Differences in insectivore bird diets in coffee agroecosystems driven by obligate or generalist guild, shade management, season, and year

Study sites

Our study area encompassed ~5,000 ha that was originally tropical lower montane and premontane moist and subtropical lower montane and montane wet forest. The area was developed for coffee production in the early 1900s, and the landscape is now ~90% coffee agroecosystems and ~10% small (<2 ha) forest fragments (Philpott et al., 2008b). The topography of the region is rugged, steep, and mountainous. All farms are located between 950–1,150 m asl and receive ~4,200–5,000 mm of rain annually. The region is strongly seasonal, with a wet season that typically lasts from May to October and a dry season from November to April.

We worked at five sites in three coffee farms (300-ha) in the Soconusco region of southwest Chiapas, Mexico, including Finca Belén, Finca Irlanda, and Finca Hamburgo. The five sites represented a gradient of coffee management intensity and differed in the structure of canopy vegetation and coffee plants (Mas & Dietsch, 2003; Philpott, Perfecto & Vandermeer, 2006). Based on a common classification scheme (Moguel & Toledo, 1999), TP stands for traditional polycultures, CP are commercial polycultures, and SM is a shaded monoculture dominated by Inga spp. The five sites were: (1) Belén Rustic (TP1; 15°, 15′N, 92°, 22′W), (2) Belén Production (CP1; 15°, 15′N, 92°, 23′W), (3) Irlanda Restoration (TP2; 15°, 10′N, 92°, 20′W), (4) Irlanda Production (CP2; 15°, 11′N, 92°, 20′W), and (5) Hamburgo (SM; 15°, 10′N, 92°, 19′W). TP1 is an area where the natural canopy vegetation has largely been maintained, and understory shrubs were replaced with coffee plants. TP2 is an area of active forest restoration where some native and some productive trees, including Eriobotria japonica (loquot), Manguifera indica (mango), Citrus sinensis (orange), Persea americana (avocado), and Terminalia amazonia (a timber species) have been planted with the intent of restoring the area to forest. All other sites included varying densities and diversities of shade trees, but the most common canopy trees were Inga spp., Alchornea latifolia, Trema micrantha, and Conostegia xalapensis. Vegetation data collected in 1998 at all study sites (Mas & Dietsch, 2003) and at all sites except TP2 (Philpott, Perfecto & Vandermeer, 2006) from 2000 to 2002 revealed differences between sites in canopy cover, tree species richness, vegetation structural depth and complexity, coffee bush density and height, and overall management intensity, with a general trend of increasing coffee management intensification such that TP1 < TP2 < CP2 < CP1 < SM (Table S1). All sites were certified organic with the exception of SM.

We collected data during three wet seasons and three dry seasons from 2001 to 2003. In our study sites, the typical dry season period falls between early November and late April while the typical wet season falls between early May and late October. Dry season data (when Neotropical-Nearctic migratory and resident birds were present) were collected from 11 to 25 January 2001, 29 November to 13 December 2002, and 12 to 24 March 2003. Wet season data (when Austral migratory and resident birds were present) were collected from 5 to 28 June 2001, 23 to 28 July 2002, and 19 to 28 July 2003. Although the dates differ by year, none of the sample dates fell exactly at the start or end of the typical dry or wet season for our sites.

Bird diets

We captured birds in mist-nets placed in each study site. Mist-nets (four to six 12-m, 30-mm mesh nets) were placed in the same location on consecutive days and moved after 2–5 d if capture activity dropped noticeably. Nets were placed at ground level between rows of coffee. All mist-netting was conducted between 08:30 and 12:00; on a few overcast days, mist-nets remained open until as late as 15:00. Research methods were approved by the University of Michigan’s Animal Care and Use Program Permit Number 7499 and Mexico’s Secretaría de Medio Ambiente y Recursos Naturales Permit Number 10092.

We processed all arthropod-consuming birds captured that could handle the emetic used to obtain diet samples (exceptions included small to medium-sized hummingbirds, other than Violet Sabrewings (Campylopterus hemileucurus), and raptors). However, if too many birds were captured to process them in a timely manner, we generally prioritized the most frequently captured species (Table 1) to ensure sufficient sample sizes.

We used liquid inserted into bird stomachs to induce regurgitation and stomach flushing. In 2001, we administered a 1% emetic solution of antimony potassium tartrate per 100 g of body mass (Poulin, Lefebvre & McNeil, 1994). Immediately after capture, we gave birds 0.8 cc per 100 g of body mass, although the amount administered was reduced to 0.6 cc per 100 g of body mass starting in June 2001 because it was sufficient to induce regurgitation. In 2002 and 2003, we used between 2–12 ml (depending on the size of the bird) of unflavored Pedialyte (water, dextrose, and <2% of potassium citrate, salt, sodium citrate, citric acid, and zinc gluconate) as rehydration with electrolytes is recommended for dehydrated songbirds (Welte & Miller, 2019). Some authors have suggested that stomach flushing is an invasive procedure for birds (e.g., Zach & Falls, 1976, Johnson et al., 2002, Carlisle & Holberton, 2006). However, birds in our study experienced few obvious deleterious effects while being handled. One bird died (of >900 birds processed), but ~95% of processed birds flew away immediately after processing and some individuals were recaptured (but not processed) on subsequent sample days or seasons; other birds required <5–15 min to recuperate before flying away. During all years, we introduced liquid into bird stomachs with a lubricated catheter tube inserted into a 1.0-cc or 0.5-cc syringe. We used two different sizes of tubes (5 Fr and 8 Fr feeding tubes; Kendall Company, Mansfield, MA, USA) and selected tubes based on bird size. We also collected any fecal matter left in bird-processing bags and added this to stomach content samples. We transferred stomach content and fecal samples to vials filled with 96% ethanol with a micro-spatula and stored vials for arthropod identification. Vials were stored for 1 to 2 years before identification. We clipped the tip of the outer tail feathers to assure the same individual was not processed twice. We released birds immediately after processing.

We identified all captured birds to species, and later classified birds based on migratory status, i.e., year-round residents, Nearctic-Neotropical migrants at the study site during the dry season, or Austral migrants that breed at the study site during the wet season (Billerman et al., 2020). We also classified birds by foraging guild, i.e., obligate insectivore, generalist insectivore that also eats fruit (also known as omnivores), or generalist insectivore that also eats nectar based on previously used classifications from data collected in Neotropical coffee farms (Greenberg, Bichier & Sterling, 1997). Finally, we classified species based on their primary foraging strata in coffee agroecosystems, including canopy, open scrub (birds that forage in shrubs in open areas), trunk, and understory (Greenberg, Bichier & Sterling, 1997). Only species with at least four individuals sampled were included in our analysis.

Arthropod identification

Arthropods were identified by two individuals (SML and MLB). We searched for whole bodies as well as body parts to aid in identification (e.g., heads, mandibles, elytra, and wings). Arthropods were identified to order, superfamily, family, or genus as allowed by either whole bodies or body parts. We counted the number of individuals of each family or genus per sample if we viewed entire arthropod bodies or by reconstructing the number of individuals based on combinations of component body parts. Arthropods and arthropod parts with harder exoskeletons (e.g., mandibles, heads, and elytra) are more likely to be preserved and identified from stomach content samples whereas soft-bodied arthropods such as spiders (Araneae), flies (Diptera), and butterflies and moths (Lepidoptera) may be less likely to be preserved and harder to identify from stomach samples (Rosenberg & Cooper, 1990; Sherry et al., 2016).

Data analysis

To examine possible differences in diet composition, we performed permutational ANOVAS (PERMANOVAs) with the multivariate statistical software package PRIMER-E 7.0.13 (Clarke, 1993; Quest Research Limited, 2017). Similar to other studies visually identifying prey from avian insectivores (e.g., Poulin & Lefebvre, 1996; Sherry et al., 2016), we classified dietary items into the following groups: Araneae, Auchenorrhyncha, Coleoptera, Diptera, Formicidae, Heteroptera, Other Hymenoptera (other than Formicidae), Lepidoptera, Orthoptera, and Infrequent (including infrequently encountered items such as Acari, Psocoptera, Sternorrhyncha, and Thysanoptera). We square-root transformed abundance of dietary items in each bird sample before analysis then created a Bray–Curtis similarity resemblance matrix and used PERMDISP (within PRIMER-E) to test each factor for homoscedasticity.

We ran a PERMANOVA where season, year, guild, and migratory status (resident or migrant) were fixed variables. To determine if diets of Nearctic-Neotropical migrants and residents differed during the dry season, we selected only dry season data and ran a PERMANOVA with year, guild, and migratory status as fixed variables. To determine if resident birds foraged differently in the dry and wet seasons, we selected only resident birds and ran a PERMANOVA with year, season (Dry or Wet), and foraging guild as fixed variables. To determine if coffee farm management affected avian foraging, we analyzed data for all species sampled across the three production systems with at least two samples in each production type (N = 4) during the wet season of 2001 when SM was well sampled. We ran a PERMANOVA with production management (CP, TP, and SM) as a fixed effect and bird species as a random effect.

PERMANOVAs were run for 999 permutations. Pseudo-Fs and permutational P values are reported in the text. Terms found to be significant with the main PERMANOVA model were evaluated with pair-wise comparisons where pseudo-t values are reported. We used the SIMPER procedure (in PRIMER-E) to analyze individual contributions to significant terms.

Diet descriptions

Our first research objective was to identify the major arthropod prey in the diets of obligate and generalist insectivorous birds. Among resident birds, those that were obligate insectivores consumed more spiders (Araneae) and beetles (Coleoptera), whereas generalist insectivores consumed more ants (Formicidae) and Other Hymentoptera prey that may be easier prey to find due to clumped distributions (e.g., Johnson, 2000; Sherry et al., 2016). Furthermore, Formicidae may be lower quality prey because of their chemical defenses (Eisner, Eisner & Siegler, 2005) and low nutritional value (Zach & Falls, 1978). All arthropod taxa documented in the diets of 30 of 39 bird species in our study (Araneae, Auchenorrhyncha, Coleoptera, Diptera, Formicidae, Heteroptera, Other Hymenoptera, and Lepidoptera) have also been reported as being important prey in other studies conducted in the Neotropics (e.g., Poulin & Lefebvre, 1996; Sherry et al., 2016). Notable is that several foraging studies document removal of these same arthropods solely from foraging observations (e.g., Wunderle & Latta, 1998; Jones et al., 2000; Newell et al., 2014). Moreover, excluding birds from coffee plants and shade trees in coffee farms sometimes leads to reduction in many of these same taxa compared to control plants to which birds have access (Greenberg et al., 2000; Philpott et al., 2004), but this is not always the case (Johnson et al., 2009; Karp & Daily, 2014). These results thus confirm some of what we know about the diets of insectivorous birds and emphasize the need for direct measurement of diet items, rather than inferences from bird foraging or removal experiments.

Annual variation

Our second research objective was to determine how diets might change over time or with season, and we found significant annual and seasonal differences in bird diets. We found that bird foraging guilds differed in their consumption of prey taxa from year to year and the diets of resident birds differed by guild and year. In general, significant annual effects on diets of insectivorous birds may reflect annual differences in resource availability, fluctuating precipitation, and changing migration patterns (McNamara et al., 2011; Studds & Marra, 2011). At our sites, differences in the diets of migrants and residents in the dry season were most apparent in 2003, and this was the same year we observed differences in the diets of Austral migratory YGVI and resident birds in the wet season. Differences in the diets of migrants and residents may be illustrative of behavioral adaptation to the seasonal changes in avian densities where resource competition may fluctuate annually in response to ecological conditions, such as changes in arthropod populations resulting from high precipitation (Keast & Morten, 1980; Greenberg, 1995). More specifically, shifts in avian diets over time are likely an opportunistic adaptation by both obligate and generalist insectivores to seasonal fluctuations in prey availability that may be exacerbated by wet and dry years of the ENSO cycle (Sillett, Holmes & Sherry, 2000). Significant differences between resident obligate and generalist insectivores in the wet seasons of 2001 and 2002 were absent in 2003 which may be due to precipitation differences. Precipitation data indicate that 2003 was an abundantly wet year when 3,446 mm of precipitation fell in Tapachula municipality compared to 3,057 mm in 2002 (Thornton et al., 2017). During years of less rainfall, arthropod populations may be reduced and resource competition for food resources among insectivorous birds may be high (e.g., Tilman, 1982); in wetter years, greater prey abundance may mitigate competition from interspecific dietary overlap (Sherry et al., 2016).

Although annual effects were important, diets of Nearctic-Neotropical migrants and residents were dissimilar across all years but unequal homogeneity of migrant vs resident dispersions means that we cannot confirm these differences are a result of the groups foraging differently (Warton, Wright & Wang, 2012). Poulin & Lefebvre (1996) reported little overlap in the diets of migrants and residents in Panamanian forests, specifically quantified as statistically higher consumption of Hymenoptera (ants), Coleoptera, Isoptera, Diplopoda, insect larvae, Gastropoda, Chilopoda, and smaller (0–10 mm) prey among Nearctic-Neotropical migrants and more Hymenoptera (alate ants), insect pupae, lizards, Heteroptera, Orthoptera, and Araneae, and larger (>25 mm) prey among resident birds. Further, Boyle, Conway & Bronstein (2011) found that migrants were more likely to be generalist insectivores and rely on fruit than tropical residents; a change to a more frugivorous diet might alter demand by migrant birds for particular taxa of insect prey.

Seasonality

In 2002 and 2003, diets of obligate and generalist insectivore residents differed during the dry season, but not during the wet season, and the reverse was true in 2001. In addition, season was an important factor in our PERMANOVA interacting with year and guild to influence avian diets. Seasonality of food resources and breeding requirements likely interact, so whereas generalist insectivores may consume fruits and seeds, they may need to prioritize certain arthropod taxa during the breeding season for reproductive requirements (Gill, Prum & Robinson, 2019), potentially increasing dietary overlap with obligate insectivores. An exception to this general trend is the YGVI, the Austral migratory species, that consumes fruit and insects during the wet season while breeding at our study sites (Brewer & Christie, 2020).

Unfortunately, we do not have consistent arthropod data from all seasons when bird diets were studied to compare the relative abundance in the farm to relative abundance in the diets. In general, differences in foraging behavior (Greenberg et al., 1997, 1999) and vegetative strata (Whelan, 2001; Jedlicka et al., 2006) are important factors that influence diet selection.

Shade management and ecosystem services

Our third objective was to determine how diets might be influenced by coffee agroecosystem management. Differences in diets were evident, with one-fourth of bird species in all three production systems having significantly different diets in both commercial polyculture and traditional polyculture systems. Although we did not measure fruit availability, other investigators have suggested that differences in diet in the same farms may be due to the greater availability of fruit in the traditional polyculture (Dietsch, Perfecto & Greenberg, 2007), which may have bottom-up effects on arthropod composition or result in a separate change in the arthropod communities that may affect prey availability and influence selection. With greater vegetational diversity, traditional polyculture systems may have a greater diversity and abundance of natural predators (Letourneau et al., 2009; Serrano-Davies & Sanz, 2017), leading to increased availability of spiders and presence in avian diets. Since MacArthur (MacArthur, 1958), differences in diet have been identified as pathways to coexistence of insectivorous species of birds over time and, indeed, we found evidence of dietary niche partitioning in our study. Shade management affected the diets of Yellow-green Vireos (YGVI) because arthropod composition of their diets differed depending on whether they were foraging in TP, CP, or SM. YGVI consumed more Coleoptera in TP and more Formicidae in SM systems. Nevertheless, the diets of the other three species sampled across all shade management systems did not change significantly across management systems. This is likely a result of low sample sizes across all systems. However, it may be true that generalist insectivores may not be as impacted by production changes as obligate insectivores. Some bird species may alter their foraging patterns in response to fruit resources or management shifts in coffee farms (Carlo, Collazo & Groom, 2004). Other investigators have documented differences in where and how birds forage in shaded coffee plantations (Jones et al., 2000; Bakermans et al., 2012), and that even small differences (e.g., tree height and composition) may influence the foraging behavior and diets of some species (e.g., Dietsch, Perfecto & Greenberg, 2007; Newell et al., 2014). To fully understand the implications of management shifts on bird foraging and diets, more studies are needed that explicitly link prey availability, species-specific foraging observations, and dietary analysis; such studies would help parse out cause and effect relationships.

The effect of habitat management is a vital consideration for understanding pest control ecosystem services birds may provide in agricultural landscapes (Perfecto et al., 2004; Van Bael et al., 2008; Jedlicka, Greenberg & Letourneau, 2011; Iverson et al., 2014; Milligan et al., 2016; Şekercioğlu, Wenny & Whelan, 2016). The effects of birds on populations of coffee pests are often more efficient in coffee agroforests with more diverse shade (Perfecto et al., 2004; Johnson, Kellermann & Stercho, 2010; Milligan et al., 2016). Mechanisms driving this enhanced pest control could be increases in taxonomic or functional richness of the bird community, or increased abundance of insectivores (Perfecto et al., 2004; Van Bael et al., 2008; Philpott et al., 2009; Philpott & Bichier, 2012; Milligan et al., 2016; Martínez-Salinas et al., 2016). In addition, differences in bird foraging behavior that result from differences in tree species, composition, or foraging strata present may underlie differences in bird diets (Wunderle & Latta, 1998; Dietsch, Perfecto & Greenberg, 2007; Bakermans et al., 2012; Newell et al., 2014) and, therefore, the effectiveness of bird predation on pest species in coffee agroecosystems. For this reason, we assessed if birds consumed coffee pests, or if their diets, generally, shifted with management or season. In surprising contrast to previous studies, we found no arthropod pest insects in the diets of the birds sampled. Specifically, coffee berry borers (Hypothenemus hampei; 1.2–1.8 mm long) found in the diets of migratory wood warblers (Parulidae) in Jamaican coffee farms (Sherry et al., 2016) were noticeably absent in our study. Data collected during our study indicates that coffee berry borer abundance and fruit infestation rates were low. In vacuum samples of insects from coffee plants obtained in February 2001 and August 2001, borers comprised only 10 of 835 (1.2%) of insects collected (S. Philpott, 2001, unpublished data). In addition, only ~6% of sampled fruits were found to be infested by coffee berry borers between November 2000 and June 2001 (S. Philpott, 2001, unpublished data). In other studies, borers have been described as abundant prey items (e.g., Sherry et al., 2016). Notably, the role of insectivorous birds (and bats) in limiting arthropods and coffee damage may also shift depending on landscape context (Librán-Embid, De Coster & Metzger, 2017). More work is needed to understand the impacts of coffee landscape change on bird diets, foraging, and pest predation.

Methods for diet analysis and conservation implications

Although questions about how to best study and analyze bird diets are not new (Rosenberg & Cooper, 1990), alternative methods provide complementary insights into the arthropod prey consumed by birds. Molecular scatology provides more specific identification of prey items, often to genus or species-level, but has been criticized for failing to account for individuals and relying on presence/absence data for analysis (Symondson & Harwood, 2014; but see Deagle et al., 2019). Compared to molecular techniques, either stomach-content studies using emetics (e.g., Sherry et al., 2016) or fecal dissection approaches (e.g., Burger et al., 1999) allow researchers to reliably quantify prey items. Admittedly emetics and fecal dissection recover more hard-bodied prey items (such as Coleoptera and Formicidae), but these biases extend across all samples and allow comparisons to be made. Some drawbacks of using emetics include potential to harm birds not only during sample collection, but after release (Johnson et al., 2002). Ethically, it is important to consider avoiding more invasive procedures and substituting non-invasive procedures such as fecal collection that can be used easily in molecular scatology. Studies using emetics have generally gathered quite different quantities of arthropods identified per stomach sample (5.5 in this study compared to 34 in Kent & Sherry (2020), 30 in Strong (2000) and 70 in Sherry et al. (2016)). These differences most likely reflect changes in the arthropod communities at different study sites. All three of the aforementioned studies took place in the tropical island of Jamaica, compared to this study that took place in the mountains of Southern Mexico. Bird diets will obviously be constrained by resource availability.

Finally, both emetic and fecal dissection lead to dietary items that are identified at a courser level, oftentimes to order, leading to analyses that may overestimate similarity between samples (Anderson et al., 2005). The fact that we found such striking differences between samples using course dietary categories reaffirms the inherent differences between samples in our study. However, such a course overview of diets may gloss over important distinctions of prey identity and the ecological interactions behind predator-prey dynamics. Species-level identification is necessary to understand if predators consume certain pests, for example. Because taxonomic resolution can influence results, combining molecular scatology methods with visual identification of stomach/fecal matter can provide a more comprehensive view of diets. Increasing sample sizes and combining studies from across species’ ranges will enhance our knowledge of avian diets and lead to better management for conservation and predictions of the role birds play in community-level interactions. Studies that provide baseline dietary data of avian insectivores are crucial for evaluating how community dynamics change as a result of the global decline in terrestrial insect populations (Klink et al., 2020). Although diet studies are complicated and logistically difficult, they are crucial for monitoring and conserving bird populations in the future.