Problems when camera monitoring

Cameras take photos, they do not monitor. The term ‘camera monitoring’ is fairly misleading unless the objective is purely finding out whether the species is present.
To be able to interpret the wealth of additional information collected by the camera and to work out patterns or observe behaviour it would also always be necessary to keep records about all details that can be gained from a photo. If cameras can only be deployed for a short period of time, they can only give the information what animals of what species were sighted during that period.
For more insights, long-term deployment and meticulous recording of all available data – and long-term storage of the photos – would be necessary.

Camera usage in research would be based on a hypothesis, the objectives would be clearly formulated and bias would be eliminated as far as possible. This is outside the scope of a community project. The main advantage of our camera usage was that deployment was long-term (up to 5 years) which gives a potentially broader impression of the changing situation in a habitat patch and of behavioural patterns. Results could then potentially be used for the formulation of hypotheses. 

We cannot monitor individuals and our populations are small, but monitoring is the only way of detecting and quantifying population changes in a timely manner.
However, to estimate the size of a population based on camera output, we only count the number of animals in a given period of time that is seen together in one photo.
The common scientific approach would be to count animal observations only if they were separated from the previous observation of the same species by more than 1 hour. (Bengsen, 2014a)
We chose the differing approach as ringtails are highly territorial and it seemed very likely that it was the same resident animal visiting for instance a water station several times in one night. Otherwise the risk of double counting was too high.
This would be different in a high density population such as in some areas in Busselton. The population would be large enough to have a high number of animals visit the same camera trap. In a habitat that only allows for low densities a low number of animals will inhabit the area surrounding the trap, the population will most likely be more spread out and many animals might never come close to one of our camera traps (spatial variability). Random placement would not be possible due to the shortage of available cameras.
We probably filled at least our main monitoring site to carrying capacity but our way of counting will hardly ever result in more than 3 animals in a given period of time – a mother with a young plus another adult. In some cases when some of the animals have distinctive features (such as easily recognisable spots on the tail) we have proof that this clearly leads to an underestimate. While numbers are high, results will stay similar and growth rates for populations cannot be estimated with any certainly without an attempt to identify as many individuals as possible. Varying photo quality would jeopardise identification though and add bias and probably lead to over-estimates.
Also, camera or operator failure results in no photos even though animals are present. This can also be significant in spots that are hard to camera monitor for various reasons (varying detectability). False absences  - present but not sighted - lead to underestimation of true occurrence. Our way of estimation should however be capable of showing drops in population sizes and therefore make early intervention possible – an important issue neglected in official research.

Camera technology is developing fast and it is hardly possible to keep track of all the new camera models on the market. Advances in technology suitable for monitoring small arboreal animals have unfortunately not developed as quickly.

Most wildlife studies using camera traps have targeted ground-dwelling mammals with the exception of Leadbeater’s possum.
Cameras were used at a height of about 4 m,  facing bait stations. An interesting find was that if the camera was rotated  90 degrees so that the motion-detection bands followed the line of the tree trunk instead of going across it, significantly more images were taken. (Harley et al, 2014)
Objects that move horizontally are easier to spot if the camera is placed as intended, however animals running up a tree trunk or otherwise approaching in a straight line, are then often missed as they do not trigger the camera.
Rotating  the camera would necessitate some rain protection for it though.  

The variables that might negatively influence the number of photos and their quality are far too many to even try to avoid all at the same time.
The quality of the camera, its orientation and placement which is often dictated by the environment (trees), deployment time, the size of the animal species, the use (or not) of attractants, the density of the habitat, weather conditions and human factors all influence heavily the output of the camera.

Hot weather conditions reduce the trigger sensibility even if the sensor level is ‘high’ as the differential between ambient temperature and body temperature of animals is small. Temperatures above 35 degrees lead to unreliable results for ringtails. Summer is a season when spotting returns high numbers but cameras usually show a significant reduction in sightings.

Very cold conditions also influence camera performance, but scarcity of photos could also be the result of animals not moving around as much as under warmer conditions. 

High winds or storm lead to swaying of trees and heavy movement of foliage. Even small branches that repeatedly get in the focus of the camera can worsen photo quality to a degree that animal identification to species level is not possible.
However, in some cameras storm in cold conditions leads to no footage at all. 

Rain or a rapid drop in temperature (hot day, cold night) cause high levels of condensation, which at times obscure the image and might also damage the electronics.
Wet fur also masks the heat of the animal and leads to false negatives.  

False negatives (animal present but no photo) can best be validated with the use of an additional camera or a person spotting.

Cameras have generally improved over the last years. Older cameras often took a photo every few minutes (sometimes breaks of up to 5 minutes) so that it was obvious that most of the activity was missed. Lures might be depleted in hot/very cold or rainy weather without any footage of animals taking them.

Subsequent photos:

Vertical movements are rarely ever shown. This can be tested by slowly approaching in a straight line – there might be only one or two photos.

Monitoring of a release cage is difficult as the leaf or shade cloth cover (for privacy and protection) in addition to the metal mesh will block warmth of the animal and lead to very few – if any – photos. If the camera is focussed on the open/uncovered part of the aviary, footage might only show animals in the cage but they never seem to go in or come out because the cameras’ detection bands only cover a small horizontal area. Subsequent photos might seemingly show the same animal which has not moved significantly – but the time stamp might reveal that minutes, hours or even a day have passed.
A roof that has soaked up the warmth of the day can ‘swallow’ particularly small animals and not trigger photos.
A clear indication for a number of false negatives is for instance if a brushtail possum approaches but a ringtail arrives.

Subsequent photos:

 

 

The number of false positives (photos taken with seemingly no movement) is usually higher when using inexpensive cameras.
Tree/leaf and even grass movements in the camera’s field of vision and affixing the camera to a thin, swaying object trigger false positives as long as there is some warmth in motion. A source of heat close by, e.g. a roof that warms up, and the tiniest movement will trigger masses of photos even when the most expensive cameras are used.
Wide panels such as a cage roof often produce false positives particularly if close-up lenses are used.

The rising or setting sun/moon can also be involuntarily documented - usually several days in a row at about the same time.

As minimising false triggering would often necessitate removal of vegetation, which we do not recommend, false positives have to a certain degree to be accepted as an unavoidable nuisance.

Cameras are too deceiving to reliably develop a body condition index. Impressions can change between consecutive photos that show the same animal from a slightly different angle. Neither the gender nor a full pouch can usually be seen.

The size of an animal is hard to estimate as it seems to change with the distance of the moving target. The angle, the lenses, the distance from the camera, all influence the perceived size of the animal.  Only mother/small offspring teams show a clearly definable difference in size. If 2  animals are situated in different distances from the camera, the closer animals will always look far bigger than the one further away. This also makes identification of an animal less likely.

 

Identification / Photo Quality
Identification even just to species level is often only possible when a series of photos is taken in quick succession, by taking videos which show the animal from several angles or by putting more than one camera at a spot and then comparing photos.

Professional researchers claim that it is possible to identify nondescript species (e.g. antechinus or rat or mouse) but this would rely on the setup of cameras and experience of the observer and his/her close attention to detail. (Dundas et al, 2014a)
This would certain not apply if only partial images are taken unless – in the case if the western ringtail – the fragment shows the ears or the tail.
Tails are so conspicuous that even a tail tip is at times enough
to allow a definite species ID.

If the animal is close to the camera (approximately 1 m) photo quality is usually most satisfying. However fast movement will then lead to nothing but blur.

Even a thin tree trunk close to the camera and in its target zone can lead to a focus on that trunk and a highly  overexposed photo.

 

 

Camera placement at spots where animal activity can be expected – trails, water stations, release cages – greatly enhances the probability of detection, but will also impose bias and even create their own sets of problems.

Waterers as target areas need to be protected from predators, but those protective measures (leaf cover) will lead to false positives.  

Animals released from a cage are more likely to come back to it. However focussing a camera on to a cage roof which warms up on a hot day will fail to produce footage if the heat of the animals is similar to the temperature of the roof but might produce endless footage of shadows crawling over the roof.
Sometimes a slightly higher or lower positioning of the camera can severely influence the number of photos. The camera’s detection band that must be traversed by the animal to trigger a photo seems to vary in all cameras – and usually we don’t know where the animal will appear. This is less critical when dealing with ground-dwelling species as they will most likely pass the camera but in a tree we could have several possums and no photos if they do not traverse the space that triggers the photo.

For cage monitoring, the best option is to see the uncovered strip on the upper part of the cage, the roof and a little above.This will capture most movements particularly if the animal is not seen entering the cage, it will still be visible inside.
A high positioning will lead of more false positives as movement in distant trees is recorded but without any chance of identifying animals in the distance.

The best positioning depends on the objectives of the investigating and whether footage from a longer time span should be compared. Even a slightly changed angle (happens frequently when exchanging sd cards) can jeopardise results. A tree animals tend to climb down can be just visible at the edge of the photo and allow for evaluation whether they go to ground a lot or not – until a slight shift of the camera lets the animal ‘disappear’ without further information.

Using lures also greatly enhances detectability of species. As the animals stay longer and move less the number of photos is far higher which aids identification.

 

Length of deployment
A major aim of our studies is to examine ongoing use of a site by that species over a longer period of time which makes long deployment necessary.

However, wildlife will always vary across time and space – water stations will for instance be more highly frequented by mother/offspring teams and release cages are more often visited in the early stages after release. Absence of our released populations can be either because the animals are truly dispersed or because they are present but not detected for a variety of possible reasons.
Detecting generally elusive species would also have higher chances of success with a long camera deployment time. Elimination of presence through camera observation output seems just as questionable to me as through spotting surveys. At least camera observation will only cause minimal disturbance and my experience is that animals get used to the camera being there and do not avoid them in the long run.

 

 

CAM influences on ringtails
Even the most expensive cameras are far from covert. Electromagnetic waves shorter than violet and longer than red are invisible to humans and most animals. Snakes can detect infrared radiation but not truly ‘see’ it. Depending on the wavelength of the LEDs (mostly between 850 nanometre and 940 nm) even people might see a light glimmer and all cameras emit some audible sound. However, the critical point is whether the target animal responds to light or sound by changing its behaviour.

Interpretation/use of photos
How useful any footage will be depends on clear objectives.
The best case scenario for ringtail observation with a high chance of identification of individuals and their behavioural patterns would consist of deployment of 2 still cameras and 1 set to video so that a wide area is covered. The output from such a setup is usually surprisingly varied.

When monitoring predators (ground dwellers) and a target species (arboreal) simultaneously, good camera placement is hardly possible and a general impression of animals present is all that can be achieved.
As ringtails are nocturnal, it is tempting to set cameras to night operation only to save battery power, however 24-hour operations can give some additional insights in what is happening at the site in the day. Even sightings of ringtail possums in broad daylight are far more common than expected.

Comparison between cameras
The use of trapping/monitoring cameras is rapidly expanding, but all cameras have their inherent flaws and imperfections. Various brands and models have particular strengths and weaknesses and it depends on the aim of the study which camera is best suited. However, investigating which of the masses of cameras on offer would be the best for a particular project is already a major task.

It is hard to identify constraints in particular devises and to evaluate whether they could affect the outcomes of a particular investigation. The literature gives lots of information about shortcomings (Meek et al, 2014) but finding out how these may affect individual results is up to the critical researcher.

An interesting piece of research compared the results of 2 camera brands for the use in occupancy modelling.  Conclusions about key environmental factors influencing mammal presence would have been very different for each camera type if used exclusively. (Swan et al, 2014)

In our experience, the performance between all cameras varied markedly – even between cameras of the same brand and model – and performance changed with cameras aging. Trigger intervals, which we regard as one of the most important features, was in all cameras apart from Reconyx much higher than the manufacturer claimed.

Lenses and sensor arrangements seem to differ between brands and even models but usually there are no clear specifications given in the operating manual.

Cameras are as imperfect as observers as people, they are just limited in different aspects.
The question whether a higher number of cameras or a few expensive ones will provide better insights is a personal decision. The main flaw in cheap cameras in our experience is that they produce mountains of photos due to high numbers of false positives and it is a slow and tedious process to go through them, particularly if manpower is very limited.

In the literature Reconyx is the preferred brand and regarded as the most sensitive and reliable camera on the market, the ‘gold standard’.
Reconyx clearly produces the highest quality images. If high quality photos are the most important aim, Reconyx is the best of all cameras we use.
The Reconyx range has a unique detection zone and the target animal is required to move within a horizontal and vertical zone in order to trigger the camera.  Those zones are described in the manual which helps placement. I could not find any description of the sensor arrangements in manuals of any other camera brand.
Placement needs to be so that sensors cover the area where possums most often appear (either on the roof, in the cage or at a certain tree level). A walk test mode helps determine the active motion detection zones but unfortunately it is a completely different situation if a large animal like myself walks in front of the camera or if a small animal like a possum moves in a tree. How do you use the function on ‘mid-tree-level’? As we never know where a ringtail will appear, correct placement is just as hard as with other cameras and movements of  small animals are often not recognised even though Reconyx cameras are able to pick up smaller heat signatures than other brands.

In a comparison between Scoutguard and Reconyx, the latter however recorded more species per site and was more effective at detecting small and medium sized mammals. (Swan et al, 2014)

Reconyx has a very fast trigger speed (e.g. Reconyx HC600 0.203 s), which minimises the time between detection and image capture and increases detection probability. ‘Slow’ trigger speed can result in images being taken after the target animal has already left the field of view.
However, fast trigger speeds may be unnecessary if the animal will be present for some time (e.g. at a feeding or water station).

Trigger intervals are also shorter than in all other brands we use. We mainly set the camera to taking 3 photos in a row, which is usually done in 3 seconds, however 2 photos per second are achievable (rapid fire mode), while all other cameras at best take a photo every 3-6 seconds. The question is however, whether we really want 300 photos of an animal grooming in front of the camera. 

If movement is ongoing, the next series will start straight after the one before. Strangely the professional PC 900 seems to always finish the series of 3 even if the animal vanished after the first picture and there is no ongoing movement.

As the camera is so fast and of such good image quality taking 3 or even 10 photos in a row increases the chance of identifying an animal by picking up certain distinguishing features. Also,  movement can be observed this way even though Reconyx has no video mode. Videos do deplete batteries very quickly.

Reconyx also produces far less false positives than other brands and rarely any series of photos with no visible movement.  According to the literature false triggers account for less than 25% (Urlus et al, 2014) which still seems quite high in my view.
In my own experience this percentage is likely to increase with age. After 2 years in almost constant operation, the number of false positives increased notably.
I have however observed the same issue in other brands (checked with second camera).

I would argue that Reconyx’s  reliability is not as high as claimed as there are usually less photos (showing the target animal) than taken by Bushnell Nature View cameras when used simultaneously as a ‘control’.
In a test in late December (heat!) masses of photos were taken by both cameras but most of the true sightings were only covered by Nature View.

Reconyx cameras are very heavy and therefore hard to put tightly in place, it does not help that the tripod socket is at the back of the camera instead of at the bottom as in other cameras. 

The main obstacle is however the high price. The camera is very expensive (approximately double the price of other well-working brands) which is hard to justify if only ringtail presence is researched. For many laypeople who want to display their results on social media, the lack of a video mode is also negative.
Ease of use and Reconyx are contradictions – these are clearly cameras for people who love reading and re-reading manuals.

The most used cameras in our work are Bushnell. The range of models is huge and every few years cameras are taken off the market and replaced by newer models.
Some notable improvement has been made in the last years particularly regarding trigger speeds. Our early Bushnell Trophy Cam had a trigger speed of 1.344 seconds while the new model NatureView is down to 0.2 seconds.
However, unfortunately the image quality of night photos has deteriorated with every new model.

Our now cameras of choice are Bushnell HD Nature View 119439/440 and 740 as we had the best results on all sites with this type of camera.
They are easy to set-up and use, have good trigger speeds and long battery lives. The detection rates are the highest of all cameras we have used and the supplied close focus lenses (46 and 60 cm) allow good, non-blurry photos even if an animal comes very close to the camera.
False positives at night are rare and even rarer in the day time.
However, for a camera in the mid-price range (approx. $500) the night image quality is so bad that at times e.g. a phascogale could be mistaken for a possum. Day photos are however of good quality. 

 

Images are also frequently under-exposed, making analysing pictures difficult. This is particularly bad in day photos when a ringtail is out during the day and easily missed. However, it is a trade-off for avoiding frequent over-expose in night photos.

In particular the newest model (740) produces night photos that are too grainy to see features such as a pouch even if the body part is directed at the camera.
If the object is close to the camera and no close focus lens is used, any movement will lead to further deterioration of the quality.

Nature View cameras seem to suffer from frequent time/day stamp issues. The stamp was dysfunctional in several of our cameras – one from the outset, several others developed the problem after a few months. Some cameras also stopped working with the change to a new year and would need to be re-set.  If cameras are bought over the internet from the US, claiming warranty is difficult. Most attempts of contacting the company directly to sort out any problems were ignored.
However, Bushnell was the only company we had dealings of this kind with and it would therefore be unfair to claim that their customer service is worse than the service by others – we only know about the very inadequate service by Bushnell.

Cameras we mainly use for predator monitoring are the small and cheap Acorn Ltl-5210A & Acorn Ltl-6210MC. Image quality particularly in the day time is great and quite satisfactory at night. These cameras are absolutely adequate for ‘backyard’ use.
The main downside is low sensitivity, particularly in the 5210 model and therefore low reliability. Compared to other cameras the number of false negatives and false positives was extremely high. Animals tend to vanish quicker than the camera can take a photo. The only 8 batteries also deplete very fast and the camera would then almost exclusively take day photos. 

In summary, there are cameras and there is marketing! All cameras have issues and I do not share the enthusiasm about Reconyx’ reliability – at least not for small arboreal creatures.

Expectations are generally too high when cameras are bought. They are not more neutral than observers – even cameras of the same brand and model are ‘individuals’ and not all react the same.

Observation ‘lines’ are harder to establish with cameras than with human observers and it is close to impossible to establish them at the same height, angle, distance, etc.  We would have to build an artificial environment for the cameras to function according to scientific rules. Often enough in official research this is attempted but I would argue that this would add more bias, not less.