Study site
Our study was conducted at the Nevado de Colima National Park (NCNP) in Jalisco, Mexico. Nevado de Colima is an inactive high-altitude volcano (4260 m a.s.l.) located at the western end of the Trans-Mexican Volcanic Belt (19° 33′ 45″–19° 30′ 40″ N, 103° 36′ 30″–103° 37′ 30″ W; INEGI 2011). The climate in the region is highly seasonal (CONANP 2006, 2017). Our study site was located at 3194 m a.s.l. and consisted of subalpine scrublands (dominated by plants of the genus Salvia, Ribes, and Festuca), some scattered alders (Alnus) on exposed ridges, and pine and fir forests (Pinus and Abies) located in ravines (Schondube 2012). The most important flowering plants that hummingbirds feed on include Salvia elegans, S. gesneriflora (Lamiaceae), Ribes ciliatum (Saxifragaceae), Senecio angulifolius (Asteraceae), and Penstemon roseus (Plantaginaceae) (Schondube 2012).
Fieldwork
We sampled hummingbirds three times over a one-year period. Our sampling corresponded with the three climatic seasons in the region, including (1) a rainy season from June to October, (2) a cold-dry season from November to February, and (3) a warm-dry season from March to May (CONANP 2006, 2017). We selected this sampling scheme because weather conditions and the availability of floral nectar and arthropods varied widely among these three seasons (CONANP 2006, 2017). We sampled during May and September 2016, and February 2017. During each sampling period, we captured hummingbirds using 10 mist-nets (12-m long, 24-mm mesh) for three consecutive days. Mist-nets were opened at sunrise and closed after 6 h. Net rounds were conducted every 5 min. We located the mist-nets within a 2 ha area, and their location remained constant during the study. We identified all captured birds and recorded their age and sex using plumage characteristics and bill striations (Williamson 2001; Howell 2002; Russell et al. 2019). Additionally, for the Mexican Violetear, the only species that did not present a clear sexual dimorphism in plumage at our study site we used wing chord length and bill tip serrations to determine their sex (Rico-Guevara et al. 2019). Data on wing chord differences among sexes for this species was obtained from a 30-year hummingbird banding program located in the same region. We define females as those individuals with a wing chord < 60 mm, and males as those individuals with a wing chord > 63 mm (Contreras-Martínez personal communication).
Stomach content analysis
We obtained the stomachs of hummingbirds collected as part of an independent stable isotope project conducted at our study site (n = 6 in May 2016, n = 8 in September 2016, and n = 37 in February 2017). That study collected blood, liver, pectoral muscle, and bones to extract collagen, and allowed us to use the stomachs. The remaining feathers, skulls, and tongues were deposited at the collection of the Functional Ecology Laboratory of IIES, UNAM. Samples were collected with permission from the Secretaría de Medio Ambiente y Recursos Naturales, Mexico (SGPA/DGGFS/712/2767/14). All collected birds were humanely euthanized by carefully placing their heads inside a small vial that contained a cotton ball soaked in isoflurane, following the guidelines to the use of wild birds in research (Fair et al. 2010), and their stomachs were placed in plastic vials with saline solution (0.8% NaCl) and frozen in liquid nitrogen until processed in the laboratory. The time between hummingbird capture in the nets and stomach freezing was less than 20 min. Because soft arthropods require 3–4 h to be digested completely by hummingbirds (Remsen et al. 1986), this time frame allowed us to quantify stomach arthropod content at the moment of capture. The species sampled were: Mexican Violetear (Colibri thalassinus), Amethyst-throated Mountaingem (Lampornis amethystinus), White-eared Hummingbird (Basilinna leucotis), Rivoli’s Hummingbird (Eugenes fulgens), Broad-tailed Hummingbird (Selasphorus platycercus), and Rufous Hummingbird (S. rufus). The number of individuals collected at each season differed due to variation in capture rates among seasons, and due to restrictions on collecting permits (maximum of 10 individuals per species per season).
We analyzed hummingbird stomachs in the lab to determine the number of grit particles they contained. Stomachs were thawed and dissected, and their contents removed. We quantified grit particles by carefully separating them from the arthropod remains in hummingbird stomachs using a stereoscopic microscope (AmScope, 7–45 × binocular stereo zoom microscope). We described the color and shape, and weighed and measured grit particles. To determine their size, we determined the grit area by taking a picture of each grit particle on top of a millimetric grid. Images were analyzed using ImageJ (National Institute of Health, NIH Version v1.32j). Because the role of grit as either a grinding agent or nutritional supplement depends upon its hardness and solubility in the digestive tract (Meinertzhagen 1954; Myrberget et al. 1975; Gionfriddo and Best 1999), we determined grit hardness. We did this by pressing each grit particle twice with fine-point reverse action tweezers. This kind of tweezer allowed us to generate a standard pressure on the grit particle and separate them into two hardness categories: hard (did not break) and soft (did break into smaller pieces).
Physical and chemical characterization of grit particles
We used scanning electron microscopy (SEM) and energy dispersive X-ray (EDX) spectroscopy (Tabletop Microscope Hitachi, Model TM3030Plus) to perform the physical and chemical characterization of the grit particles. Because soft grit broke into tiny particles, we were only able to analyze the hard grit particles. Of these, due to the limitations imposed by the method, we were able to analyze only the largest hard grit particles (diameter of 0.5–1 mm, n = 5). Analyses were carried out in the Microanalysis Laboratory of the Geophysics Institute, UNAM.
Ingested arthropod biomass and chitin content
We separated all arthropod remains into those that were identifiable (arthropods partially fragmented) and those that were not (very fragmented arthropods) using a stereoscopic microscope. We identified recognizable arthropods to the taxonomic level of order following Triplehorn and Johnson (2005). To determine the biomass of ingested arthropods (g dry mass), we dried both identifiable and unidentifiable arthropod samples at room temperature for 3 h and then weighed them using an analytical balance (Ohaus Adventurer, capacity/readability of 110 g × 0.001 g). Chitin content (percent dry mass) of the different identified arthropod orders was obtained from Rothman et al. (2014). We estimated the mean chitin content of arthropods by averaging the percent chitin content of the different arthropod orders present.
Data analysis
We used different analyses to test our main hypotheses on the use of grit particles by hummingbirds. First, we compared the number of grit particles present in hummingbird stomachs among species using a Kruskal–Wallis test (Zar 1999). We used a non-parametric test because our data did not present a normal distribution. In view of the fact that we did not find grit particles in the stomachs of hummingbirds during the warm-dry season, and due to the small sample size of the rainy season, this analysis was limited to our data for the cold-dry season.
Second, we used a generalized linear model (GLM) to assess whether the number of grit particles (response variable) varied among seasons, and between age and sex classes. In this model, we include season (warm-dry, rainy, and cold-dry seasons), age (adult and juvenile individuals), and sex (female and male individuals) as categorical explanatory variables. We select for this analysis the data of all 25 male and 21 female individuals sampled, excluding data from the 5 individuals whose sexes were unknown. Due to our small sample size, we created a 0–1 binary response variable in which 0 represented the absence of grit and 1 represented the presence of grit, and fitted our model with a Binomial distribution and a logit link function. Additionally, we used the adjusted maximum likelihood estimator for reducing biases of the Binomial logistic regression parameters (following Firth 1993).
Third, because the two response variables of arthropod ingestion (i.e. the biomass of arthropods ingested and their chitin content) presented different distributions, we ran two GLMs with the number of grit particles as the explanatory variable, whereas the response variables differed. For GLM 1, we used the biomass of ingested arthropods (g dry mass) as the response variable and a normal distribution with an identity link function. For GLM 2, we used the chitin content of arthropods ingested by hummingbirds (percent dry mass) as the response variable, and a Poisson distribution with a Log link function. In both models, we only included the data from those individuals whose stomachs presented grit particles (n = 12 for GLM 1, and n = 8 for GLM 2; Additional file 1: Table S2).
Finally, we performed some tests to search for differences in the arthropod content and grit particles characteristics between individuals of both sexes. Since we only had a male individual, we use the data of the females to construct a confidence interval to compare it against the male values using a one-sample t-test for those variables with normal distributions (number of grit particles and the biomass of ingested arthropods), and a Wilcoxon test for non-normal distributed data (size of grit particles; Sokal and Rohlf 1995). We conducted all analyses using JMP version 9.0 (SAS Institute Inc.). Values are provided as means ± SD.