Pallid bands in feathers and associated stable isotope signatures reveal effects of severe weather stressors on fledgling sparrows

Sample collection and feather measurements

During August 27–28, 2013 we captured juvenile Grasshopper Sparrows (Ammodrammus savannarum) by mist-net within a 29.3 ha grassland unit at the United States Department of Agriculture’s Grazinglands Research Laboratory (GRL; N35.555, W98.041) near El Reno, Oklahoma. As part of a standard sequence of morphological measurements, we scored each juvenile’s tail as having pallid bands (evident as reduced pigmentation and/or structural weaknesses; Fig. 1A) or having apparently normal feathers. We photographed the entire tail, removed a single outer rectrix, and stored the feather in a labeled coin envelope.

In the laboratory, we photographed each sampled feather adjacent to a 0.5 mm-scaled ruler, under fixed light sources, and against a white grid paper background. From these pictures we extracted measures of vane length, width of the pallid band, and distance from the pallid band midpoint to the feather tip (all to nearest 0.1 mm) using the program imageJ (Rasband, 2014). Nutritional deficiency can not only induce pallid bands, but also ultimately reduce overall feather growth (Murphy, King & Lu, 1988; Grubb, 2006), therefore we conducted a t-test to compare vane length of normal feathers against those containing pallid bands (vane length was scaled against the first-order principle-component score of overall body size extracted from each individual’s wing chord, tarsus, and head-bill lengths). To determine if there were age-related patterns in pallid band size and position, we also evaluated the pairwise correlations between all three measures (pallid band feathers only) using Pearson’s product-moment tests. All statistical analyses of feather/body morphology were conducted using the base functions available in R (R Core Team, 2014).

Observed pigmentation deficiencies in pallid bands often coincide with malformed feather barbules (Michener & Michener, 1938; Wood, 1950; Prum & Williamson, 2001; Møller, Erritzøe & Nielsen, 2009). Therefore, we examined and photographed select feathers under a dissecting microscope equipped with a digital camera. We qualitatively noted the physical attributes of barbules within pallid bands relative to: adjacent regions, normal juvenile rectrices, and feathers freshly-replaced during the late-summer post-juvenile molt that had been collected at a later date (September 11, 2013).

Local disturbances and grasshopper sparrow breeding phenology

Specific to the May 31, 2013 weather event, we determined the probable size of hailstones that impacted our study site by consulting National Weather Service storm reports, including: ground observation reports made through the Severe Hazards Analysis & Verification Experiment (SHAVE; Ortega et al., 2009), and hail estimates derived from weather radar data using the Maximum Estimated Size of Hail (MESH) model (Witt et al., 1998; Stumpf, Smith & Hocker, 2004). Additionally, we examined local-scale meteorology from the ‘ELRE’ Oklahoma Mesonet station at the GRL (http://www.mesonet.org/index.php/sites/site_description/elre) to determine if there were any other weather anomalies for May–July of 2013 relative to 1999–2012 norms that may have impacted Grasshopper Sparrow foraging abilities. Finally, we interviewed GRL staff regarding the management history of the study area to determine if any anthropogenic disturbance may have occurred within that grassland paddock during the 2013 breeding season.

To determine whether the timing of the May 31 storm could have realistically affected such a large proportion of the juvenile cohort at our study site, we examined existing data on the species’ nesting phenology in Oklahoma. One of the authors (DLR) had collected these data for 149 Grasshopper Sparrow nests 200 km to the northeast in Washington and Osage Counties, Oklahoma from 1992 to 1996 as part of a separate study by the GMSARC (see Rohrbaugh et al. (1999)) for the nest searching methodology. Our preliminary analysis indicated that clutch size significantly declined over the course of the breeding season (see Fig. S1). Therefore, we restricted our use of these data to 104 nests with both final clutch size plus dates of clutch initiation (either directly observed or extrapolated based on incubation stage at discovery), and classified the clutches according to their expected status on May 31. For Grasshopper Sparrows, a typical clutch will hatch approximately 14 days after the first egg is laid and the young fledge after 11 days in the nest (Vickery, 1996). According to this schedule, by the afternoon of May 31, clutches initiated prior to May 7 would likely have fledged, clutches initiated May 7–17 would likely be nestlings, and clutches initiated May 18–31 would still be eggs (Table 1). We used a z-test implemented in R (R Core Team, 2014) to compare the proportion of juveniles captured during late August 2013 that showed synchronous pallid bands against the proportion of a typical cohort which would have been expected to have hatched prior to May 31 and were fully independent by late August.

Stable isotope analysis

To prepare feathers for analysis, we first cleaned each with dilute detergent and then a 2:1 chloroform–methanol solution (Paritte & Kelly, 2009) and dried the samples at room temperature. We then sectioned the distal end of each feather into four or five 0.25–0.40 mg portions (at least two sections grown after (i.e., proximal) and all sections preceding the pallid band (Fig. 1B)) and packed each section into a 3.5 × 5 mm tin capsule for insertion into an autosampling tray. We contained the pallid band, if present, within only one of these sections. We performed stable isotope analyses using a Thermo Delta V Plus isotope ratio mass spectrometer (Thermo Scientific, Bremen, Germany) connected to a Costech ECS 4010 elemental analyzer (Costech Analytical Technologies, Valencia, California, USA) in the Stable Isotope Laboratory at the University of Oklahoma School of Geology and Geophysics. We measured isotopic ratios for nitrogen (δ¹⁵N) and carbon (δ¹³C) to the nearest per mil (‰), and included as in-run standards multiple samples from the same feather of an adult Brown-headed Cowbird (Molothrus ater).

We used R (R Core Team, 2014) and package ‘lme4’ (Bates, Maechler & Bolker, 2012) to analyze the relationship between δ¹⁵N and, separately, δ¹³C relative to the presence or absence of a pallid band using generalized linear mixed models (GLMM; Bolker et al., 2009). We conducted separate GLMM analyses: (a) among groups according to pallid band presence, (b) among sections of individual feathers according to pallid band presence, and (c) among two-section regions of individual feathers that were grown during and immediately following pallid band development (i.e., the pallid band region). We included in each analysis a variable indicating the presence of a pallid band for each individual, section, or region (respectively) as the only fixed effect, and bird ID and feather section number as random effects in a random intercepts model. We examined residual plots for deviations from homoscedasticity or normality, and obtained p-values by likelihood ratio tests of each respective full model (i.e., pallid band as a fixed effect) versus its reduced/null model (i.e., without pallid band as a fixed effect; Bolker et al., 2009). Finally, we used the partimat function in the R package ‘klaR’ (Weihs et al., 2005) to calculate and visually represent a principal components analysis of δ¹⁵N and δ¹³C among feather sections grouped according to whether the source feather had a pallid band. The dividing function was based upon a linear discriminant function analysis, and an error rate was calculated based on the number of sections that were incorrectly assigned (i.e., positioned on the opposite side of the dividing line from its group mean).

Pallid band incidence and feather characteristics

Grasshopper Sparrow juveniles captured from the GRL population displayed a very high incidence of synchronous pallid bands in the tail (11 of 25 individuals; 44%). Among the 25 juveniles examined, we noted fault bars on the wing feathers for only one individual, and these were the narrow, “true” fault bar type and were slightly asynchronous (i.e., not uniformly positioned; see Fig. S2). The pallid bands we observed among juvenile Grasshopper Sparrow rectrices were up to 3.4 mm wide, aligned across all of the tail feathers, and showed modestly-reduced pigmentation and barbule density (Figs. 1A and 3). We found that mean (±s.d.) vane length of the sampled rectrices, corrected for overall body size, did not differ between feathers having pallid bands (41.0 ± 1.1 mm) versus those sampled from normal-appearing individuals (41.0 ± 1.6 mm; t = 0.853, p = 0.406). Among feathers having pallid bands, the mean distance from band midpoint to feather tip was 8.8 ± 5.0 mm and the mean pallid band width was 1.8 ± 0.9 mm. The distance of pallid band midpoints to feather tips, as a proportion of vane length, ranged from 0.06 to 0.42 with a mean of 0.21 ± 0.12 (s.d.). Pallid band width significantly increased as a function of midpoint distance from feather tip (r = 0.928; t = 7.45; p < 0.001), but was not related to vane length (r = 0.018; t = 0.05; p = 0.958). The midpoint locations of pallid bands were likewise not correlated with vane length (r = − 0.025; t = − 0.075; p = 0.942). We observed that the structure of feather barbules was moderate-to-severely degraded in pallid bands, in extreme cases showing large sections of unhooked or entirely missing barbules (Fig. 3).

Local disturbances and breeding phenology

Based on data from both weather radar and the ‘ELRE’ Oklahoma Mesonet station at the GRL, we estimated that hail fell over the GRL during approximately 16:00–16:20 local time on May 31, 2013. Hailstone diameters estimated from weather radar data using the MESH model ranged from 4.45 to 5.72 cm at the GRL (Fig. 2). These estimates likely accurately reflect the actual hailstone sizes, since nearby ground-based observations of maximum hailstone sizes reported through SHAVE consistently matched or modestly exceeded the MESH estimates (see Fig. S3).

Local meteorological data also indicated that during the period when juveniles that were captured in August would have likely been in the nest or recently fledged (i.e., May 1–July 13, 2013) the amount of precipitation recorded at the GRL exceeded the site’s 1999–2012 mean by 22.4 cm, representing 150% of the normal rainfall for this period. Perhaps not surprisingly, this exceedance was primarily caused by a 12.1 cm downpour during the May 31 storm. No other daily precipitation totals exceeded 4.1 cm during this period. There were no other obvious stressors associated with the peak Grasshopper Sparrow reproductive period. At no point between May 1 and July 13, 2013 did the maximum or minimum temperature depart more than ±5 °C from 1999 to 2012 means. Other possible stressors, such as local land use and management, were similarly unlikely. During 2013, our grassland study site at the GRL was grazed lightly by cattle (∼1 per 10 ha) and the area was not managed with herbicides, pesticides, or mowing treatments during May–July (S Coleman, USDA, pers. comm., 2013)

The median date of clutch initiation observed during GMSARC surveys in Oklahoma from 1992 to 1996 was June 4, though when temporal variation in clutch size was accounted for, this median became May 31. Based on those surveys, we also estimated that among Grasshopper Sparrow offspring that would have been fully independent by late August (provided all equally hatched and survived; n = 394), 25.6% would have hatched or fledged and 30.7% would have been eggs in the nest as of May 31 (Table 1). The former value was significantly lower than the percentage of juveniles captured at the GRL in August 2013 showing pallid bands across their rectrices (44%; n = 25; z = − 2.03; p = 0.042).

Stable isotopes

The mean stable isotope ratios in feathers showing pallid bands were 2.2‰ lower with respect to δ¹⁵N but 1.2‰ higher for δ¹³C when compared to normal appearing feathers (see Table S1). After correcting for variation among individuals and feather sections, the difference was significant for δ¹⁵N [χ²(1) = 5.90, p = 0.015] but not for δ¹³C [χ²(1) = 1.55, p = 0.213]. The discriminant function analysis partition plot indicated that only 19.4% of sections were misassigned relative to their true group (pallid band or normal; Fig. 4).

Sections containing pallid bands did not differ from other parts of the feather for either δ¹⁵N [χ²(1) = 1.10, p = 0.295] or δ¹³C [χ²(1) = 0.35, p = 0.553]. However, if we paired data from sections containing pallid bands with those subsequently grown (i.e., immediately proximate), then relative to the remaining sections, this pallid band region contained significantly elevated δ¹⁵N [χ²(1) = 4.81, p = 0.028], while δ¹³C remained undifferentiated [χ²(1) = 1.96, p = 0.162]. On average, the presence of a pallid band was associated with an increase in δ¹⁵N of about 0.37 ± 0.16 (S.E.)‰, which falls well above the 0.09‰ error rate calculated from independent runs of our Brown-headed Cowbird feather standard.

Stable isotopes as records of environmental stressors

Nitrogen isotopes within the pallid band region were significantly enriched in ¹⁵N relative to other parts of the same feathers. These spikes in δ¹⁵N support our predictions and are consistent with increased muscle catabolism as part of the stress response that produced the pallid bands. There was also notable among-individual variation both in the slope and magnitude of δ¹⁵N and δ¹³C along the feather vane, which suggested differences in diets during development consistent with young originating from different nests. This lends confidence that the sampled individuals suitably represented the population variation.

The magnitude of differences between feathers from juveniles possessing normal feathers versus juveniles displaying pallid bands for both δ¹³C (−1.2‰) and δ¹⁵N (2.2‰) would be consistent with trophic or successional (i.e., C4 to C3 plant community) shifts that one might expect among individuals growing feathers at different times during a temperate breeding season (Kelly, 2000). More specifically, if we assume that juveniles without pallid bands were reared after the large scale disturbance, then the data are consistent with a scenario in which the diets provisioned to Grasshopper Sparrow nestlings and fledglings shifted as the season progressed. Increased δ¹⁵N in sampled rectrices of normal appearance could be attributed to ingestion of insect prey items from higher trophic levels (e.g., larger, omnivorous grasshoppers), and the difference in δ¹³C could result from an increasingly C3 plant base (Hobson, 1999; Kelly, 2000; West et al., 2006). This trophic progression may even be evident within individual feathers where δ¹⁵N naturally increases from tip-to-root, as observed in this study and by Symes & Woodborne (2011) in White-bellied Sunbirds (Cinnyris talatala).

Pallid bands as a response to stress

High rates of fault bars have been reported in other species (Hawfield, 1986; Bortolotti, Dawson & Murza, 2002; Møller, Erritzøe & Nielsen, 2009), but such accounts appear entirely restricted to the narrow (i.e., ≤1 mm) “true” fault bars that we earlier distinguished from pallid bands. The incidence of pallid bands within a population appears to be much rarer (Michener & Michener, 1938; Erritzøe & Busching, 2006). In fact, only 1.5% of 271 Grasshopper Sparrow juveniles captured at 22 other sites in Nebraska, Kansas, and Oklahoma during 2013–14 had pallid bands similar to those observed in such prevalence at the GRL in 2013 (WA Boyle, 2013–14, unpublished data). The underlying cause of the high incidence of pallid bands at the GRL was, therefore, likely a rare event that was sufficiently intense and widespread to affect much, if not all, of the local Grasshopper Sparrow breeding population; most probably the severe weather outbreak of May 31, 2013.

The emergence of pallid bands appears to follow a strong linear pattern between position along the length of the feather and width. If pallid bands in our study population all represented the same period of growth, then this pattern would indicate that feather growth linearly increased as feathers became longer. However, Elderbrock, Kern & Lynn (2012) found that although width of growth bars down the vane of individual feathers in juvenile Eastern Bluebirds (Sialia sialis) did vary substantially, these differences were randomly located and did not linearly increase with position down the feather vane. Instead, the authors noted that growth bar width was disconnected from the rate of feather elongation, which was constant during the development of individual feathers. Rather than feather growth increasing with age, our observed patterns in pallid band width relative to distance from the tip more likely indicate that the duration of the stress response had linearly increased as a function of chick age. This inference agrees with prior studies showing that older chicks are more prone to an extended stress response due to increased development of the hypothalamic-pituitary-adrenal axis (Sims & Holberton, 2000; Sockman & Schwabl, 2001; Wada, Hahn & Breuner, 2007). Moreover, the increasingly-independent fledglings are regressively provisioned by the adults while simultaneously experiencing greater metabolic demands and, therefore, are usually closer to starvation than younger chicks (Blem, 1975). In at least three prior studies of age-related patterns and true fault bars this relationship was reversed (i.e., older chicks displayed fewer fault bars; Machmer et al., 1992; Jovani & Tella, 2004; Jovani & Diaz-Real, 2012) again suggesting that the two types of fault bars fundamentally differ in causation.

Ecological and evolutionary implications of pallid bands

The narrow and abrupt feather malformations we previously referred to as “true” fault bars are prone to breakage (Sarasola & Jovani, 2006), which affects flight performance (Murphy, Miller & King, 1989; Norberg, 1990; Jovani et al., 2010) and may explain why Goshawk (Accipiter gentilis) prey had significantly higher-than-average incidences of fault bars relative to a random sample of the same species from the same general area (Møller, Erritzøe & Nielsen, 2009). Jovani & Blas (2004) argued that, in light of this, true fault bars should be less likely to occur among feather tracts that are strongly tied to flight performance, because evolutionary pressures would have favored individuals that were capable of inherently allocating greater resources to feathers critical for flight (i.e., the “fault bar allocation hypothesis”). Though pallid bands may be less likely to represent a breakage risk, they do contain structural weaknesses that could affect flight performance and durability (Fig. 3). Among the Grasshopper Sparrows examined in this study, the appearance of pallid bars exclusively in the rectrices with only one example of true fault bars in the remiges (see Fig. S2) suggests that the fault bar allocation hypothesis may apply to pallid bars as well, and is worthy of further investigation. It may be the case that ground nesting birds like Grasshopper Sparrows, which are frequently exposed to environmental stressors (Nice, 1957; Ricklefs, 1969), would benefit by coupling fault bar/pallid band allocation with their complete late-summer post-juvenile molt (Pyle, Jones & Ruth, 2008) as an adaptive response to stress during early development (Pap et al., 2007).

Biological relevance of severe weather

Local environmental conditions dictate whether species survive and reproduce successfully in any given area. Lack (1966) suggested that extremes in local environments would be strongly responsible for limiting species, as these would present the most grievous of stressors. Severe weather is a prime example, as it can cause widespread mortality and is known to necessitate specific local adaptations within biological communities (Wingfield, 1988; Newton, 2007). Severe weather impacts on birds have generally been studied in association with large, widespread, and relatively long-lasting events such as cold snaps, hurricanes, blizzards, and regional storm fronts (Whitmore, Mosher & Frost, 1977; Wiley & Wunderle, 1993; Brown & Brown, 1998; Newton, 2007; Frederiksen et al., 2008; Rittenhouse et al., 2010; Streby et al., 2015). However, intense but relatively localized perturbations, such as large hail or tornadoes associated with severe thunderstorms, have primarily received attention only after the most obvious and widespread destruction (e.g., Gates, 1933; McClure, 1945; Smith & Webster, 1955; Narwade et al., 2014). Yet, severe thunderstorms can vary in their degree of impact across species and, thus, may represent a particularly relevant factor in regulating biological community composition. Considering that much of the American Great Plains experiences severe thunderstorms and hail annually (Doswell, Brooks & Kay, 2005; Cecil & Blankenship, 2012; Cintineo et al., 2012) and that hailstones as small as 1 cm diameter can destroy eggs and injure adult birds (Hanford, 1913; Graul, 1975; Hume, 1986; Narwade et al., 2014), these weather events are likely to have profound ecological and conservation relevance to grassland species, especially ground-nesting birds. This is especially true during the key breeding period of April–July, when vulnerable adults, eggs, and young face the peak of the severe thunderstorm season.

In our study, Grasshopper Sparrow juveniles with ‘normal’ rectrices appear to show the stable isotope signatures of hatching after May 31. Our analysis indicated that such individuals were significantly less abundant than expected based on the species’ nesting phenology in Oklahoma (i.e., 56% in 2013 versus 74.4% in 1992–96). This could be due to slight climatic differences between study areas, such as the date of last spring freeze (Oklahoma Climatological Survey, 2014), or because of 20-years of climate change. Alternatively, the lower proportion of late-hatching young in 2013 could have reflected an actual net loss among the latter half of the 2013 cohort, in this case ∼25% relative to the 1992–96 demography. It is entirely reasonable to expect that hailstones which exceeded 4 cm in diameter could have led to widespread destruction of eggs (an estimated 30% of a typical cohort on May 31), or that this stressor could have disrupted late-season nesting attempts (e.g., adults directly killed or abandoned the area after the storm).

Scientifically assessing the biological impacts of severe weather might be viewed as being limited by the ability to predict in advance where storms will strike, precluding before-and-after comparisons. However, every year there are likely field studies occurring at points throughout regions where severe weather is likely to strike (Doswell, Brooks & Kay, 2005; Cecil & Blankenship, 2012; Cintineo et al., 2012) and, so, a broadly-distributed network of researchers could readily be assembled to study large-scale ecological consequences of severe weather. We call for the ecological research community to take advantage of severe storm events as they occur at their research sites, opportunistically sampling biologically informative data required to assess the nature and magnitude of such stressors on animal communities. In the face of a changing climate and the expected shifts in severe weather regimes (Trapp et al., 2007; Goodess, 2013) there is a need to expand our knowledge of the current and future ecological impacts of severe weather events (Jentsch, Kreyling & Beierkuhnlein, 2007), so that we may work toward mitigating losses among vulnerable species and biological communities.