woodland areas, with higher vegetative diversity, than woodlots (Dean 1999, Swanson et al. 2003, Gentry et al. 2006). In addition, natural riparian woodlands are adjacent to rivers or streams, which may provide subsidies of aquatic arthropods, such as Chironomidae, Trichoptera, Ephemeroptera, and Plecoptera to terrestrial ecosystems (Butakka et al. 2014, Hamid and Rawi 2014, Muehlbauer et al. 2014, Wesner et al. 2020. Because aquatic arthropods differ from terrestrial arthropods in fatty acid composition of body tissues (Twining et al. 2018(Twining et al. , 2019(Twining et al. , 2021a and dietary fatty acids can influence performance (Pierce and McWilliams 2004, Twining et al. 2016, del Rio and McWilliams 2016, such aquatic subsidies could influence migration and stopover biology in birds. Thus, these differences between corridors and woodlots might result in between-habitat variations of arthropod abundance and diversity. If differences in arthropod abundance and community composition exist between the two habitats, these differences may impact stopover habitat suitability and the distribution of woodland migrants between habitats, because migrants appear capable of responding to differences in food resources (Graber and Graber 1983, Wang et al. 1998, Rodewald and Brittingham 2002, Tietz and Johnson 2007. In addition, delineation of differences in arthropod abundance/biomass and stopover habitat suitability could help focus conservation efforts on preserving the highest-quality woodland stopover sites. Previous studies suggest not only similar occurrence of migrant birds in the Missouri River riparian corridor and nearby woodlot habitats, but also similar functional habitat quality. For example, Liu and Swanson (2014a) compared plasma metabolite levels of landbird migrants at both corridors and woodlots, including both plasma triglyceride (TRIG; indicator of fat deposition rates) and β-hydroxybutyrate (BUTY; indicator of fat catabolism), which have been routinely used to measure fat deposition rates of birds during migration. Liu and Swanson (2014a) found that both plasma TRIG and BUTY levels suggested similar fat deposition rates of woodland migrant birds between corridors and woodlots. Between-habitat comparisons of stress physiology of landbird migrants to examine a potential effect of habitat quality on stress suggested similar plasma corticosterone (the main stress hormone in birds) levels in woodland migrant birds at corridors and woodlots (Liu and Swanson 2014b). To quantify woodland bird stopover behavior, such as stopover duration, movement rates, and temporary home range, Liu and Swanson (2015) conducted an experimental relocation study for Yellow-rumped Warblers (Setophaga coronata) during fall migration using radio-telemetry. Liu and Swanson (2015) found that Yellow-rumped Warblers showed similar stopover duration, movement rates, and temporary home range between corridors and woodlots.
Given similar fattening rates (Liu and Swanson 2014a), roughly similar bird abundances and community composition (Swanson et al. 2005), space use and stopover durations (Liu and Swanson 2015) in Missouri River riparian corridors and nearby farmstead woodlots, we might expect birds are foraging on similar prey in these two forest types, despite possible differences in insect community composition. Alternatively, birds maybe foraging on different prey species that are similar in nutritional value in terms of energetics and fatty acid content (Swanson et al. 2005, Wesner et al. 2020. The latter hypothesis predicts that similar fattening rates in the two habitats result from factors other than arthropod prey abundance or biomass, such as different arthropod communities with different nutritional value, different prey preferences, or differences in abilities of migrant birds to find arthropod prey between the two habitats. We address these hypotheses in the present study by assessing arthropod diversity and abundance at corridors and woodlots, examining dietary patterns of migrant landbirds, comparing dietary vs. habitat arthropod abundance/biomass and communities, and exploring trends between prey abundance and previous data (Liu and Swanson 2014a) on migrant energetic condition at the same study sites and during the same years. To do so, we sampled arthropods and collected fecal samples from woodland migrants at both habitats during spring and fall migrations. The comparisons of arthropod prey resources and diet composition of woodland migrants in this study complement previous studies in this study system (Dean 1999, Swanson et al. 2003, Liu and Swanson 2014a, 2014b to provide a comprehensive assessment of stopover habitat quality at natural riparian corridor and anthropogenic woodlot habitats in the woodland-restricted northern prairie region.

Study site
We conducted this study in southeastern South Dakota, where woodlands cover about 4% of the total land area (Castonguay 1982). The riparian corridor of the Missouri River is the most extensive of these woodland habitats. Missouri River riparian corridor woodlands are mostly dominated by cottonwoods (Populus deltoides), with some later successional tree species, such as green ash (Fraxinus pennsylvanica), elm (Ulmus spp.), and hackberry (Celtis occidentalis) also present (Hesse 1996, Swanson et al. 2005, Dixon et al. 2012. Woodlots are smaller (0.7-3.5 ha; Swanson et al. 2005), more isolated, and less vegetatively diverse than corridor woodlands (10.1-42.3 ha in the present study) and cover about 1% of the total land area in southeastern South Dakota (Castonguay 1982). Woodlot study sites are comprised primarily of elm, mulberry (Morus alba), box elder (Acer negundo), hackberry, and green ash (Swanson et al. 2003, Gentry et al. 2006. Both habitat types are embedded in a landscape matrix primarily of row-crop agriculture (Newberry and Swanson 2018). Corridors retain more forest area within the landscape than woodlots, but because corridor woodlands are narrow and rarely extend beyond 0.5 km from the Missouri River, the landscape surrounding both habitats is still primarily row-crop agriculture (Gentry et al. 2006). We sampled three study sites each from both corridor (all along the Missouri River) and woodlot habitats (Fig. 1) in Clay County, South Dakota, from among those sampled by Dean (1999), Swanson et al. (2003Swanson et al. ( , 2005, Gentry et al. (2006), and Liu andSwanson (2014a, 2014b).

Food abundance and bird diet
We collected arthropods and bird fecal samples from corridors and woodlots during spring (20 April to 2 June 2010-2012) and fall (20 August to 30 October 2010-2011) migratory seasons. We sampled each of the six study sites once per week during both seasons, rotating through the sites, with woodlot and corridor sites sampled on alternating days. We collected all samples before 12:00 CDT and only on days without rain and with winds < 30 km/h.

Arthropod sampling
We simultaneously applied branch-clipping, sticky-trap, and pitfall-trap methods to collect arthropods. Arthropod sampling locations were randomly chosen within a 200-m radius surrounding the locations where we set up mist nets to catch migrants concurrently. The branch-clip method is an effective method for investigating prey abundance for birds that feed on arthropods in foliage (Kuenzi and Moore 1991, Johnson 2000, Carlisle et al. 2012. Each day prior to branch-clipping, we generated three series of random numbers to determine which tree branches to collect in the field. One series of numbers (between 0 and 200 m) was used to determine the distance from the mist net. The second series of numbers (between 0 and 360°, with 0° as north) was used to determine the direction of sample collection from the mist net. If branch-clipping locations determined based on these two random numbers were outside of woodland habitat, we added 90° to the random direction until we arrived at a location within the woodland habitat. We collected a branch-clip sample from the tree nearest to the random location. The last series of numbers (from 0 to 10 m) was used to determine the height from which branches were collected. Branch-clip sampling was conducted from 10:00 to 11:30 CDT. A total of ten branch-clip samples were collected from two locations (five samples from each location) during a single sampling period at each study site. One branch-clip sample was collected at the randomly selected central location and the other four were collected 10 m from the selected center location in each of the four cardinal directions (Carlisle et al. 2012). To obtain a branchclip sample, we placed a cloth bag over a terminal branch (samples were approximately 1 m in length) at a randomly selected height (measured with a drawstring marked at 0.5 m intervals) ranging from ground level to 10 meters. The bag was quickly sealed with the drawstring to prevent arthropods from escaping before clipping the branch into the bag (Johnson 2000). After insecticide treatment, the bag was shaken to release arthropods from the branch before removing the branch and inspecting it for remaining arthropods. The bags were brought back to the lab and arthropods were placed into 75% ethanol in collection vials.  Sticky traps were used to collect flying insects because branchclipping may not effectively capture these insects (Kuenzi and Moore 1991). Sticky traps were prepared from 30.5 x 30.5 x 0.6 cm plywood strips. We spread non-toxic, non-drying sticky attractant (i.e., Tanglefoot, Contech Enterprises, Victoria, BC, Canada) onto both sides of the plywood strips. We deployed sticky traps at sunrise and collected traps before noon. During a sampling period, five traps were hung on the branches (1-3 m high) of trees at locations extending from the edge to the interior of woodlands across the location where we set up mist net to catch birds concurrently. The direction (between 0 and 360°) and the first location (between 0 and 100 m) for sticky traps were randomly chosen as described previously for branch-clipping. The remaining four sticky traps were hung following the same randomly chosen direction at 20 m intervals. After each sampling period, traps, with insects still attached, were brought back to the lab where insects were identified and placed into collection vials with 75% ethanol. We removed the attractant from traps and spread new attractant for the next sampling period.
Pitfall traps were used to sample ground-dwelling arthropods Whitmore 1990, Carlisle et al. 2012). We did not conduct pitfall-trapping in the spring of 2010 but did so during all other seasons. The direction (between 0 and 360°) and the first location (between 0 and 50 m) for pitfall traps were randomly chosen as described previously for branch-clipping. Following the chosen direction, on the first day of each sampling season, five metal cans (15.8 cm diameter, 17.6 cm depth) were buried flush with the ground at each study site, about 50 m apart. These locations extended from the edge to the interior of woodlands at the location where we concurrently captured birds. A thin layer (1-2 cm depth) of soapy water was poured into the bottom of each can to facilitate trapping of arthropods. Pitfall-trapping was also conducted from sunrise until shortly before noon each day. The soapy water was replaced at the beginning of each pitfalltrapping session. At the end of each sampling session, all arthropods were placed into vials with 75% ethanol.
Collected arthropods in all samples were identified, counted, and body lengths from head to the tip of abdomen were measured to the nearest 0.1 mm using the built-in ruler on the dissecting microscope (Triplehorn and Johnson 2005). Based on published length and dry weight relationships (Rogers et al. 1977, Calver and Wooler 1982, Ganihar 1997, the biomass for each arthropod group was estimated. When no published length-weight relationships were found or the length of the arthropods fell outside of the range of the published length-weight relationship, we measured the biomass directly after 48 hours in a drying oven at 55 °C for Opiliones (daddy longlegs), Isopoda, Diplopoda> (millipedes), Chilopoda (centipedes), Ephemeroptera, Plecoptera, Neuroptera, and Trichoptera (Rogers et al. 1977).

Bird fecal sample collection and prey identification
Metabolic studies suggest that small birds need to fast for only 3-4 hours before they became postabsorptive (e.g., Brenner and Malin 1965). Thus, digestive retention times are short enough that we can be reasonably sure that items detected in the fecal samples were eaten at or near the study sites. Arthropods remaining in bird fecal samples are also strongly correlated to the true diet, with only minor bias against small and soft-bodied arthropods, so fecal samples can be used to non-invasively represent bird diets (Calver and Wooller 1982, Ralph et al. 1985, Carlisle and Holberton 2006. Concurrently with arthropod sampling efforts, we captured migrant birds with 9-m, 36-mm mesh nylon mist nets. Fecal samples were collected by holding birds in wax-paper lined boxes following capture. Fecal samples were then transferred to 75% ethanol-filled collection vials for preservation until later analyses. To identify diet components (e.g., arthropod body parts), fecal samples were teased apart in glass dishes under a binocular dissection microscope (Leica EZ6). The arthropod body parts in fecal samples were sorted, counted, and identified to taxonomic order, suborder, and family, when possible. The lengths of arthropods were estimated by measuring the length of body parts found in fecal samples and then calculating body length based upon a ratio of body part length to body length using reference samples of whole arthropods collected during sampling efforts. The number of arthropods of each prey type was conservatively estimated by body parts counted in fecal sample (e.g., seven legs in a fecal sample were counted as two arthropods). The biomass for each taxon was either directly weighed (see above) or estimated following published length-weight relationships (Rogers et al. 1977, Calver and Wooler 1982, Ralph et al. 1985, Ganihar 1997, Hódar 1997. Fruits in fecal samples were counted from individual seeds to provide information on fruits eaten by migrant birds.

Statistics Arthropod counts and biomass
To examine relationships among arthropod counts and biomass with predictor covariates, we developed a suite of five models for each season and sampling method to determine the best combination of random and fixed effect variables (see Table A1.14 for models). We used combinations of period and year as fixed and random effects and study sites as a random effect (to account for repeated measures) in models. We also ran models with no random effects. Other predictor variables in models included habitat (corridor vs. woodlot), plant type (branch-clip sampling only, see Tables A1.12 and A1.13), average daily temperature, and average daily wind speed. Average daily temperatures and wind speed for Vermillion, Clay County, South Dakota, were downloaded from the Weather Channel (https://weather.com/ weather), which records temperature measurements every hour and wind speed measurements every 2 minutes and averages these measurements over a 24-h day. For models of biomass, we developed generalized linear mixed models (GLMM) and generalized linear models (GLM) for models with and without random effects, respectively. For models of arthropod counts, we used negative binomial distributions to account for potential overdispersion in the data. For these models, we used the "glmer. nb" functions of R-package "lme4" (version 1.1-28; Bates et al. 2015) for GLMM models and the "glm" function of R-package "stats" (version 3.6.2; R Core team 2020) for the GLM model with no random effects, and used maximum likelihood approaches for estimating variable effects for all models. We compared each suite of models using AIC, and selected models with the lowest AIC scores for subsequent analyses. For biomass analyses, we used GLM and GLMM models with a Gaussian distribution, because these data were continuous in nature. We also used GLM (for models with no random effects) and GLMM approaches, using the same distributions and R packages as above, to test for seasonal variation in arthropod abundance and biomass. For these analyses we pooled spring and fall data for each sampling method and used combinations of plant species (branch-clip samples only), period, and year as fixed and random effects and study sites as a random effect (to account for repeated measures) in models. We also ran models with no random effects. We included average temperature and average wind speed as predictor variables in all seasonal models and used AIC to evaluate model fit and selected models with the lowest AIC scores for subsequent analyses. For all models, we considered models with a ΔAIC of < 2 as competitive with the top model.

Arthropod community diversity
We calculated the Gani-Simpson's Index (D) of arthropod diversity from pooled data from all three sampling methods for each sampling day and calculated mean values of D from all daily samples for both corridors and woodlots separately. We then applied a Student's t-test to compare mean arthropod diversity between the two habitats (Jost 2006). We also calculated Morisita's Index Horn 1966) to compare between-habitat similarity of pooled arthropod taxa from all years for both spring and fall seasons. The value of Morisita's Index ranges from 0 (0% overlap) to 1 (100% overlap). If arthropod taxa showed strong betweenhabitat overlap, we used a contingency table test with Chi-square (χ²) to determine if the frequencies of arthropod taxa differed between corridors and woodlots.

Diet composition
We also calculated Morisita's Index to estimate between-habitat similarity of arthropod communities in bird diets and used a contingency table test with Chi-square (χ²) to compare the frequency of arthropod taxa in the diets (pooled across all years) of individual bird species at corridors and woodlots, when we obtained sufficient samples from each habitat (n ≥ 10; Carlisle et al. 2012). If the relative frequency of arthropod taxa in bird diets differed significantly between habitats, we tested for differential consumption of prey for individual bird species at corridors and woodlots separately (that is, by pooling diets across individuals within each habitat). If there was no significant between-habitat difference in frequencies of occurrence, we pooled diets of individuals sampled from both habitats to test prey consumption vs. abundance. For those species where we obtained ≥ 10 fecal samples from both habitats combined, we pooled diets of birds from both habitats to test prey consumption vs. abundance (see Appendix 1, "Diet composition versus prey abundance").
We grouped migrants into foraging guilds (Tables A1.6-A1.11) as described previously (Simberloff and Dayan 1991, Martin and Finch 1995, Liu and Swanson 2014a, 2014b. We pooled arthropods collected from sticky traps to assess prey consumption vs. abundance for the flycatcher guild, from branch clips for the foliage-gleaning guild, and from pitfall traps for the groundforaging guild (Martin and Finch 1995). For each foraging guild, we followed Johnson (1980) to compare the average difference between ranks of presence in the habitat (counts and biomass) and dietary arthropod counts and biomass. This allowed us to test relative consumption of arthropod groups in the diet against expectations derived from presence in the habitat. Although we recognize that our arthropod sampling methods imperfectly measure arthropod prey availability to birds, we still compared arthropods sampled from the habitats with those in bird diets to allow for future hypotheses on which arthropod taxa might be particularly important to migrant birds in these habitats.

Arthropods and plasma metabolites associated with fattening
To investigate the possible effects of arthropods on the energetic condition of migrant birds, we examined correlations between arthropod abundance or biomass with those for plasma triglycerides (TRIG, a measure of fattening rate) and plasma βhydroxybutyrate (BUTY, a measure of fat mobilization). We obtained plasma TRIG and BUTY data from Liu and Swanson (2014a), who measured plasma metabolite concentrations from migrant landbirds captured concurrently with arthropod and fecal sampling from the same study sites. We hypothesized positive correlations between arthropod abundance and plasma triglycerides and negative correlations between plasma βhydroxybutyrate and arthropod metrics (Schaub andJenni 2001, Bairlein 2002). We used GLMM models with TRIG or BUTY as dependent variables (separate models for each metabolite) with separate models for both arthropod counts and biomass and separate models for spring and fall. GLMM models included year as a random effect and arthropod metrics (counts and biomass) for branch-clip, sticky-trap and pitfall-trap sampling methods, average temperature, time since sunrise (min), period (three equal time periods in spring and four in fall, as described above), habitat (corridors vs. woodlots), and foraging guild (as described above), as predictor variables. We employed a Gaussian distribution for these models, because the plasma metabolite data were continuous in nature, and used maximum likelihood approaches for estimating variable effects.
All statistical analyses were performed with R (version 3.6.2; R Core Team 2020). Statistical significance was accepted at P < 0.05. Results are presented as means ± SE.

Arthropod abundance and diversity
We collected a total of 7187 arthropods in spring (2652 at corridors and 4532 at woodlots). All collected arthropods in spring were from 54 and 53 discrete taxa at corridors and woodlots, respectively (Table A1.1). Arthropod diversity combined across all three sampling methods was similar between corridors (D = 0.89 ± 0.10) and woodlots (D = 0.87 ± 0.10) in spring. Morisita's Index (C H = 0.91) suggested strong, consistent similarity of arthropod communities between corridors and woodlots in spring.
We collected 10,466 arthropods in fall (4309 at corridors and 6157 at woodlots). All collected arthropods were from 59 and 62 discrete taxa at corridors and woodlots habitats, respectively (Table A1.2). Ceratopogonidae and Odonata were collected only from woodlots and Berytidae and Plecoptera were collected only from corridors (Table A1.2). Similar to spring, arthropod diversity combined across all three sampling methods was similar between corridors (D = 0.79 ± 0.09) and woodlots (D = 0.87 ± 0.06) in fall. Morisita's Index (C H = 0.90) suggested strong, consistent similarity of arthropod communities between corridors and woodlots in fall.

Diet Composition
We collected 123 and 205 fecal samples from 8 and 10 migrant bird species from various foraging guilds in spring and fall, respectively, to analyze diets (Tables A1.4 (Table  A1.5).
Migrant birds consumed some arthropod taxa at higher or lower levels than expected based upon presence (see Appendix 1 and Tables A1.6-A1.11 for details). Interestingly, comparisons based on arthropod counts or biomass sometimes gave opposing results regarding the relative importance of diet items, although some generalizations are evident. The migrant community as a whole consumed more Formicidae, Lepidoptera, and Carabidae in spring and Auchenorrhyncha, Sternorrhyncha, Carabidae, Lepidoptera, and Nemetocera in fall than expected based on counts or biomass.
Aquatic insects were generally not highly represented in migrant diets at either corridors or woodlots at either season (Tables A1.4 and A1.5). For spring migrants, Chironomidae, Trichoptera, Ephemeroptera, and Plecoptera contributed only 3.4 and 9.7% of total counts of arthropods in migrant diets at corridors and woodlots, respectively. Similarly, aquatic insects were not frequently observed in migrant diets during fall migration, given that Chironomidae, Emphemeroptera, Trichoptera, and Odonata comprised only 9.4 and 8.7% of total arthropod counts at corridors and woodlots, respectively. Thus, aquatic subsidies were apparently not numerically important contributors to migrant diets at our study sites, even at corridor habitats near the Missouri River.  We observed a total of 32 seeds in bird diets, and 68.8% of the seeds were found in birds from corridors. Almost two-thirds (65.6%) of these seeds occurred in the diet of a single species, the Yellow-rumped Warbler (Table A1.5). Of the seeds consumed by Yellow-rumped Warblers, 80.9% occurred in diets at corridors. Poison ivy (Toxicodendron radicans) seeds (n = 9) were observed in the diets of Yellow-rumped Warblers only at corridors, where poison ivy is much more abundant than at woodlots (personal observation). The remaining seeds came from the diets of Warbling Vireos (Vireo gilvus), White-throated Sparrows (Zonotrichia albicollis), and Dark-eyed Juncos (Junco hyemalis; Table A1.5). The 32 seeds included dogwood (Cornus spp.), wild grape (Vitis spp.), poison ivy, and grasses (

Arthropod abundance vs. energetic condition
For spring GLMMs for plasma metabolites, the only significant effector of triglyceride levels was time since sunrise, which was significantly positively associated with triglyceride levels for analyses of arthropod biomass and nearly significantly positively associated with triglyceride levels for analyses of arthropod abundance (Table A1.15). Plasma β-hydroxybutyrate levels in spring were significantly positively correlated with period for migratory periods 2 and 3 and significantly negatively correlated with the flycatcher guild for both arthropod abundance and biomass (Table A1.15). These data signify that plasma βhydroxybutyrate levels increased as the spring migratory season progressed and that flycatchers had lower β-hydroxybutyrate levels than other foraging guilds. In addition, spring plasma βhydroxybutyrate levels were significantly positively associated with arthropod abundance, but not biomass, from sticky-trap captures (Table A1.15), signifying higher plasma βhydroxybutyrate levels with higher arthropod numbers collected by sticky traps.
Plasma triglyceride levels in fall were significantly positively correlated with arthropod abundance and biomass from sticky trap collections (Table A1.15). Fall plasma β-hydroxybutyrate levels were significantly positively correlated with period for period 2 for analyses of both arthropod abundance and biomass (Table A1.15), signifying that plasma β-hydroxybutyrate levels increased as the fall migratory season progressed from late August to mid-September. Time since sunrise was also significantly negatively correlated with plasma β-hydroxybutyrate levels in fall for analyses of arthropod abundance and nearly significantly negatively correlated for analyses of arthropod biomass (Table  A1.15).
Neither branch-clip nor pitfall-trap arthropod abundance or biomass had any significant effect on plasma metabolite levels at either season (Table A1.15). In addition, habitat (corridors vs. woodlots) had no significant effects on plasma metabolite levels in either spring or fall (Table A1.15), suggesting that relationships between plasma metabolites and potential effectors were similar in both habitats.

Arthropod abundance and diversity
Our data suggest that corridors and woodlots provide roughly similar standardized arthropod abundance and diversity for migrant birds during both spring and fall migrations. Some differences in standardized arthropod counts or biomass occurred between corridors and woodlots, including spring branch-clip, spring sticky-trap, fall pitfall-trap counts and fall branch-clip biomass, which were greater in woodlots than in corridors, and fall branch-clip counts, which were greater in corridors than in woodlots. Despite these differences, the majority of comparisons of arthropod abundance and biomass did not identify differences between habitats and differences, when present, were not in consistent directions. Thus, no clear pattern of variation in arthropods between corridors and woodlots relevant to migrant bird food resources was apparent. Furthermore, Simpson's and Morisita's indices also suggest similar arthropod diversity and a strong overlap of arthropod communities between these two habitats, at least at the taxonomic levels of orders, suborders, and dominant families. ). Thus, avenues exist for direct effects of aquatic subsidies in the diets of migratory birds on overall migratory performance, potentially including stopover biology. At our study sites, all Ephemeroptera and Plecoptera collected, as well as over 90% of Trichoptera collected were from corridor study sites. However, over 50% of Chironomidae were collected at woodlots in both spring and fall. Overall arthropod abundance was not elevated at corridor compared to woodlot study sites and aquatic arthropods contributed < 5% in both spring and fall to the total number of arthropods collected at these sites. Thus, aquatic arthropod subsidies do not appear to be a major supplement to terrestrial arthropod communities at riparian corridor sites in this study and seem unlikely to markedly alter migration performance in the study region for most terrestrial riparian birds.

Diet composition
Dietary arthropod data also suggested that corridors and woodlots provide similarly suitable stopover habitats in both spring and fall migrations. Migrant birds fed on a variety of arthropods and all arthropods consumed by individual bird species were observed at both corridors and woodlots. Morisita's Index also indicated substantial overlap in dietary arthropods consumed in both habitats. The dominant dietary arthropods in this study are consistent with Dean (1999), who documented that Coleoptera and Hemiptera were the most frequently observed dietary items in both spring and fall at riparian corridor woodlands in southeastern South Dakota. Hemipterans were also the most important dietary arthropod taxon during fall migration for migrants in shrub and woodland habitats in Idaho (Carlisle et al. 2012 Thus, migrants at our study sites tended to forage on generally similar arthropod taxa compared to migrants at other stopover sites in North and Central America. Similar to Dean (1999), we did not observe a large proportional abundance or consumption of arthropod larvae, such as lepidopteran larvae, at our study sites. This similarity to the results of Dean (1999) occurred despite the different sampling methods in the two studies, as Dean (1999) used emetic sampling, which should be less biased against soft-bodied insects than the fecal sampling that we used in this study. Lepidopteran larvae may form the dominant ingested arthropod prey item by woodland migrants at some other stopover sites where the abundance of caterpillars was high (e.g., Graber and Graber 1983, Moore and Wang 1991, Yard et al. 2004. For example, Graber and Graber (1983) and Moore and Wang (1991) found that lepidopteran larvae contributed over 46% of total arthropod counts. However, Lepidoptera did contribute the largest proportion of biomass of bird diets at our study sites in both spring (22.3%) and fall (21.8%), but this was primarily because of the large size of ingested adult, rather than larval, Lepidoptera.
If aquatic subsidies of arthropod resources are important to migrant birds, we should expect migrants to feed more frequently on aquatic insects at corridors than at woodlots. Chironomidae were eaten by all bird species, but chironomids and other aquatic insects were eaten at low proportions relative to all arthropods in both spring and fall. The low consumption of aquatic insects by most bird species at both habitats and seasons suggests that aquatic arthropods were not important dietary components at either corridor or woodlot study sites for most bird species. Similarly, Dean (1999) found low availability and a relative lack of use of aquatic arthropods, Ephemeroptera, Trichoptera, and Odonata along river corridors in southeastern South Dakota. The minor importance of aquatic insects in bird diets in this study might be due to a rapid reduction in aquatic arthropod abundance with distance from the riverbank (Triplehorn and Johnson 2005, Chan et al. 2007, Muehlbauer et al. 2014, because Missouri River riparian corridors at our study sites were often wider than 0.5 km, or to dietary preferences for terrestrial arthropods. Our data suggesting a minor importance of aquatic insects in migrant diets are consistent with some other studies of migrant diets. For example, Yard et al. (2004) found that aquatic arthropods contributed only 9% of total counts of dietary arthropods along the Colorado River in Arizona. Poulin and Lefebvre (1996) and Green (1981, 1983) found that Ephemeroptera, Trichoptera, and Odonata were available in low numbers and were rarely eaten by migrating landbirds at stopover sites in Panama and western France, respectively. Exceptions to the finding that aquatic insects are not important components of migrant diets are the studies of Dallman and Smith (1995), Smith et al. (2007), and Ewert et al. (2011), who documented that Chironomidae (Diptera, Nematocera) was an important arthropod for landbird migrants at stopover sites along the shore of Lake Huron in early spring when the abundance of Chironomidae was high. A similar dietary use of on an ephemeral abundance of chironomids might be a possible explanation for the higher proportion of Chironomidae in the diets of early spring migrant Yellow-rumped Warblers at woodlots, where we collected the largest numbers of chironomids.
In this study, we focused on insectivorous migrant landbirds and did not quantify fruit abundances at our study sites, but we did identify fruit seeds in bird diets. For the 32 total seeds that we observed, more than half were in the diets of fall migrant Yellowrumped Warblers, a late-arriving fall migrant species in southeastern South Dakota (Tallman et al. 2002, Swanson et al. 2003. Fruit consumption in Yellow-rumped Warblers may be associated with a rapid decrease of arthropod prey availability (Berthold 1976). Consistent with this argument, we found that arthropod prey availability rapidly decreased from mid-September to mid-October concurrent with the arrival of Yellowrumped Warblers.
We found that most individual bird species showed similar frequencies of arthropods in diets between corridors and woodlots in both spring and fall. However, Yellow-rumped Warblers in fall and Swainson's Thrushes in spring showed between-habitat differences in dietary arthropod frequencies.
These differences might be associated with variation in arthropod abundance between the two habitats. For example, counts of Polyphaga were higher and Formicidae lower at woodlots than at corridors, and, correspondingly, Swainson's Thrushes consumed larger proportions of Polyphaga and smaller proportions of Formicidae at woodlots.

Arthropod abundance vs. energetic condition
Of the arthropod sampling methods employed in this study, only arthropod metrics from sticky-trap samples were significantly correlated with plasma metabolite levels related to refueling rates in migratory birds. Plasma triglyceride levels were positively correlated with both arthropod abundance and biomass from sticky traps in fall, but not spring. The positive relationship between plasma triglyceride levels and arthropods collected by sticky traps in fall is consistent with high flying insect numbers contributing to high fattening rates in fall migrants. Flying insect abundance and biomass were both lower in fall than in spring in the present study, so periods with higher flying insect availability in fall were associated with higher refueling rates at this season but not in spring, perhaps because flying insect abundance exceeded threshold levels promoting fattening throughout the spring migratory season. Plasma β-hydroxybutyrate levels were positively correlated with arthropod abundance, but not biomass, and only in spring. The positive relationship between plasma βhydroxybutyrate levels and high abundance of flying insects in spring is difficult to explain because high fattening rates, from migrants feeding on abundant flying insects, should lead to reduced plasma β-hydroxybutyrate levels (Guglielmo et al. 2005, Liu and Swanson 2014a). Sticky-trap arthropod biomass did not show a similar relationship with plasma β-hydroxybutyrate levels in spring, perhaps suggesting that flying insects in spring were relatively small and low-quality diet items at this season, so inclusion of these insects in diets might result in relatively low fattening rates. The significant negative correlations of plasma βhydroxybutyrate levels with the flycatcher guild in spring for both models of arthropod abundance and biomass, however, reveals that the flycatcher guild, which feeds predominately on flying insects, had lower plasma β-hydroxybutyrate levels than other foraging guilds in spring, so the suggestion that flying insects represent low-quality diet items in spring is not fully supported by the data in the present study.
Habitat (corridors vs. woodlots) was not a significant predictor of plasma metabolite levels for GLMM models in the present study, consistent with the conclusion of Liu and Swanson (2014a) that corridors and woodlots provided similarly suitable habitat for refueling for migrants. Given that branch-clip and pitfall-trap arthropod metrics were not significantly related to plasma metabolite levels in the present study and that plasma TRIG and BUTY levels in these habitats were at levels consistent with successful refueling (Liu and Swanson 2014a), these data collectively suggest that arthropods were sufficiently available to migrants at both corridors and woodlots to successfully refuel at these riparian stopover sites. The sole exception to this conclusion is the finding that low flying insect numbers in fall may act to negatively impact refueling rates for migrants in both habitat types.
Time since sunrise was significantly or nearly significantly positively related to plasma triglyceride levels in spring for both arthropod abundance and biomass models, consistent with a general trend of morning fattening in migratory or wintering birds (Guglielmo et al. 2005, Swanson and Thomas 2007, Smith and McWilliams 2010, Mandin and Vézina 2012). Time since sunrise was not significantly positively related to TRIG levels in fall, consistent with lower refueling rates during fall compared to spring in these habitats (Liu and Swanson 2014a), but was significantly or nearly significantly negatively correlated with plasma BUTY levels in fall, suggesting a reduction in reliance on mobilized fat during stopover as the morning progressed, consistent with other studies (e.g., Guglielmo et al. 2002, Smith and McWilliams 2010. The absence of a significant association between plasma TRIG and time since sunrise in fall may be a function of the lower refueling rates in these habitats in fall compared to spring (Liu and Swanson 2014a). Foliage-dwelling arthropod counts and biomass were higher in fall than in spring, but arthropod counts from both pitfall and sticky traps were higher in spring than in fall. Arthropod biomass from sticky traps was also higher in spring than in fall, although biomass from pitfall traps was similar between seasons. Liu andSwanson (2014a, 2014b) found that foliage-gleaner and groundforaging guilds showed both higher plasma TRIG and/or baseline CORT levels in spring relative to fall. Thus, between-season differences in foliage-dwelling or flying arthropod counts and biomass were not reflected in similar between-season differences in fattening rates or stress hormone levels for migrant birds at our study sites. Arthropod counts from pitfall traps, however, were higher in spring than in fall, which is consistent with higher rates of fattening for ground-foraging birds in spring.

CONCLUSIONS
Taken together, these data suggest that arthropod prey were sufficiently diverse and abundant in both corridors and woodlots for migrants to replenish fuel stores during stopover, and further confirm that corridors and woodlots provide similarly suitable stopover habitats for woodland migrants during both spring and fall migration in the northern prairie region where woodland habitats are scarce (Liu and Swanson 2014a, 2014b. These results suggest that the conservation of even small woodland parcels, such as the isolated woodlot sites in this study, will benefit woodland birds by providing suitable stopover habitats during migration and thereby positively impact populations of these birds. Further, our data suggest that Hemiptera, Hymenoptera, Coleoptera, Diptera, and Lepidoptera are important arthropod taxa for migrant landbirds at our study sites, so conservation strategies focusing on increasing flying and foliage-dwelling arthropod populations, particularly of these arthropod taxa, might be especially beneficial to migrant birds.

Variation in arthropod abundance among plants
We used a GLMM approach with plant species as a predictor variable (see Methods section for details) to investigate variation in arthropod abundance and biomass among plants for foliagedwelling arthropods collected from corridors and woodlots in spring and fall.

Comparisons of arthropods within seasons
During spring, period showed a significant positive effect in GLMMs for branch-clip samples for both arthropod counts and biomass (Tables S12, S13, Fig. 3), indicating that foliage-dwelling and flying arthropod abundance increased as the spring season progressed. Per-unit counts and biomass of arthropods collected by pitfall traps, however, were not significantly affected by period during spring (Tables S12, S13, Fig. 3). During fall, period was a significant negative effector for per-unit counts of arthropods sampled by branch-clip and pitfall trap methods, indicating a reduction in foliage-and ground-dwelling arthropods from late-August to late-October (Tables S12, S13, Fig. 4). Period also showed a significant negative effect on arthropod biomass in late fall for branch-clip samples, but not for sticky or pitfall trap samples (Tables S12, S13, Fig. 4). Indeed, arthropod biomass from sticky trap samples actually increased in late fall (Tables S12, S13), driven by changes in 2010 but not 2011 (Fig. 4).

Variation in arthropod abundance among plants
The spring and fall GLMMs showed differences among plants for arthropod counts and biomass (Tables S12, S13). Plants supporting high arthropod counts in spring included common buckthorn, elm (American and Siberian pooled), eastern red cedar and honeysuckle. Low arthropod counts occurred on cottonwood and "other" trees and shrubs (the "other" category included crabapple Malus spp., sumac Rhus spp., and willow Salix spp.). Arthropod biomass was low for a number of plant species in spring, including black walnut, box elder, buckthorn, cottonwood, green ash, honey locust, white mulberry, and the "other" category. Fall arthropod counts were high on box elder and eastern red cedar and low on cottonwood. Arthropod biomass in fall showed less variation among plant species, with only lilacs having relatively high biomass.

Spring
Based on both arthropod counts and biomass, all flycatcher, foliage-gleaner, and ground-foraging guilds consumed more Lepidoptera, including both adults and larvae, than expected (Table S6 -S8). Both flycatcher and foliage-gleaner guilds showed lower consumption of Coleoptera and Hymenoptera than predicted from abundance and biomass and both flycatcher and groundforaging guilds consumed Diptera at lower levels than expected based on abundance and biomass (Table S6 -S8). The foliage-gleaner guild consumed more Diptera and less Araneae and Hemiptera than expected based on abundance and biomass (Table S7). Based on arthropod biomass, flycatchers consumed more Hemiptera and fewer Aranae than expected. Based on arthropod abundance, flycatchers showed more consumption of Aranae and lower of Hemiptera (Table S6). Based on biomass, the foliage-gleaner guild consumed Anthocoridae, Auchenorrhyncha, and Apocrita more than expected. Based on abundance, the foliage-gleaner guild consumed these same arthropod taxa less than expected (Table S7). The ground-foraging guild showed more consumption of Araneae and less of Coleoptera based on biomass. Based on abundance, this foraging guild consumed Araneae less and Coleoptera more than expected. (Table S8).

Fall
Based on counts and biomass, flycatcher, foliage-gleaner, and ground-foraging guilds consistently consumed more Lepidoptera and fewer Aranea than expected (Table S9 -S11). Both foliage-gleaner and ground-foraging guilds consumed Diptera and Nematocera more than expected based on counts and biomass (Table S9 -S10). The flycatcher guild, however, consumed Diptera and Nematocera less than expected based on counts and biomass (Table S10). Flycatchers consumed Ephemeroptera, Neuroptera, Orthoptera, Psocoptera, and Trichoptera less than expected based on both counts and biomass (Table S9). Foliage-gleaners consumed Neuropotera, Orthroptera, Psocoptera, and Trichoptera more than expected based on abundance and biomass (Table S10). Based on biomass, the flycatcher guild consumed Hymenoptera and Apocrita more than expected (Table S9). The foliage-gleaner guild consumed Coleoptera and Sternorrhyncha more than expected based on biomass (Table S10). However, both flycatcher and goliage-gleaner guilds consumed less of these same arthropod taxa than expected based on arthropod abundance (Table S9 -S10). The ground-foraging guild consumed fewer Hemiptera and Coleoptera, including both Adephaga and Polyphaga, than expected based on biomass (Table S11). Based on arthropod abundance, the ground-foraging guild consumed more of these arthropod taxa than expected (Table S11).

SUPPLEMENTARY DISCUSSION
The frequency of occurrence of some arthropod taxa varied between habitats, which could account for the different between-habitat trends in counts and biomass if arthropod groups with larger body masses were more prevalent at corridor sites. In spring, the mean size (biomass/individual) of Areane, Coleoptera, and Hemiptera were larger at corridors than at woodlots, but these arthropod orders showed lower occurrence at corridors than at woodlots (Table S1). In fall, Coleoptera, Diptera, Araneae, and Hymenoptera showed larger average size but lower occurrence at corridors than at woodlots (Table S2). These variations might be associated with differences in composition of canopy tree species between corridors and woodlots (Dean 1999, Schaffers et al. 2008). Both Dean (1999) and Tallamy and Shropshire (2009) suggested that arthropod abundance varied among tree species. Similarly, we also found that foliage-dwelling arthropod abundance and biomass greatly varied among plant species (SI: RESULTS, Tables S3, S12, S13). These plant species composition differences (Dean 1999, Swanson et al. 2003, Gentry et al. 2006) together with the arthropod abundance variation among plant species (Table S3, S12, S13) contribute to the different between-habitat trends for arthropod counts and biomass.

Arthropod variation among years, between within seasons and with plant species
Prey abundance showed among-year variation, being lower in spring 2011 than in springs of 2010 and 2012, and lower in fall 2011 than in fall 2010. These among-year variations might be driven, at least partially, by differences in bird abundance (Moore and Wang 1991, Tulp and Schekkerman 2008, Ewert et al. 2015). Indeed, Schwenk et al. (2010) and Moorman et al. (2012) suggested that arthropod density was negatively related to bird abundance. Bolduc et al. (2012) documented that terrestrial arthropod abundance was higher in mesic than in xeric habitats. Perhaps more likely, these differences could be related to weather variation, as 2011 represented a very wet spring and summer, with a large flood on the Missouri River where substantial riparian forest was inundated for much of the summer; 2010 and 2012 were more average weather years (Munes et al. 2015).
We found that prey abundance also varied between and within seasons. Dean (1999) found that foliage-dwelling arthropod abundance was higher in fall than in spring at riparian woodland sites in southeastern South Dakota and suggested that the increase in foliage-dwelling arthropods in fall might be due to the increase of habitat availability (more foliage in fall than in spring) and arthropod recruitment over the summer. In this study, we also found higher abundance and biomass of foliage-dwelling arthropods in fall than in spring. In contrast, we found that flying insect abundance was higher in spring than in fall. Opposite seasonal patterns of flying and foliage-dwelling arthropod abundance might result from the increased density of foliage in fall reducing long-distance flights and, thereby, sticky-trap captures (Dreistadt et al. 1998). The within-season variation in prey availability in this study was also similar to Dean's (1999) findings that arthropod abundance increased with the progression of spring and decreased with the progression of fall. This within-season variation is likely related to changes in temperature and plant growth but might also partially result from changes in bird abundance at our study sites (Dean 1999, Swanson et al. 2003, Lill and Marquis 2004, Høye and Forchhammer 2008. Cottonwood is the dominant tree species in riparian forests along the Missouri River (Dixon et al. 2012) and had generally low arthropod counts and/or biomass at both seasons. This could partially explain the lower arthropod counts in corridors relative to woodlots in spring documented in this study. Invasive or non-native species, including eastern red cedar (native, but expanding its landcover in the region; Greene and Knox 2014, Meneguzzo and Liknes 2015, Illeperuma and Dixon 2021), buckthorn, lilac, honeysuckle, and hemp did not have consistent negative effects on arthropod abundance or biomass at our study sites, suggesting that they may provide some value for migrant woodland birds, at least for supporting arthropod prey. Lilac was the only plant species supporting relatively high arthropod biomass in fall and eastern red cedar also had high arthropod counts, although not biomass, at both seasons. Honeysuckle had high arthropod counts, although not biomass, in spring. The low nutritional value of honeysuckle fruits tends to make honeysuckle-dominated habitats poor quality stopover habitats in fall (Smith et al. 2013, Smith andHatch 2020). Nevertheless, arthropods do not appear generally low on honeysuckle plants (Smith et al. 2013) and habitats dominated by honeysuckle appear to allow mass gain during spring stopover (Smith and Hatch 2016).

Differences between arthropod abundance and dietary consumption
The migrant community as a whole consumed more Formicidae, Lepidoptera and Carabidae in spring and Auchnorrhyncha, Sternorrhyncha, Carabidae, Lepidoptera, and Nemetocera in fall than expected based on counts or biomass. Some other studies of bird diets also found strong selection of Coleoptera, Lepidoptera, Hemiptera, and Hymenoptera (e.g. Robinson and Holmes 1982, Poulin and Lefebvre 1996, Yard et al. 2004, Wolfe et al. 2009, Carlisle et al. 2012. Although some arthropod taxa were consumed or avoided by most individual bird species in the present study, some individual bird species showed different patterns of consumption of certain arthropod taxa than the overall migrant community, probably because of divergent foraging habits of individual species (Holmes and Robinson 1988, Poulin et al. 1994, Yard et al. 2004, Wolfe 2009, Steele et al. 2010, Carlisle et at 2012. Carlisle et al. (2012) documented that fall landbird migrants in Idaho preferred Coleoptera, Hemiptera (Auchenorrhyncha, Cercopidae) and Lepidoptera. In eastern North America, Moorman et al. (2007) suggested that landbird migrants preferred Coleoptera and Lepidoptera in spring and Coleoptera in fall at stopover sites in South Carolina. The foliage-gleaning migrant community in the Moorman et al. (2007) study consumed proportionately more Diptera, Hemiptera, and Lepidoptera (spring) than their abundance in both spring and fall. Thus, Lepidoptera and Coleoptera appear to be generally preferred arthropod prey for migrants across broad geographic areas in North America, with Diptera and Hemiptera also often preferred diet items.
Birds also showed variations in patterns of arthropod consumption relative to abundance in the habitat when expressed as counts vs. biomass. These variations are likely due, at least in part, to variation in the size (biomass) of arthropod prey items. For example, the average biomass of Araneae (17.64 mg) was 7.4-fold larger than that of Adephaga (2.37 mg) and the groundforaging guild in spring consumed more Adephaga than expected based on counts but fewer Adephaga than expected based on biomass, with the opposite trends for Araneae. Consequently, including analyses of both counts and biomass of arthropod prey items is important for a comprehensive picture of migrant diets for conservation purposes (Poulin andLefebvre 1996, Yard et al. 2004).
SUPPLEMENTARY TABLES TABLE S1. Spring relative arthropod community composition from branch-clip, sticky-trap, and pitfall-trap samples at corridors (CO) and woodlots (WO). Data represent counts of arthropods sampled (n) and percentages of total arthropod composition within a habitat (%). Taxa included are those with n  5.                S3. Arthropod abundance from branch-clip samples collected from various plant species at corridors (CO) and woodlots (WO) in both spring and fall. Data represent arthropod counts per branch (n/b), percentage of total branch-clip samples (%b), and arthropod abundance (abundance = total counts of arthropod from a plant species multiplied by the percentage of that plant species (Dean 1999, Swanson et al. 2003).   .5 (28.0) 7.9 (4.3) 6.7 (6.8) 9.0 (37.9) 4.9 (8.7) 6.7 (6.5) 9.9 (13.3) 6.2 (3.5) 10.0 (15.7) 6.6 (7.2) Hemiptera 12.0 (20.6) 12.5 (14.2) 13.1 (9.1) 6.6 (8.8           Table S11. Results of comparisons of presence in the habitat vs. dietary consumption for arthropod taxa for fall ground-foraging migrant birds. Positive scores represent greater consumption than expected and negative scores less consumption than expected based on presence in the habitat. Abbreviations as in Table S6 and S5.  Table S13. Statistics from GLMMs showing the best-fit model results (all models presented had AIC values > 2 lower than the next competing model, except for fall branch clip samples where a model with site as a random effect had a ΔAIC < 2; significance of variables did not change between these two models, so only the best-fit model is presented) for arthropod biomass (mg/branch or mg/trap hr), with habitat (corridor vs. woodlot), site (six study sites), year, plant species (branch clip sampling method only), period, average temperature, and wind speed as predictor (or random, see Methods) variables. Statistics from GLMM showing the best-fit model results for, with site, year, plant species (branch clip sampling method only), period, average temperature, and wind speed as predictor variables. Plant species are the same as in Table S12.  Table S15. Statistics from GLMM models with plasma triglycerides (TRIG) and βhydroxybutyrate as dependent variables, year as a random effect and arthropod abundance or biomass (arthropod metric) by the different sampling methods, average temperature, habitat (corridors vs. woodlots), temporal migration period (sequential equal periods over the migration seasons), time since sunrise (min), and foraging guild as predictor variables.