If you have ever tried to take a ‘freeze-frame’ photograph your pet leaping in a trick move, a bird flicking past your feeder, or your kid at a sporting event, you have probably looked at more than a few blurry photographs.  Getting a crisp photograph of a body in motion requires a faster shutter speed to provide a sharp image of what is happening at the instant the photo is snapped.

A blurry photograph of a mother tamandua carrying its baby.

Using GPS tags to study animal movement can be compared to taking snapshots of behaviors we can’t observe with our own eys – scientists connect a series of locations taken by an animal’s GPS collar to create a picture of how the animal uses its environment.  The schedule of the GPS unit sets the resolution of the picture we get about the animal movement.  More frequent fixes (e.g. every few minutes) give a high resolution image of where the animal goes while less frequent fixes (e.g. every few hours) are analogous to a blurry photograph. But as anyone who has ever used a handheld GPS unit knows, they chew through batteries like nobody’s business if they collect fixes constantly.  Unfortunately, wild animals won’t change their collar’s batteries when they run out, so scientists have to make the most out of one battery, and therefore face a dilemma in how to program the collars they will use on animals: infrequent GPS sampling will make the collar last longer, but give a fuzzy picture of movement paths; more frequent GPS sampling gives sharp paths but dramatically shortens battery life.  This is an especially difficult problem when animals spend long stretches of down time in tree cavities, burrows, or thick vegetation where a GPS collar has little hope of successfully connecting with satellites and wastes even more battery power in futile attempts to record locations on schedule.

What is needed is a flexible GPS schedule tied to the behavior of the animal wearing the collar.  A more active animal would trigger the GPS unit to record locations more often, while a resting animal would signal the GPS unit to record locations less often.  What kind of sensor monitors the moment-by-moment behavior of wild free-living animals in any type of habitat?  An accelerometer—a matchbook-sized wireless device that measures the change in speed of an animal’s body over time as it moves through its environment.  Accelerometers are in everything from vehicle airbags to Nintendo Wii handsets and over the past few years they have exploded onto the scene in the world of animal movement research.  They are an ideal sensor for linking animal behavior to the location recording schedule of a GPS collar because they are low-cost, can easily be incorporated into a standard GPS collar and they sample movement behavior every few seconds without using much battery power.  So in theory a researcher can have the best of both worlds: a long-lasting GPS collar that gives sharply focused pictures of the paths animals take as they move around their habitats.

Testing the performance of just such a collar is the subject of a paper we recently published in Wildlife Society Bulletin (Brown et al 2012).  We worked both with fisher in upstate New York and Tamandua anteaters on Barro Colorado Island in Panama, testing two types of GPS: one that recorded locations on a fixed schedule every 15 minutes and the other with a flexible schedule that recorded locations every 2, 15 or 60 minutes based on accelerometer-measured movement behavior.  The accelerometer-informed collars performed considerably better than the traditional GPS collars: they attempted 74% more locations per day and had 62% higher location success rates, which means that on days animals were more active, GPS collars recorded more locations thus providing more detailed movement paths.  At the same time they spent 28% less time searching for satellites and recorded 67% fewer locations when animals were at rest, reducing the overall amount of battery power used for each unique location and lengthening the lifespan of the collar.  Ultimately the accelerometer-informed GPS collars produced more information about animal movement for a given battery size and study period when compared to traditional fixed-schedule collars.  This technological development is a boon for researchers and potential study animals alike: ecologists get higher quality data with little additional cost per collar and can instrument fewer study animals for shorter periods of time than they would using traditional collars because the snapshots of daily movements are so much clearer.  Currently only two companies (e-obs (used by our study) and Telemetry Solutions) produce accelerometer-informed GPS collars, but as word gets around, those scientists studying animal movement ecology are sure to appreciate the value of this novel tool.

map of fisher GPS data

Map showing high resolution 'picture' of the life of Phineas the Fisher as recorded by an adaptive GPS collar. Click the map to zoom in on the map at Movebank.org

Reference:  Brown, D.D., LaPoint, S., Kays, R., Heidrich, W., Kuemmeth, F. and M. Wikelski.  2012.  “Accelerometer-Informed GPS Telemetry: Reducing the Trade-Off Between Resolution and Longevity” Wildlife Society Bulletin 36(1):139-146.

 

Written by Danielle Brown

Advertisements

Biologist love islands. It’s true, we really do. Animals isolated on oceanic islands inspired both Charles Darwin and Alfred Wallace to independently come up with the theory of evolution by natural selection, and continue to be a focus of research today. Thus, it is surprising that so little is known about the fauna of Central America’s largest island, Coiba Island. The lack of research is even more surprising when you consider that the island has a unique form of what may be North America’s most studied wild mammal, the White-tailed Deer, living on it.

View from Coiba

view from Coiba

Approximately 27 miles off of Panama’s Pacific coast, Coiba was formed over 20,000 years ago, providing enough time for substantial genetic isolation to occur on any species able to colonize it. To date, 17 terrestrial mammal species, excluding 25 bat species, have been identified on the island. Nine of these were likely to have been introduced within the last century, leaving only nine native terrestrial mammals, five of which are endemic. The only proper survey of the mammals of Coiba Island was conducted by J.H. Batty in 1902, and for good reason (Olson 2008). Research on Coiba has certainly been limited by the penal colony established on the island from 1919-2004, where Panama sent some of its most notorious criminals and gang members such as those of “Los Hijos de Dios” (The Children of God), and “Los Chuckys” (named after the possessed doll of the 80’s horror flick “Child’s Play”). Vivid tales of violence, torture, and political murder during the dictatorship of Torrijos and Noriega filtered back to the mainland instilling a general fear of the island. While rough on the people living on Coiba, this penal period provided protection to the island’s forest and approximately 80% of Coiba’s forest still remains today (ANAM 2009). With the prison now closed, Coiba is gaining a reputation as a popular tourist destination and is protected as a World Heritage Site and National Park.

But what animals patrol underneath the forest canopy? Are the rumors of big cats existing on the island true? Is it possible that, somewhere within the 50,000 hectare island, mammals remain that have yet to be documented? These mysteries were the motivation behind our new camera trap pilot survey of the island’s mammals

In May of 2011, I deployed ten cameras along a trail on the Northeast corner of Coiba Island to capture a snapshot of the diversity, abundance, and activity patterns of its terrestrial mammals. In total, only four mammal species were documented: the Coiban Agouti, Panamanian White-throated Capuchin, Coiba Island White-tailed Deer, and Black-eared Opossum. Compared with our similar surveys on mainland Panama, the agoutis and capuchins of Coiba Island were very common; the agoutis were photographed 2-7 times more often than typical mainland sites and the monkeys 10-100 times more. This information alone is extremely interesting and may suggest two things: 1. There is a general lack of native predators on the island, causing an expected spike in smaller mammal populations. 2. Coiba’s capuchins may come to the ground more frequently than those of the mainland.

One of the shiest animals I photographed was the Coiba Island White-tail Deer (Odocoileus virginianus rothschildi). Our camera trap photos show that they walked in front of our cameras twenty-eight times, or about once every 10 days, and that they are active throughout the day with an apparent peak around dusk (17:00 – 19:00). These photos generated the most attention from Coiba’s park rangers back at the ANAM dining hall, the only place at the station with both a table and electricity during the day. Their interest came as a bit of a surprise since I saw deer so frequently in the mornings while hiking on Coiba, but reminds me why they are the most popular mammal on the continent. And these are special deer – they look like no other white-tailed deer in the world.

Coiba Island White-tail Deer (Odocoileus virginianus rothschildi)

This male, being kind enough to model, provides us with a nice look at the various characteristics of the head:
A. Notice the white markings on his muzzle and around the eyes. B. A perfect example of his dark facial markings. Notice the reddish “crest” on the forehead. C. Here the coloration of the chin and throat are easily viewable.

Oldfield Thomas first described the Coiba Island White-tail Deer in 1902 while documenting the private collection of Walter Rothschild, the very same specimens that J.H. Batty collected during his trip to Coiba earlier that year (Olson 2008, Thomas 1902, Allen 1904). The most distinguishing characteristic, aside from their isolated range, is their size; much smaller than Panama’s mainland white-tail species O.v. costaricensis. Additionally, adult O. rothschildi have much darker coats and the white spots of fawns tend to be more inconspicuous. Are their coats more cryptic in response to hunting, was there once a significant non-human predator on the island that influenced this potential adaptation, or is this simply the result of a founder effect?

Average dimensions of the animal are unknown although both Allen and Thomas included measurements in their descriptions. I say this because Thomas’ measurement of the head and body (1120mm) differ from Allen’s (avg. 2287mm) by over 1000mm. In addition, the measurements described by Allen were supplied by Batty, whose information often prompted complaints from Allen, and at times was unreliable or completely fictitious (Olson 2008). Both Allen and Thomas collected measurements from the skulls of six individuals (avg. length 205mm; avg. breadth 86mm). Even then both Thomas and Allen agree that this is the smallest in its genus, humorously describing it as a, “tiny little deer…”

There is obviously much to learn about this little guy as well as the rest of the mammals of Coiba. Plans to continue and expand the camera trap operation are underway. Proper measurements, as well as blood, tissue, and fecal samples also need to be collected. These can help, as they say, set the record straight, and provide valuable behavioral, genetic, and physiological data for future research.

Despite its horrific past, Coiba’s future is has huge potential. The number of scientific research projects on and around Coiba is steadily increasing, such as the recent submarine expedition of Hannibal Bank. Tourism is also increasing with visitors attracted to the unparalleled experience of scuba diving in one of the world’s richest marine biodiversity hotspots. Be sure to visit Coiba if you have the chance. Spend some time diving. Get a tour of the historical prison. Maybe even catch a glimpse of the tiny Coiba Island White-tail Deer. I know you’ll have a great time. After all, biologists are not the only ones who love islands…

By Zach Welty

 

Allen, J.A. 1904. “Mammals from Southern Mexico and Central and South America.” Published by order of the Trustees, American Museum of Natural History. v20.

ANAM. 2009. “Plan de Manejo del Parque Nacional Coiba.” Compiladores JL Maté, D Tovar, E Arcia,Y Hidalgo, STRI.

Olson, Storrs L. 2008. “Falsified Data Associated with Specimens of Birds, Mammals, and Insects from the Veragua Archipelago, Panama, Collected by J. H. Batty.” American Museum Novitates 3620(1):1. Retrieved (http://www.bioone.org/perlserv/?request=get-abstract&doi=10.1206%2F592.1).

Thomas, O. 1902. “On Some Mammals of Coiba Island, Off The West Coast of Panama.” Novitates Zoologicae 9:135-137.

We have a new paper out in the journal Methods in Ecology and Evolution:

http://onlinelibrary.wiley.com/doi/10.1111/j.2041-210X.2011.00183.x/abstract#.Tvt1QWJ0ql0

One of the biggest problems studying seed dispersal is that seeds are hard to follow! With traditional seed tagging methods, many seeds simply cannot be found, especially when they travel long distances. One method researchers have used is to limit their search to a reasonable distance and track all the seeds within that radius. The problem with this method is that it tells one little/nothing about long distance dispersal. Given that long distance dispersal is believed to be particularly important, this lack of information is problematic. Patrick Jansen and colleagues developed a statistical method to reconstruct the shape of the tail of the seed dispersal distribution using the shape of the distribution inside the search radius. We thought this was a cool method and had the potential to be used more widely. Unfortunately, this method had never been tested with empirical data. Fortunately, we had a perfect dataset to test the method because we used radio-transmitters to track seeds in our project. We also decided to give it a name: the Censored Tail Reconstruction method (CTR). In general, we found that the CTR method worked exceptionally well at recreating the long-tail of the seed dispersal distribution. On the other hand, this method was highly sensitive to which mathematical function was used in the method, and what percentage of seeds were ‘overlooked’ by researchers searching for seeds. The upshot of our paper is that it is possible to calculate fairly accurate dispersal kernels using censored data collected with traditional low priced tagging methods. The caveat to this is that researchers need to be certain that they are able to find the vast majority of seeds within their search radius, and they must choose the most appropriate mathematical function for use in the CTR method (using AIC selection). We think this method should be widely adopted, especially by researchers who cannot afford tons of radio-transmitters.

In our first field season we noticed that some seeds traveled so far that they must have moved out of the home range of an agouti. This led us to wonder whether the massive seed re-caching behaviors we observed were the result of one agouti caching its seeds many times or agouti thieves moving seeds from one territory to the next.

To solve this question we needed to be able to identify individual agoutis, which otherwise all look alike. After substantial live-trapping effort we were able to catch >20 agoutis in one portion of the island (in and around the 25 ha plot). We individually marked each agouti we caught with a distinctive radio-collar, ear-tag and/or freeze brand. Then, we put a motion-sensitive camera next to a buried seed where we knew the “owner” and recorded every agouti (or other species) that walked by. This allowed us to determine whether the cached seeds were being dug up and moved by their owners, or by thieves. We are still crunching the numbers from these experiments but in the process discovered a cool video that shows just how crafty agouti thieves can be.

Many rodents bury seeds in times of plenty to save them for lean times, and these hidden food caches are critical to their survival. Rodents will go to great lengths to protect their seeds from potential thieves. For example, Michael Steele and colleagues found that if a squirrel wanted to bury a seed but was being watched, it would often behave “deceptively” by making fake caches and making caches behind trees so the observer could not see the cache being made. While I would have loved to do some similar experiments with the agoutis, it would have been hard to observe the agoutis without scaring them away. However, if you run enough camera traps for enough time you eventually record some surprising clips.

In this case, I didn’t notice what was going on the first time I saw this video, and only recently realized how it shows the subtle dynamics between different agoutis sharing an area. The seed that this camera was monitoring was previously cached by an agouti named Tracy, who can be identified by a small diagonal white freezebrand mark on her body. In the video you can see Tracy come to uncover her seed four months after she originally placed it there (we had a camera there the whole time). Given the high rates of cache movement, Tracy was lucky her seed lasted that long and her strategy of saving food underground for the low food season seems to be a good one. Unfortunately for her, as soon as she begins digging up her cached seed she gets chased away by another more dominant agouti who steals her seed. Even sadder is the fact that Tracy came back to the old cache location a few seconds later to see if it was still there (it wasn’t). While this isn’t fair to Tracy, the other dominant agouti got a free meal on the cheap. The degree to which our agoutis used this kleptoparastic strategy is unclear, but given the behaviors seen in this video, we think it might be a good idea if agoutis used the same sorts of deceptive behavior found in squirrels.


If you get excited about sensor networks, yagi antenna, live data streams, and agoutis, then you’ll love our new paper in the Computer Journal describing the technical details of the Automated Radio Telemetry System (ARTS) that helped us track animals and seeds in Panama.

 

Photo: ARTS tower rises above the tropical rainforest canopy of BCI to help researchers track the movement and activity of animals and seeds, sending data back to the lab in real-time.

ARTS tower

With our kitchens and restaurants it’s been easy for us humans to settle into a comfortable eating routine of a morning breakfast, mid-day lunch, and evening dinner.  Animals, on the other hand, have to hunt down their meals out in the wild.  For agoutis, and other ground-dwelling frugivores, this means walking around waiting for fruit to fall out of trees and land on the ground.  That leads to a simple question – when do fruits fall out of trees?

As far as I can tell, this question has never been studied – who wants to sit around and watch fruits slowly fall down all day, and night?  Yet this is a basic bit of natural history information that could be quite important when considering the survival strategies of agoutis, in particular, and daily rhythms in the forest, in general.

Camera traps are a perfect tool for this question, and our student Vivian Mass set some aimed at the fruits of Astrocaryum palm trees.  For this to work she had to find another tree at just the right distance away from the targeted palm-fruits and then get a camera trap up there, which she did with the help of ladders and some canopy access climbing by Daniel and Alejandro.  This was a bit of a side project from Vivian’s main thesis, and the data remained unanalyzed until this spring when high school student Tessa (Taz) Holliday joined our team as an intern from the Emma Willard School.

Taz looked through each picture and noted if any fruits had fallen off since the last picture.  In some cases it was obvious – a big howler monkey was on the tree picking fruits, eating the fleshy part, and then dropping the nut down to the ground, where agoutis and other terrestrial critters could find them.  In other cases fruits would fall off without any animal intervention, just because they were ripe.  In total Taz noted 450 seed fall events from 8 different Astrocaryum trees, with slightly less than half being dropped by animals.

Looking at the fruit fall over the entire fruiting period (figure 1) it was obvious that arboreal animals had a huge effect on when a fruit would fall.  A few trees that had no arboreal animals visit would drop a few fruits a day for 2-3 weeks.  Once an animal visited they would drop all, or nearly all, of the fruits in one sitting.

Fig 1. Timing of fruit-fall over four weeks for 6 Asrocaryum trees.

These arboreal animals also had an impact on what time of day the fruits hit the ground. The monkey- and parrot-fed fruits fell in the day while the kinkajou-fed fruits fell at night (figure 2).  Overall, this led to more fruits falling during the afternoon than you might otherwise expect (figure 3b), although with only 4 trees in this analysis (the time lapse photos didn’t work for 4 trees), we should be careful in what we conclude from it.

Fig 2. Time of day that fruits fell out of 4 Astrocaryum trees

Even more surprising was considering what time the fruits fell when no animals knocked them off. This is tree-behavior, when do they ‘want’ their fruits to fall to the ground.  Surprisingly, this showed a strong trend to more fruits in the late night (midnight-5am, figure 3a).  We know from our agouti tracking that this is the most dangerous time for agoutis to be out foraging, we have found quite a few agoutis killed by ocelots in this time period.  Are the trees trying to temp the agoutis out for an early breakfast?

Fig 3. Expected and observed times of day that fruits fell out of trees, with and without arboreal animals (significantly different than random p<0.0001)

Lets call this the “Machiavellian plant behavior hypothesis”.  Long-lived agoutis might be bad for trees because they remember where they buried all the seeds and come back and eat them year after year.  However, if dropping seeds more at night led to a higher turnover of agoutis nearby due to ocelot predation, it might also lead to a better chance that their cached seeds survived, since the “new agouti on the block” wouldn’t know where the “recently deceased” buried the seeds. This Machiavellian plant behavior seems a long shot, but this preliminary result that Taz teased out of data collected earlier by Vivian suggests it might be worth following up on.

On Barro Colorado Island we have the rich agoutis and the poor agoutis. The rich agoutis live in fancy neighborhoods with lots of palm trees, which make nuts for them to eat. The poor agoutis live in areas with few palm trees, the agouti ghettos, with less food. Rich or poor, all agoutis bury seeds in scattered underground caches as an insurance policy for the end of the rainy season, when there is almost no freshly produced fruit. Of course, the trees would rather the agoutis don’t come back later and eat the seeds, giving the seeds a chance to germinate and grow into a new generation of trees. These buried seeds are more likely to be dug up in the agouti ghettos, where seeds are more valuable because they are so scarce.

One of the ways that a cached seed might avoid being eaten by an agouti is if the agouti which buried the seed dies. This leads to the question: do predators like ocelots help palm trees by killing agoutis before they have a chance to go back and dig up palm seeds?   Related to this, is the question: are agoutis living where food is scarce more likely to get killed because they must work harder for their food, thus take more risks?

These are the questions asked by Willem-Jan Emsens, a Masters student at Wageningen University. He studied activity patterns, refuge use, and space use of agoutis in relation to predation risk. He compared rich and poor agoutis: animals that differed strongly in the amount of Astrocaryum fruits they had in their home range. As any loyal reader of the Agouti Enterprise knows, Astrocaryum fruits are the most important food resource to agoutis.

Willem-Jan used the Automated Radio Tracking System (ARTS) to track the animals 24/7 and determine the home range size of the radio-collared agoutis. This showed that poor agoutis had larger home ranges than rich agoutis. Not surprisingly, poor agoutis seem to need a larger area to find enough food. Therefore, they may take more risks of getting caught by an ocelot.

Willem-Jan also took his radio-tracking receiver and tromped around BCI at night, when agoutis are sleeping, to find out exactly where agoutis had their refuges. He found agoutis used three types of sleeping sites: dense vegetation, burrows, and hollow logs. Each agouti had a 2-8 different places they could sleep on different nights.

Agouti RefugeTypes

Agouti RefugeTypes

Once he identified the sleeping sites, Willem-Jan put camera traps at burrow entrances to see exactly when they came and went. Agoutis never left their refuge before sunrise, but were more variable in the time they went to bed. Most agoutis retired to bed around sunset, but 13% came back at night. The cameras also caught some amazing footage of ocelots coming to these sleeping sights and looking for agoutis. You can actually see the ocelots looking into logs, presumably scaring the crap out of an agouti, but then leaving empty handed. Ocelots were the only predators that visited the refuges, and they visited refuges more often than random locations. It seems that ocelots knew where the agoutis were sleeping dropped by hoping to catch one near the entrance to their burrow.

Some of the agoutis that Willem-Jan was studying were killed by ocelots. By analyzing the ARTS data, he was able to determine the exact time the agouti was killed. Willem-Jan found that agoutis are most susceptible to predation around dusk and dawn, which is also when they overlap most in activity with the nocturnal ocelots. This may also be tied to their entrances and exits from dens. If ocelots know these locations, they might wait nearby to catch a commuting agouti.

One of the surprising results came from comparing the location of the refuges with the overall space use of agoutis. We presumed that agoutis would act like many other burrowing animals, using the refuge as a ‘central place’ and conducting feeding bouts more near the hiding hole than far from it. But this was not the case. Maybe the fear of a lurking ocelot is enough to make an agouti keep its kitchen and bedroom in different parts of its home range.

Finally, combining his work with data collected by an earlier student on the project, Lennart Suselbeek, we were able to compare the risks taken by the rich and poor agoutis. The ghetto agoutis had to be more active to find food in their enlarged home ranges, and accomplished this by getting up earlier and going to bed later. This is risky behavior when your neighborhood is patrolled by nocturnal ocelots. All of this makes the agoutis nervous, but is quite fine with the Astrocaryum trees, who root for the agoutis to burry their seeds and then the ocelot to make them disappear.