INTRODUCTION

Traditional behavioural studies employ direct observation as their primary method (e.g., ^{Altmann 1974}) and rely on finding and following animals, the pillars of behavioural studies (^{Boonstra et al. 1994}). However, the success in performing these tasks depend on the species’ biology. For instance, solitary, small, nocturnal, and arboreal species (e.g., some lemurs; ^{Oliver et al. 2020}) may be difficult to find. Conversely, other primate species are large, terrestrial and live in large groups that inhabit open grassland (e.g., most anthropoids; ^{Dixson 1998}), hindering counting or individual identification (^{Plumptre et al. 2013}).

In general, forested areas invoke several limitations regarding access and overall visibility of arboreal species (^{Whitesides et al. 1988}; ^{Barker & Pinard 2001}). For this reason, there is an emerging literature with ingenious methods to overcome these obstacles, such as camera traps (e.g., ^{Pebsworth & Lafleur 2014}), passive acoustic monitoring (e.g., ^{Kalan et al. 2015}), and drones (e.g., ^{Koh & Wich 2012}; ^{Melo 2021}). Although promising, each method still has its caveats, like short recording length (e.g., camera traps; ^{Head et al. 2012}); lack of behavioural context (e.g., acoustic monitoring; ^{Kalan et al. 2015}); limited battery life and struggle with low-light conditions (e.g., drones; ^{Koh & Wich 2012}; ^{Gade & Moeslund 2014}).

Thermal cameras (TCs) are able to capture objects that emit radiation (i.e., electromagnetic, infrared or thermal radiation; 8–13.5 µm) accurately and non-invasively (^{Kaplan 1993}; ^{McCafferty 2007}; ^{Lahiri et al. 2012}). TCs convert temperature into images even during low light conditions and for extended periods of time (e.g., ^{Hughey et al. 2018}). Because animals naturally lose heat, they are detectable by TCs, expanding their application to a myriad of subjects, from disease diagnosis to reproductive processes and behaviour (^{Cilulko et al. 2013}; ^{Zhang 2020}). Although TCs are not a novel technique (e.g., ^{Boonstra et al. 1994}), it has gained popularity and is becoming increasingly affordable (^{Graveley et al. 2020}).

TCs allowed the observation of rare and elusive species, providing new and invaluable insights on species’ behaviour and ecology (^{Hristov et al. 2008}; ^{Burke et al. 2019}; ^{Melo 2021}). For instance, nocturnal species may present diurnal behaviour (e.g., Lemur catta, ^{Donati et al. 2013}), particularly in the tropics. The main explanations are to escape from heat, from predators or to exploit new niches (^{Hill 2006}; ^{Donati et al. 2013}). However, even common or diurnal animals show unrecorded behaviours when observed with TC (e.g., ^{Hristov et al. 2008}). One example is the thermoregulatory behaviour (e.g., howler monkeys; ^{Thompson et al. 2017}) and the nocturnal behaviours such as sleeping location in diurnal species (e.g., gibbon; Nomascus hainanus^{Zhang 2020}).

Howlers (Alouatta spp.) are diurnal Neotropical monkeys, considered one of the best studied Neotropical primate genera (^{Bicca-Marques et al. 2020}). Despite the abundance of studies, few have reported their nocturnal behaviour (^{Dahl & Hemingway 1988}; ^{Drubbel & Gautier 1993}; ^{Byrne & Da Cunha 2006}; ^{Peker et al. 2009}). These few studies reported important information such as howling at night (^{Byrne & Da Cunha 2006}), occasional births (^{Peker et al. 2009}) and foraging, although the latter based on the sound of arboreal movement, without direct visualization (^{Dahl & Hemingway 1988}). The brown howler monkey (Alouatta guariba) inhabits a large portion of the Atlantic Forest (^{Jerusalinsky et al. 2021}). There are two currently recognised subspecies with striking differences regarding their extinction risk. The northern brown howler monkey (A. g. guariba) is rare and critically endangered, while the southern brown howler monkey (A. g. clamitans) is considered common throughout its distribution. However, the species suffered severe population decline due to the yellow fever outbreak, classified as Vulnerable in 2020 (^{Jerusalinsky et al. 2021}). Given the gains provided by the TCs (see above), even for diurnal species, this species would be benefited from increased knowledge of their behaviour while serving as a model for TCs tests, enhancing our understanding of their sister subspecies.

Here we report the use of thermal imaging in the observation of this free-ranging Neotropical primate, providing new data from its biology and contributing to its behavioural knowledge and fieldwork routine. We believe that TCs may be helpful in similar field-works applied for this genus or others, particularly those regarding the behaviour of forests-dwelling or nocturnal species.

MATERIALS AND METHODS

The study took place at the Parque Estadual Carlos Botelho (PECB), São Paulo, Brazil, with approximately 38000 hectares, included in one of the largest Atlantic Forest remnants in the world (^{Pisciotta 2002}; ^{Ribeiro et al. 2009}). This biome is a non-seasonal ecosystem regarding rainfall (annual mean of 1440 mm) and temperature (mean 19.7 ºC), although with marked fruiting seasons (^{Morellato etal. 2000}), and a canopy height of up to 40 m (^{Cullen et al.2001}).

The results of this study were part of a larger project investigating several aspects of brown howler monkey behavioural endocrinology and ecology (e.g., ^{Fuzessy et al. 2020}). We followed three groups between November 2017 and December 2018 (with a pilot field work in February 2017) and no animal was radio collared. Habituation of animals was adequate (observation was within, approximately, 20 m), except for two shy individuals that fled from the observers when approached.

We collected behavioural data and faecal samples concomitantly for hormonal analysis and seed dispersal experiments. Collecting the first faecal sample of the day is a standard protocol for hormonal analysis because, usually, these samples are more concentrated in hormones (^{Heistermann 2010}). Hence, we stayed with the animals for two hours after sunset every day and arrived at the sleeping tree one hour before sunrise to ensure that animals did not change sleeping sites and that we did not miss the morning voids. However, dense vegetation hindered direct observation before sunrise or after sunset. In our study site, sunrise varied from 05:50 h, during summer, and 06:50 h, during winter, while sunset was between 20:50 h, during summer, and 17:30 h, in winter.

Distance from animals was measured using a hypsometer Nikon Laser Forestry Pro (Nikon Forestry Pro WaterProof 6x21 6.0º#057222). As for the thermal imaging device, we used LTO-Thermal Tracker (Leupold^®Tracker Thermal Viewer Black 172830, 6x Digital Zoom Thermal Imaging Monocular, hereafter “TC”) that measures relative, not absolute, temperature (i.e., signature temperature) among objects, and displays an image on a liquid crystal display (LCD). LTO-Tracker is portable, weighs approximately 290 g, measures 15 cm x 3 cm and costed USD 565.99 in October 2017. The device presents six colour pallet modes: black, white, red, green, hi-white, and hi-black (Fig. 1A-F). The following results refer to the black mode.

The first field seasons (in February and November 2017) were completed without the tracker, allowing comparison among performances. We conducted four series of trials comparing performances with and without the TC. Trial I consisted of direct observation (detection and follow-up); Trial II was group census (location and counting); Trial III aimed to detect fresh biological samples; and in Trial IV, we tested the TC on other animals (homeothermic and of smaller size). We directly pointed the device at animals for the first trial when we knew their location (i.e., we saw them with an unaided eye). As for the second series of trials, we mimicked survey/line transect by scanning nearby trees when we did not know the animals’ location. For the third series of trials, we assessed whether the device detected fresh faeces or urine by examining the ground right below where the animals had defecated/urinated. Finally, for the fourth trial, we mostly wanted to understand the detectability limit, pointing towards random animals whenever we saw one.

For each trial, we determined how much of the animal’s body was exposed to the device (expressed in quarters, i.e., 0 % - not visible, 25 % - one limb visible, 50 % - two limbs visible, 75 % three limbs visible, and 100 % - completely visible). We also recorded if the weather was sunny or cloudy (no intermediates), daytime, and distance. If we failed in a trial, we tried to identify the reason, considered as caveats.

RESULTS

Trial I (direct observation without the TC)

All groups regularly moved from their original sleeping branches or trees overnight (42 % of the times, travelling up to 80 m far). When they moved to a different tree, we were able to find groups again in only one occasion by listening to their movement, recording the urination or defecation. Howlers showed nocturnal activity in 53 % of the days, confirmed by noise and the use of flashlight, staying becoming active for about one hour before sunrise or after sunset. In one instance, the dominant male of one group was seen sprinting after sunset, and we lost the group because we were unable to follow them in complete darkness. We could not find them again in the following days. Nocturnal activity was more frequent in the rainy/summer season (60 %), and the most prolonged hours of activity were in November.

Trial I (direct observation with the TC)

If the group moved overnight, we were able to scout nearby trees and quickly locate howlers before sunrise 100 % of the time (Fig. 2 A and B). Additionally, we could successfully identify and observe individuals 50 m away. Animals were still detectable at 100 m but lacked definite shape. We could observe groups after sunset and their nocturnal activities included play, latrine, scent marking, foraging or general movement. In one of the occasions, the individuals who were playing were the two shyest ones, and we managed to observe them for forty minutes. In another troop, one rover individual approached the group after they were set to sleep, and the group repeatedly chased it away for an hour. During the following days of diurnal observation, we identified this individual as an immigrating female. Additionally, we witnessed interspecific interactions (e.g., with southern muriquis Brachyteles arachnoides, and black capuchins Sapajus nigritus) before sunrise.

Trial II (census without the TC)

We could only locate groups by scouting their territories and preferred trees because howling is infrequent since there was no radio-collared animal. Location could take up to 15 days, highly dependent on movement or sounds of movement.

Trial II (census with the TC)

We employed the TC to locate primate groups of unknown location for one week, simulating a survey/line transect census. These attempts were unsuccessful due to dense foliage, heat, and the need to search in a wide area. However, if we knew where the group was, we counted individuals more quickly, although we did not perform this counting systematically. Counting difference between a naked eye and the TC was particularly noticeable for muriqui and capuchin groups, as they live in troops of more than 10 individuals.

Fig. 1 Colour palettes available in the LTO-Thermal Tracker were: A) black; B) white; C) red; D) green; E) hi-white; and F) hi-black. Warmer element colours varied according to the colour palette. 

Trial III (biological sample location without the TC)

Animals defecated at high branches, mincing and scattering samples before hitting the ground. Moreover, the stool was camouflaged with the forest floor. Hence, in the first two months without the TC, we could not always find or gather the minimum amount of faeces for hormonal analysis.

Trial III (biological sample location with the TC)

With the tracker, the samples stayed warm long enough (a few minutes) for us to quickly locate them (Figure 2C), even if it was beneath a leaf or a branch. We increased the number of samples collected from 26 (without the TC) to a mean of 59 samples per month (with the TC). The tracker was particularly useful for smaller samples (e.g., from young individuals) as the stool pieces were nearly indistinguishable from their surroundings.

Trial IV (detecting other animals)

We conducted tests on two frogs (Figure 2E), two snakes (Figure 2F) and one freshwater tortoise (Figure 2G). They were indistinguishable from their surroundings. We successfully observed medium-and small-bodied animals such as one tinamou bird Tinamus solitarius (Figure 2H), one squirrel (Figure 2I), one hummingbird (Figure 2J), and one rodent (supplementary video). Despite the success, it is important to state that small animals generally moved too quickly to be observed for long periods.

Fig. 2 Examples of thermal tracker use in black mode in the field. A) howler monkey individual sleeping on a tree branch; B) howler monkey individual sleeping on a tree branch approximately 100 m away; C) faeces on the ground; D) heat causing colour aberrations; E) frog on a hand; F) snake on the ground; G) fresh water tortoise on a hand; H) tinamou bird on the ground; I) squirrel on a branch; J) humming bird on a branch; K) half a pen under plant litter; and L) a battery almost fully covered by plant litter. 

Caveats

We provide a comparative overview of all these results in Table 1. Weather and time of the day strongly influenced tracker effectiveness. Sunny days caused many colour aberrations on the LCD screen due to branches and leaves reflecting or radiating heat. The dark patches shapes at the LCD screen could be easily mistaken with an animal silhouette (Figure 2D). The tracker was most effective during the night or when lower environmental temperatures occurred, such as cold or rainy days, emphasising the howlers’ silhouettes.

DISCUSSION

Howler monkeys are among the most well-known groups of Neotropical primates, but their nocturnal activity remains largely unknown (^{Dahl & Hemingway 1988}; ^{Drubbel & Gautier 1993}; ^{Byrne & Da Cunha 2006}; ^{Peker et al. 2009}). Here we report the effective use of a thermal tracker, to observe a richer behavioural repertoire, even during their nocturnal activity. In general, TC was especially beneficial in low light conditions, drastically improving our ability to eyewitness rare and subtle interactions.

The TC was most beneficial for direct observations, enabling us to witness intraspecific encounters before sunrise and after sunset, which were rare during daytime. Rare behaviours, such as social interactions, immigrations, and interspecific encounters, are challenging to observe and detect (e.g., ^{Bolin 1981}; ^{Dias & Rodríguez-Luna 2006}; ^{Haugaasen & Peres 2008}; ^{Bowler & Bodmer 2009}), especially during the night time.

Table 1 Combined results of all trials. Efficiency means that the thermal tracker was useful in a given situation (first column), and its efficiency was classified as low/null, medium and high. 

We observed rover individuals lurking the groups at night, and the group frequently howled towards visitors or neighbouring groups in the morning. One study reported howlers howling at night (French Guianan howlers, ^{Drubbel & Gautier 1993}), and one study described nocturnal activity for Belizean howlers (^{Dahl & Hemingway 1988}). Night howling is still poorly understood, and the park staff have said that they have listened to the howlers at night, although very rarely. Based on our findings, perhaps these night vocalizations could be directed to visi- ting individuals, although we did not observe night howling during these nocturnal interactions with rovers.

We were able to observe shy individuals (and even shy species; e.g., rodents), possibly because we stood still and observed them non-invasively (i.e., no noise or external light). The presence of an observer (“observer effect”; ^{Carpenter 1934}) has always been under great discussion since it can modify animal behaviour (^{Tuyttens 2014}) or cause stress (^{Cañadas-Santiago 2020}). Therefore, we considered that thermal imaging could minimise human presence and disturbances.

Another major advantage was reducing time searching for groups if they changed trees overnight. Losing groups can be particularly problematic for studies that combine behavioural data with hormonal analysis. Collecting the first faecal sample of the day is a standard protocol for hormonal analysis because, usually, these samples are more concentrated in the morning (^{Heistermann 2010}). Therefore, researchers should find their study individuals before they wake up, one of the main high points of the TC. The ability to quickly locate fresh stool on the forest floor is another major asset. Designed to detect fresh blood on the ground, the tracker was able to show exactly where the fresh stool was. Moreover, the tracker can be particularly useful for young individuals since they produce tiny droppings at unpredictable moments (^{Inoue et al. 2007}), unlike adults that usually defecate in latrines (e.g, ^{Fuzessy et al. 2020}). Because fresh samples are essential for endocrine and seed dispersal studies (^{Estrada & Coates-Estrada 1991}; ^{Hodges & Heistermann 2003}; ^{Wehncke et al. 2004}; ^{Fuzessy et al. 2020}), increasing the number of samples identified by the TC can greatly boost these types of studies.

Among the caveats, we experienced common problems such as the inability to see through dense foliage (^{Boonstra et al. 1994}). Dense vegetation is still impossible to overcome with thermal imaging, and it is its main drawback. Moreover, since this specific thermal camera measures relative temperature among objects, it was ineffective when observing animals whose temperature did not differ greatly from their surroundings, such as frogs, snakes, and tortoises, or when trying to locate homeothermic animals in the middle of the day.

Neotropical primates are mostly diurnal but present a set of traits that reduce their visibility and tracking success. They are small and agile (e.g., ^{Fedigan et al. 1988}; ^{Fedigan & Jack 2012}), hindering tracking success. Additionally, all Neotropical species are arboreal, and many live in densely forested areas, an environment with decreased visibility and with a greater distance from observer (^{Glander et al. 1991}). Because forests are filled with obstacles, forest-dwelling primate species remain understudied, with considerable knowledge gaps in their behaviour and general biology. In our study area, the Atlantic Forest, visibility was poor due to steep terrain and dense understory, preventing us from understanding the context of many behaviours (as already emphasised by ^{Glander et al. 1991}). Additionally, night-time movements occasionally resulted in losing howler groups for a whole campaign, greatly reducing data collection. Hence, thermal camera is useful not only for observing nocturnal species, but also diurnal species during lowlight conditions.

CONCLUSIONS

Neotropical and nocturnal primate behaviour remains understudied due to their cryptic nature and the habitat in which they live. Any method that enhances the opportunity to observe animals with minimal stress is encouraged. A thermal tracker can locate animals and allow behavioural observation in low-light conditions with minimal disturbance. It also allows quick location of stool, a major contribution to studies that depend on this type of sample, such as endocrine/seed dispersal studies. Thermal imaging devices are becoming increasingly affordable and portable and can easily be integrated into commonly used field equipment. We advocate for its employment as an invaluable tool for studying Neotropical primate species.