Study Design Recommendation for a Park

Study Design > Park

Land is often managed in protected areas ranging from 1-1000km2.  Typical questions asked by managers include:

  • What animals live there?
  • Which habitats do they prefer?
  • How do management actions affect their populations? 
  • How do their numbers change over time? 


These objectives can be met with arrays of camera traps set in a pattern called stratified random.


Spatial Configuration-

We suggest that cameras be run in a stratified random pattern.  To do this first figure out what environmental factors you are most interested in comparing (e.g. habitats, distance from trails, level of human use, distance from park edge).  You would divide the feature into categories (e.g. low, medium, high density housing) and these categories would be called strata. You want to sample  evenly across strata, so you either assign an equal number of cameras locations to each strata or assign them based or their relative abundance (large strata get a lot of camera locations and rare strata get few camera locations). .   Once you decide the number of samples you have to decide where to place the cameras within the strata.  Either select locations randomly or place a grid over the strata and place a camera in a set number of grid cells.  The standard intervals of a grid will still be somewhat randomized in comparison to other environmental factors - the important thing is to not try to find the absolute 'best' place to run a camera.  Using a regular or randomized sample scheme can help you get a representative sample of the animals in your study area.  If your work is exploratory, or you have no prior stratifications, a simple grid across your study site will help ensure you have a representative sample of the area by placing one camera in each grid cell.

The next question is how far apart to place the cameras.  If cameras are close to each other their results are likely to be similar simply because of close proximity, making their results 'spatially autocorrelated'.  These non-independent samples will be more difficult to analyze (although there are some new techniques that can take spatial dependence into account during analysis).   Setting a minimum distance between cameras will help reduce this problem.  Exactly what this distance should be remains an active area of research.  Separating them by the diameter of a typical home range of the largest species in your area would be independent, but logistically difficult.  We have analyzed results within parks in the Eastern USA and found that white-tailed deer, red fox and striped skunks had no spatial autocorrelation in the pattern of their detections at distances >50m, the smallest interval.  Other smaller-bodied species (i.e. squirrels, chipmunks, cottontail, and raccoons) were spatially correlated at distances <500m, while some medium and most larger species (i.e. bobcat, turkey, bear, coyote) were correlated at maximum ranges from 500-1000m.  In all of these cases, the autocorrelation was relatively slight, so accounting for it later in the analyses is a viable option.  Our general recommendation is to separate cameras by as much as you can, with 500-1000m spacing being best, and anything >200m probably being acceptable.    Finally, some analysis techniques actually require spatial autocorrelation (e.g. spatially explicit capture recapture) and spacing for those studies should be closer to allow recaptures of individual animals on multiple cameras.



There are two basic approaches to monitoring a park: taking a snapshot in time, or constant monitoring.  Snapshots, like a bioblitz, should be focused in one season to minimize seasonal variation.  These short but intense surveys can then be repeated annually if long-term trends are of interest. Constant but less-intensive sampling over a longer period is also a viable option for parks with year-round education programs or dedicated volunteers.


How many cameras do I need?

Unfortunately, this question is still being debated by researchers. We hope new research will help us refine these suggestions soon.  Rowcliffe et al. (2008) and Kays et al (2010) both found that precision of estimating the detection rate for a species at one site increased quickly up to 20 camera locations, and then continued improving more slowly after that.  However, both of those analyses were for fairly common species, so that more camera locations would be needed to get an accurate measure  of rarer species at a given site.  We don't know exactly how many are needed, but we generally aim for 25-50 camera sites per sampling objective (i.e. per strata, per treatment, per site).  You may need more cameras for more data-intensive analyses like capture-recapture. 

We recommend running cameras for 3-4 weeks at one location and then moving the camera to a new location to increase overall sample size.  The more sites you have sampled the better, and 3-4 weeks usually gets a good sample of the animals at any site (longer time is needed for sites with less animal activity overall or if you are targeting a particularly rare species).  Thus, this discussion will focus on how many camera locations needed, which is typically one camera run at one location for 3-4 weeks.

For example, if you wanted to evaluate the difference in mammal populations across three habitat types, you'd want to set cameras in at least 25 locations per habitat. This could be 25 cameras set once, or 5 cameras each moved 5 times (with minimum 200m spacing between sites).

Note that this recommendation is not dependent on the size of a study area.  No one has evaluated how the size of a study area impacts study design.  There is no recommendation of some minimum camera density to survey a given area.  Instead, we consider what the research question is and sample sufficient locations in the areas to be compared.


Example questions and recommended analyses

Results from this type of survey will allow you to describe the medium and large mammals and terrestrial birds that use a given area.  Basic diversity metrics cashow the number of species, the species accumulation curve, the Shannon Diversity index, and the relative detection rate for the whole community (all available through eMammal analysis tools).

If cameras are set in a stratified random pattern without bait, the raw detection rate (photo sequences/day) can be used to quantify the intensity of use and compare the spatial distribution and habitat preferences of species.  A sequence is a set of photographs that are linked by a short time interval between images. For example, an animal passing in front of most cameras will leave a series of photos that constitute a single sequence. Sometimes where to break a sequence is not obvious, like when an animal leaves the view of the camera but then a reappears shortly after. The length of time between images needed to create a new sequence is often set by researcher.  

 The percentage of sites that detected a species can be used to show the spatial spread of a species over an area.  In both cases, simple comparisons across treatments (e.g. on/off trail) are possible, as are more sophisticated  models evaluating multiple variables.  Occupancy models that account for imperfect detection are now common with camera traps, using each day (or group of days) as separate sampling intervals.

If animals are individually recognizable, individual capture histories can be constructed to analyze  to estimate animal abundance and/or density.  There are a variety of experimental approaches to estimate abundance and density without recognizing individuals, but all are still problematic.

Temporal patterns can also be evaluated by combining the time-stamps from capture events for a single species across a study area (available through eMammal automated analysis). This allows you to identify times when a species is active, and you can compare patterns across areas or with other species to examine relationships (e.g. competitors, predator-prey). 

Scroll to the bottom for an example study! For other studies, see these papers on the effects of Hunting and Hiking on wildlife in parks and on the presence of invasive cats in parks. 


Volunteer Recruitment

Most parks and nature preserves are centers for outdoor recreation and thus volunteers can probably be recruited from local hikers or hunters.  Many parks have friends groups that include dedicated volunteers.  We have also had success partnering with local Master Naturalists groups.  Camera checks can also be integrated into educational activities of nature centers on the property, involving visitors in the camera trapping experience.



We recommend Reconyx brand camera traps as they are the most reliable, but generally any medium-priced camera is suitable.  Reconyx are the most expensive at about $500 each (including batteries, memory cards and locks).    Figure out how many sample locations you need and over what time you will be collecting the data to estimate how many cameras you need.  For example, a park looking to get 50 camera locations sampled over one year would need to buy about 5 cameras and rotate them across locations every month.  A park wanting to get a snapshot evaluation comparing two areas would need at least 25 samples in each (50 total) over one season (3 months) requiring 17 cameras (each used at 3 locations). Please also take note of the eMammal service charges to cover the cloud-computing expenses of uploading pictures.



We would love to hear your thoughts and questions on this advice! Please contact us at

Example Study - South Mountain, NC

Here we present one example study designed to evaluate the effect of hunting and hiking on wildlife communities.  We selected pairs of parks that were similar in habitat and geography, but had different hunting regulations.  In this case, South Mountain State Park had no hunting while South Mountain Game Lands allowed hunting.  We also stratified cameras on/near/far from trails to evaluate the effect of hikers on animals. We aimed for 25 cameras locations in each treatment (on/near/far) for a total of 75 camera locations per park.  We worked with volunteers to place all cameras for ~3 weeks before moving them to new locations. 

Figure 1. Map showing the locations of camera traps run in the adjacent South Mountains Gamelands (hunted) and State Park (not hunted). 

Figure 2. Number of species detected by camera traps in the two parks.


Figure 3. Species accumulation curve for South Mountains Gameland, showing a leveling off after sampling about 16 sites.


Figure 4. Species accumulation curve for South Mountains State Park levels off less than for the game lands, suggesting some additional species might be detected with more sampling effort. 



 Figure 5. Average detection rate for animals (other than deer) at South Mountains State Park (not hunted) and Gamelands (hunted).


Figure 6. Average detection rate for deer at South Mountains State Park (not hunted) and Gamelands (hunted).



Figure 7. The probability of site occupancy of a given species at South Mountains State Park (not hunted) and Gameland (hunted). Occupancy models account for imperfect detection.


Figures 8 and 9. Daily activity patterns for coyotes and deer at South Mountains State Park (top) and South Mountains Gameland (bottom).  Coyotes are more strictly nocturnal, and deer more diurnal, at South Mountains Gameland.