Positive Density Dependence Promotes Host Persistence in the Face of Infectious Disease

ABSTRACT

Sociality offers benefits to species that can enhance their fitness. However, pathogen transmission can be higher in larger groups, potentially negating the advantages of group living. Despite the important paradoxical effects of population density on disease impacts and recovery, the competing effects of density remain unexplored. Here, we examine the response of a social bat species to pathogen invasion by comparing the effect of colony size on disease impacts during the summer (disease‐free period) and winter (disease period). During pathogen invasion, larger winter colonies initially experienced relatively higher declines than smaller colonies. Conversely, summer colony size positively influenced colony growth immediately following pathogen invasion and during recovery, suggesting that Allee effects may be important in population resilience. Our results show that hosts faced with a novel pathogen may experience both benefits and costs of group living, and balancing these competing effects could impede evolutionary selection pressure toward asociality.

Outcomes of disease outbreaks can be complicated when introduced into populations of social species. The fungal disease white‐nose syndrome (WNS) causes mass mortality in bats who commonly form groups in summer and winter. We found that bat colony size significantly influenced growth rates in response to the arrival of WNS, where overall both summer and winter colonies declined because of the disease; however, summer colonies showed positive density dependence whereas colonies in winter showed negative density dependence, suggesting a complicated relationship between disease invasion and host response.

1. Introduction

Anthropogenic activity has increased the emergence of novel infectious diseases (Jones et al. 2008). Over the past several decades, pathogen introductions have subsequently resulted in mass mortality events and severe population declines for numerous taxa, including amphibians impacted by chytridiomycosis (Lips 2016), waterfowl affected by avian influenza (Swayne and Suarez 2000) and bats from white‐nose syndrome (WNS, Hoyt et al. 2021). Host sociality is frequently a key component in determining initial spread and transmission (Altizer et al. 2003), and it can also influence population recovery as hosts become rare (Courchamp et al. 1999).

Species form large groups for a variety of reasons and often depend on conspecifics for predator avoidance, thermoregulation and cooperative care of young (Gregory and Jones 2010; Ward and Webster 2016), which may make them subject to inverse density dependence, also known as Allee effects, where fitness decreases with decreasing population density (Courchamp et al. 1999). Allee effects can significantly complicate outcomes when infectious disease is introduced since it can exacerbate disease impacts and result in critical threshold relationships that drive populations to extinction; however, Allee effects may also help mediate disease impacts and buffer extinction risk (Hilker et al. 2009; Thieme et al. 2009). For example, infected individuals in larger groups may experience less severe impacts from disease, as in the case of Canis lupus infected with Sarcoptes scabiei (Almberg et al. 2015), whereas transmission of Mycobacterium suricatae in Suricata suricatta was higher in smaller groups, often causing group failure when they fell below a critical size threshold (Duncan et al. 2020). Animals that exhibit highly social behaviour may be more vulnerable to infectious diseases, as higher densities and contact rates can increase transmission (Altizer et al. 2003). For pathogens that cause mortality and population declines, greatly reduced population sizes can increase population vulnerability, enhancing the possibility of Allee effects and extirpation. Therefore, for diseases with density‐dependent transmission, theory predicts that pathogen spread should be higher in more dense populations; but simultaneously, that larger populations may be less likely to go extinct (Deredec and Courchamp 2006). However, these paradoxical effects of population density on host populations affected by disease have not been investigated empirically.

WNS is a disease of hibernating bats that is caused by the fungus Pseudogymnoascus destructans (Blehert et al. 2009; Lorch et al. 2011; Minnis and Lindner 2013). After its introduction to New York, USA in 2006, P. destructans spread across North America causing declines in multiple species of bats (Dzal et al. 2011; Frick, Pollock, et al. 2010; Langwig et al. 2012; Reynolds et al. 2016). Impacts from the disease vary by species and three formerly abundant species ( Myotis lucifugus , M. septentrionalis , and Perimyotis subflavus ) are among the most affected species (Francl et al. 2012; Frick, Pollock, et al. 2010; Hoyt et al. 2020; Langwig et al. 2012). Declines of M. lucifugus and P. subflavus in the Northeast and Midwest United States exceed 90% (Frick, Pollock, et al. 2010; Hoyt et al. 2021; Langwig et al. 2012), and M. septentrionalis has declined by over 95% (Langwig, Hoyt, et al. 2015; Reynolds et al. 2016).

Bats susceptible to WNS typically have two distinct seasonal habitats. In winter, bats hibernate underground (e.g., caves, mines, tunnels, etc.) in variable‐sized colonies (tens to hundreds of thousands) starting in fall to early spring. Transmission of P. destructans occurs through both direct host–host contact and through indirect contact with the environment (Langwig, Frick, et al. 2015; Langwig et al. 2017). Pseudogymnoascus destructans can persist in hibernaculum environments in the absence of bats, which facilitates reinfection when bats enter hibernation sites each fall (Hoyt et al. 2015; Langwig, Frick, et al. 2015). Abundance of P. destructans in the environmental reservoir is correlated with the size of bat colonies (Laggan et al. 2023). Growth of P. destructans on bats occurs only during winter as P. destructans can only grow at the cool temperatures at which bats hibernate (e.g., < 21°C; Verant et al. 2012) (Hoyt et al. 2021; Langwig, Frick, et al. 2015). Mortality occurs later in winter (January–April) (Langwig, Frick, et al. 2015; Langwig et al. 2017).

Bats that survive and emerge in spring disperse to summer colonies and can clear infection (Langwig, Hoyt, et al. 2015). Bats may move 8–> 550 km from winter to summer sites (Griffin 1940, 1970; Norquay et al. 2013) and typically exhibit a high degree of site philopatry to both seasonal roosts (Dixon 2011; Norquay et al. 2013; Schorr and Seimers 2021). For example, Norquay et al. (2013) recaptured 14% of their marked bats, but only 6% of bats recaptured (0.8% overall) were found at sites other than where they were originally marked, and the relocation rate for marked bats at summer and winter sites was 12% and 4% respectively. Estimated probabilities of site fidelity of marked bats at summer sites in Colorado ranged from 0.4 to 0.9 (Schorr and Seimers 2021). During summer, female bats form maternity colonies (ranging between tens to thousands of bats) where they give birth and raise young. Warm roosts can aid in gestation and reduce metabolic costs (Dechmann et al. 2004; Racey and Swift 1981; Ruczyński 2006); thus, thermoregulation is thought to be one of the primary bases for forming groups. Additionally, bats recovering from wing damage due to WNS infection may require torpor use to conserve energy during the healing process (Fuller et al. 2020); however, torpor can also delay reproduction (Racey and Swift 1981). Roosting in groups, particularly in spring when bats are recovering from disease, may reduce torpor requirements because larger groups may buffer against low temperatures where entering torpor would otherwise be needed. Therefore, maintaining large summer colony sizes in the face of a virulent pathogen could be essential for effective reproduction and may enhance fitness. Less is understood as to why bats form colonies in the winter; however, bats must come together to mate during fall, and thus mating opportunities could drive this relationship. However, it is also possible that underground habitat may be limited, necessitating colonies to form in relatively few suitable hibernacula. Ultimately, the transmission of P. destructans during the winter period should result in a cost to group living during winter. This suggests that bats might need to maintain large summer colonies but smaller colony sizes in the winter. Ultimately, disentangling the positive and negative effects of density on population recovery following disease‐induced declines will help identify which factors are most important in population resilience.

To understand the role of density on population dynamics following the emergence of a novel fungal disease, we assessed population trends at 39 summer colonies (disease‐free period) and 46 winter colonies (diseased period) of M. lucifugus (little brown bats) prior to and after the arrival of WNS over an 11‐year period. Our objectives were to identify the relationship between population size and growth rate in both summer and winter colonies through the epizootic and examine colony characteristics that influenced population growth rates at varying stages of pathogen invasion. We predicted that larger summer colonies would experience less severe impacts following WNS arrival due to the benefits of thermoregulation in larger colonies. Conversely, we predicted that large winter colonies would experience more severe declines after the arrival of WNS due to increased pathogen transmission.

2. Materials and Methods

2.1. Study Sites and Data Collection

We investigated population growth rates pre‐ and post‐WNS infection at 39 M. lucifugus summer colonies and 46 winter colonies in Wisconsin, USA (Figure 1; Figure S1). The study area for this project encompassed Wisconsin, northern Illinois and the Upper Peninsula of Michigan. Wisconsin and Michigan are two of the Great Lakes States located in north central United States. Summer roost colony sites included bat houses, attics in houses, barns, and bridges (Figures S1 and S2). We selected summer sites that had at least one emergence survey prior to the estimated arrival of WNS to the site and at least one survey after WNS arrival. Visual emergence counts were conducted primarily by property owners and volunteers in summer months (May 15 through August 31) from 2010 to 2021 by counting the number of bats emerging from the roost in the evening following Kunz et al. (2009). Surveyors positioned themselves such that they could observe the bats emerging from the main exit (e.g., the bottom of the bat house, or the vent of a building). Surveys generally started at sunset and continued until either the bats were no longer exiting or darkness precluded accurate counting. High temperatures in summer roosts generally ranged from 2°C below ambient to 21.5°C above ambient high temperatures during the summer residency period May–August. Mean summer colony counts ranged from 0 to 3895 bats. Winter colony surveys were conducted by entering the site and directly counting the number of individual bats present in hibernaculum passages of caves and abandoned mines, railroad tunnels and beer caves in late winter (Hoyt et al. 2015; Langwig et al. 2012). Temperatures in winter hibernacula ranged from 1.26°C to 13.5°C. Winter colony counts ranged from 0 to 128,020 bats. We analysed a total of 323 summer emergence counts and 473 winter survey counts.

FIGURE 1.

Yearly bat population growth rates in relation to arrival of white‐nose syndrome in summer (A) and winter (B) and map of study sites (C). (A) Population growth rates at 39 M. lucifugus summer maternity colonies and (B) 46 winter hibernation colonies for 5 years prior to WNS and 7 years post‐WNS detection at local winter hibernation sites. Years are colour coded by phase: Pre‐WNS in dark blue, arrival in pink, decline phase in green and post‐WNS in gold. Red dots indicate mean, and whiskers show 95% confidence limits. Dotted line at 0 indicates stability and rates above/below 0 indicate growing/declining populations. (C) Map of summer (red) and winter (dark blue) study sites in Wisconsin, Michigan and Illinois. Hashed counties indicate counties where P. destructans was found or full WNS infection was confirmed. For visualisation purposes, four log₁₀ lambda values in the winter dataset and three in the summer data outside of the −1−1 plot bounds are not included in Figure 1 but are represented in the analyses.

2.2. Determining the First Year of WNS at a Site

We assigned infection to winter colonies on the basis of direct sampling for P. destructans. There was a substantial effort to determine the timing of P. destructans arrival into Midwest hibernacula with surveillance efforts by the Wisconsin Department of Natural Resources, United States Geological Survey, and Hoyt et al. (2020). This included several years of sampling and surveys prior to the arrival of P. destructans, ensuring accurate timing of arrival. Myotis lucifugus do not spend the summer in the same sites where WNS infection takes place during winter, instead making regional migrations to summer roost sites in spring. Therefore, to assign arrival years to summer colonies, we estimated the arrival year on the basis of the year of P. destructans detection in the closest large hibernaculum (> 500 bats) and assigned that as the year a maternity colony became WNS‐affected. We used only large winter colonies to prevent small hibernacula (< 100 bats) that likely housed few bats at each summer roost in our study from biasing the arrival time of P. destructans at summer sites. For example, if summer site A was closest to a hibernaculum where bats were first infected in 2016, summer site A was assigned 2016 as the year of WNS arrival.

2.3. Estimating Yearly Population Growth Rates

For summer colonies, we calculated population trends prior to, during, and after WNS arrival at the site by using data from emergence counts conducted each summer. When more than 1 survey was completed within a summer, we averaged the counts. Our results were qualitatively similar when we considered only counts from before juveniles were believed to be volant (pre‐volancy; first flight of juveniles; Figure S3) or only post‐volancy counts (Figure S4) so we used mean counts from each year because it provided the most data points. We assumed counts conducted May to early July were only adults since we estimate juvenile volancy occurred beginning in mid‐July in our region (Huebschman 2019). We estimated yearly population growth rate, λ, for each roost site by adjusting for the number of years between counts:

$$\lambda =\sqrt[t_x]{\left(\frac{N_i}{N_{\left(i-1\right)}}\right)}$$

Here, N _i is the count of year N _i , N _i‐1 is the most recent count prior to year N _i , and t _x is the number of years between the counts. This results in a λ for year N _i . Population growth rates in winter sites were estimated identically by using single estimates of counts from late hibernation, as winter sites were only routinely surveyed once per year.

2.4. Statistical Analysis

We examined yearly population growth rates and arrival of WNS using general linear mixed‐effects models with the package glmmTMB in R v4.0.3 (Brooks et al. 2017). In these models, we treated the roost site as a random effect and years since WNS arrival interacting with variables (WNS phase: log₁₀ population size and log₁₀ distance to water) to allow for differential trajectories as the disease established. We used log₁₀ population size, distance to major water bodies, type of roost, number of roosts (i.e., number of bat houses at one locality), distance to major hibernaculum and age of the roost as interacting fixed effects (Tables S5–S8). We added 1 to all counts in order to log₁₀ population sizes. Yearly population sizes for summer sites were determined by taking the mean from emergence counts for each year and the count from the previous spring for winter sites. Distance to the closest major hibernaculum and major water body was calculated in ArcGIS Desktop v10.6.1 (ESRI 2018).

At each site, we split count data for years since WNS arrival into pre‐ (years < 0, where 0 was the year of P. destructans invasion [arrival]), declines (years 1 and 2 after initial invasion) and post (years > 2 after initial invasion) WNS phases. For example, for a site infected in 2016 (WNS arrival phase, year 0), years before 2016 would be < 0 (pre‐WNS), 2017–2018 would be years 1–2 (declines phase) and 2019 and later would be years 3+ (post‐WNS arrival phase). Year zero was omitted from summer analyses because bats at maternity colonies may come from different winter hibernation sites (Davis and Hitchcock 1965; Norquay et al. 2013) which may not have been infected in the same year, and thus bats at summer colonies during the arrival may or may not have come from WNS‐positive sites. Most hibernation sites in Wisconsin were infected over a three‐year period from 2014 to 2016, and so by omitting year 0 from analyses at summer sites, we increased the likelihood that bats were coming from affected winter colonies. We calculated the log₁₀ of all predictor variables to normalise and standardise data.

3. Results

3.1. Temporal Trends and Disease Impacts

Prior to the arrival of WNS, summer colonies in our study were stable on the basis of bat numbers from emergence counts (Figure 1). WNS arrived in hibernacula in the study area during winters over a 3‐year period (2014–2016). After P. destructans invaded local hibernacula, causing population declines, growth rates at summer roosts depended strongly on years since pathogen detection, with years 1–3 showing the lowest population growth rates (Figure 1A). Summer colonies declined from pre‐WNS counts on average by 77.7% (±0.02) by year 3 of WNS infection. After initial declines from WNS in years 1 and 2 after disease arrival, growth rates at colonies increased significantly in the following years during the post‐WNS arrival phase (post‐WNS phase compared to the decline phase coef: 0.28 ± 0.037; z‐value: 7.722, p = < 0.001). By year 4 post‐WNS arrival, summer colonies approached stabilisation (log₁₀ λ = 0; Figure 1A). WNS had largely identical effects on winter colonies, causing declines, and counts began to stabilise and increase in winter sites by year 5 post‐WNS arrival (following summer of year 4) (Figure 1B; Table S1).

3.2. Effect of Colony Size on Population Dynamics

We used an interactive linear mixed effects model with log₁₀ annual population growth rates as our response variable, and WNS phase (pre‐WNS, decline, and post‐WNS arrival) interacting with log₁₀ population size to assess whether the effect of population size differed by disease phase on both summer and winter colonies. Prior to the arrival of WNS, summer population size had no effect on yearly population growth rates (log₁₀ colony size: pre‐WNS phase coef: 0.072 ± 0.091, z‐value = 0.785, p = 0.43; Table S2); however, during both the decline and post‐WNS phases, summer yearly population growth rate was significantly positively correlated with log₁₀ population sizes such that smaller colonies experienced more severe declines from WNS (log₁₀ colony size: decline phase coef: 0.25 ± 0.06, z‐value = 4.183, p = < 0.001; log₁₀ colony size: post‐WNS phase coef: 0.25 ± 0.061, z‐value = 4.09, p = < 0.001; Figure 2; Table S2). The effect of population size on growth rates of summer colonies during the declines and post‐WNS arrival phases was significantly different than the effect of population size on growth rates during the pre‐WNS phase (log₁₀ colony size: declines phase coef difference from pre‐WNS phase: 0.175 ± 0.076, p = 0.02; log₁₀ colony size: post‐WNS phase coef difference from pre‐WNS phase: 0.177 ± 0.066, p = 0.007), suggesting a shift to positive density dependence in summer colonies after the arrival of WNS.

FIGURE 2.

The effects of population size on colony growth rates. General linear mixed model of the interaction of log₁₀ yearly population size by phase of infection on log₁₀ population growth rates at (A) 39 M. lucifugus summer maternity colonies and (B) 46 winter hibernacula. Points in panel (A) are log₁₀ of mean colony size from summer counts (each summer colony is counted multiple times over the season), and points in panel (B) are single counts from hibernacula surveys. Points indicate individual colony counts, and lines indicate the model fits (Table S2). Line type indicates significant positive or negative slopes (solid = significant, dashed = non‐significant). Prior to arrival of WNS, there was no statistically clear effect of mean population size on population growth rate in summer or winter. However, during the decline phase (year 1–2 post‐WNS occurrence) and post‐WNS phase (year 3+ post‐WNS occurrence), larger mean colony sizes in summer had higher population growth rates than smaller colonies. In winter, log₁₀ population size had an inverse relationship during the arrival and decline phases, where larger colonies experienced more severe declines than smaller colonies.

At winter sites, smaller colonies had faster growth rates prior to the arrival of WNS, which may be consistent with less room for growth in colonies that were approaching carrying capacity (largest colony 128,000 bats, log₁₀ colony size: pre‐WNS phase coef: −0.05, z‐value = −1.995, p = 0.0461). After the arrival of WNS during the decline phase, larger colonies experienced more severe declines than smaller colonies (log₁₀ colony size: decline phase coef: −0.07 ± 0.032, z‐value = −2.319, p = 0.02). During the post‐WNS phase, as WNS established and population growth rates stabilised, colony size had little effect on population growth rates but did not differ significantly from the effects during the decline phase (log₁₀ colony size: post phase coef difference from decline phase: 0.063 ± 0.04, z‐value = 1.58, p = 0.114; Figure 2). Effects of colony size appeared to be largely independent of other known ecologically important variables, as bat population size in hibernacula was uncorrelated with hibernaculum temperature (variable‐level p = 0.046) and did not improve models of the relationship between population density and population change (variable‐level p = 0.063).

3.3. Impact of Ecological Variables on Colony Recovery

To understand the ecological factors that might also influence the growth of summer bat colonies, we used similarly structured linear mixed effects models with log₁₀ distance to water as a fixed effect. Similar to population size models, annual growth rates showed no relationship with log₁₀ distance to water prior to WNS occurrence (log₁₀ distance to water: pre‐WNS phase coef: 0.024 ± 0.037; z‐value = 0.643, p = 0.52; Table S4). During the decline phase, log₁₀ distance to water had weak effects on yearly population growth rates such that colonies further from water experienced lower growth rates than those closer to water (log₁₀ distance to water: decline phase coef: −0.073 ± 0.035, z‐value = −2.078, p = 0.037; log₁₀ distance to water: post‐WNS phase coef: −0.026 ± 0.029, z‐value = −0.901, p = 0.37; Figure 3; Table S4). We found no relationship between distance to water and overall colony size (correlation coef = −0.07, p = 0.21). We also examined additional roost characteristic variables such as age of roost, type of roost, distance to closest major hibernaculum, and number of roosts in one locality (bat houses), but none had statistically clear effects on summer maternity colony population dynamics (Figures S5–S8; Tables S5–S8). We found a statistically unclear relationship between summer colony size and closest major winter hibernaculum colony size; however, as summer colonies are comprised of bats from multiple winter hibernacula, the link between summer and winter colony size is likely to be weak.

FIGURE 3.

Population growth rates in relation to distance to water. General linear mixed model of the effects of the interaction of log₁₀ distance to water by phase of infection on log₁₀ population growth rates at 39 M. lucifugus summer maternity colonies. Points indicate individual mean colony sizes annually, and lines indicate the model fits (Table S4). Line type indicates significance (solid = significant slope, dashed = non‐significant).

4. Discussion

We found that larger summer colonies had higher population growth rates than smaller colonies during both declines and post‐WNS phases following WNS arrival. This is counter to observed declines from WNS in winter sites where larger colonies experienced lower growth rates than smaller colonies during the initial years of pathogen invasion (decline phase), but there was little effect of colony size during the post‐WNS arrival phase. Accelerated growth at large summer colonies suggests a behavioural and survival advantage to roosting with many other individuals in summer. Large colonies and clustering behaviour can help roosts warm quickly and retain heat. Warmer roosts can decrease gestation and maturation times (Racey and Swift 1981; Tuttle 1976) and reduce metabolic costs (Dechmann et al. 2004; Law and Chidel 2007; Ruczyński 2006), which may be especially important in spring for bats returning from winter hibernacula in poor condition from WNS infection (Fuller et al. 2020). Thus, large summer colonies (warmer roosts) may have supported faster individual recovery from WNS after emergence in spring and higher reproductive success, resulting in a quicker return to stabilising and even positive growth rates. The selection of large summer colonies to take advantage of thermoregulatory benefits while simultaneously choosing small winter colonies to potentially reduce the impacts of P. destructans is evolutionarily complicated for a long‐lived animal like bats. Our results suggest that although WNS is highly seasonal, with infection occurring in winter, density effects in summer roosts where WNS does not manifest may be playing an important role in the recovery of M. lucifugus . For a susceptible‐infected‐susceptible (SIS) disease system like WNS, models predict that total populations will equilibrate at smaller sizes because chronic infections reduce the population for an extended period of time, but that Allee effects can mediate disease impacts by slowing group extinctions (Brandel et al. 2021). Our results suggest that positive density dependence observed in summer colonies during the WNS outbreak ultimately prevails over disease dynamics at winter sites, promoting survival and recovery of M. lucifugus , though colonies may still stabilise at smaller sizes than pre‐WNS populations.

We posit that positive density dependence in summer colonies may be largely due to the thermoregulatory benefits of grouping; however, we lack the data to determine whether other aspects of sociality may also play a role in the recovery of M. lucifugus from WNS. Although sociality in most bat species is poorly understood, colonies are common and may form because of limitations of suitable habitat, physiological demands, and benefits from communication (Kerth 2008). It is possible that, in addition to thermoregulatory benefits, more individuals in colonies may promote better communication about things like insect hatches in spring, increasing foraging success, similar to a finding of C. lupus infected with S. scabiei benefitting from higher hunting success in larger groups (Almberg et al. 2015).

Whether WNS has density‐dependent or frequency‐dependent transmission remains unclear; however, our study shows some effect of host population density on population declines—positive density dependence in summer and negative density dependence in winter during the first years after WNS arrival (Figure 2). Some relationship between host density and transmission of P. destructans has been shown in Midwest United States bat hibernacula where environmental loads of P. destructans shed into the environment were higher in sites with larger M. lucifugus colonies (Laggan et al. 2023). As well, a study using ultraviolet (UV) dust as a surrogate for P. destructans showed within‐site transmission of the UV dust to be density‐dependent (Hoyt et al. 2018). In both cases, transmission of P. destructans varied among bat species, but evidence suggests there may be a density‐dependent transmission component for infection of M. lucifugus, which could have played a role in declines in winter hibernacula in our study, although more research is needed to understand the important results. Ultimately, because P. destructans persists in the environment for long periods of time (> 10 years in the absence of bats; Grimaudo et al. 2022; Hoyt et al. 2014), there is unlikely to be a critical community size at which no transmission will occur.

The rate of recovery for some of our colonies appeared to be higher than can be accounted for with reproduction alone (e.g., one site had a low count of 30 bats in 2017 and 90 bats in 2019). Though the percentage of female M. lucifugus giving birth is high (90%–99%), first‐year overwinter survival of juveniles can be low under some circumstances (Frick, Pollock, et al. 2010; Frick, Reynolds, and Kunz 2010) (20%–45%) and overall winter survival was greatly reduced during WNS invasion (Kailing et al. 2023). A doubling of summer colony estimates within 2 years post‐WNS declines could be due to bats concentrating in fewer, high‐quality roost sites; however, a study of M. lucifugus demographics at summer roost sites in the Northeastern United States revealed that not only did recruitment increase in years since WNS invasion, so too did the reproductive rate of yearling females (Ineson 2020). Increases in recruitment and reproduction post‐WNS impacts coupled with immigration to high‐quality roosts may help account for the return of some summer colonies to near pre‐WNS numbers within just a few years. Female M. lucifugus generally have high philopatry to summer roost sites; however, immigration to non‐natal summer roosts can occur (Norquay et al. 2013; Schorr and Seimers 2021). Such permanent immigration is difficult to document in bats, but it is known in at least one unpublished instance in Wisconsin. Given the sometimes large distances travelled between summer and winter roost sites as well as high winter site fidelity (Norquay et al. 2013), it is unlikely that female M. lucifugus shifting among summer roost sites would alter winter site bat density since relatively scarce hibernation sites likely house bats from many summer roosts. Summer roost shifting that may occur would be more likely to be to sites still within the area served by the same hibernaculum.

Though the effect was modest, our results generally suggest that summer roosts closer to permanent water experienced less severe declines from WNS than those further away, particularly during initial declines after WNS invasion. Myotis lucifugus is known to feed on soft‐bodied aquatic insects like Diptera (Burles et al. 2008; Clare et al. 2011; Whitaker 2004). Roosting in sites close to permanent water may reduce daily commuting costs to foraging areas, and the spring activity of aquatic Diptera (Bouchard and Ferrington 2009; Soszyńska 2004) may provide a crucial food source for bats who often emerge from hibernation in poor condition because of WNS infection (Fuller et al. 2020; McGuire et al. 2017).

Although WNS caused substantial declines in population growth rates of M. lucifugus summer colonies for several years after invasion, the arrival of the disease only resulted in the extirpation of a single M. lucifugus summer colony in our dataset. Growth rates at our 39 colonies began approaching stability as early as year 4 post‐WNS arrival. Early predictions of M. lucifugus extinction were made within a few years of the discovery of WNS and mass mortality events (Frick, Pollock, et al. 2010), which may have biased suspected extinction risk since years 2 and 3 saw the highest declines, and growth rates approach stabilisation starting year 4 (Hoyt et al. 2021; Ineson 2020; Langwig et al. 2012) (Figure 1). As was suggested on the basis of long‐term monitoring of a New York summer colony (Dobony and Johnson 2018), post‐WNS population trends of M. lucifugus summer colonies in Wisconsin were only apparent many years after the arrival of WNS.

Populations of M. lucifugus in our study area as well as in the northeastern United States are persisting despite repeated infection with WNS primarily due to host resistance (Langwig et al. 2017). However, our study suggests that survival and persistence in this species may also have a behavioural component where positive density dependence in summer is playing a role in recovery. The exact mechanism that causes resistance remains unclear, and whether this mechanism may be heritable has important implications for the species in other regions. Such persistence may not be found in populations elsewhere within this species' range such as in the Southeast United States where colony sizes were much lower than in the Northeast and Midwest United States, even before the arrival of WNS (Frick et al. 2015) and possibly had lower genetic diversity on which selection could act (Wilder et al. 2015). Whether WNS leads to extirpation of M. lucifugus in other regions remains to be seen; however, many winter and summer colonies in Wisconsin show signs of trajectories toward recovery.

Since WNS' arrival in winter sites varied, and bats at summer colonies likely came from multiple winter hibernacula (Davis and Hitchcock 1965; Norquay et al. 2013), summer roosts during WNS invasion likely comprised a mix of affected and unaffected individuals. Despite not knowing the exact year of infection of all bats in summer colonies, our growth rates in response to years since estimated WNS arrival mirrored the yearly patterns in growth rates at winter colonies, where arrival of P. destructans was known (Hoyt et al. 2020, 2021) (Figure 1). This suggests that, despite some variation in the precise year that summer colonies could be considered affected because of individuals from multiple hibernacula, the arrival of WNS at summer roosts occurred in nearly the same years as declines in the region. The similar declines in winter and summer colonies in response to the arrival of P. destructans also suggest that in regions where winter hibernacula are unknown or difficult to access, colony dynamics at summer roosts could be used to estimate the arrival of P. destructans and its impacts on bat populations in the area.

Predicting the consequences of infectious diseases on social species can be challenging because of the potential unforeseen negative costs of asocial behaviour in typically colonial species, which is further complicated by strongly seasonal transmission or seasonal differences in sociality. In particular, density‐dependent diseases may have high transmission and mortality during one season, but effects may be mitigated by positive density dependence in another season. For bats affected by WNS, seasonally varying costs of sociality may present trade‐offs that enhance pathogen transmission during winter but also promote higher survival and recruitment of young during summer. Understanding the drivers of social behaviour, and the consequences of asociality, has been difficult to assess, as exemplified by the recent SARS‐CoV‐2 pandemic where isolation successfully reduced transmission and averted deaths but also increased anxiety and depression (Hsiang et al. 2020; WHO 2022). Generally, a better understanding of the important relationships between social behaviour and pathogen dynamics is critical for predicting and controlling population declines and facilitating population recovery.

Author Contributions

H.M.K., K.E.L. and J.R.H. conceptualised the project, analysed data and wrote the manuscript with input from all authors; H.M.K. and J.P.W. collected the summer data; K.E.L., J.R.H., H.M.K. and J.P.W. collected the winter data; all authors contributed critically to draft revision.

Ethics Statement

Work was conducted under the IACUC of Virginia Tech (17–180).

Peer Review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/ele.70203.

Supporting information

Data S1: ele70203‐sup‐0001‐Supinfo.docx.

Kaarakka, H. M. , Hoyt J. R., White J. P., and Langwig K. E.. 2025. “Positive Density Dependence Promotes Host Persistence in the Face of Infectious Disease.” Ecology Letters 28, no. 9: e70203. 10.1111/ele.70203.

Data Availability Statement

Data and code used for these analyses are available in Dryad: https://doi.org/10.5061/dryad.zgmsbccp7.