Study Area

The study area encompassed the northern extent of the Lesser Prairie-Chicken's distribution in Kansas and Colorado and included 4 study sites spread among the Mixed-Grass Prairie (Red Hills, Clark), Short-Grass Prairie/CRP Mosaic (Northwest), and Sand Sagebrush Prairie (Colorado, Clark) ecoregions (McDonald et al. 2014; Figure 1). Although the Colorado study site occurred within the Sand Sagebrush Prairie ecoregion, this site was predominantly composed of Conservation Reserve Program grassland (CRP) and cropland on the border of Prowers and Baca counties. Dominant grasses, forbs, subshrubs, shrubs, mean annual precipitation, and soil texture varied among study sites (Appendix Table 5). For example, subshrubs (e.g., Gutierrezia sarothrae and Amphiachyris dracunculoides) were more abundant than forbs in northwest Kansas and more abundant than shrubs at the Red Hills study site (Appendix Table 5). Forbs were predominantly Salsola tragus and Kochia scoparia, which were 2 of the top 3 most abundant forbs at all sites, excluding the Red Hills.

FIGURE 1.

Extent of study area as determined by minimum convex polygons (shown in red) of VHF- and GPS-marked Lesser Prairie-Chickens in western Kansas and eastern Colorado, USA, 2014–2015. Study sites in Gove and Logan counties, Kansas, were combined for analyses and are referred to as “Northwest.” The study site on the edge of Comanche and Kiowa counties, Kansas, is referred to as “Red Hills.” The estimated current distribution of Lesser Prairie-Chickens is indicated by hatch marks (Hagen and Giesen 2005).

Sample Collection

We collected fecal samples during the brooding period (May–September) and winter (November–March) from Lesser Prairie-Chickens captured at leks between early March and mid-May using walk-in funnel traps and drop nets (Haukos et al. 1990, Silvy et al. 1990). We sexed the birds on the basis of plumage coloration, length of pinnae, and tail pattern (Copelin 1963). We marked female Lesser Prairie-Chickens with either a 15 g VHF transmitter or a 22 g GPS satellite PTT transmitter. We obtained locations for each VHF-marked female 3–4 times wk⁻¹, whereas females marked with GPS PTT transmitters accrued 8–10 locations day⁻¹, contingent on available daily solar energy. GPS locations were recorded every 2 hr during the day, with a 6 hr gap between 2300 and 0500 hours.

During the brooding season, we collected fecal samples from marked hens and chicks (separate vials for each) during brood capture and weekly flush counts occurring within 1 hr of sunrise (2–98 days old). We classified fecal samples as either chick or adult samples on the basis of their relative size differences. During winter and early spring (December–March), we collected fecal samples (≥1 pellet) at roost sites. Fresh fecal samples that were still moist and appeared to have been dropped the previous night were placed in 20 mL vials using small plastic sampling spoons to minimize DNA contamination. Vials labeled with the date, unique bird ID, and coordinates of the collection location were stored in a freezer at field sites and at Kansas State University before being shipped frozen overnight for laboratory analyses.

Sequencing

We extracted Genomic DNA from fecal samples using the PowerSoil-htp 96-well Soil DNA Isolation Kit (MO BIO Laboratories, Carlsbad, California, USA). For arthropods, we amplified a fragment of the Folmer region of the cytochrome oxidase I (COI) gene using arthropod-specific primers (Bohmann et al. 2011, Zeale et al. 2011). To determine the contribution of plants to diets, a portion of the chloroplast trnL intron was PCR-amplified from each genomic DNA sample using the c and h trnL primers (Taberlet et al. 2007), but modified to include appropriate barcodes and adapter sequences for Illumina multiplexed sequencing. The barcodes used were 12 base pair (bp) error-correcting barcodes unique to each sample (Caporaso et al. 2012). Each 25 μL PCR reaction was mixed according to PCR Master Mix specifications (Promega, Madison, Wisconsin, USA), with 2 μL of genomic DNA template. For trnL, the thermocycling program used an initial step at 94°C for 1 min, a final extension at 72°C for 2 min, and the following steps cycled 36 times: 1 min at 94°C, 30 s at 55°C, and 30 s at 72°C. For COI, the thermocycling program used an initial step at 94°C for 5 min, a final extension at 72°C for 10 min, and the following steps cycled 45 times: 30 s at 94°C, 45 s at 45°C, and 45 s at 72°C. We cleaned amplicons from each sample and normalized them using SequalPrep Normalization Plates (Life Technologies, Carlsbad, California) before pooling them for sequencing on a MiSeq (Illumina, San Diego, California) running the 2 × 150 bp chemistry.

Assignment of Reads to Arthropod Genera

For COI reads indicating arthropod taxa, we demultiplexed sequences using “prep_fastq_for_uparse.py” (Leff 2018). Read 2s were used for downstream analysis, due to higher-quality scores. Sequences were filtered and operational taxonomic unit (OTU) picking was performed using the UPARSE pipeline (USEARCH 7). Quality filtering included trimming sequences to the expected amplicon length (158 bp—only for 250 bp reads), filtering by quality score (maxee value of 1.5), removing sequences below the minimum expected amplicon length (90 bp), and removing singletons. We clustered sequences de novo at 99% similarity for OTU picking. We performed taxonomy assignment in QIIME, using the hierarchical naive Bayesian classifer RDP, retrained with a custom reference database curated from the Barcode of Life Database (version 3). Taxonomy was assigned at 99% similarity, with a 50% confidence threshold. We further filtered sequences to remove non-arthropod sequences by removing sequences that were not resolved to at least the family level. All samples with <10 COI reads were excluded from analysis for arthropods in diet.

We calculated the percentages of all sequences assigned to a given OTU for each sample. This is referred to as RRA (relative read abundance; Kartzinel et al. 2015). For COI, an average of 9.67% of all sequences were matched to genera in the order Diptera, almost exclusively during summer. Due to observations of contact between fecal material and dipterans, we assumed that dipteran DNA entered fecal material through secondary contact after defecation and before collection. Therefore, we excluded all dipteran reads from analyses. We limited assignment of OTU to genera present among all study sites as estimated from arthropod sweep-net survey (see details below).

Assignment of Reads to Plant Taxa and Functional Groups

Sequences were demultiplexed for trnL using a Python script (available from  https://github.com/leffj/helper-code-for-uparse/blob/master/prep_fastq_for_uparse_paired.py). Paired end reads were then merged using “fastq_merge” pairs (Edgar 2010). We used “fastx_clipper” to trim primer and adaptor regions from both ends ( https://github.com/agordon/fastx_toolkit) because merged reads often extended beyond the amplicon region of the sequencing construct. Sequences lacking a primer region on both ends of the merged reads were discarded. Sequences were quality trimmed to have a maximum expected number of errors per read of <0.1, and only sequences with >3 identical replicates were included in downstream analyses. BLASTN 2.2.30+ was run locally, with a representative sequence for each OTU as the query and the current National Center for Biotechnology Information (NCBI) nucleotide and taxonomy database as the reference. The tabular BLAST hit tables for each OTU representative were then parsed so that only hits with >97% query coverage and identity were kept, using the “usearch7” approach (Edgar 2013, Craine et al. 2015). The NCBI genus names associated with each hit were used to populate the OTU taxonomy assignment lists. All samples with <50 trnL reads were excluded from analyses of trnL RRA (Kartzinel et al. 2015). We estimated OTU-specific RRA and defined a representative genus for each OTU to describe composition in diet. We used the representative genera when summarizing OTU composition in diets. For example, OTUs were from species in genera similar to Ambrosia. We limited plant genera within OTU to those detected during extensive vegetation surveys among sites (Appendix A).

For trnL, an average of 4% of sequences was from Pinus (range: 0–51%). Because of the unlikelihood of Pinus biomass being consumed and the presence of Pinus DNA in the blanks, the one OTU that matched with Pinus species was removed from the dataset. For trnL, among the top 10 OTUs, OTU 23 did not match at 97% levels for coverage and identity for any species in the NCBI database. However, OTU 23 matched at 100% coverage and 95% identity with a Chenopodium species in the NCBI database and was considered a species similar to Chenopodium for the purposes of this study.

Functional group assignments

Because OTUs often encompassed multiple genera, we grouped RRA from different plant genera into functional groups including forbs, shrubs, subshrubs (mostly Gutierrezia), legumes, grasses, crops (not including alfalfa), and alfalfa. Placing genera into each functional group presented challenges because the OTUs frequently encompassed genera indicative of multiple groups (see below). However, linking plant foods consumed to specific functional groups was necessary to allow for comparisons among sites and to make direct connections to the utility of landscapes with an agricultural component. In some instances, OTUs that included genera related to both grass and crop as well as shrub and subshrub functional groups included repeat values and, therefore, added values could surpass 100%. For example, 17 of 33 OTUs that identified either grass or crop foods included both crop and native grass genera (e.g., Triticum and Elymus); 2 of 45 OTUs of genera including shrub, subshrub, and forb species included representatives of >1 functional group (e.g., Artemisia and Ambrosia); and 1 of 5 OTUs for genera of legumes included both cultivated and native species (e.g., Medicago and Vicia). To overcome functional-group overlap within OTUs, we constrained the use of crop and shrub foods to instances when each land cover type occurred within 48 hr home ranges; and we used the Bayesian approach, similar to regional assignments in Royle and Rubenstein (2004), to estimate RRA for each functional groups using identity values as a prior probability:

We estimated an adjusted RRA for each functional group (RRA_fg = k) by estimating the average identity value (I_g) among genera within an OTU and then dividing I_g by the sum of identity values for functional groups within each OTU. We then multiplied the quotient by the RRA estimated for each OTU (RRA_{OTU = j}). The adjusted RRA accounts for the probability that each read is from a particular functional group based on the identity value. The identity value is a measure of the match between the OTU detected in the fecal sample and genus-specific reference sequence.

Evaluation of Sampled Taxonomic Richness

To examine whether sample sizes were sufficient to detect all arthropod and plant foods used by Lesser Prairie-Chickens at each study site, we used species accumulation curves depicting the relationship between number of OTUs and number of fecal samples. Species accumulation curves were generated in the R package “vegan” with the “specaccum” function, and the “Lomolino” function was used to describe the curves (Oksanen et al. 2015). From the function, we estimated an asymptote and the number of OTUs achieving a midpoint of the asymptote. We also estimated extrapolated species richness using the function “poolaccum” within package “vegan” following Chao (1987).

Arthropods

A total of 75 arthropod OTUs were identified in diets of Lesser Prairie-Chickens using COI analyses. Results from OTUs encompassed 4 classes: Insecta (63), Arachnida (9), Collembola (1), and Malacostraca (1). Among these 4 classes, 12 orders and 50 families were represented. Twenty-eight of the genera were Lepidoptera, 7 Araneae, and 6 Hemiptera (Appendix Table 6). On average, 35% of the RRA was from Lepidoptera, 26% from Orthoptera, 14% from Araneae, and 13% from Hemiptera (Appendix Figure 8 and Appendix Table 7).

Sweep-net transects indicated that arthropod communities varied among study sites. Orthoptera had the greatest percent biomass among taxa at each site (Clark = 90.2%, Red Hills = 71.5%, Northwest = 73.1%, and Colorado = 46.5%), followed by Lepidoptera, Phasmatodea, and Coleoptera (Appendix Figure 9). Lepidopterans comprised >4 times more of the arthropod community biomass in Northwest and Colorado sites than in the Red Hills site and 1.6 times more than in the Clark study site.

Beta regressions suggested no differences among Lepidoptera, Orthoptera, Hemiptera, and Araneae composition in diets between the brooding period and winter (winter β = 0.054 ± 0.303, 0.269 ± 0.293, 0.210 ± 0.265, −0.265 ± 0.279, respectively; brooding period as reference intercept). However, average reads per sample were fewer in the winter than in the brooding period for all sites except Colorado (Appendix Table 7). Given our sample size, the power of detecting a difference at an 85% CI was 0.24, 0.43, 0.47, and 0.56, respectively. Chick and adult diets during the brood-rearing period did not differ in consumption of Lepidoptera, Orthoptera, Araneae, and Hemiptera (chick β = 0.013 ± 0.403, 0.205 ± 0.386, 0.122 ± 0.388, −0.199 ± 0.370, respectively). Beta regressions also indicated no differential consumption of foods by age for Lepidoptera, Orthoptera, Araneae, and Hemiptera (age of chick days β = −0.004 ± 0.00779, 0.00732 ± 0.00788, −0.000999 ± 0.007839, −0.00218 ± 0.00700, respectively). There was an indication of more complicated nonlinear trends in the consumption of Lepidoptera and Orthoptera with minimal use of Lepidoptera after 40 days of age and greater consumption of Orthoptera when chicks surpassed 40 days of age (Figure 2).

FIGURE 2.

Scatter plots fitted with least squares (red) and locally weighted scatterplot smooth lines (blue) to depict patterns in the composition of Orthoptera (A, B) and Lepidoptera (C, D) in the diets of Lesser Prairie-Chicken chickens during the brood-rearing period of 2014 in Kansas and Colorado, USA. Days encompass May 27, 2014, to August 29, 2014; ages of chicks depicted range from 2 to 98 days.

The lack of variation among periods and ages is further indicated by stronger model support for land cover (Native Prairie, CRP, cropland) and site-based covariates for Lepidoptera and Araneae, which suggest that variation in arthropod diet consumption is more influenced by landscape characteristics than by temporal factors (Table 2). For Orthoptera during brood rearing, the model including date as a covariate was ranked highest but was equally parsimonious (ΔAIC_c < 2) with the native grassland, crop, and CRP models, and its beta coefficient overlapped zero at the 85% CI (Table 2). The combined effect of spatially related covariates in predicting the composition of each order during both brood rearing and winter carried an average model weight of 72% (Tables 2 and 3).

TABLE 2.

Results of beta regression model for the consumption of Lepidoptera, Orthoptera, and Araneae by Lesser Prairie-Chickens in Kansas and Colorado, USA, during the brood-rearing period (June–September) of 2014. K is the number of parameters, AIC_c is Akaike's Information Criterion adjusted for small sample size, ΔAIC_c is the difference in AIC_c compared to the smallest value, and w_i is model weight. Models with beta coefficients not overlapping zero at the 85% confidence interval are in bold.

TABLE 3.

Beta regression model results for the consumption of Lepidoptera, Orthoptera, and Hymenoptera by Lesser Prairie-Chickens in Kansas and Colorado, USA, during winter 2014–2015. K is the number of parameters, AIC_c is Akaike's Information Criterion adjusted for small sample size, ΔAIC_c is the difference in AIC_c compared to the smallest value, and w_i is model weight. Models with beta coefficients not overlapping zero at the 85% confidence interval are in bold.

Spatial variation in dietary composition was indicated by RRA among sites (Figure 3). During the brood-rearing period, presence of native grassland had the greatest influence on arthropod diet composition among Lepidoptera, Orthoptera, and Araneae but carried, on average, 30% of model weight (Table 2), which suggests that several variables were likely influential. The contribution of Lepidoptera in diets during the brood-rearing period decreased in native grassland (native grassland β = −0.657 ± 0.405; Table 2). Consumption of lepidopterans was 2.12× less in native grassland in comparison to cropland (23.2 ± 6.00% vs. 49.2 ± 11.8%; Figure 4). Similarly, the categorical native grassland covariate was the best predictor of the consumption of Araneae, based on AIC_c, and the beta coefficient did not overlap zero at the 85% CI (native grassland β = 0.559 ± 0.379). Araneae contributed 65× more to diets in native grassland than in other cover types and was rarely consumed in cropland (26.2 ± 7.02% vs. 0.04 ± 0.004%; Figure 4). For Orthoptera, the model including native grassland as a covariate was not informative (native grassland β = 0.154 ± 0.361). Despite not providing a statistically meaningful difference, point estimates for Orthoptera RRA was 21.7 ± 6.50% in native grassland vs. 12.7 ± 6.71% in other cover types. Hemiptera contributed relatively equally to diets among Lesser Prairie-Chickens using CRP grassland, native grassland, and cropland (Figure 4).

FIGURE 3.

Relative readable abundance (RRA; proportion) of DNA within Lesser Prairie-Chicken fecal samples matching barcodes similar to arthropod orders (A) Lepidoptera, (B) Orthoptera, (C) Araneae, and (D) Hemiptera, grouped by study site. Fecal samples were pooled among study sites in Clark County, Kansas (Clark); Gove and Logan counties, Kansas (NW); Kiowa and Comanche counties, Kansas (RH); and Prowers and Baca counties, Colorado (CO), USA, and were collected during summer 2014 (hatch to 98 days old) from brooding females and chicks.

FIGURE 4.

Relative readable abundance (RRA; proportion) of DNA within Lesser Prairie-Chicken fecal samples matching barcodes similar to arthropod orders (A) Lepidoptera, (B) Orthoptera, (C) Araneae, and (D) Hemiptera, grouped by land cover type where collected. Land cover types included cropland, Conservation Reserve Program grassland (CRP), and native working grassland (native grassland). Fecal samples were pooled among study sites in Clark County, Kansas; Gove and Logan counties, Kansas; Kiowa and Comanche counties, Kansas; and Prowers and Baca counties, Colorado, USA, and were collected during summer 2014 (hatch to 98 days old) from brooding females and chicks.

In winter, Lepidoptera, Orthoptera, and Hymenoptera (most likely galls) contributed most to arthropod-based food for Lesser Prairie-Chickens (Appendix Figure 8 and Appendix Table 7). Of the top 4 orders contributing to winter diets, Orthoptera was the only order that changed (decreased) as the winter progressed, which was significant at the 85% CI (day since start of period β = −0.035 ± 0.0131). Among sites, Clark birds had the greatest percentages of Orthoptera in their winter diet when compared to all other sites, and this was significant at the 85% CI (51.7 ± 12.6% in Clark vs. 18.3 ± 7.7% in Colorado vs. 0% in Red Hills and Northwest; Clark β = 1.86 ± 0.613).

Plants

Metabarcoding of fecal samples indicated that Lesser Prairie-Chickens consumed foods encompassing 2 classes (Magnoliopsida and Liliopsida), 19 orders (predominantly Asterales, Poales, and Fabales), 30 families, and 90 genera. A total of 235 OTUs were found to represent ≥1% of the plant diet for a given bird at a given time. In contrast to the assignment of OTU to specific arthropod taxa, trnL OTUs were not genus specific and, on average, comprised 4.15 ± 4.79 genera, ranging from 1 to 28 potential genera that were present at all study sites combined. Of the 235 recorded OTUs, 70 represented ≥10% of the diet for ≥1 of the samples. The most abundant OTUs were from species in genera similar to Ambrosia (27% OTU-specific RRA of all reads), followed by species in genera similar to Lactuca or Taraxacum (10%), Medicago (6%), and Triticum (5%).

For the brood-rearing period, the 10 most abundant OTUs included species similar to Ambrosia (16.2%), Lactuca (8.5%), Triticum (5.5%), Chenopodium (4.3%), Physalis (3.9%), Commelina (3.1%), Trifolium (1.8%), and Elymus (1.4%). Ambrosia and Triticum were represented by 2 separate OTUs as part of the top 10 most abundant summer OTU foods. During winter, the 10 most abundant OTUs consumed included species similar to Ambrosia (21.0%), Lactuca (5.6%), Medicago (4.8%), Triticum (4.4%), Bromus (1.1%), Oenothera (0.9%), Elymus (0.7%), Sorghum (0.6%), and Chenopodium (0.6%). Triticum was represented by 2 separate OTUs as part of the top 10 most abundant winter OTUs.

Functional groups

Home ranges (48 hr) averaged 45.06 ± 44.50 ha during the nonbreeding season and 11.17 ± 8.84 ha during brood rearing for GPS-marked birds. We then used the maximum-size home ranges of nondispersing GPS-marked individuals during each time period to estimate home ranges for VHF-marked Lesser Prairie-Chickens. Home ranges for VHF birds were derived from the higher-resolution GPS-marked bird data because GPS locations were obtained frequently enough to generate 48 hr home ranges. Maximum home range sizes during the nonbreeding and brooding periods were 191.52 ha and 32.83 ha, respectively, from which we derived 781 m and 323 m buffer distances around VHF fecal collection locations to account for all potentially used food sources.

In both the brood-rearing and winter periods, forbs were the predominant plant-based food source (winter 53.7 ± 3.7%, brooding 60.67 ± 5.5%; Appendix Figure 10). Differences in the overall use of functional groups among the winter and brood-rearing periods were minimal. However, subshrubs (e.g., Gutierrezia spp.) and grasses contributed 1.5 times (43.4 ± 3.7% vs. 29.8 ± 5.7%) more to Lesser Prairie-Chicken diets during winter than during brood rearing (winter β = 0.564 ± 0.220, 0.287 ± 0.195). By contrast, there was no difference in the consumption of forbs, legumes, shrubs, and crops between periods (brooding β = 0.198 ± 0.230, −0180 ± 0.209, 0.222 ± 0.175, −0.265 ± 0.185, respectively).

We assessed differences among all sites separately for each period. Within the brood-rearing period alone, foods in the forb, grass, and legume functional groups did not differ among sites. Shrub- and subshrub-based foods contributed more to diets during the brood-rearing period in the Red Hills and northwest Kansas compared to Clark and Colorado (Red Hills β = 1.82 ± 0.782, Northwest β = 0.769 ± 0.430, Clark β = 1.22 ± 0.779, Colorado β = 0.836 ± 0.444). Crop-based foods provided a greater contribution to brood-rearing diets in Colorado compared to other sites (β = 3.67 ± 0.509).

During winter, grass composition in diets varied among sites. More grasses were consumed during winter at the Northwest study than at the Clark study site (23.0 ± 2.6% vs. 11.0 ± 1.7%; β = 0.855 ± 0.289; Figure 5). Shrub foods contributed more in winter at the Red Hills study site than at Clark (β = 0.908 ± 0.391). Crop foods contributed more in winter to diets at the Northwest site than at Clark (β = 0.443 ± 0.288). Last, subshrub foods contributed more in winter to diets at the Northwest and Red Hills study sites than at Clark (β = 0.836 ± 0.445, 1.22 ± 0.779, respectively; Figure 5).

FIGURE 5.

Adjusted relative readable abundance (RRA; proportion) of DNA within Lesser Prairie-Chicken fecal samples matching barcodes indicative of plant functional groups, including forbs, grasses, legumes, shrubs, crops, and subshrubs, grouped by study site. Fecal samples were collected from study sites in Clark County, Kansas (Clark); Gove and Logan counties, Kansas (NW); Kiowa and Comanche counties, Kansas (RH); and Prowers and Baca counties, Colorado (CO), USA, during winter 2014–2015 (November–March).

After screening for differences among periods and sites, we focused on winter diets, using a multimodel inference approach, because Lesser Prairie-Chickens predominantly consume plant material during winter (Jones 1963). Models including spatially related covariates carried, on average, 99% of model weight (AIC_c weight; Table 4). The top-ranking predictor for forb diet composition was occurrence in alfalfa and crop fields (Table 4). Forbs were consumed less in winter by Lesser Prairie-Chickens using alfalfa fields and crop fields in general (β = −1.57 ± 0.467; identical beta values for alfalfa and crop). Forbs were more readily consumed in native grassland and CRP (Figure 6). The proportion of grass in diets was best predicted by site (Table 4; see differences above), with use of native grassland ranking second among models (native grassland β = 0.386 ± 0.238). Birds using alfalfa and crop fields had the greatest relative proportion of legumes in their diet (β = 4.60 ± 0.507 for both alfalfa and crop). All fecal samples collected in cropland were collected in cultivated alfalfa, which confirms that birds can use alfalfa fields in winter as a food source. Shrubs contributed more to the diets of Lesser Prairie-Chickens using native grassland than to those in other cover types (native grassland β = 1.55 ± 0.254; Table 4). The relative diet composition of subshrub appears to be most strongly influenced by use of crop fields, with consumption of subshrub lower in cropland (β = −1.38 ± 0.454).

TABLE 4.

Akaike's Information Criterion adjusted for small sample size (AIC_c), difference in AIC_c compared to the smallest value (ΔAIC_c), and model weight (w_i) for beta regression models explaining winter plant diets of Lesser Prairie-Chickens in Kansas and Colorado, USA, 2013–2014. K is the number of parameters. Models with beta coefficients not overlapping zero at the 85% confidence interval are in bold.

FIGURE 6.

Adjusted relative readable abundance (RRA; proportion) of DNA within Lesser Prairie-Chicken fecal samples matching barcodes indicative of plant functional groups, including forbs, grasses, legumes, shrubs, crops, and subshrubs, grouped by land cover type. Land cover types included cropland, Conservation Reserve Program grassland (CRP), and native working grassland (native grassland). Fecal samples were pooled among study sites in Clark County, Kansas; Gove and Logan counties, Kansas; Kiowa and Comanche counties, Kansas; and Prowers and Baca counties, Colorado, USA, and were collected during winter 2014–2015 (November–March).

Evaluation of Sampled Taxonomic Richness

Among all sites, the arthropod species accumulation curve achieved an estimated asymptote at 156 OTUs, which suggests that we didn't sample all available forage; the midpoint for achieving an asymptote was estimated at 105 fecal samples (Figure 7). The extrapolated species richness at the OTU level (based on Chao 1987) was 101. The plant species accumulation curve achieved an estimated asymptote at 282 OTUs, which suggests that we sampled nearly all used plant forage at the OTU level. The midpoint for achieving the asymptote was estimated at 17 fecal samples (Figure 7). The extrapolated species richness at the OTU level (based on Chao 1987) was 262.

FIGURE 7.

Species accumulation curves for plants and arthropods estimated using the R package “vegan” (Oksanen et al. 2015), depicting the relationship to number of operational taxonomic units (OTUs) detected in Lesser Prairie-Chicken fecal samples collected during brood rearing and winter, 2014–2015, in Kansas and Colorado, USA. Lomolino curves: plants 282.7/[1+17.1^log(2.3/x)]; arthropods: 156.0/[1+105.3^log (2.25/x)].

Arthropods in Lesser Prairie-Chicken Diets

The greater consumption of Lepidoptera in this study than was found in past research is likely a product of both the limited detection of soft-bodied prey using traditional methods and inclusion of study sites that have a strong row-crop agriculture component. Lesser Prairie-Chickens are known to consume lepidopteran larvae, yet the results of previous research suggest minimal consumption of Lepidoptera in comparison to Orthoptera (Davis et al. 1980). The traditional use of fecal dissection may not be effective in detecting lepidopteran larvae (e.g., butterfly and moth caterpillars). No study using fecal dissection identified Lepidoptera as a prey item for Lesser Prairie-Chickens (Jones 1963, Doerr and Guthery 1983). Only studies that examined crop contents have reported consumption of lepidopteran larvae (Crawford and Bolen 1976, Suminski 1977, Smith 1979, Davis et al. 1980, Riley et al. 1993). However, not all studies examining crop contents have explicitly identified Lepidoptera as a food item, and foods from this order may be clumped as “other insects” (Olawsky 1987), making comparisons among other studies challenging. Overall, the soft-bodied nature of caterpillars likely makes them easier to digest and subsequently harder to detect using traditional dissection approaches (Trevelline et al. 2016). DNA metabarcoding may be the least biased tool for comparing dietary composition among soft- and hard-bodied prey.

In addition to palatability, use of lepidopteran larvae during the brood-rearing period may be related to the ease of capture by a small, 13–35 g chick. Lepidopteran larvae would be easy for Lesser Prairie-Chicken chicks to obtain when occurring within reach on the ground or in shorter vegetation. It is possible that soft-bodied larvae from other orders (e.g., Coleoptera) could also be consumed when available. Although we didn't expect a greater consumption of Lepidoptera than of Orthoptera by Lesser Prairie-Chicken chicks, we predicted that chicks would be restricted to smaller arthropod prey of limited mobility (following optimal diet theory; Suminski 1977, Sih and Christensen 2001). The use of lepidopteran larvae by Lesser Prairie-Chicken chicks supports this prediction. The potential dietary selection of lepidopterans further identifies the necessity of matching life histories among predator and prey. The life-history strategies of arthropod species may largely determine their importance as a prey item.

Although Lepidoptera were used as a food source among all land cover types and sites, specific lepidopteran genera were used in agricultural landscapes. Diets of Lesser Prairie-Chickens during the brooding period were largely supported by the genera Euxoa and Dargida. These 2 genera comprise several known agricultural pest species, including army cutworms (Euxoa auxiliaris). Dietary use of cutworms by Lesser Prairie-Chickens was also detected in fall by Crawford and Bolen (1976) in fragmented sand shinnery oak prairie. Consumption of agricultural pests provides evidence of one ecological service provided by Lesser Prairie-Chickens that could be used to gain conservation support in private working landscapes throughout their distribution (Wenny et al. 2011).

In contrast to the prevalent consumption of Lepidoptera in their northern range, the predominant use of orthopteran foods by Lesser Prairie-Chickens is well supported by other published research (Jones 1964, Suminski 1977, Davis et al. 1980, Doerr and Guthery 1983). The difference in predominant foods (Orthoptera vs. Lepidoptera) may be a result of spatial variation among study areas, in addition to potential biases in detecting soft-bodied prey using traditional methods. Even within the present study, we detected substantial variation in diets among study sites. The greater consumption of orthopterans at the Clark study site could be driven by the limited availability of lepidopterans and an increased abundance of grasshoppers in the genus Melanoplus at the Clark site (Appendix Figure 9; D. A. Haukos et al. personal observation). Melanoplus was the main genus of orthopterans used as a food across all sites. At the Clark study site, Melanoplus sanguinipes was substantially more abundant, and the roosting and morning basking of this species on bare ground may make it an easily obtainable prey item for Lesser Prairie-Chickens (Pfadt 1994, D. A. Haukos et al. personal observation).

The similar consumption of Orthoptera by Lesser Prairie-Chickens using grassland compared to cropland or CRP also doesn't provide any indication of difference in use of Lepidoptera vs. Orthoptera in grassland. Although Orthoptera composition was greatest in grassland, the RRA of Orthoptera was nearly identical to that of Lepidoptera in native grassland. Because RRA data are proportional among arthropod orders, an estimate close to 25% (split among 4 main foods) within one cover type would suggest that individuals using that cover type have more diverse diets. Although the split among the 4 orders was not perfectly uniform, Lesser Prairie-Chickens that used native grassland consumed a more diverse arthropod diet, which contrasts with our hypothesis that Lesser Prairie-Chickens would specialize on Orthopteran prey. Lesser Prairie-Chicken broods using native grassland may be opportunistic predators when diets are assessed during 0–90 days of age (Davis et al. 1980).

Despite the fact that brood diets appeared to be opportunistic when examining the brooding period as a whole, there was some indication of a nonlinear transition from Lepidoptera- to Orthoptera-dominated diets as chicks surpassed ∼40 days of age. We were particularly interested in diet during the first few weeks of a Lesser Prairie-Chicken's life. Knowledge of factors influencing survival during the first 21 days can be crucial for understanding what drives overall population growth rates (Hagen et al. 2009, McNew et al. 2012, Lautenbach 2015). The finite rate of population growth (λ) among prairie grouse and other galliformes has consistently been shown to be sensitive to variation in the 0- to 21-day-old survival bottleneck (Tympanuchus spp.; Wisdom and Mills 1997, Sandercock et al. 2008, Hagen et al. 2009, McNew et al. 2012, Taylor et al. 2012). Food availability may be particularly important for survival through this life stage, as indicated by strong variation in the mass of chicks and by observations of dead, undepredated chicks that may have died from starvation or thermal stress (Lautenbach 2015). Knowledge of the effects of food availability on chick survival is largely limited to inference from a closely related species within the subfamily Tetraoninane, the Greater Sage-Grouse (Centrocercus urophasianus). Sage-grouse chick survival can increase with the availability of Lepidoptera, slender phlox (Phlox gracilis), and total forb cover (Gregg and Crawford 2009). The influence of food availability on chick survival may contrast with the remainder of a grouse's life when there is strong support that predation poses the greater survival risk (Bergerud and Gratson 1988). However, if food availability drives passage through the most influential life stage and survival bottleneck, even if only lasting up to 21 days (the first 7 days may be most influential; Lautenbach 2015), the influence of food availability may be paramount and materialize in population level trajectories at much broader scales.

Comparative Nutrient Values of Lepidopterans and Orthopterans

Lepidopteran and orthopteran foods both provide greater concentrations of protein than any plant-based foods at the nutrient level (Lassiter and Edwards 1982, Savory 1989, Rumpold and Schlüter 2013). Protein in arthropod foods are also likely more digestible than that in plants (Stiven 1961, Savory 1989). On average, orthopterans can provide a food source that is 61% protein and 13% fat, whereas lepidopterans are 45% protein and 27% fat (Sugimura et al. 1984). Among protein estimates, there is interspecific variation and differences in digestibility. Furthermore, assimilation of protein from chitin-rich orthopterans and soft-bodied lepidopterans may be similar amid differences in nutrient composition (Sugimura et al. 1984). Mineral and amino acid composition provided by the 2 families appears to be similar, with variation among prey species (Rumpold and Schlüter 2013).

The Need for Ancillary Data

The potential benefits of using DNA metabarcoding to understand diets of wildlife species are numerous, but the current utility of the method hinges on ancillary data that can be used to constrain and evaluate the completeness of reference databases. We were unable to distinguish among certain plant foods that were from grass and crop functional groups using the primers we selected. The addition of 48 hr home range data allowed for greater inference on the use of cultivated foods. Additionally, reference DNA sequences for species that did not occur at any of the field sites sometimes matched sequences in fecal samples. To avoid inaccurate predictions, we constrained possible food sources to those detected during vegetation and arthropod surveys. The amplification of plant and arthropod DNA in only a proportion of the samples may be a problem unique to Lesser Prairie-Chickens and, potentially, other grouse species. For example, DNA was successfully amplified in all fecal samples from Louisiana Waterthrush (Parkesia motacilla), in 100% of bison (Bison bison) fecal samples, and in 74% of fecal samples from bats (Bohmann et al. 2011, Craine et al. 2015, Trevelline et al. 2016).

Plants

The predominant use of forbs as a food source during both brood-rearing and winter periods highlights the need to maintain disturbance regimes that support healthy forb populations (Hagen et al. 2004). Forbs provided a critical habitat component for Lesser Prairie-Chickens as food resources, even though they often comprised <10% of the available vegetation.

We detected greater RRA of forbs during brood rearing and winter, with specific forbs showing greater use during specific periods. During the brood-rearing period, forbs consumed by Lesser Prairie-Chickens were largely from Chenopodium- and Abutilon-like species. Chenopodium album (lamb's quarters) was present at all field sites during summer. The leaves of C. album are known to be palatable and high in calcium, which may be particularly important for growing Lesser Prairie-Chicken chicks (Adedapo et al. 2011). The use of Abutilon-like species may indicate consumption of Callirhoe involucrata (purple poppy mallow) or Sphaeralcea coccinea (scarlet globemallow), both of which were present at all sites and actively growing during the brood-rearing period (D. A. Haukos et al. personal observation). Leaves of S. coccinea are high in vitamin A, calcium, and protein and can be selected as food by scaled quail (Callipepla squamata; Ault et al. 1983, Arthun et al. 1992). Although documentation of C. involucrata as food for grassland birds is limited, the plant has adequate phosphorus and crude protein content (Odocoileus virginianus; Everitt and Gonzalez 1981). It also functions as a known larval host and food source for several butterflies (Fernandez-Canero and Gonzalez-Redondo 2010, Scott 2014). Observations were made of several caterpillar larvae on the receptacles of C. involucrate flowers at the Clark study site during the brooding period (D. Sullins personal observation). The presence of Abutilon-like plants in Lesser Prairie-Chicken diets could be from either direct or indirect consumption mediated through lepidopteran herbivory. The presence of arthropod foods can be attained only by first providing necessary host plants.

Outside of the brooding period, plant matter becomes particularly important in Lesser Prairie-Chicken diets during winter and spring as available forage decreases, thermoregulatory needs are maximized, and stored energy becomes particularly important with approaching lekking and nesting seasons (Haukos and Zavaleta 2016). Winter diets of grouse are often limited to only a few items that can provide sustenance—typically high in fiber, low in nutrient content, and requiring longer digestive tracts to process (Moss 1983). In the present study, the greater consumption of forbs compared to all other functional groups suggests a reliance on noncultivated foods in the northern portion of the Lesser Prairie-Chicken range. Use of forbs by Lesser Prairie-Chickens contrasts with grouse of more ancestral Arctic and boreal origins that largely consume woody vegetation during winter (Schmidt 1936, Moss 1983, DeYoung and Williford 2016) but is consistent with comparatively greater predation of “weed seeds” by pinnated grouse (e.g., Greater Prairie-Chickens [Tympanuchus cupido]) in comparison to Sharp-tailed Grouse (T. phasianellus; Schmidt 1936). Forb DNA was nearly absent from fecal samples collected in cropland, which suggests that current use of herbicides may reduce the availability of forbs in cropland.

Although forbs were dominant plant foods used by Lesser Prairie-Chickens during brood rearing and in winter, the relative importance of crops, shrubs, legumes, and subshrubs as food sources increased from brood rearing to winter. The amount of grass consumed remained the same, in contrast to the results of Jones (1963), who documented a slight increase in grasses consumed during winter. The increased use of shrubs and subshrubs may be related to the persistence of shrub- and subshrub-based foods during winter. Broom snakeweed was present at all study sites. This subshrub maintains green basal leaves longer into the fall and winter compared to other plants in the region, thus providing a persistent source of leafy green vegetation (Ralphs and Wiedmeier 2004). Broom snakeweed is a known food for Lesser Prairie-Chickens and has protein and nutrient content similar to green grass, but numerous secondary metabolite compounds make broom snakeweed challenging to digest (Jones 1963, Davis et al. 1980, Ralphs and Wiedmeier 2004). Although subshrubs such as broom snakeweed may not be easy to digest, they may provide a food source, persistent throughout the winter, for which grouse have evolved advanced digestive systems to procure nutrients, as indicated by seasonal changes in gut morphology (Olawsky 1987, Sedinger 1997, Donaldson et al. 2006).

Shrub-based foods can be important for Lesser Prairie-Chickens (Jones 1964, Crawford and Bolen 1976, Suminski 1977, Olawsky 1987, Riley et al. 1993) and other grouse (Patterson 1952, Remington and Braun 1985). Most research indicating that shrubs are important for Lesser Prairie-Chickens has focused on the use of sand shinnery oak where available in Texas and New Mexico, USA (Suminski 1977, Olawsky 1987, Riley et al. 1993). Sand sagebrush, sumac (Rhus spp.), willow (Salix spp.), and cottonwood (Populus spp.) have also provided food for Lesser Prairie-Chickens (Schwilling 1955, Jones 1963, 1964). The increased use of shrub-based foods during winter corresponded with the increased consumption of sand sagebrush from December to February in northwest Oklahoma, USA (Jones 1963).

Outside of using persistent winter foods in the form of shrubs and subshrubs, cultivated crops can be used by Lesser Prairie-Chickens (Salter et al. 2005). Use of cultivated legumes during winter was largely restricted to the Clark study site, where the OTU containing alfalfa (Medicago spp., 100% identity and coverage) was consumed 1.95× more than the next leading OTU containing Triticum-like species. Cultivated alfalfa was available at the Clark study site and was consumed by Lesser Prairie-Chickens in distinct cropland areas. The use of alfalfa cropland at this site may explain differences in space use among regions (Robinson 2015).

Diversity and Food Stability

The greater diversity of forage in native working grassland may be key to food and nutrient stability in Lesser Prairie-Chickens. Lesser Prairie-Chickens occur in a region with the greatest variability of net primary productivity in the Great Plains (Sala et al 1988, Grisham et al. 2016). In such a variable environment, population viability may be more influenced by a stable presence of foods from year to year than by an abundance at any one time. Various arthropod and plant taxa can boom or bust in response to years of above-average precipitation or drought, and therefore food stability may be linked to a diversity of forage (Haglund 1980, Tilman and Downing 1994, Gutbrodt et al. 2011, Craine et al. 2013). Our results indicated that native working grassland provided forage for Lesser Prairie-Chickens, in addition to providing cover for reproduction and adult survival (Hagen et al. 2013). However, in some landscapes it is possible that the presence of small-scale row-crop agriculture adjacent to grassland could diversify food options (Rodgers 2016).