Animals are expected to select a breeding habitat using cues that should reflect, directly or not, the fitness outcome of the different habitat options. However, human-induced environmental changes can alter the relationships between habitat characteristics and their fitness consequences, leading to maladaptive habitat choices. The most severe case of such nonideal habitat selection is the ecological trap, which occurs when individuals prefer to settle in poor-quality habitats while better ones are available. Here we studied the adaptiveness of nest box selection in a tree swallow (Tachycineta bicolor) population breeding over a 10-year period in a network of 400 nest boxes distributed along a gradient of agricultural intensification in southern Québec, Canada. We first examined the effects of multiple environmental and social habitat characteristics on nest box preference to identify potential settlement cues. We then assessed the links between those cues and habitat quality as defined by the reproductive performance of individuals that settled early or late in nest boxes. We found that tree swallows preferred nesting in open habitats with high cover of perennial forage crops, high spring insect biomass, and high density of house sparrows (Passer domesticus), their main competitors for nest sites. They also preferred nesting where the density of breeders and their mean number of fledglings during the previous year were high. However, we detected mismatches between preference and habitat quality for several environmental variables. The density of competitors and conspecific social information showed severe mismatches, as their relationships to preference and breeding success went in opposite direction under certain circumstances. Spring food availability and agricultural landscape context, while related to preferences, were not related to breeding success. Overall, our study emphasizes the complexity of habitat selection behavior and provides evidence that multiple mechanisms may potentially lead to an ecological trap in farmlands. 

Study area

The study area included 40 farms distributed along a gradient of agricultural intensification covering approximately 10,200 km² in southern Québec, Canada. Three land cover types dominated the study system: low-intensity agricultural fields (i.e. hay, alfalfa (Medicago sativa), clover (Trifolium spp.), and pastures, henceforth referred to as “forage crops”), intensive agricultural fields (i.e., annual row crops mainly composed of corn (Zea mays), soybean (Glycine max), and wheat (Triticum spp.)) and forest.

Each farm included 10 identical nest boxes mostly arranged in a row along field margins and separated by at least 50 meters to limit intra and interspecific competition (see Ghilain and Bélisle 2008 for further details on the study system). A Thermochron iButton device was fixed on the outside of one nest box on each farm to record hourly ambient temperature (model DS1922L; Embedded Data Systems, Lawrenceburg, Kentucky, USA). A pluviometer collected precipitation data (millimeters of rainfall) on each farm. We used the mean daily temperature and mean daily rainfall between May 1 and May 15 to characterize spring climate. 

 

Nest monitoring

We monitored nest boxes every other day from 2009 to 2018 starting in the first week of May. We recorded the occurrence of nest materials, the laying date (first egg), and the number of eggs, hatchlings and fledglings. Monitoring ended when all nestlings had fledged on a given farm, which occurred between June 15 and August 5 during the study. Nest boxes were cleared of any nest material and/or dead nestlings every year in October.

 

Preference

Preference for each nest box was estimated for each year according to the occurrence of a laying event (at least one egg laid) and settlement date (Julian date at which nesting material was first observed).

 

Because some early settlement dates were left-censored given that some boxes already contained nest material at the first visit (45% of all boxes), settlement dates were classified as either “early” or “late” with respect to the annual median settlement date. The category “early” included boxes with settlement dates preceding or equal to the annual median, which comprised nearly all (91.3%) left-censored dates. Overall, the average difference between the annual mean settlement dates categorized as “early” and “late” was 10.4 ± 2.9 days (mean ± SD). Nest boxes occupied by other species were excluded from analyses (N = 964 boxes between 2009 and 2018). Such exclusions were made possible, even in the absence of a laying event, because the material and shape of nests are very species specific. We are thus confident that the vast majority of nests included in the study were initiated by tree swallows. The ordinal preference variable featured three categories:

1: No laying event

2: Laying event and late settlement

3: Laying event and early settlement

 

Habitat quality

We used two proxies of reproductive success: (1) the number of hatchlings produced in a nest box and (2) the proportion of hatchlings that successfully fledged in a nestbox (i.e., fledging success).

 

Landscape context

We characterized landscape habitat composition by measuring the relative cover of forest, perennial forage crops, as well as of water bodies and wetlands, within radii of respectively 100 m, 5 km and 10 km. We assessed landscape habitat composition up to the 500-m scale on a yearly basis in the field by visually identifying cultures and marginal habitats and delineating them using orthophotos (1:40,000). Characterization beyond the 500-m scale was based on a mosaic of yearly georeferenced classified optical and radar satellite images taken between 2011 and 2018 (pixel resolution 30 m × 30 m; Agriculture and Agri-Food Canada (AAFC) 2018). Only the year 2018 was used to assess water cover at the above range of scales because it showed better accuracy than the data of previous years (AAFC, 2020), and because the cover of large water bodies, as those covered by the data we used, did not vary significantly across years (e.g., median between-year correlation of yearly water cover between 2011 and 2018 was 0.90 at the 10-km scale).

 

Food availability

Two passive insect traps were installed on each farm around the first and second third of the nest box transect. Traps consisted of ~4-L yellow buckets placed 1.5 m above ground. They were filled with ~2 L of salty detergent solution to reduce surface tension and slow the growth of bacteria and fungi. Two transparent plexiglass screens were mounted perpendicularly to one another above each bucket to intercept flying insects (see Bellavance et al. 2018 and Garrett et al. 2021a for details). We collected the content of each trap on every visit to a farm (i.e., every other day) and conserved arthropods in 70% ethanol until processing. We sorted samples by removing arthropods unlikely to be preyed upon by tree swallows (i.e., bumble bees (Bombus spp.: Hymenoptera), June bugs (Phyllophaga spp.: Coleoptera), large spiders (Araneae, > 0.5 cm body width), and caterpillars (Lepidoptera); Bellavance et al. 2018). The rest of the sample was dried at 50ºC for at least 48 hours before being weighed (Adam Equipment, model AAA250L, ± 0.0001 g). The mean daily dry biomass of arthropods collected between May 1 and May 15 was used as a proxy of yearly food availability on a given farm at the time of nest site selection.

 

Heterospecific social information

House sparrows (Passer domesticus) are tree swallow’s main nest-site competitors in our system and they initiate breeding before swallows return from their wintering grounds (Robillard, Garant and Bélisle, 2013). We evaluated the use of heterospecific social information through the number of nest boxes occupied by house sparrows on each farm in the current year. Nest boxes and house sparrow nests were visited every other day concurrently to tree swallow monitoring. Occupancy was determined by the presence of at least one egg, and only first clutches observed in each box were included since a nest box is rarely used by another species once house sparrows have built a nest therein.

Conspecific social information

We defined two sources of social information regarding the future breeding success that an individual could expect to experience on a given farm: the density of tree swallows that bred on a farm during the previous year and the mean number of fledglings obtained by those breeders.