Introduction

An organism’s reproductive phenology is a central aspect of its biology, and knowledge of mating seasons is thus often critical for human efforts at conservation, management, or biological control. In birds, for example, events associated with reproductive timing, such as migration dates, plumage changes, and courtship and parental behavior, are a primary focus of field guides, conservation and citizen science initiatives (Sullivan et al. 2009), and research (Pardieck et al. 2020). The emphasis by researchers on reproductive phenology in birds and other taxa contrasts with our knowledge of ants, whose mating season information is often sparse or absent in otherwise detailed species accounts and identification resources (Ellison et al. 2012).

This lack of information is due to the difficulty of observing ant mating. Mating in ants is usually mediated through specialized flying reproductive castes (in nearly all ant species males and/or young queens fly, Hölldobler and Wilson 1990; Peeters and Ito 2001; Helms 2018) which lead temporarily solitary lives and fly through the atmosphere (Markin et al. 1971; Shik et al. 2013; Helms et al. 2016b). Many of the behaviors associated with reproduction in ants—mating aggregations or lekking, mate choice, and copulation—may thus occur during flight unseen by humans. Mating flights also occur during only a brief period of an ant’s lifetime, sometimes lasting only a few minutes (Helms and Godfrey 2016), after which the males die and the queens shed their wings, never to fly again (Peeters and Ito 2001). The production of reproductive offspring occurs years to decades later, when mature colonies rear males from unfertilized eggs and rear queens using sperm cells stored from that initial mating event at the beginning of life (Hölldobler and Wilson 1990). The sequence of events during the mating seasons of ants thus runs in reverse relative to that of many other organisms—instead of starting with mating and ending with offspring production, ant mating seasons begin with the maturation of reproductive offspring and end when those offspring leave the nest to mate. Prior to emerging for mating flights, reproductive castes are cloistered within nests, making developmental and physiological changes in preparation for mating invisible to casual observation in the field. The aerial nature of ant mating, the narrow window of time in which to observe it, and the small size of ants all combine to make direct observation of mating virtually impossible for many species.

When reproductives leave the nest, however, the resulting mating flights can be conspicuous, sometimes involving millions of ants flying across hundreds of square miles (Markin et al. 1971). In these cases, ant mating flights may show up on weather radar (Royal Meteorological Society 2021), disrupt sporting events or other human activity (Waldstein 2017), and even alter the chemistry and productivity of water bodies through the mass deposition of winged ant corpses (Kannowski 1971; Carlton and Goldman 1984). Nocturnally flying ants often swarm to artificial lights, and mating flights of building-inhabiting species may even take place indoors (Hansen and Klotz 2005). Predicting the occurrence of ant mating has nevertheless proved elusive for most species, and we likewise remain unaware of how ant mating phenology has been altered due to climate change or other anthropogenic impacts (Parr and Bishop 2022; see Chick et al. 2019 for advanced mating seasons in urban heat islands). These gaps are due mostly to a scarcity of basic information for most species on when mating occurs and to a lack of general theory describing the evolution and ecology of ant mating phenology.

It is difficult to generate or test hypotheses about mating season evolution because we lack a comprehensive treatment of ant mating phenology for any sizeable region. One early hypothesis (Hölldobler and Wilson 1990) posited that the scramble nature of mating in many ant species, combined with the fitness costs of wasteful interspecific mating attempts, selects for temporal segregation of mating seasons among species within a community. But intensive surveys of mating phenology at multiple sites (Talbot 1965, 2012; Baldridge et al. 1980; Woyciechowski 1987; Kaspari et al. 2001a, b; Torres et al. 2001; Dunn et al. 2007; do Nascimento et al. 2011; Dean and Dean 2018; Donoso et al. 2022) have consistently failed to find evidence of temporal segregation of mating seasons through the year. To the extent that temporal segregation occurs in ant mating, it is likely restricted to partitioning flights throughout the 24-h day (Talbot 1963; Kannowski 1959; Haddow et al. 1966; Dunn et al. 2007), as is frequently observed in ant foraging (Albrecht and Gotelli 2001). Indeed, the extremely broad mating seasons of some species, such as potential year-round reproduction in Solenopsis invicta (Tschinkel 2013; Dean and Dean 2018) and Azteca instabilis (Helms and Kaspari 2014), preclude strong seasonal partitioning in many communities.

Ant mating seasons instead appear to be strongly associated with climate or weather, although the conditions vary among species (Kannowski 1959; Talbot 1963, 2012; Markin et al. 1971; Baldridge et al. 1980; Boomsma and Leusink 1981; Noordijk et al. 2008; Donoso et al. 2022). The importance of climate in regulating ant mating is captured by the latitudinal synchronicity hypothesis (Kaspari et al. 2001b; Dunn et al. 2007), which emphasizes the role of thermal variability in limiting the range of dates over which mating can occur. In this view, consistent temperatures at low latitude sites allow mating to occur over a larger fraction of the year, resulting in the evolution of a greater range of phenologies. At high latitude sites, in contrast, cold winters limit reproductive activity to the warmer parts of the year, resulting in a pulsed pattern where many or most ant species mate at the same time, sharing one synchronous brief mating season. The latitudinal synchronicity hypothesis was proposed based on paired comparisons of two temperate sites (Talbot 1965, 2012; Dunn et al. 2007) with one tropical site (Kaspari et al. 2001b) and has yet to be tested across a large region. The scope and precise nature of the relationship between latitude and mating phenology thus remain unclear. There are also likely life history correlates to ant mating phenology, as species must match the timing of mating to the availability of nesting requirements. Arboreal or plant nesting ants, for example, require host plants of appropriate size or developmental stage in which to start a colony (Tschinkel 2002; Bruna et al. 2011). Likewise, queens of socially parasitic ants, which infiltrate or usurp existing colonies, tailor their life cycles around those of their hosts (Buschinger 2009). Species may also vary in the amount of time males or queens lead a solitary existence outside the natal nest, which may alter the timing of maturation, flight, mating, and nest founding relative to other events in the life cycle (Shik et al. 2013). Finally, the dependence of mating on environmental conditions suggests a sensitivity of ant phenology to climate change, but this prediction has not been tested.

Studying museum collections may help address phenological questions while circumventing the difficulty of observing behavior (Dunn and Winkler 1999; Suarez and Tsutsui 2004; McLean and Guralnick 2021). I use this approach to answer several questions. (1) When do ants mate and how does that vary with latitude and elevation? (2) Are ant mating seasons more tightly aggregated at higher latitudes? (3) Have ant mating seasons shifted in response to climate change? (4) Do the observed relationships differ in exotic versus native species? The results highlight the sensitivity of mating phenology to geography and climate.

Materials and methods

To study ant mating phenology, I compiled museum records from the largest open access database of ant specimens—AntWeb (2020, data collected March through August 2020). I searched all identified ant specimens in the database collected in the contiguous U.S. from the nineteenth century through 2019 and located every specimen that was a mature reproductive adult in mating condition and that was associated with a collection date and location. I considered a specimen to be in mating condition if it was a mature male or a winged female, as determined from collection notes, specimen descriptions, or photographs. For each specimen meeting these criteria, I recorded the species, date, latitude, longitude, and elevation. I treated each such record as one observed mating event, even if multiple specimens were collected from the same event. For specimens from traps that collected ants over a range of dates (for example, a malaise trap left open for a week), I conservatively used the trap starting date as the date of the mating event. In cases where the collecting location was described (e.g., a town, county, or mailing address) but latitude, longitude, or elevation data were not given, I used online mapping resources to assign geographic coordinates and elevation to the described locations. I further avoided multiple countings of single mating events by excluding specimens collected within 0.01 degree of latitude or longitude and at the same elevation of another mating event on the same day. I also excluded records that occurred under laboratory or indoor conditions.

To test the relationship of ant mating phenology to geography and climate, I performed a multiple regression of mating record date on latitude, elevation, and calendar year among all ants taken together. I complemented this analysis of continuous variation by sorting all mating records into three roughly equal latitudinal bands (< 32°, 32° to 40°, and > 40°) and testing for differences in median mating date among low, middle, and high latitude ant assemblages. I used a similar two-pronged approach to test predictions of the latitudinal synchronicity hypothesis. I first used two-tailed F-tests to compare mating date variability among ants in the low, middle, and high latitude bands described above. I complemented this with a higher resolution method in which I grouped all mating records into 25 non-overlapping 1-degree latitudinal bands (median n = 79 records per band, range 18–359) and used linear regression to test whether coefficients of variation in mating date decreased with increasing latitude. To avoid the possibility of overall results being driven by variation within a few well-sampled species, and given the absence of a species-level phylogeny for the hundreds of ants in this study which would otherwise allow for phylogenetically independent analyses, I repeated mating phenology analyses within individual species. To test for species-level effects, I examined specific mating phenology for each of the 27 most common species (those with at least 20 mating records in the database). I used a similar approach as for all ants taken together, using multiple regression to test relationships between mating date and latitude, elevation, and calendar year.

I performed all analyses in R (R Core Team 2020). I used Kruskal–Wallis tests for comparisons of non-parametric data and present results as medians with interquartile ranges (IQR). I checked for potential confounding effects of variable sampling effort in two ways. First, I tested whether the number of ant specimens in mating condition changed through time relative to all national ant records in the database. Specimens in mating condition consistently made up about 2% of all U.S. records in the database throughout the study period (linear regression through origin for years 1879–2019, excluding 2001–2003 for which U.S. summary data were not available, mating records = 0.022 × U.S. ant records, P < 0.0001, F1,137 349.1, r2 = 0.72), indicating no major shifts or bias in sampling effort for or against reproductive ants. Second, I tested whether researchers had extended or shortened their collecting seasons over the study period. I limited this analysis to the year 1926 and later, which is the longest continuous period for which there were at least 50 U.S. ant records each year (and which contains 98.4% of all mating records in our dataset). Throughout this period there was no change in either the earliest or latest ant collecting records through time (linear regression excluding 2001–2003; start dates: P = 0.15, F1,89 2.085, r2 = 0.02; end dates: P = 0.15, F1,89 2.078, r2 = 0.02). To a first approximation, over the past century collectors have apparently been finding ants throughout the calendar year in places ants are active (median start date 3 January, interquartile range 10 days; median end date 20 December, IQR 17.5 days), precluding large-scale collector effects on mating phenology results.

Results

I compiled 2927 ant mating records from across the contiguous U.S. representing 501 species, 71 genera, and 9 subfamilies (Fig. 1). Sampling was reasonably comprehensive across the study area, and despite local variability in geographic coverage there were no obvious biases across the entire data set. The median number of mating records per species was 2 (interquartile range 5), with a minimum of 1 (166 species) and a maximum of 109 (Nylanderia faisonensis) (Fig. S1a, Table S1). Many species were observed in only one year, such that only 317 species (~ 63%) had mating observations from two or more years, with a median timespan (time elapsed from oldest to most recent observations) of 46 years [IQR 52, maximum of 122 years in Camponotus vicinus (92 records), Formica dakotensis (2 records), and Tapinoma sessile (74 records)]. Over 300 additional species listed in the AntWeb database as occurring in the contiguous U.S. had no mating records in the database (nearly 40% of total species). Coverage at the genus level was likewise uneven, with about a third of genera (32%) represented by 3 or fewer mating records, and an overall median of 13 records per genus (IQR 34.5, maximum of 405 records in Camponotus carpenter ants) (Fig. S1b). Many exotic species were represented (201 mating records from 45 species), but their temporal coverage was less than half as extensive as for native species (median timespan of 20 years (IQR 38) for exotic species versus 49 (IQR 52) for natives, Kruskal–Wallis P = 0.007, χ2 = 7.18, df 1), likely reflecting the recent arrival or documentation of many exotic species, perhaps paired with a general underrepresentation of exotic species in museum databases.

Fig. 1
figure 1

Ant mating records across the contiguous U.S. (2927 mating records from 501 species) used for analysis. Darker points indicate overlapping records

Full size image

As expected from general relationships between geography and climate, areas at higher latitudes or elevations experienced mating records later in the year (Figs. 2, S2). Average ant mating dates occurred about a day later in the year for each 100 m elevation or degree latitude (0.9 days per 100 m, 1.2 days per degree), and consistent with predicted impacts of climate change have also been advancing by 0.1 day per year (elevation linear regression P < 0.0001, latitude P < 0.0001, year P = 0.00281, F3,2923 44.24, total r2 = 0.04 for all three variables). Average mating dates did not differ between native and exotic ant species after accounting for latitude, elevation, and year effects (ANOVA F1,2922 0.7084, P = 0.4).

Fig. 2
figure 2

Ant mating seasons occur later and are more tightly aggregated at higher latitudes. Points and boxplots show median and quartile mating record dates

Full size image

When grouped into three bands of roughly equal latitudinal spans, about half of the mating records in the U.S. occurred at mid-latitudes (between 32° and 40°, n = 1,473 records, 50.3% of the total, 53% of relative land area), and a quarter each at low (< 32°, n = 682 records, 23.3% of records, 14% of land area) and high latitudes (> 40°, n = 772 records, 26.4% of records, 33% of land area) (Figs. 1, 2). As seen in the regression results, the median mating date for ants at high latitudes occurred over a month later than at low and mid-latitudes (median date of 9 August IQR 54 days versus 7 July IQR 95 at low latitudes and 4 July IQR 77 at mid-latitudes, Kruskal–Wallis P < 0.0001, χ2 = 151.63, df 2, post hoc Holm-adjusted Dunn’s test P = 0.0288 for low and mid-latitude comparison, P < 0.0001 for other paired comparisons, Figs. 2, S2). Consistent with the latitudinal synchronicity hypothesis, ant mating seasons at higher latitudes were also more tightly aggregated during the warm months. Among the three broad latitudinal bands, mating dates at the highest latitudes were roughly only half as variable as those at the lowest (coefficients of variation in mating dates were 0.39 at low latitudes, 0.34 at mid-latitudes, and 0.23 at high latitudes, F-test P < 0.0001 for all pairwise comparisons of mating date variation among latitudinal bands, low-mid F681,1472 1.4538, low–high F681,771 2.3775, mid-high F1472,771 2.3775, Figs. S3, 2). When examined across one-degree latitudinal windows, ant mating seasons showed a strong linear decrease in duration, and increase in synchronicity among species, with latitude, such that the coefficients of variation in mating date decreased by about 0.01 per degree of latitude (linear regression P < 0.0001, F1,23 27.83, r2 = 0.55, Fig. 3). In this relationship, mating seasons at the lowest latitudes are nearly three times as variable as those at the highest latitudes (predicted coefficient of variation 0.45 versus 0.16).

Fig. 3
figure 3

Ant mating seasons are more synchronous at high latitudes. y-axis shows coefficients of variation of mating date for records within each one-degree latitudinal band on the x-axis

Full size image

I analyzed 27 ant species individually—24 natives and 3 exotics (any species with n ≥ 20, Table S2). Consistent with predicted climate change impacts, a fourth of native species (6 species) have advanced their mating seasons an average of 0.9 ± 0.6 days earlier per year (Figure S4). One native species (Crematogaster emeryana) delayed its mating season by 1.0 day per year. This is a species of semiarid habitats in the southwestern U.S. that mates during late summer and fall. Its altered mating season may thus track later arrivals of cool weather under recent climate change. The 3 exotic species (the fungus-gardening ant Cyphomyrmex rimosus, Argentine ant Linepithema humile, and Asian needle ant Brachyponera chinensis), in contrast, showed no changes in mating phenology over time (Fig. S5).

Discussion

This study represents the first large-scale description of the geography of ant mating phenology. Although mating records are sparse for most species and genera, the data affirm the link between climate and phenology. Mating seasons occur later in the year at higher elevations and latitudes. As predicted by the latitudinal synchronicity hypothesis, ant mating seasons are also shorter and less variable at higher latitudes, likely due to the colder winters associated with more variable climates. Consistent with predicted responses to climate change, the mating seasons of all ants taken together have been occurring about 1 day earlier per decade, which is equivalent in magnitude to an ant community moving 100 m downward in elevation or 1 degree toward the equator in latitude. Among the 25% of native species that demonstrated phenological advances, the transition is ten times as stark—an average of 1 day earlier each year. The phenological changes observed here appeared less common than geographic range shifts in arthropods (only 25% of ant species showing earlier mating seasons versus 2/3 to 3/4 of arthropod taxa demonstrating poleward or upward shifts in range, Parmesan et al. 1999; Hickling et al. 2006; Chen et al. 2011) but more drastic in impact. The 1-day advance in mating phenology per decade among all ants—equivalent to adopting the mating season of a population 100 m lower in elevation—is an order of magnitude larger than typical elevational range shifts of only ~ 10 m per decade (Hickling et al. 2006; Chen et al. 2011), highlighting the role of phenological shifts in allowing organisms to rapidly respond to climate change. These rapid recent changes in mating season also raise the possibility of phenological mismatch between the timing of ant reproduction and the availability of resources they require or provide to other organisms.

Ant mating flights represent major pulses of available nutrients, due to the large numbers of nutrient-rich queens and males that are produced and their high mortality. Mating ants are a major food source for a wide range of animals in terrestrial, aquatic, and atmospheric food webs (Talbot 1963; Baldridge et al. 1980; Carlton and Goldman 1984; Helms et al. 2016b). For some birds, ant mating flights may be the predominant food source for developing nestlings (Helms et al. 2016a; Goffová et al. 2021), and several species specialize on eating flying ants throughout their lives (Hespenheide 1975; Law et al. 2017; Goffová et al. 2021). These nutrient fluxes can have effects that propagate through food webs beyond the direct role of flying ants as food. Ant queens and males, for example, may concentrate mercury or other contaminants in their tissues and transfer them among food webs through predation during flights (Helms and Tweedy 2017). Limiting nutrients released from flying ant corpses, or excreted by fish after feeding on them, may also spark bursts of algal productivity in aquatic ecosystems (Carlton and Goldman 1984). Shifts in the timing of ant mating flights have the potential to alter the occurrence or magnitude of all these interactions, with possible repercussions for the organisms that rely on them.

Over ecological time scales, changes in ant mating phenology may be driven by either developmental or behavioral mechanisms. Warmer ambient temperatures may directly speed the production and development of larval queens and males within ant nests (Porter 1988; Hölldobler and Wilson 1990; Chick et al. 2019; Parr and Bishop 2022). Mature queens and males, and the colonies that produce them, may also wait for specific environmental conditions or cues to trigger mating flights (Kannowski 1959; Talbot 1963; Hölldobler and Wilson 1990). Both types of mechanism allow ant colonies to calibrate the timing of mating flights to climate and weather regimes. The earlier arrival of warm weather in spring, for example, could lead to earlier mating seasons both by speeding the production and development of larval ants and by advancing the occurrence of environmental triggers for mating flights. It remains unclear to what extent the phenological changes observed here arise from either of these ecological mechanisms or from rapid evolutionary changes (Diamond et al. 2018).

Distinct phases of the ant mating season—such as the rearing of queens and males versus the timing of mating flights—may also be differentially affected by environmental variation. The development time of offspring, for example, may rely on ambient temperature, whereas the precise timing of flights may be determined by precipitation or wind patterns. Species also vary in the relative timing of these events through the mating season, such that queens and males of some species may fly immediately after maturing, whereas those of other species remain in the nest for a period before flying. In extreme cases, queens or males may even overwinter in their natal nest before flying with the arrival of warm weather in the spring, as seen in some Camponotus carpenter ants (Hansen and Klotz 2005). Similar variation may occur within a species or even a single nest. Colonies of the fire ant Solenopsis invicta, for example, practice a mixed strategy in which some queens fly soon after maturing and others overwinter in the nest to fly the following year (Tschinkel 1996; Helms and Godfrey 2016). A better understanding of these complexities, as well as comparison of phenology data collected specifically during flights (AntFlights 2022; AntNupTracker 2022), will help disentangle the impacts of environmental changes on ant reproductive activity.

Despite broad temporal, spatial, and taxonomic coverage, this study highlights gaps in collections data, which mirror the reality of our general knowledge of ant phenology. Few data were available for most species, and nearly 40% of regional ant species were not captured at all by the dataset. Low sample sizes were exacerbated by incomplete metadata for specimens. While searching through digitized records, for example, I was unable to extract data from a substantial number of specimens that were described as queens without clarifying whether they were winged unmated queens (thereby indicative of mating phenology) or were older queens within established colonies (not informative for phenological studies). With more descriptive metadata or images, many of these specimens would have provided additional mating records. At least some well studied exotic species were under-represented in the data, possibly because collectors consider those specimens less valuable for museum collections or because they are less likely to occur in habitats sampled by collectors (Arnan et al. 2021). The fire ant Solenopsis invicta, for example, was represented by only 6 mating records, despite its mating phenology being more extensively documented in the literature than perhaps any other ant (Markin et al. 1971; Tschinkel 2013). As recent arrivals, exotic species also had shorter timespans available for analysis. Finally, the museum data had limited utility for teasing apart the relative timing of flight within the broader reproductive season, probably due to the difficulties of collecting ants in the act of mating. Based on collection notes, no more than ~ 20% of the specimens were collected during flight. Many of these issues could be resolved by future research efforts that integrate citizen science initiatives (AntFlights 2022; AntNupTracker 2022), passive aerial sampling of flying ants, and archiving and digitization of specimens.

The patterns in ant mating phenology outlined here among several hundred species in one region represent a small step toward understanding the mating seasons of each of the world’s more than 16,000 ant species (AntWeb 2020). The results showcase the variability of ant mating seasons with taxonomy and geography, as well as their sensitivity to anthropogenic impacts like climate change. Further efforts to study ant phenology at large temporal and spatial scales will continue to open new avenues for developing and testing hypotheses about the evolutionary ecology of phenology and for predicting the impacts of future climate scenarios.