- ANTS Litter Surface Temperature : A Driving Factor Affecting Foraging Activity in Dinoponera lucida ( Hymenoptera : Formicidae )

Foraging requires appropriate decisions related to the time spent in exploration and choosing the appropriate time to maximize net amount of energy gained, foraging efficiency, and survival, or to minimize the risk of starvation (Emlen, 1966; MacArthur & Pianka, 1966; Ydenberg, 2007). These decisions are subject to constraints that limit viable alternatives, such as time spent and the energy cost of the activity (Emlen, 1966; MacArthur & Pianka, 1966; Perry & Pianka, 1997; Ydenberg, 2007). Decisions may be related Abstract Dinoponera lucida is a poneromorph ant endemic to the Atlantic Forest of Brazil. The species is classified as endangered in Brazil’s Red List due to its peculiar reproductive biology and high habitat fragmentation. Herein, we characterize D. lucida foraging activity and response to litter surface temperature in a lowland forest remnant in south-eastern Brazil. The mean flow of workers at nest openings was 3.8 ± 0.6 per hour, mean foraging trip was 14.2 ± 2.2 min, and mean foraging distance was 3.8 ± 0.4 m. The time spent per foraging trip and litter surface temperature were positively correlated. Flow of workers at nest openings was higher with mean temperature of litter surface between 21.0 and 27.0 °C. Our results show that D. lucida has a diurnal foraging activity related to habitat temperature. Our data contribute to the knowledge about the ecology of D. lucida and support the hypothesis of optimal food foraging regulated by habitat temperature. In addition, the better understanding of D. lucida activity patterns can assist on conservation planning of this endangered and endemic ant. Sociobiology An international journal on social insects


Introduction
called giant ants because reach up to 40 mm (Kempf, 1971;Lenhart et al., 2013;Escárraga et al., 2017). Among these species, Dinoponera lucida Emery, 1901 is endemic to Brazil's Atlantic Forest from the states of Espírito Santo, eastern Minas Gerais and southern Bahia (Peixoto et al., 2010;Simon et al., 2020). Currently, only 12% of the area originally covered by Atlantic Forest has forest remnants, mostly small and isolated across the human-altered matrix (Fundação SOS Mata Atlântica & Instituto Nacional de Pesquisas Espaciais, 2019). Dinoponera lucida is a forest-specialist and currently recorded only in forest remnants. The area of geographic range of the D. lucida was estimated at 156 km 2 (Instituto Chico Mendes de Conservação da Biodiversidade & Ministério do Meio Ambiente, 2018). Habitat fragmentation causes isolation of populations, since the species is unable to colonize over long distances, leading to a clumped distribution pattern over the geographic range (Instituto Chico Mendes de Conservação da Biodiversidade & Ministério do Meio Ambiente, 2018). New colonies of D. lucida are generated by fission of extant colonies, increasing inbreeding and reducing genetic diversity, which has led to local extinctions (Mariano et al., 2008;Campiolo et al., 2015), a genetically fragmented distribution (Mariano et al., 2008;Resende et al., 2010;Simon et al., 2016), and endangered (EN) species status on the Brazilian Red List (Ministério do Meio Ambiente, 2014; Instituto Chico Mendes de Conservação da Biodiversidade & Ministério do Meio Ambiente, 2018). Thereby D. lucida conservation is intrinsically linked to the preservation and recovery of the Atlantic Forest and the connection between its forest remnants.
How microclimate affects the ecology of D. lucida remains poorly understood. For instance, D. lucida responds to regional temperature and rainfall pattern (Simon et al., 2020); populations settle mainly in areas with higher annual mean temperatures, lower annual temperature range and less precipitation during the driest month (Simon et al., 2020).
Microclimatic variables such as temperature, rainfall, and humidity are known to influence the realized niche of insects, and can do so by influencing behavior and foraging strategy (Willmer, 1982;Colinet et al., 2015;Pincebourde & Suppo, 2016;Simon et al., 2020;Welch et al., 2020). Dinoponera lucida increase foraging activity in mild temperature associated with high elevations (Peixoto et al., 2010). The pattern of foraging activity of D. lucida has an optimal temperature range in areas surrounding the nest (Peixoto et al., 2010). Inside the forest, microclimatic variables are influenced by canopy cover, edge effects, evapotranspiration and other factors related to vegetation and habitat heterogeneity (Magnago et al., 2015;Stangler et al., 2015;Boehnke et al., 2017). Thus, an appropriate approach to assess the foraging activity of D. lucida should consider microclimate conditions. Dinoponera lucida commonly prey upon epigeic ectotherm taxa, and these prey respond to temperature, which alters its density and availability (May, 1979;Colinet et al., 2015;Pincebourde & Suppo, 2016;Welch et al., 2020). We expect that D. lucida will adapt foraging time to maintain an optimal strategy (Norberg, 1977;Bernstein, 1979;Azevedo et al., 2014;Medeiros et al., 2014).
Herein, we characterized the foraging activity and the response to litter surface temperature by D. lucida. We expected that daily variation in air and litter surface temperatures to influence the foraging activity of D. lucida. Also, we expected more workers to forage during the optimal temperature range, due to physiological constraints and prey availability.

Study site
We carried out this study in the Vale Natural Reserve (VNR; 22,711 ha; 19º 09' 05" S, 40º 04' 15" W), Linhares, northern Espírito Santo state, south-eastern Brazil. VNR has flat relief and altitude between 28 to 65 m (Peixoto et al., 1995). VNR is an important remnant of the Tabuleiro Forest, which is a specific formation of the Atlantic Forest (Kierulff et al., 2014). VNR region has tropical and humid climate with two dry months (Alvares et al., 2014). The annual mean temperature is 24.3 °C and the annual mean rainfall is 1,214 mm (Kierulff et al., 2014).

Sampling
We conducted active visual survey in a 30 m × 30 m plot to locate D. lucida workers (19º 09' 14.5" S, 40º 04' 14.0" W) from September 29 to October 2, 2017. We used sardines as bait along the worker's route (Silvestre et al., 2015). We located three nests by following workers carrying the bait back to the nests.
We estimated height, width of the nest openings and canopy coverage over the nest openings with the aid of photos analyzed at ImageJ 1.52p software (Schindelin et al., 2012;Schneider et al., 2012). We neither disturbed the workers nor excavated the nests.
In the center of the sampling area, we measured air temperature (T A ) and relative humidity (RH) at 1.5 m high using a Kestrel portable weather device (model 3000); and litter surface temperature (T S ) using a Tekpower digital infrared thermometer (model DT-8380). T A , RH and T S were recorded every 30 min from 6 h to 18 h, which is the expected foraging period of D. lucida (Peixoto et al., 2010). It rained from 6 h to 9 h on the third day of sampling. Flow of workers ceased during the rain. The data affected by rain were not considered for the quantitative analyzes.
We observed foraging workers exiting the nests and searching for food in the surrounding litter. We recorded inflow and outflow of workers for 15 min in each nest per hour (Peixoto et al., 2010). We considered inflow (F IN ) as the number of workers entering the nest and outflow (F OUT ) as the number of workers exiting nests to forage. We considered total flow (F T ) in the nest openings as the sum of flows of workers. Total flow is colinear with F IN and F OUT . Using the 15 min measured data, we estimated F T per hour in each nest.
We used focal-animal sampling (Altmann, 1974) and followed the workers individually during foraging activities from the time workers exited until they returned to the nest. We remained at least 2 m away from the observed worker to avoid interference of behavior.
We recorded duration of foraging trip (t f ) as the time spent searching for food items from their exit to their return to the nests. We used t f as a proxy to estimate foraging efficiency (η ~ t f -1 ). We considered a returning worker those that were observed collecting a food item (successful trip) or those that started moving toward the nest even without food item (unsuccessful trip). The efficiency rate of foraging was calculated as number of workers returning to the nest with food divided by total number of workers returning to the nest (Giannotti & Machado, 1991;Medeiros et al., 2014). We defined reached distance (D f ) as the maximum distance reached by a worker during a foraging trip, measured as a straight line from the nest opening to the returning point. We identified collected food items visually or from zoomed photos to the lowest possible taxonomic level.

Data analysis
First, we perform Pearson correlation analyzes between environment variables (T S , T A , and RH), t f -T S , and t f -D f . Because T A is positively correlated with T S , we used T S for description of the behavioral patterns studied, since T S represents the microclimate variable related to the microhabitat of D. lucida workers. We established the curves with the variation of the foraging activity (F T ) as function of T S and time of day. A fourth-order polynomial fit was used in the nonlinear regression model to describe the bimodal pattern between F T and time of day. We tested if there was a relationship between F T and T S applying a Pearson's Chisquared test (χ 2 ) from a six-categories of T S (from 19 to 31 ºC, with regular intervals of 2 ºC, as showed in Fig 2b) defining a contingency table. The relative frequencies of F T (R f ) were calculated for each hour of the day and each T S category as a proportion of F T in each hour sampled or T S range category divided by the total F T of all hours or all T S categories, as described by equation 1: To calculate the relative frequencies of F T , we calculated the means of F T of all nests to represent the foraging activity in the sampling area. We applied linear regression analysis to describe the relationship between t f and T S . We used Student's t test to assess differences for each T S category between successful and unsuccessful trips. We used Graphpad Prism 7.00 (GraphPad Software, 2017) to perform statistical analyses. A p-value ≤ 0.05 was considered significant. We presented the variables in terms of mean ± standard error of the mean (SE).

Characteristics of the sampling area and nest openings
The three monitored nests were beneath understory vegetation roots in a forested area (900 m 2 ) with ground totally covered by leaf litter. The nest density was 33 nests/ha. The mean distance between nests was 11.3 ± 1.4 m, showing a triangular shaping on the ground (nest 1 to 2: 8.5 m, nest 1 to 3: 12.5 m and nest 2 to 3: 13.0 m). Mean canopy cover on the nest's openings was 79 ± 4 % (nest 1: 74%, nest 2: 86% and nest 3: 78%), which characterizes the sampling area as a mosaic of light and shade. The nests had an elliptical entrance with mean height of 20.4 ± 1.0 mm (nest 1: 18.5 mm, nest 2: 21.5 mm and nest 3: 21.3 mm) and mean width of 41.1 ± 1. mm (nest 1: 43.1 mm, nest 2: 39.7 mm and nest 3: 40.5 mm).

Environment variables
The T A during the foraging activity was 27.2 ± 0.5 ºC and T S was 24.5 ± 0.6 °C (Table 1). T A and T S were positively correlated (r = 0.89, p < 0.01, n = 26). Highest temperatures were recorded in the middle of the day (Fig 1). The RH during the foraging activity was 82.6 ± 2.1 %, with lower RH in the same hours of higher temperatures (Fig 1). T A and RH were negatively correlated (r = -0.80, p < 0.01, n = 26). T S and RH were also negatively correlated (r = -0.81, p < 0.01, n = 26). Nest flow, foraging activity time, distance, and influence of temperature Dinoponera lucida foraging activities were done between 7 h and 17 h. Dinoponera lucida exhibited a bimodal pattern of foraging activity (r 2 = 0.26, p < 0.01, AICc = 209.6, df = 58) (Fig 2a). The first F T peak was between 8 h and 10 h. The second F T peak was recorded between 14 h and 17 h.

Collected food items
We recorded 48 food items collected by D. lucida workers, of which, 94% were animals and 6% vegetation items (Table 2). All food items were solid. We observed workers capturing macroinvertebrates with high mobility (such as spiders, grasshoppers, 78%), low mobility (insect larvae and gastropods, 13%), and no mobility (insect pupae, 9%). These percentages about mobility of macroinvertebrates consider only identified animal food items.

Discussion
Dinoponera lucida presented bimodal cycle on daily foraging activity with a higher peak in the morning. The foraging activity was higher in a range of litter surface temperature between 21.0 and 27.0 ºC, and successful foraging trips were mostly during lower litter surface temperatures (24.6 ± 0.4 ºC) in comparison to unsuccessful trips (26.8 ± 0.8 ºC). Similar patterns were found for D. lucida and Dinoponera quadriceps Kempf, 1971, both showing bimodal foraging cycles and higher peak in the morning than in the afternoon (Peixoto et al., 2010;Medeiros et al., 2014). In contrast, Dinoponera longipes Emery, 1901 seems to be mainly nocturnal, but workers are also active during the day (Morgan, 1993). In addition, Dinoponera gigantea (Perty, 1833) also has a bimodal pattern of foraging activity, but with higher peaks at dawn and dusk (Fourcassié & Oliveira, 2002). Finally, D. lucida likely reduces foraging in response to temperature cues to minimize foraging activity during times of low prey availability (Norberg, 1977;Bernstein, 1979).
Our data suggest that D. lucida limits foraging according to the activity pattern of prey, suggesting an optimization strategy driven by prey availability (Emlen, 1966;MacArthur & Pianka, 1966;Norberg, 1977;Perry & Pianka, 1997), with litter temperature serving as a proxy for successful foraging likelihood rather than being a direct influence on D. lucida foraging behavior.
Foraging attempts had no relationship with time spent or distance reached, suggesting that the foraging success depends on choosing the appropriate time. It seems that these ants optimize their foraging behavior by relying on the chances to find a food resource instead of covering large areas or traveled distance during the foraging trip. Our data on patterns of flow of workers of D. lucida in foraging was similar to other populations (Peixoto et al., 2010). The increase of time spent in foraging trip could be related to search and capture efficiency, supporting the hypothesis of optimal food searching (Emlen, 1966;MacArthur & Pianka, 1966;Norberg, 1977;Perry & Pianka, 1997). The positive correlation between time spent in foraging trip and litter surface temperature suggests the existence of an optimal   temperature range for the foraging activity. However, it is not clear that the workers use temperature as a proxy. It seems that temperature is an effective proxy for prey availability. In addition, as temperature is highly correlated with time of day, it may be that they simply use time of day as a proxy for prey availability. It also seems possibly that foraging activity is directly related to prey availability, with ants foraging more when preys are abundant and less as prey availability decreases. As D. lucida is an eusocial species, some decisions about the time to forage could be related to communication between the nestmates (Leonhardt et al., 2016). Further studies could assess these questions assertively with prey density data combined with the other sampled data. According to our data D. lucida was predominantly carnivorous and might be categorized as a hunting and collecting species (Almeida & Queiroz, 2015). We did not observe predation of other ant species, even when they were abundant near a worker in the foraging activity. Similarly, ants were not in the diet of D. quadriceps (Araújo & Rodrigues, 2006). Ants, however, have been observed in low frequency in the diet of other species of Dinoponera (Fourcassié & Oliveira, 2002;Peixoto et al., 2010). Considering the number of prey items, ants comprise 2% of D. lucida diet (Peixoto et al., 2010) and 4% of D. gigantea diet (Fourcassié & Oliveira, 2002). The capture of animals with high mobility (78% of the total macroinvertebrate items) indicates that D. lucida performs active hunting besides collection of food items. This performance was also observed in other species of Dinoponera, including capture of relative larger prey such as small lizards (Sousa & Freire, 2010;Ribeiro et al., 2011;Carvalho et al., 2012). Dinoponera lucida seems to have a more diverse diet than we recorded, which may be related to our short sampling period. Specifically, we identified a relatively low frequency of plant items (6%) when compared to 16% from other population of D. lucida (Peixoto et al., 2010), 22% from D. gigantea (Fourcassié & Oliveira, 2002) and 30% from D. quadriceps (Araújo & Rodrigues, 2006). Workers of D. quadriceps (Araújo & Rodrigues, 2006) and D. gigantea (Fourcassié & Oliveira, 2002) collect mainly animal food items, mostly arthropods, as we observed for D. lucida. It is possible that the diet of D. lucida and others Dinoponera vary seasonally according to the availability of food resources in their territories. Therefore, additional studies on the diet composition of D. lucida may be useful to better describe its ecological role in the Atlantic Forest.
Our results were only focused on three nests. We recommend further study to assess populations level responses to microclimatic factors across the species range. The density recorded herein, however, was similar to the reported by Peixoto et al. (2010), but far from describing a continuous and homogeneous distribution of D. lucida nests in the forest. Nests of D. lucida occur in an aggregate distribution (Peixoto et al., 2010), a consequence of its reproduction mode by colony fission (Mariano et al., 2008;Campiolo et al., 2015). During the active search for nests, we found vast areas within the Atlantic Forest in VNR without the presence of D. lucida workers. This is aligned to the endangered conservation status of the species by Brazilian Red List (Ministério do Meio Ambiente, 2014; Instituto Chico Mendes de Conservação da Biodiversidade & Ministério do Meio Ambiente, 2018), which does not seem to be particularly abundant, even where adequate conditions and resources exist.
Our data support the hypothesis of optimal foraging regulated by habitat temperature. It would be a direct response by prey species and an indirect response by D. lucida as they may respond more directly to prey availability. Time spent in foraging trip and litter temperature were positively correlated. Flow of workers was higher in an optimal litter surface temperature range, with more successful foraging trips and more food items (mainly macroinvertebrates) collected in the same temperature range. The better understanding about D. lucida activity patterns can assist in the planning of study activities, such as monitoring and inventory of this endangered and endemic species.