Abstract

Group-living organisms offer a unique perspective on how environmental gradients influence geographic distributions, as not only the properties of individuals, but also those of their groups interact with the environment to determine a species range. In turn, the ranges of group-living organisms should provide insights on the conditions that favor group versus solitary living. Here we show that rain intensity and predation by ants, factors postulated to exclude subsocial Anelosimus spiders from the lowland tropical rainforest, are greater in this habitat than at higher elevations. We further show that experimentally excluding these factors increases the survival of subsocial Anelosimus colonies when transplanted to the lowland rainforest, but not at their native higher elevation range. While providing a rare experimental test of the simultaneous importance of abiotic and biotic gradients on species range limits, these results provide direct evidence that adverse environmental factors may prevent solitary living and require group living in certain environments.

INTRODUCTION

The geographic ranges of group-living animals provide an opportunity to examine both the traditional roles of biotic and abiotic gradients and the less studied roles of individual and group level properties on niches and ranges. In turn, the ranges of group-living organisms should provide insights on the conditions that favor group versus solitary living. Group living, for instance, is thought to be favored where ecological challenges, such as intense predation, strong competition, and unpredictable environments, make solitary living untenable (Emlen 1982; Avilés 1999). The eusocial mole rats (Family: Bathyergida), for example, occur in extremely arid areas where foraging costs are prohibitive for solitary individuals, whereas solitary congeners occur in areas with greater precipitation and resource abundance (Jarvis et al. 1998). Likewise, cooperatively breeding birds have been found to be associated with greater climatic variability (Jetz and Rubenstein 2011). Other studies have found that abiotic and biotic gradients associated with latitude and elevation may influence arthropod sociality (Purcell 2011; Majer et al. 2013; Guevara and Avilés 2015).

Demonstrating a causal association between environmental variables and the range of particular social systems requires that such variables be manipulated through factor exclusion, transplantation, or both. Only a handful of such experiments have been conducted (e.g. Plateaux-Quénu et al. 2000; Fernández Campón 2008; Purcell and Avilés 2008; Zammit et al. 2008; Keiser et al. 2015), but none involving the simultaneous manipulation of abiotic and biotic factors. Here we used a group-living species in the spider genus Anelosimus to experimentally test the role of specific abiotic and biotic factors on the geographic distribution of spider social systems. The genus Anelosimus includes solitary, subsocial, and social species, where both subsocial and social species build dense, basket-shaped, 3D webs that house adults and offspring (Agnarsson 2006; Avilés 1997; Avilés and Guevara 2017): single families, in the case of the subsocial species, and multiple families, in the case of the social ones. In subsocial species, the offspring cooperate while young, but disperse before reaching adulthood (Yip and Rayor 2014), whereas in social species the offspring mature and mate within the natal nest, giving rise to large groups of highly inbred individuals (Avilés 1997; Lubin and Bilde 2007; Avilés and Purcell 2012). This system is ideal to examine the environmental factors influencing spider sociality because subsocial and social Anelosimus have contrasting ranges—social species are concentrated at low to mid-elevation wet tropical habitats, whereas subsocial species are absent from the lowland tropical rainforest, but extend into higher elevations and latitudes, overlapping with social species in areas of intermediate characteristics (Avilés et al. 2007; Guevara and Avilés 2015).

Avilés et al. (2007) and Guevara and Avilés (2015) argue that these contrasting distributions are the result of 2 separate processes: According to the “prey size” hypothesis, the absence of social species from higher elevations and latitudes may be attributed to a paucity of large insect prey in those areas (Guevara and Avilés 2007; Powers and Avilés 2007), as large insects are required to address scaling challenges of the spiders’ 3D webs (Yip et al. 2008). According to the “predation risk” and “rain intensity” hypotheses, on the other hand, absence of subsocial Anelosimus species from the lowland tropical rainforest may reflect strong disturbance in this habitat due to high predation and intense rains (Purcell and Avilés 2008). The latter are factors that would endanger the mother and damage the dense and presumably expensive webs the offspring depend on. The latter 2 hypotheses are the focus of this study.

Predation, in particular by ants, appears to be greater in the tropics (Schemske et al. 2009) and, within the tropics, at lower elevations, particularly in the lowland rainforest (Janzen 1973; Samson et al. 1997; Bito et al. 2011). Greater rates of predation may favor group living for several reasons. Maternal death during offspring development may leave the offspring unattended and favor nest sharing by multiple females (Avilés et al. 2007; Jones et al. 2007). Larger social groups are also housed in larger nests, which allow more spiders to be away from the periphery where predation risk is greater (Rayor and Uetz 1993). Individuals in larger colonies would also benefit from decreased individual risk due to predator dilution (Uetz and Hieber 1994) and from early warning and predator detection (Uetz et al. 2002). The intense and frequent rains of the rainforest, on the other hand, may cause frequent damage to the dense, and presumably costly 3D Anelosimus webs (Avilés et al. 2007). Individuals in larger social groups would benefit from sharing the costs of web building (Riechert et al. 1986) and from a faster recovery following intense rain events.

To test the “predation risk” and “rain intensity” hypotheses, we 1) quantify differences in ant predation threat and rain intensity across an elevational gradient, and 2) use a fully factorial rain and predator exclusion experiment to determine the effect of these factors on colonies of a subsocial Anelosimus species when transplanted both within its native range, as a control, and to the lowland tropical rainforest. We show that both rain intensity and predation by ants increase monotonically with proximity to the rainforest and that the exclusion of these factors improves survival of subsocial Anelosimus colonies when transplanted to the rainforest, but not at their native higher elevation site. These findings thus shed light on the factors that may be responsible for subsocial Anelosimus naturally occurring at higher elevations and latitudes, but being absent from the lowland rainforest where congeneric species that form large social colonies thrive.

MATERIALS AND METHODS

Rain intensity and predation threat across elevation

We obtained rain intensity data for years 2011–2014 from NASA’s Tropical Rainfall Measuring Mission Project, TRMM (http://disc.gsfc.nasa.gov) for 20 locations along 2 elevational gradients (228–2100 m) in Eastern Ecuador (Online Supplementary Material Table 1). After processing the raw data from TRMM’s 3B43v6 product (see Online Supplementary Material) we obtained monthly rain rate averages (mm/h), which we classified into quarters (February–April; May–July; August–October; November–January) based on similarity in precipitation regimes across months. We first analyzed the data for the north and south transects together, but opted for separate analyses upon encountering a significant transect × elevation interaction (F1,254 = 128.8, P < 0.0001; Online Supplementary Material Table 2). Data for each transect was then analyzed with a linear model that included elevation, year, quarter, elevation2, and their interactions, and used the Akaike Information Criterion, corrected for finite sample sizes (AICc), to choose the most parsimonious model that best explained the data.

We additionally manually obtained rain intensity and predation pressure data at 4 of the sites included in the above transects, all in the Napo Province: the Jatun Sacha Biological Station, near Tena (S 1.07° W 77.61°, 400–430 m elevation); areas along the road to Loreto, near Guagua Sumaco (S 0.71° W 77.59°, 1000–1200 m); areas along the Baeza–Tena road, near Cocodrilo (S 0.65°, W 77.88°; 1780–1950 m); and the Yanayacu Biological Station, near Cosanga (S 0.60°, W 77.87°, 2100 m). These are areas where Anelosimus spp of different social systems occur, with Jatun Sacha containing only social species (A. eximius, A. domingo, and A. rupununi); Sumaco, both social and subsocial (A. eximius and A. elegans, respectively); Cocodrilo, social and subsocial (A. guacamayos and A. elegans, respectively); and Yanayacu, only subsocial (A. baeza, which occasionally may also be found at Sumaco and Cocodrilos) (Avilés et al. 2007).

At each location, we used rain gauges to manually measure the rate of rainfall (mm/h) during bouts of intense rain taking place during the time we were at each field locality (n = 14 over 54 days, at 400 m elevation; n = 20, over 57 days, at 1000 m; n = 2, over 3 days, at 1800; and n = 4, over 3 days, at 2100 m, rain gauge 5 inch/12 cm standard model). After averaging the rainfall amount from all gauges used for a given rain event (1–2 used), we performed a regression of the average rain rate by elevation, weighted by the number of replicates per rain event.

To understand predation threat, mainly from crawling predators, ants, in particular, at each of the 4 locations we used as baits high-protein drained tuna (June–July 2013 and 2014) and live spiders (June–July 2014). In both cases, baits were affixed to the vegetation at 1–2 m from the ground and at least 3 m from each other, in areas similar to those occupied by the local social or subsocial spiders. Sets of 8–15 tuna baits (drained tuna packed into 1.5 mL centrifuge tubes) were checked at least every 10 min for up to 2 h or until each was found. Drained tuna baits are commonly used to attract predatory or scavenging ants (Bito et al. 2011). Spider baits consisted of live spiders of the local species glued onto small pieces of vegetation (Krazy Glue®, Westerville, OH, USA)—A. eximius at 400 m and 1000 m; A. elegans and A. guacamayos at 1780–1950 m; A baeza at 2100 m. Sets of 9–20 spiders were set up either in the morning (8:30 to 10:15) or early evening (16:30 to 18:20) and monitored at least every 2 h during the day, but not overnight. Spider baits were monitored for 9–17 h at the lowest elevation (400 m), where baits were found quickly, but up to a full day (22–27 h) at higher elevations, where they were found slowly. We noted baits as “discovered” if there were ants on the bait (body of the spider, rim or within the tube) or if parts of the body, particularly the abdomen or legs, were missing. We also noted baits discovered by wasps, as wasps are also predators of spiders (Uetz and Hieber 1994, Rayor 1996, Blamires et al. 2013). For both types of baits, we performed Kaplan–Meier survival analyses (JMP 10.02) of the time until baits were found, with elevation as the explanatory variable and baits that were not found during the observation period considered right censored.

Subsocial spider colony performance at local and transplant sites

To test the hypothesis that intense rains and high predation are at least partly responsible for the absence of subsocial Anelosimus from the lowland tropical rainforest, we performed a rain and predator (crawling predators) exclusion experiment on transplanted nests of the subsocial A. elegans. We chose this species as it has populations at elevations (900–1200 m) (Avilés et al. 2007) low enough to overlap with the lowland rainforest social species A. eximius. We thus expected it would be physiologically better able to withstand lowland tropical rainforest conditions when transplanted to this habitat than an alternative subsocial species, A. baeza, which typically occurs at higher elevations in this area of Ecuador.

The experiment involved transplanting nests of A. elegans to the lowland tropical rainforest (Jatun Sacha, 400 m elevation) and, as a control, within its native mid-elevation site (Wawa Sumaco, 1000 m), hereafter referred to as Lowlands and Mid-elevation. On the transplanted nests at both sites, we used a blocked fully-factorial rain and predator exclusion design where No Rain treatments had a tarp (100 cm × 64 cm) placed above the nest and No Predator treatments had a disk (diameter: 12 cm) covered in Tanglefoot (Tanglefoot Company, Grand Rapids, MI, USA) surrounding the stem of the plant where the nest was found. We confirmed the effectiveness of the treatments by placing rain gauges both under and outside the tarps and by recording the presence of ants and other crawling predators on protected vs. unprotected nests. If rain and predators were partly responsible for preventing this subsocial spider from colonizing the lowland rainforest, we expected that excluding these factors would increase the survival and quality of transplanted nests to this habitat, but would have no effect at the native mid-elevation habitat where the species naturally occurs. Our units of analyses were thus the individual nests, blocked in groups of 4 (one nest per treatment), with rain and predators being the factors tested, rather than any particular location or elevation.

Nests of A. elegans were collected from sites between 900–1100 m elevation along the “Via Loreto” road near Wawa Sumaco (see coordinates above). To standardize substrates across treatments and localities, and minimize disturbance to the spiders. we clipped intact nests along with their substrate and installed them on potted plants of an Asteracea commonly used by the spiders (Baccharis trinervis Pers., C. Cerón, pers. comm.). Once the colonies had incorporated the new substrate into their nests, we were able to transplant them to their new location with minimal disturbance. We visually inspected the content of the nests and grouped them in blocks of 4 based on their size and the predominant age category of the spiders they contained (singe adult female; juveniles instar 1–2; juveniles instar 3–4; subadult 1; subadult 2 individuals). Colony size (range 1–83) was the maximum number of individuals seen in the nests at any time during the census period, excluding offspring eclosing from newly laid egg sacs.

For the transplants we chose sites as those occupied by naturally occurring colonies of the local Anelosimus species (A. eximius, in the lowlands, A. elegans, at the mid-elevation). Nests within blocks were placed at least 5 m from each other and from any existing natural nest and randomly assigned to one of the treatments. We had a total of 63 nests transplanted to the lowlands, in 16 blocks (one of the blocks was missing the No Rain & Predator treatment), and 60 nests transplanted within the native mid-elevation site, in 15 blocks. Due to a limited population of nests, we reused 8 nests originally transplanted within the native habitat that had high survival. The previous procedures apply to these colonies, with 4 transplanted within the mid-elevation habitat and 4 to the lowlands.

To prevent transplant failure, we closely monitored the nests during the first 3 days after establishment, returning to them any spiders that may have prematurely dispersed. The colony population was then censused every 2 to 3 days for 28 days. Censuses took place at night when spiders were out of their refugia and thus easier to score. We counted spiders by instar, noting any new or eclosed egg sacs, and checked the areas surrounding the nests for any newly founded colonies (propagules) product of dispersal and colonization. We also quantified the amount of new webbing added to the capture area and basket of the nests (see legend Figure 2, for details) as an indicator of the health of the colony. New webbing is easy to identify due to its clean, white color compared to older webbing in the original nest. We estimated colony survival time as the difference between the time a nest was transplanted and the first of 3 consecutive checks with zero spiders.

We performed parametric survival analyses with a Weibull distribution, with an initial full model containing as factors location, colony age structure (the 5 categories listed above, ordered), rain exclusion, predator exclusion, and their subsequent 2-, 3- and 4-way interactions. Colony ID, nested within block, was included as a random effect in this and other models. We used the Akaike Information Criterium, corrected for finite sample sizes (AICc), to determine the set of factors and their interactions that most parsimoniously explained the data, followed by likelihood ratio tests to determine the significance of each of the terms in the reduced model (Online Supplementary Material Table 2). As there was a significant interaction between location and one of the treatments (location × rain interaction: df = 1, X2 = 6.40, P = 0.01) and between location and age structure (df = 4, X2 = 12.11, P = 0.016), we ran separate analyses per location. For these we maintained age structure in the model given its significance in explaining the patterns (df = 4, X2 = 34.7, P < 0.0001; Online Supplementary Material Table 2). We did not include colony size in models with age structure, as the 2 variables were highly correlated (colonies with older instars contained fewer individuals, P < 0.0001, R2 = 0.79). Instead we ran separate analyses where colony size, rather than age structure, was the covariate. Colonies that survived to the end of the observation period were coded as right-censored. Data on the amount of new webbing, coded on a qualitative scale from 1–4 (see legend Figure 2), were analyzed with a linear mixed effects model, with location, log group-size, rain exclusion, and their interactions as fixed effects, and Colony ID nested within block as the random effect. Data points were weighted by the number of days colonies were observed.

RESULTS

Rain intensity and predation threat across elevation

Satellite rain intensity data—quarterly averages for years 2011–14—showed a general pattern of rain rate (mm/h) decreasing from low to high elevations for all quarters and both transects, with a slight peak at the second (~400 m), rather than the lowest elevation (Figure 1a and b). Elevation was the most important factor explaining rain rate patterns at the southern transect (Logworth 21.17, df = 1, P < 0.0001 vs. 6.78, df = 3, P < 0.0001, for the next factor, Online Supplementary Material Table 3). As there was greater variability among quarters and years at the northern transect, here elevation was second in importance to a year-quarter interaction (Logworth 11.20, df = 1, P < 0.001, for elevation, vs. 12.53, df = 9, P < 0.0001, for the interaction; Online Supplementary Material Table 3). Our manually collected data (June–July 2013–2014) for 4 localities along the northern transect (arrows in Figure 1a) was consistent with a pattern of monotonically decreasing rain rate between 400 m and 2100 m elevation (R2 = 0.14, F1,38 = 6.3, P = 0.017; Figure 1c, inset).

Figure 1

(a and b) Rainfall rate (mm/h) data from NASA’s Tropical Rainfall Measuring Mission Project, TRMM (http://disc.gsfc.nasa.gov) for 2 elevational gradients in Eastern Ecuador (see Online Supplementary Material for individual localities). Shown are quarter averages (February–April; May–July; August–October; November–January) for years 2011–2014. Arrows show localities where predation threat and transplant experiments were conducted. (c, insert) Manually collected rainfall rate data for June–July 2013, 2014, for a total of 40 rain events. Mean ± standard deviation shown. (c and d) Kaplan-Meier survival analyses showing the probability of (c) tuna baits and (d) live spider baits remaining undiscovered overtime by ants and other predators. Number of baits used at each location are as follows: 400 m (tuna n = 153; spider 61), 1000 m (tuna n = 110; spider 40), 1800 m (tuna n = 87; spider 40), 2200 m (n = 50; spider 34).

Figure 1

(a and b) Rainfall rate (mm/h) data from NASA’s Tropical Rainfall Measuring Mission Project, TRMM (http://disc.gsfc.nasa.gov) for 2 elevational gradients in Eastern Ecuador (see Online Supplementary Material for individual localities). Shown are quarter averages (February–April; May–July; August–October; November–January) for years 2011–2014. Arrows show localities where predation threat and transplant experiments were conducted. (c, insert) Manually collected rainfall rate data for June–July 2013, 2014, for a total of 40 rain events. Mean ± standard deviation shown. (c and d) Kaplan-Meier survival analyses showing the probability of (c) tuna baits and (d) live spider baits remaining undiscovered overtime by ants and other predators. Number of baits used at each location are as follows: 400 m (tuna n = 153; spider 61), 1000 m (tuna n = 110; spider 40), 1800 m (tuna n = 87; spider 40), 2200 m (n = 50; spider 34).

Matching our predictions, both the tuna and spider baits were discovered by predators significantly faster at lower elevations (Log-Rank: X2 = 119.5, for tuna, 86.3, for live spiders, df = 3, P < 0.0001) (Figure 1c and d). Specifically, after 60 min of being set only 50% of the tuna baits remained at the 400 m elevation, whereas >90% remained undiscovered until the end of the observation period at the 2 highest elevations (Figure 1c). Spider baits took longer to be discovered, ~14 h at 400 m vs. 17 h at 1000 m, for a 50% discovery, whereas >50% still remained after 24 h at the 2 higher elevations (Figure 1d).

Subsocial spider colony performance at local and transplant sites

Location and colony age structure were the main factors explaining the persistence of transplanted colonies in the reduced global model (lowest AIC) that considered the 2 locations simultaneously (Online Supplementary Material Table 4). We found that, whereas about 80% of the transplanted colonies still persisted after 25 days at the native mid-elevation site, most were gone by this time in the lowlands (Figure 2), a difference that was highly significant (df = 1, X2 = 12.7, P = 0.0004). The reduced global model also contained significant interactions between location and at least one of the treatments (location-rain protection interaction: df = 1, X2 = 6.4, P = 0.01) and with age structure, plus an interaction between the treatments (Online Supplementary Material Table 4). When looking at the locations separately (Table 1), based on AIC values, only age structure was necessary to explain persistence at the mid-elevation, whereas in the Lowlands a model containing the 2 treatments, their interaction, and age structure was the minimal model required (Table 1).

Figure 2

(a–d): Survival probability of colonies in each treatment group at the native mid-elevation and lowlands transplant locations. Overall survival was lower at the Lowlands (b) than at the Mid-elevation sites (a). In the Lowlands, colonies with at least No Rain and/or No Predators (crawling) had higher survival than control colonies (Rain and Predators), but there were no differences among treatments at the mid-elevation. (c) and (d): The average rating of new web added to each nest against its log Group Size in the Lowlands (d) and at the Mid-elevation sites (c). Regression lines show all No Rain (solid line) and Rain (dashed line) treatments combined. Lowlands (n = 63) and Mid-elevation (n = 60). Rating scale is as follows: 1) no new visible prey capture lines or webbing on the basket, 2) less than 5 new prey capture lines, no visible change to the basket, 3) more than 5 prey capture lines, some new webbing on the basket, but not a full basket, 4) significant new and dense prey capture lines and new basket and/or moved totally to a new plant.

Figure 2

(a–d): Survival probability of colonies in each treatment group at the native mid-elevation and lowlands transplant locations. Overall survival was lower at the Lowlands (b) than at the Mid-elevation sites (a). In the Lowlands, colonies with at least No Rain and/or No Predators (crawling) had higher survival than control colonies (Rain and Predators), but there were no differences among treatments at the mid-elevation. (c) and (d): The average rating of new web added to each nest against its log Group Size in the Lowlands (d) and at the Mid-elevation sites (c). Regression lines show all No Rain (solid line) and Rain (dashed line) treatments combined. Lowlands (n = 63) and Mid-elevation (n = 60). Rating scale is as follows: 1) no new visible prey capture lines or webbing on the basket, 2) less than 5 new prey capture lines, no visible change to the basket, 3) more than 5 prey capture lines, some new webbing on the basket, but not a full basket, 4) significant new and dense prey capture lines and new basket and/or moved totally to a new plant.

Table 1

Results of parametric survival analyses (Weibull distribution) of the outcome of a fully factorial rain and predator protection experiment on colonies of the subsocial spider Anelosimus elegans when transplanted within their native mid-elevation site and to the lowland rainforest. Bold terms form part of the model with the lowest AICc (delta-AIC[listed-minimal] = 20.9, for the Mid-elevation; 10.5, for the Lowlands).

Location Source LogWorth NParm DF LR-X2 P > X2  
Mid-elevation (native habitat) age structure 4.00 ++++++++++ 4 4 22.97 0.0001  
predators × age structure 1.32 +++ 7.92 0.0477  
Predators 0.65 ++ 1.47 0.2250 
rain × predators 0.37 0.64 0.4231  
Rain 0.30 0.46 0.4990 
rain × age structure 0.02  0.64 0.9584  
Lowlands (transplant) Rain 2.08 ++++++ 1 1 6.97 0.0083  
age structure 2.07 ++++++ 4 4 13.62 0.0086  
rain × predators 1.91 +++++ 1 1 6.26 0.0124  
predators × age structure 1.63 ++++ 9.47 0.0237  
rain × age structure 0.53 4.90 0.2978  
Predators 0.19  1 1 0.21 0.6467 
Location Source LogWorth NParm DF LR-X2 P > X2  
Mid-elevation (native habitat) age structure 4.00 ++++++++++ 4 4 22.97 0.0001  
predators × age structure 1.32 +++ 7.92 0.0477  
Predators 0.65 ++ 1.47 0.2250 
rain × predators 0.37 0.64 0.4231  
Rain 0.30 0.46 0.4990 
rain × age structure 0.02  0.64 0.9584  
Lowlands (transplant) Rain 2.08 ++++++ 1 1 6.97 0.0083  
age structure 2.07 ++++++ 4 4 13.62 0.0086  
rain × predators 1.91 +++++ 1 1 6.26 0.0124  
predators × age structure 1.63 ++++ 9.47 0.0237  
rain × age structure 0.53 4.90 0.2978  
Predators 0.19  1 1 0.21 0.6467 

Factors listed in order of importance within each of the sites based on likelihood ratio tests that compare the full model, as shown, with a model lacking the term in question. LogWorth is calculated as (−log10(P)), where the P values were derived from the Likelihood-ratio chi-square test. Models shown lack the 3-way interaction, which was not significant at either site, but maintain other non-significant terms for the purpose of comparison between sites. “^” denotes terms with contained effects above them.

In the lowlands, colonies with at least rain or predators excluded (i.e. treated colonies combined) had higher survival than controls (df = 1, X2 = 7.0, P = 0.008), an effect that was absent at the mid-elevation (df = 1, X2 = 0.94, P = 0.33; Figure 2). When the treatments were analyzed separately (Table 1), only rain exclusion was significant on its own at the lowlands. At this site there was, however, a significant interaction between the treatments (df = 1, X2 = 6.26, P = 0.01), as the effect of predator removal could only be detected in colonies that were covered from the rain; those exposed to the rain did so poorly that also being exposed to predators could not make things any worse.

That colony age structure was the primary factor explaining colony persistence at the mid-elevation (Table 1) is consistent with dispersal being the process responsible for the dissolution of colonies at this site. This is also shown by the fact that transplanted nests at the native range were increasingly more likely to disappear as offspring reached older instars (Figure 3). At the lowlands, in contrast, colonies of all age structures failed (Figure 3), as would be the case if extinction, rather than dispersal, had been responsible for colony disappearance at this site. Consistent with these observations, 40 newly established nests were detected surrounding 17 colonies that dispersed at the native range, whereas no newly founded nests were detected surrounding transplanted colonies in the lowlands.

Figure 3

Failure probability at 14 days for colonies of the subsocial Anelosimus elegans of different age structures (1 = singe adult female; 2 = juveniles instar 1–2; 3 = juveniles instar 3–4; 4 = subadult 1; 5 = subadult 2 individuals) and all treatments combined when transplanted within their native mid-elevation site or to the lowland rainforest. Shown are means ± confidence intervals.

Figure 3

Failure probability at 14 days for colonies of the subsocial Anelosimus elegans of different age structures (1 = singe adult female; 2 = juveniles instar 1–2; 3 = juveniles instar 3–4; 4 = subadult 1; 5 = subadult 2 individuals) and all treatments combined when transplanted within their native mid-elevation site or to the lowland rainforest. Shown are means ± confidence intervals.

With colony size (log-transformed), rather than age structure, in the models, larger colonies had a significantly diminished probability of failure at both locations, but more dramatically so in the lowlands (effect of log-colony size: Wald-X2 = 10.3, P = 0.0013, at the mid-elevation, and X2 = 27.6, P < 0.0001, in the lowlands). In the lowlands larger colonies exposed to the rain were significantly less likely to fail than smaller ones (interaction rain x colony size in the lowlands: Wald-X2 = 4.3, P = 0.039), whereas no significant interaction (P > 0.05) was found at the native mid-elevation site.

We found that location, rain exclusion, log group size, and the interaction between location and rain exclusion significantly affected the quantity of new webbing visible in the nests (Figure 2c and d and Table 2). At both locations, the amount of new webbing increased with colony size. At the native mid-elevation site, however, colonies exhibited the same amount of new webbing regardless of whether they were covered from the rain, whereas in the Lowlands protected colonies exhibited significantly more new webbing (Figure 2c and d). There was further a trend for larger colonies exposed to the rain in the Lowlands to build almost as much new webbing as protected colonies of all sizes (Figure 2d).

Table 2

Results from a linear mixed effects model with the fixed effects of location, log group size, and rain exclusion and their interactions on the quantity of new webbing built by spider colonies. Shown (P values in bold) model with the lowest AICc (delta AICc[full-minimal] = 7.47).

Effect DF, DF den F P 
Location 1, 79.0 7.3 0.008 
Log group size 1, 84.2 37.3 <0.0001 
Rain exclusion 1, 80.0 8.9 0.004 
Location: rain exclusion 1, 80.1 8.9 0.004 
Effect DF, DF den F P 
Location 1, 79.0 7.3 0.008 
Log group size 1, 84.2 37.3 <0.0001 
Rain exclusion 1, 80.0 8.9 0.004 
Location: rain exclusion 1, 80.1 8.9 0.004 

Colony ID nested within block was the random effect.

DISCUSSION

The range and geographical distribution of animal social systems should provide clues as to the factors that favor group over solitary living, whereas manipulation of such factors should provide evidence of their causal role in shaping sociality. Using a combined factor-exclusion and transplant experiment we test the “predation risk” and “rain intensity” hypotheses as explanations for the absence of subsocial species of the spider genus Anelosimus from the lowland tropical rainforest, where social congeners thrive. We show that ant predation risk and rainfall intensity intensify with proximity to the lowland tropical rainforest and that experimentally excluding these factors improves the survival of subsocial spider colonies transplanted to the lowland tropical rainforest, but not at their native higher elevation site.

We first show quantitatively that rate of rainfall (mm/h) and the abundance of predatory ants increase with proximity to the rainforest (Figure 1), corroborating preliminary data obtained by Purcell and Avilés (2008). We note that we focused on rate of rainfall as the variable to be measured, rather than total annual rainfall, as we expect that a given quantity of rain falling over a short period of time would have greater force, and thus be more damaging to webs, than the same amount of rain falling over an extended period of time. Our results on bait encounter rate by ants add to the growing body of evidence that ant abundance decreases with elevation (Samson et al. 1997; Bruhl et al. 1999; O’Donnell et al. 2010), which in our case is evidence of greater risk of ant predation in the lowland rainforest as we targeted predatory ants.

Consistent with the hypothesis that high predation rates and strong rains may prevent the colonization of the lowland tropical rainforest by subsocial Anelosimus species, the survival of transplanted colonies in this habitat was not only significantly lower than at the native mid-elevation habitat, but also significantly improved by the exclusion of these factors (Figure 2a and b). Furthermore, the fact that these factors did not have an effect at the native habitat, either in terms of colony survival or quantity of webbing (Figure 2), is evidence that neither high risk of predation nor strong rains are limiting within the native range. The fact that survival of protected colonies in the lowlands was still lower than in the native habitat (Figure 2a and b), however, suggests that factors unaccounted for, such as flying or jumping predators, which would not have been excluded with our predator exclusion method, may further impact survival in the lowland rainforest. Colony age structure, which correlates with dispersal timing in subsocial spiders (Avilés and Gelsey 1998), was the main explanatory variable for the disintegration of colonies at the native mid-elevation site, but only a secondary factor in the Lowlands (Figure 3), demonstrating that factors above and beyond dispersal led to colony collapse at the latter habitat. Furthermore, whereas dispersing colonies at the native site produced successful propagules, dispersing individuals in the Lowlands failed to become established.

Studies on social Anelosimus in the lowland rainforest are consistent with the hypothesis that living in groups may be protective against conditions that make solitary living untenable for some types of organisms in certain environments. Thus, the social A. eximius and A. domingo inhabit the lowland rainforest where they form colonies containing hundreds, thousands or tens of thousands of individuals housed in dense 3D webs. These species may also establish nests as solitary females (Avilés 1997), with the vast majority of these, however, failing (9 out of 10 in A. eximius, Avilés and Tufiño 1998), and only colonies above a certain size having positive growth (Avilés and Tufiño 1998; Hart and Avilés 2014). Consistent with the expectation of a protective role of larger colony sizes against adverse conditions in the lowland rainforest, larger colonies of the transplanted subsocial species maintained better webbing and survived longer than smaller ones (Figure 2), whereas single dispersers failed. Colony size and dispersal tendencies for species in the genus Anelosimus are thus likely to be under particularly strong selection in this habitat, as these traits appear to modulate the impact of the environment. This is further suggested by the fact that Anelosimus species at higher elevations, where rain is less intense and predators are fewer, maintain smaller colony sizes and occur more frequently as solitary individuals, both within (Purcell and Avilés 2007) and among species (Avilés et al. 2007). There is thus ample evidence to suggest that large groups must be maintained for Anelosimus spiders to inhabit the lowland tropical rainforest, but that selection for large groups is relaxed away from the rainforest.

Data from other spider genera are consistent with the hypothesis that group living may be protective against adverse environmental conditions and that those conditions may involve strong rains and/or strong predation. A number of other social spider genera—Agelena, Aebutina, Parasteatoda (formerly Achaearaneae), Theridion, and Tapinillus—also occupy wet tropical environments where both of these factors likely play a role (Avilés and Guevara 2017). Riechert et al. (1986), for example, showed that small Agelena consociata colonies in the rainforests of Gabon are more likely to go extinct following periods of strong rains and that individuals were preyed upon within an hour of being artificially released. Social species in the genus Stegodyphus, however, occupy dry savanna habitats (Majer et al. 2013) where strong rains are unlikely to be of major importance. Instead, predation, in particular by ants, appears critical in these habitats. Henschel (1998) and Keiser et al. (2015), for instance, showed that group-living Stegodyphus in Namibia are better able to withstand attacks by predatory ants than solitary individuals. Furthermore, data by Bilde et al. (2007) suggest that, on average, the lifetime reproductive success of individuals in very small colonies is below replacement value in populations in Namibia. Solitary Stegodyphus, however, do occur in areas where social species are found. Majer et al. (2013) analyzed the distribution of group- and solitary-living Stegodyphus in Africa and Asia and showed that social species are absent from areas of low vegetation productivity where subsocial species occur, but that the latter do overlap with social species in the more productive areas. It would be interesting to determine, however, whether solitary species are missing from certain areas where social species occur or whether they have adopted alternative strategies to cope with adverse conditions in those areas. In this context, it is interesting that 2 solitary Stegodyphus in South Africa have been found as kleptoparasites in colonies of Stegodyphus dumicola (Wickler and Seibt 1988). The association between the social species and high vegetation productivity, on the other hand, appears to respond to a greater availability of insects and perhaps also larger insect sizes in these areas (Majer et al. 2013).

As studies of range limits and range shifts burgeon, social animals can offer a unique perspective on how the environment influences niches and distributions. Here we present experimental evidence of the simultaneous role of abiotic and biotic factors on the range limit of a social animal. That congeneric social relatives of our study species do occur in environments that seem out of bounds to less social ones further suggests that group-living species can modulate the effects of the environment through their unique demographic properties and thus allow colonization of areas that may be unsuitable to solitary animals.

SUPPLEMENTARY MATERIAL

Supplementary data are available at Behavioral Ecology online.

Funding

This work was supported by James S. McDonnell Foundation (220020227, USA) and NSERC Discovery grants (NSERC CANADA RGPAS 446016-13 and RGPIN 261354-13) to L.A., with additional funding from a UBC Zoology Graduate Fellowship and a NSERC BRITE Fellowship to C.H.

Data accessibility: Analyses reported in this article can be reproduced using the data provided by Hoffman and Avilés (2016).

For insightful comments throughout this project, we thank A. Angert, P. Fernandez-Fournier, A. González, J. Guevara, C. Harley, J. Jankowski and R. Sharpe. C Hoffman collected the field data, with assistance from M.A. Leclerc, P. Fernandez-Fournier, M. Robertson, and E. Calvache. Special thanks to J. Guevara for obtaining and processing NASA’s remote sensing data on rain rate. We thank the Department of Biology of the Pontificia Universidad Católica del Ecuador for sponsoring our research in Ecuador, the Jatun Sacha Biological Station, Yanayacu Biological Station, and the Sumaco National Reserve for logistic support in the field, and the Ministerio del Ambiente del Ecuador for research permits.

REFERENCES

Agnarsson
I
.
2006
.
A revision of the New World eximius lineage of Anelosimus (Araneae, Theridiidae) and a phylogenetic analysis using worldwide exemplars
.
Zool J Linn Soc
 .
146
:
453
593
.
Avilés
L
.
1997
.
Causes and consequences of cooperation and permanent- sociality in spiders
. In:
Choe
JC
Crespi
BJ
, editors.
The evolution of social behaviour in insects and arachnids
 .
Cambridge
:
Cambridge University Press
. pp.
476
498
.
Avilés
L
.
1999
.
Cooperation and non-linear dynamics : an ecological perspective on the evolution of sociality
.
Evol Ecol Res
 .
1
:
459
477
.
Avilés
L
Gelsey
G
.
1998
.
Natal dispersal and demography of a subsocial Anelosimus species and its implications for the evolution of sociality in spiders
.
Can J Zool
 .
76
:
2137
2147
.
Avilés
L
and
Guevara
J
.
2017
.
Sociality in spiders
. In:
Rubenstein
D
Abbot
P
, editors.
Comparative social evolution
 .
Cambridge (UK)
:
Cambridge University Press
. pp.
188
223
.
Avilés
L
Purcell
J
.
2012
.
The evolution of inbred social systems in spiders and other organisms: From short-term gains to long-term evolutionary dead ends?
Adv Stud Behav
 .
44
:
99
133
.
Avilés
L
Tufiño
P
.
1998
.
Colony size and individual fitness in the social spider Anelosimus eximius
.
Am Nat
 .
152
:
403
418
.
Avilés
L
Agnarsson
I
Salazar
PA
Purcell
J
Iturralde
G
Yip
EC
Powers
KS
Bukowski
TC
.
2007
.
Altitudinal patterns of spider sociality and the biology of a new midelevation social Anelosimus species in Ecuador
.
Am Nat
 .
170
:
783
792
.
Bilde
T
Coates
KS
Birkhofer
K
Bird
T
Maklakov
AA
Lubin
Y
Avilés
L
.
2007
.
Survival benefits select for group living in a social spider despite reproductive costs
.
J Evol Biol
 .
20
:
2412
2426
.
Bito
D
Burwell
CJ
Kitching
RL
.
2011
.
Predator pressure, herbivore abundance and plant damage along a subtropical altitudinal gradient
.
Mem Queensl Mus
 .
55
(
2
):
451
462
.
Blamires
SJ
Hou
C
Chen
LF
Liao
CP
Tso
IM
.
2013
.
Three-dimensional barricading of a predatory trap reduces predation and enhances prey capture
.
Behav Eco Sociobiol
 .
67
:
709
714
.
Bruhl
C
Mohamed
M
Linsenmair
KE
.
1999
.
Altitudinal distribution of leaf litter ants along a transect in primary forests on Mount Kinabalu, Sabah, Malaysia
.
J Trop Ecol
 .
15
:
265
277
.
Emlen
ST
.
1982
.
The evolution of helping. I. An ecological constraints model
.
Am Nat
 .
119
:
29
39
.
Fernández Campón
F
.
2008
.
More sharing when there is less: insights on spider sociality from an orb-weaver’s perspective
.
Anim Behav
 .
75
:
1063
1073
.
Guevara
J
Avilés
L
.
2007
.
Multiple techniques confirm elevational differences in insect size that may influence spider sociality
.
Ecology
 .
88
:
2015
2023
.
Guevara
J
Avilés
L
.
2015
.
Ecological predictors of spider sociality in America
.
Glob Ecol and Biogeogr
 .
24
:
1181
1191
.
Hart
T
Avilés
L
.
2014
.
Reconstructing local population dynamics in noisy metapopulations- the role of random catastrophes and Allee effects
.
PLoS One
 .
9
(
10
):
e110049
.
Henschel
JR
.
1998
.
Predation on social and solitary individuals of the spider Stegodyphus dumicola (Araneae, Eresidae)
.
J Arach
 .
26
:
61
69
.
Hoffman
CR
Avilés
L
.
2016
.
Data from: rain, predators and spider sociality: a manipulative experiment
.
Dryad Digital Repository
 . http://dx.doi.org/10.5061/dryad.m78b6.
Janzen
DH
.
1973
.
Sweep samples of tropical foliage insects: effects of seasons, vegetation types, elevation, time of day, and insularity
.
Ecology
 .
54
:
687
708
.
Jarvis
JUM
Bennett
NC
Spinks
AC
.
1998
.
Food availability and foraging by wild colonies of Damaraland mole-rats (Cryptomys damarensis): implications for Sociality
.
Oecologia
 .
113
:
290
298
.
Jetz
W
Rubenstein
DR
.
2011
.
Environmental uncertainty and the global biogeography of cooperative breeding in birds
.
Curr Biol
 .
21
:
72
78
.
Jones
T
Riechert
S
Dalrymple
S
Parker
P
.
2007
.
Fostering model explains variation in levels of sociality in a spider system
.
Anim Behav
 .
73
:
195
204
.
Keiser
CN
Wright
CM
Pruitt
JN
.
2015
.
Warring arthropod societies: Social spider colonies can delay annihilation by predatory ants via reduced apparency and increased group size
.
Behav Processes
 .
119
:
14
21
.
Lubin
Y
Bilde
T
.
2007
.
The evolution of sociality in spiders
.
Adv Stud Behav
 .
37
:
83
145
.
Majer
M
Svenning
JC
Bilde
T
.
2013
.
Habitat productivity constrains the distribution of social spiders across continents - case study of the genus Stegodyphus
.
Front Zool
 .
10
:
9
.
O’Donnell
S
Kaspari
M
Kumar
A
Lattke
J
Powell
S
.
2010
.
Elevational and geographic variation in army ant swarm raid rates
.
Insectes Sociaux
 .
58
:
293
298
.
Plateaux-Quénu
C
Plateaux
L
Packer
L
.
2000
.
Population-typical behaviours are retained when eusocial and non-eusocial forms of Evylaeus albipes (F.) (Hymenoptera, Halictidae) are reared simultaneously in the laboratory
.
Insectes Sociaux
 .
47
:
263
270
.
Powers
KS
Avilés
L
.
2007
.
The role of prey size and abundance in the geographical distribution of spider sociality
.
J Anim Ecol
 .
76
:
995
1003
.
Purcell
J
Avilés
L
.
2007
.
Smaller colonies and more solitary living mark higher elevation populations of a social spider
.
J Anim Ecol
 .
76
:
590
597
.
Purcell
J
Avilés
L
.
2008
.
Gradients of precipitation and ant abundance may contribute to the altitudinal range limit of subsocial spiders: insights from a transplant experiment
.
Proc Biol Sci
 .
275
:
2617
2625
.
Purcell
J
.
2011
.
Geographic patterns in the distribution of social systems in terrestrial arthropods
.
Biol Rev Camb Philos Soc
 .
86
:
475
491
.
Rayor
LS
Uetz
GW
.
1993
.
Ontogenetic shifts within the selfish herd : predation risk and foraging trade-offs change with age in colonial web-building spiders
.
Oecologia
 .
95
:
1
8
.
Rayor
LS
.
1996
.
Attack strategies of predatory wasps (Hymenoptera: Pompilidae; Sphecidae) on colonial orb web-building spiders (Araneidae: Metepeira incrassata)
.
J Kans Entomol Soc
 .
69
:
67
75
.
Riechert
SE
Roeloffs
R
Echternacht
AC
.
1986
.
The ecology of the cooperative spider Agelena consociata in Equatorial Africa (Araneae, Agelenidae)
.
J Arachn
 .
14
:
175
191
.
Samson
DA
Rickart
EA
Gonzales
PC
.
1997
.
Ant diversity and abundance along an elevational gradient in the Philippines
.
Biotropica
 .
29
:
349
363
.
Schemske
DW
Mittelbach
GG
Cornell
HV
Sobel
JM
Roy
K
.
2009
.
is there a latitudinal gradient in the importance of biotic interactions?
Annu Rev Ecol Evol Syst
 .
40
:
245
269
.
Uetz
GW
Hieber
CS
.
1994
.
Group size and predation risk in colonial web-building spiders : analysis of attack abatement mechanisms
.
Behav Ecol
 .
5
:
326
333
.
Uetz
GW
Boyle
J
Hieber
CS
Wilcox
RS
.
2002
.
Antipredator benefits of group living in colonial web-building spiders: the “early warning” effect
.
Anim Behav
 .
63
:
445
452
.
Wickler
W
Seibt
U
.
1988
.
Two species of Stegodyphus spiders as solitary parasites in social S. dumicola colonies (Areaneida, Eresidae)
.
Verb Naturwiss Ver Hamburg
 .
30
:
311
317
.
Yip
EC
Powers
KS
Avilés
L
.
2008
.
Cooperative capture of large prey solves scaling challenge faced by spider societies
.
Proc Natl Acad Sci USA
 .
105
:
11818
11822
.
Yip
EC
Rayor
LS
.
2014
.
Maternal care and subsocial behaviour in spiders
.
Biol Rev Camb Philos Soc
 .
89
:
427
449
.
Zammit
J
Hogendoorn
K
Schwarz
MP
.
2008
.
Strong constraints to independent nesting in a facultatively social bee: quantifying the effects of enemies-at-the-nest
.
Insectes Sociaux
 .
55
:
74
78
.

Author notes

Address correspondence to L. Avilés. E-mail: laviles.ubczool@gmail.com
Handling editor: Jonathan Pruitt