Introduction

Optimizing foraging efficiency is essential to an organism’s survival. The Optimal Foraging Theory proposes that organisms forage in a way that enhances their fitness (Pyke 1984). Several studies have shown that Neotropical ants seem to optimize foraging by incorporating apparent obstacles (e.g., vines, lianas, roots and fallen twigs, branches, and logs) in their foraging trails even when the incorporation of these objects increases the length of their foraging trail (Farji-Brener et al 2007, Clay et al 2010, Holt & Askew 2012). These substrates provide ants with smoother and less complex tridimensional surfaces than the substrates where foraging trails usually occur (e.g., soil or tree-trunk bark), thus increasing travel speed to and from food sources (Farji-Brener et al 2007, Holt & Askew 2012). The factors influencing the selection of various substrates as fast travel surfaces are poorly understood, but research on Atta ants has started to clarify the circumstances in which these objects become part of foraging trails.

Atta ants are important foragers in tropical forests that may reduce annual leaf production per area by as much as 17% (Cherret 1989). These ants forage for leaves in the forest understory and bring them to their nest to cultivate fungal colonies that the ants consume (Cherret 1989). Atta ants travel on cleared trails, known as trunk-trails—maintained by cutting and clearing plants on the forest floor (Shepherd 1982). A single trunk-trail can travel along the forest understory up to 2 or 3 km (Cherrett 1968, Howard 2001) and be up to 30 cm wide (Lewis et al 1974a). Fallen branches are notably prevalent in trunk-trail systems. For example, in the forests of Costa Rica and Panama, 17–30% of the trunk-trail was found to be composed of fallen branches (Howard 2001, Mogollón & Farji-Brener 2009). Atta ants deviate from the main direction of the trunk-trail to include fallen branches (Farji-Brener et al 2007), suggesting an active use of this substrate.

Very few studies have evaluated whether walking on fallen branches increases travel time of Atta ants. One study showed that laden Atta ants indeed travel faster across fallen branches than on the ground (Farji-Brener et al 2007). A follow-up study showed, however, that ants walk more slowly when they ascend to or descend from a fallen branch than on the forest ground (Mogollón & Farji-Brener 2009). Specifically, Mogollón & Farji-Brener (2009) report that ants lose 1.5 s of travel time per 20 cm segment when accessing and departing a branch. The authors propose two potential causes for the slowdown. The first cause is the difficulty of ascending the branch due to gravity. The second cause is a bottleneck that creates congestion, occurring in periods of high ant flow, especially when the branch is narrower than the trunk-trail on the ground. In the study by Mogollón & Farji-Brener (2009), branches were on average 50% narrower than the trunk-trail on the ground, explaining why congestions were common in this case. The height of the branch (i.e., the vertical distance traveled by ants from the ground to the top of the horizontal branch laid on the ground and vice versa, see Fig 1) was not measured by the authors. The height of the branch is obviously important because it determines the distance traveled by ants at the access and departure points of a branch and therefore determines the speed of the ants at these segments (Fig 1).

Fig 1
figure 1

Side view of experimental setup of timed trail and branch segments in a pacific premontane wet forest at 1300 m in elevation in Monteverde, Costa Rica. The trail and branches were segregated as follows: (1) a 20-cm segment of trunk-trail before the fallen branch, (2) a 20-cm segment centered over the fallen branch at its access point, (3) a 20-cm segment in the center of the branch, and (4) a 20-cm segment centered over the fallen branch at the point of departure. All segments chosen were clear of debris and the same laden ant was timed across each segment. The variable height is represented above as the vertical distance traveled by the ants from the ground to the top of the horizontal branch laid on the ground and vice versa.

Full size image

The height of the branch may be one of the most important variables affecting travel times up and down a branch. Lewis et al (2008) manipulated the steepness of 1-m long trail sections in the field and detected much higher transport rates in downhill and horizontal trails compared to uphill trails in laden Atta ants. Moreover, transport rates increased sharply with the decline of the trail when laden ants were placed on manipulated trails in the field (Lewis et al 2008). Lewis et al (2008) ruled out that trail congestion caused the differences in the observed speed. The study by Lewis et al (2008) was not conducted in the context of branch incorporation into foraging trails, but it suggests that the height of a branch may affect travel times uphill independently of trail congestion. The study mentioned above also suggests that departing a branch does not necessarily increase travel time. Even if congestion plays a role in travel time and creating bottlenecks (Mogollón & Farji-Brener 2009), the ants may take longer to access or depart a branch (i.e., decrease their speed) only when congestion exists (i.e., when the branches are significantly narrower than the ground trail). Ultimately, branch height may also determine whether a branch is too tall to be incorporated into the foraging trail. To the best of our knowledge, no study has evaluated how branch height influences (1) travel times and speed and (2) the incorporation of fallen branches into foraging trails in Atta or any other ant species.

In this study, we timed the same Atta cephalotes (L.) laden ants as they ascended, traveled across, and departed from branches of varying widths following the protocol of Mogollón & Farji-Brener (2009). In addition, the height of each branch was measured to evaluate the effect it might have on speed (i.e., distance traveled divided by the time elapsed). This is because the height of the branches adds length to the trail and therefore has to be considered when studying the potential benefits of traveling on branches versus traveling on the ground. Additionally, we offered branches/logs of varying widths and heights to the ants to evaluate whether they naturally avoid fallen branches/logs with dimensions that actually increase travel time.

Material and Methods

Study site

The study was conducted on a farmland in Cañitas, Costa Rica, on a mid-elevation pacific slope, located in the Monteverde area (10°31′43″N,84°82′50″W). The area is a premontane wet forest at around 1300 m in elevation that receives a mean annual rainfall of 2–4 m. Data were recorded from April to May 2014, at the beginning of the rainy season in the Monteverde area (Clark et al 2000). However, this year showed an extended dry season with unusually high temperatures and little rain during April and May compared to previous years (J.C-C. pers. observ.). The farmland consisted of pasture bordered by secondary growth-regenerating forest. Nests of A. cephalotes were scattered both in the pasture and in and around the forest edge.

Ant travel times

Six large A. cephalotes nests were located in and around the forest edge and in the pasture to test if traveling up and down branches already incorporated into foraging trails is slower than traveling on a trail or across a branch. For each nest, four fallen branches parallel to the ground were selected along any point in a nest’s main trunk-trail. Following Farji-Brener et al (2007) and Mogollón & Farji-Brener (2009), ants were timed in relation to branch width and height across four 20-cm segments along a trunk-trail on the ground: (1) preceding a fallen branch, (2) at the access point of a fallen branch, (3) across the branch, and (4) departing the branch (Fig 1). The ants were undisturbed during this observation. The substrate on the ground was mostly dirt cleared of leaf litter by the ants. We selected branches that were greater than 60 cm in length because it was difficult to accurately time ants on shorter branches. Data were collected between 0800 and 1200 hours and on sunny days because the ants decreased their activity in the presence of rain. In addition, the width (i.e., the branch’s extent from side to side) and height of each branch (Fig 1) segment were measured after the ants were timed. Width was measured at the narrowest part of the branch segment and height was measured at the highest point on each of the branch segments relative to the trunk-trail. The angle of access and departure on the branches was close to 90° for the majority of branches. We calculated whether the proportion of branches that were narrower (or wider) than the trunk-trail (as an indicator of potential trail congestion) was different than expected by chance (i.e., 50% chance) using a proportion’s test. Branch height was added to the length of the trail segment at access and departure points (segments 2 and 4; Fig 1) to calculate speed in centimeters per second.

To examine the contribution each segment has on both ant travel time (s) and speed (cm/s), these variables were compared between segments 1, 2, 3, and 4 (Fig 1) using a repeated measures analysis of variance implemented as a general linear mixed model (GLMM) with colony, branch, and ant identity as random effects (Zuur et al 2009) (one GLMM for each variable) using the R package nlme in R 3.02. Independent contrast post hoc pairwise comparisons were performed using the R package gmodels when tests were significant.

Branch selection experiment

To examine the effect of branch size on branch selection, one large and active A. cephalotes nest was selected and examined over the course of a week between the hours of 0800 and 1200. The width of one main trunk-trail was measured by choosing a segment around 10 m from the main nest that seemed representative of the visible trail before and after it. In order to determine if fallen branch width and height affect branch selection for the trunk-trail systems, two different branch treatments were created (Fig 2). The width treatment consisted of (1) five branches that were half the width of the trunk-trail and 90 mm tall (this height is within the range of heights usually found in the field; see results below) and (2) six branches nearly equal in width to the trunk-trail (several cm on either side of the branch allowed the ants the option to not choose the branch) and 90 mm tall. The height treatment consisted of (1) five branches that were nearly as wide as the trunk-trail but 250 mm tall (this height is above the range of the vast majority of heights seen in the field; see results below) and (2) five branches that were nearly as wide as the trunk-trail and only 90 mm tall. In this way, ants were offered the choice between branch sizes either potentially beneficial or detrimental to their travel times, and the only variable changing in the width treatments was width and in the height treatments was height. The choice to use five or six branches was based on the availability of branch sizes around the study site. All branches had relatively smooth surfaces and were not previously used by other leaf cutter ants.

Fig 2
figure 2

Experimental setup of height and width treatments for the branch selection experiment in a pacific premontane wet forest at 1300 m in elevation in Monteverde, Costa Rica. The boxes above represent fallen branches or logs, and measurements are given for the height and width of each branch. The length of all branches was standardized to 800 mm. Each treatment has two variations. For the height treatment, ants were offered a choice between branches nearly as wide as their trunk-trail but either 250 or 90 mm tall. For the width treatment, ants were offered a choice between branches 90 mm tall but either nearly the width of their trunk-trail or half of the trunk-trail’s width. For widths nearly equal to the trunk-trail, several centimeters were left on either side of the branch to allow ants the opportunity not to use the branch.

Full size image

To construct the tall branches in the height condition (Fig 2), unused branches located around the farm were cut flat on either side so that they could be stacked to 250 mm. The topside of the uppermost branch in the stack was not manipulated to ensure branch-like conditions. All branches were cut to 800 mm in length and at a 90° angle on both sides. This controlled for length and ensured the ants experienced the full effect of the branch height. Starting from the beginning of the trunk-trail at the nest entrance, five half-width branches and six near full-width branches were placed directly in the middle of two to three trunk-trails in the same nest. This was done gently and carefully to minimize disturbance to the ants. We did not observe any detrimental effects of this branch placement at any point during this experiment. After 24 h, it was recorded whether the ant column had reformed over the branch or reformed around it to avoid it. If the ant column only climbed partway up a branch but did not utilize its entire height, this was counted as a “No.” All branch conditions were then removed, and ants were given 30–60 min to readapt to the changes before methods were repeated for height treatment branches (Fig 2)—no branch was used more than once. The entire experiment was replicated again at the same nest but on a different trail of the same width, and results were analyzed using Fisher’s exact tests. At the end of the experiment, care was taken to reset the trunk-trails back to their original state.

Results

Ant travel times/speed

Traveling up or down a branch took a different amount of time than traveling on the ground or across a branch (GLMM, F = 278.9, df = 3, and 1167, p < 0.001; independent contrast post hoc test, all comparisons p < 0.001). It took nearly 50% less time to travel on segments across fallen branches than it did to travel on segments of trail on the ground (Fig 3a). It took slightly more time (~6%) to travel up the branch than it did to travel down, but accessing and departing the branch was 60–90% slower than walking across it. Both traveling up and down the branch was about 14% faster than walking across the trail on the ground, despite that the distance traveled was larger as the ants went up and down the branch. Compared to the time it takes to walk 20 cm of cleared trail segment on the ground, ants gained on average 2 s traveling up a branch, 9 s to travel across a branch, and 3 s departing the branch. Differences in time were caused by differences in speed (GLMM, F = 590.7, df = 3, and 1167, p < 0.001; independent contrast post hoc test, all comparisons p < 0.0001). The ants increased their speed when accessing and departing a branch nearly fourfold compared to when they traveled on the ground, and 2.5 times compared to when traveling across (Fig 3b).

Fig 3
figure 3

Travel time (a) and speed of Atta cephalotes (b) across four segments of trail and branch on farmland in a pacific premontane wet forest at 1300 m in elevation in Monteverde, Costa Rica. Ants were timed across four 20-cm cleared segments (see Fig 1): (Trail) a segment of trail preceding the fallen branch, (Up) centered over the access point of the branch, (Across) in the center of the branch, and (Down) centered over the departure point of the branch. Mean travel times for ants across the four segments of trunk-trail and branch differed significantly and differed significantly from each other (n = 390 ants). Error bars represent one standard error.

Full size image

Heights from segments 2, 3, and 4 (Fig 1) were averaged alongside widths from the same segments to describe the height of branches regularly incorporated into the trunk-trail by the ants. The majority of branches (22/24) used by the ants as part of their trails were around 10–120 mm in height and width. The remaining branches were 200–360 mm in either height or width on average. For all branches, the mean height and width for segments 2 and 4 (Fig 1) were 66 and 67 mm with a standard deviation of 77 and 75 mm, respectively. Furthermore, 17 of the 24 branches observed in this study were narrower than the preceding trunk-trail, which is higher than expected by chance (proportions test 4.1667, df = 1, p = 0.04). The remaining seven branches were as wide as or wider than their preceding trails. For the larger logs and fallen branches (from 200 to 360 mm in height or width), all had broken off from the tree at a certain angle or decomposed in such a way to provide ants with a shallow and smooth access or departure point—close to 45°—that presumably would make climbing relatively easy.

Branch selection experiment

After 24-h periods, the colony avoided half-width branches the vast majority of the time but incorporated near full-width branches most of the time (Fisher’s exact test, p = 0.011; Fig 4a). The colony similarly incorporated regular-height branches the vast majority of the time and avoided tall condition branches (Fisher’s exact test, p = 0.002; Fig 4b). In cases where branches or logs were avoided, ants either avoided the objects all together or only walked partway up the branch.

Fig 4
figure 4

Number of half-width versus full-width branches (a) and number of tall branches versus regular-height branches (b) used and avoided by Atta cephalotes (L.). Data were collected on farmland in a pacific premontane wet forest at 1300 m in elevation in Monteverde, Costa Rica. In a single nest on two separate main trunk-trails, a ants were offered 10 branches of equal height but half the width of the foraging trail and 10 branches of equal height (to each other and to the half-width branches) but the same width as the foraging trail. Additionally, b ants were offered 10 constructed branches equal to the width of the foraging trail, but unusually tall (250 mm tall). Another 12 constructed branches of the same width, but of heights regularly used by the ants (90 mm tall), were also offered to the ants. Branches were placed directly in the center of the foraging trail. Yes refers to branches used, and No refers to branches unused. Distributions were likely not due to chance (p < 0.05).

Full size image

Discussion

Ant travel times

Our results support previous studies that suggest that using fallen branches decreases ant travel time. Even after including time expended when climbing up and down a branch, using a branch is still faster than using a trail. As previously suggested (Farji-Brener et al 2007), this is likely because fallen branches offer a smooth substrate for ants to walk across. It may also be that it is easier for the ants to locate their travel pheromones (Loreto et al 2013) on the branch substrate than on the ground. This is because trail pheromone may evaporate off fallen branches more slowly than on the ground (Jeanson et al 2003), which could decrease the energy costs of pheromone production and counter the cost of maintaining a longer pheromone trail.

Contrary to previous studies, our results show that accessing and departing a branch save ants time in the Monteverde region. Previous studies suggest that during higher rates of ant flow, a time bottleneck is created when fallen branches are narrower than trunk-trail (Burd & Aranwela 2003, Mogollón & Farji-Brener 2009). Our data suggest that, in Monteverde, there is no time bottleneck when accessing or departing a branch and that travel time actually decreases at these segments even though most branches observed in this study were narrower than the trunk-trail. This difference with previous studies was probably due to low levels of ant activity in the studied colonies caused by unusually warm and dry days compared to the same time period in previous years (J.C-C. pers. observ.). Foraging activity in Atta has been observed to decrease when temperatures are high and humidity is low (Lewis et al 1974b, Rockwood 1975), which would explain why the colonies sampled in our study showed relatively low activity levels, at least during the day. In any case, the results of this study show that under certain conditions the ants may actually accelerate as they ascend or descend a branch to save time. Given that humidity is usually lower in montane areas such as Monteverde compared to the lowlands of La Selva Biological station, where the Mogollón & Farji-Brener (2009) study was conducted, it seems reasonable that colonies in the highlands suffer bottlenecks less often than colonies in the lowlands.

Our study also suggests the possibility that the use of branches, at least of certain widths and heights, changes very often. Changes in ant flow seem to occur on a relatively small timescale with variations in microclimate or seasonality, like rain or temperature (Cherrett 1968, Lewis et al 1974b, Rockwood 1975). For this reason, a branch already incorporated into a trunk-trail may be avoided in the future if conditions change and ant flow increases to a rate high enough to cause a bottleneck of substantial magnitude to decrease net travel times across the branch compared to the trunk-trail. For instance, the colonies studied in Monteverde may not use the same branches at night, when ant activity is usually higher (J.C-C. pers. obs.), because temperature decreases and humidity increases.

In addition to their size, a factor that could also affect the selection of branches is their length. Mogollón & Farji-Brener (2009) predict, based on their time-saved estimates, that fallen branches used by colonies should at least be greater than 60 cm in order to be beneficial. By selecting branches that follow this rule, the benefit of walking across a branch will always outweigh the detriment of climbing up or down it. In this case, the benefit of branch length would likely overpower any branch selection status pressures caused by changes in ant flow. Our results predict that branches in Monteverde can be shorter than 60 cm because there is no cost of ascending or descending a branch—at least under the conditions observed in this study. Smaller twigs or branches shorter than 60 cm were indeed seen being used by some colonies in Monteverde.

Branch selection experiment

The fact that ants strongly prefer near full-width branches to half-width branches suggests that ants recognize the potential detriment half-width branches have on travel times. Since a smaller width may increase the number of head-on encounters with other ants, causing a bottleneck and increasing ant travel time (Burd & Aranwela 2003), it makes sense that they would not immediately use half-width branches. They may eventually use the branch if the temperature increases and/or humidity decreases (see above) because many of the branches seen in the field were half the width of the trail and of similar heights.

An alternative explanation for the result is that the ants may not have had the opportunity to truly choose between half- and near full-width branches by size only. Because we deliberately placed the near full-width branches on their trails, avoiding a near full-width branch implies the use of energy to widen or deviate the trail. If it is more energy-efficient to use the branch instead of widening the trail, the ants may not reject near full-width branches simply to avoid clearing a new trail around it. However, some observations suggest the ants actually have the opportunity to choose near full-width branches to save travel time. First, some near full-width branches (3/10) were actually avoided by walking along the several centimeters available on either side of the branch. The reason these near full-width branches were avoided and the other ones were not may depend on rates of ant flow. As mentioned above, the benefit of a branch may vary with weather and ant flow (Cherrett 1968, Lewis et al 1974b, Rockwood 1975), so these trails could have had higher rates of ant flow than other trails, thereby affecting whether or not a branch was used. Second, even before experimental manipulation, many colonies observed in this study already incorporated some branches that were close to or wider than their preceding trunk-trail. Based on our results and previous studies, these broad branches do not cause bottlenecks and offer a way to travel the same distance faster. Therefore, it is reasonable to suggest that ants choose broad branches based on these properties. This evidence seems to support that the benefit of using a near full-width branch is more apparent than the potential benefit of using a half-width branch because it saves time, even when a very wide branch falls directly on a trail, as in the case simulated in our experiment.

In the height treatments, the fact that ants selected regular-height branches more often than tall branches suggests that the benefit of walking across a regular-height branch is evident after a 24-h period. It was observed that ants would rather climb partway up the branch or walk around it than walk up its full height. Since no ants climbed the full height of the tall branches, this may indicate that the ants recognize the potential detriment the full height of the branch could have on their travel time.

In conclusion, our results suggest that branch size influences the incorporation of branches into foraging trails and that ants can alter their travel speed to compensate for the difficulty of traveling on branch slopes. Our data also suggest that branch use and ant travel time may be highly dependent on both microclimate and macroclimate. More generally, branch size should be considered when studying foraging efficiency in ants.

Our results should be considered with caution due to the short duration of the study and the unusually warm and dry days ants were observed on. Though our results may not be representative of A. cephalotes across their large geographic distribution (Mexico-Northeastern Brazil; Corrêa et al 2005), they still can offer insight to studies performed at similar sites (e.g., mountains) and/or under similar conditions (e.g., dry/warm periods). In addition, our results may also offer insight into the use of branches by A. cephalotes in future scenarios of global warming throughout the Neotropics. We make this suggestion based on two factors. First, the unusually dry and warm conditions experienced in Costa Rica when this study was conducted have actually persisted throughout the rest of 2014 and 2015 (J.C-C. pers. observ.) due to the effect of El Niño Southern Oscillation on the Northern Hemisphere (Page: 11 NOAA 2015). Second, reduced precipitation and higher than normal temperatures have been reported from Mexico to Northeastern South America (i.e., most of distribution range of A. cephalotes) during the same time period, and it is forecasted that the frequency of occurrence of these conditions is going to increase in the region in the decades to come (Magrin et al 2014). If the climate continues to change in this direction, colonies of A. cephalotes throughout the Neotropics may incorporate branches into their foraging trails in the way we observed in this study.