- Top of page
- Materials and methods
- Supporting Information
Camera trapping has become an important tool in ecological research, especially for monitoring rare and elusive species, because it can provide detailed visual information without requiring on-site human observers or physical capture. Developments and adaptations in analytical fields allow us to extract from photographic capture data information about activity patterns and behaviour (Maffei et al. 2005), species diversity/inventory (Tobler et al. 2002), occupancy (Linkie et al. 2012), abundance and density (Karanth & Nichols 2011), as well as vital rates such as survival and fecundity (Gardner et al. 2010). Like other evidence placing organisms in time and space, camera trap data originate from two processes: the ecological mechanisms determining abundance/occupancy and the process of observation determining our ability to detect an organism when it is present (Kéry 1998). Although ecological parameters are typically the focus, detectability permeates every aspect of camera trap studies and most other surveys of fauna and flora. Knowledge about detectability can help optimize study design (MacKenzie & Royle 2006), and detectability must be accounted for analytically to estimate the ecological parameter of interest (Kéry & Schaub 2011). In addition, most studies strive towards achieving high detectability because this increases cost efficiency, improves the precision of the ecological parameter estimates and, in the presence of latent heterogeneity in detection probability, also reduces bias in the estimates (Lukacs & Burnham 2007).
We can express detectability (i) as the probability of making a detection in a given time period (e.g. 1 day) or (ii) as the time until a detection is made, both conditional on presence of the focal individual or species. Although these two measures are manifestations of the same binomial process, they provide different perspectives on detectability. Camera trapping typically involves periods of inactivity, interrupted by photographic captures of wildlife. An expansive statistical field, time-to-event analysis, more commonly referred to as survival analysis, deals with such situations. As the name suggests, the measure of interest is the time until some event occurs. In the case of camera trap studies, it is the time until a photographic capture is made, and we illustrate the use of this concept to gain additional insights about detectability during multispecies camera trap studies.
Using both occupancy analysis and time-to-event analysis, we explored empirical patterns in photographic detectability over species, sites and time for three sympatric carnivore species (snow leopard Panthera uncia, red fox Vulpes vulpes and stone marten Martes foina). We asked the following questions to guide our analysis:
Q1. How do the three carnivore species differ in their photographic detectability? We predict that widespread species known to be tolerant of human activity and environmental disturbance, such as red fox and stone marten (Adkins & Stott 1998; Herr et al. 2010), are more easily detected than the snow leopard with its reputation for being elusive (Janečka et al. 2011).
Q2. What are the determinants of detectability? Site selection is a key factor during camera trapping studies targeted at carnivores, and we expect species-specific responses of detectability to site covariates, such as terrain, habitat type and the presence/absence of carnivore sign in the area. We predict that the presence of carnivore sign and application of scent lures at camera traps improve detectability (higher detection probability, shorter time to detection).
Q3. How does photographic detection proceed over time? We predict that disturbances associated with placing camera traps in the environment, as well as changes in the potency of olfactory attractants due to scent dissipation, may lead to non-constant detectability over time.
Q4. What is the quantitative effect of longer survey duration on the probability of detecting a focal species at least once at a site if it is present? Hamel et al. (2012) recommended a 30-day survey duration when multiple species are targeted in one study to allow reliable parameter estimates (conditional on the number of sites surveyed). This would entail a two- to three-fold increase compared with the camera trapping duration in our study (10–14 days), which we anticipate would substantially boost the proportion of cameras with at least one detection and consequently increase the sample size.
We discuss implications of our findings for camera trap study design and data analysis and illustrate for the first time the utility of time-to-event analysis for complementing information about detectability gained from hierarchical models based on camera trap data.