top of page
banner 7.jpg

Chapter 3: Introduction to spatial capture-recapture (SCR)

3.1 Lecture: Capture histories

A capture history contains the key data for fitting SCR models including all the information on which detector made which detection and on matched detections between detectors. Using information of matched animal detections, we build such a capture history for two examples: 

​

  1. Two detectors and ten detected animals

  2. 16 detectors and 71 detected animals

3.2 Lecture: Introduction to model fitting with R package ascr

We fit an SCR model with the function fit.ascr() and interpret the model output. Leading up to this, we introduce the type of data required for model fitting and other inputs required.

 

  • We ensure the locations of the detectors are entered in the correct format.

​

  • Using the matched detection data we build the capture history for all detected animals using ascr functions (which we did by hand in the previous lecture).

​

  • Lastly, we need to provide the information from where the animals may be detected from. We do this by creating a mask object which is essentially a grid of points across the study area from which the animals may have been detected. 

Some technical jargon explained

Abundance: The number of animals (individual abundance) or groups (group abundance) in a given region.

​

Capture-recapture: A method of estimating abundance that is based on the proportion of animals detected (or captured) on one occasion and/or detector that are redetected (or recaptured) on another occasion and/or detector.

​

Confidence interval: A random interval that includes the thing you are estimating a given percentage of the time. For example, a 95% confidence interval for abundance will include the true abundance 95% of the time. Any particular 95% confidence interval for abundance from a survey either does or does not include true abundance (there is no way of knowing), but because the interval itself is random, there is a 95% chance that it does.

​

Density of animals: Number of animals per unit area. SCR software often gives density in terms of number of animals per hectare (i.e., per 100m x 100m area). Multiply this by 100 to get density per square km. Abundance is obtained by multiplying density by area.

​

Density of calls: Number of calls per unit area, per unit TIME. Call abundance (total number of calls) in some time interval (e.g. an hour) is obtained by multiplying call density by area and by the length of the time interval.

​

​Detection unit: The thing that we record as having been detected. If calls are recorded without knowing or recording which animal (or group) the call came from, then call is the detection unit. If detection of the animal (or group) is recorded, without recording each detected calls, then animals (or groups) are the detection units. If calls are recorded as well as which animal made the call (a rare and mostly unlikely scenario) then you could treat either calls or animals as the detection unit in anaysis.

​

Detection: Hearing (acoustic survey) or seeing (visual survey) or catching (physical capture survey), or otherwise becoming aware of the presence of the thing you are looking/listening for.

​

Detection history: Detection units have detection histories, which are records of which detectors detected the units and which did not. In general each detection also has a detection time or occasion associated with it, and this forms part of the detection history. Borrowing terminology from surveys in which animals are physically captured, these are also sometimes called ``capture histories''.

​

Detection probability: The probability of observing a detection unit. For example, the probability of hearing a call if call is the detection unit, or the probability of detecting and animal (or group) if animal (or group) is the detection unit. There are various kinds of detection probabilities in SCR surveys, even for a single kind of detection unit. For example, there is the probability that a single detector detects a detection unit, and the probability that any detector in an array detects a detection unit, (a larger probability, obviously), and the probability that a detector makes a detection GIVEN the location of the animal, and the probability that certain detectors detect a detection unit while others don't (the probability of a particular ``detection history'').

​

Marking: The process of making detection units (e.g., animals) recognisable if and when they are detected again. With physical capture, this could involve physically marking captured animals (e.g., by attaching tags to them) but with acoustic and visual survey methods ``marking'' does not involve physical marking but rather just being able to recognise a detection unit that has been detected on a previous occasion and/or by a different detector.

​

Mask:  A slightly odd term for the region containing all animals or animal activity centres that could be detected (odd because ``masking'' something usually implies making it undetectable). Bayesian SCR literature tends to use the term ``state space'' instead of  ``mask''.

​

Occasion: A survey or part of a survey that occurs (nominally at least) at one point in time. Historically, capture-recapture surveys required at least two sampling occasions: one to mark animals and another to (possibly) recapture marked (and unmarked) animals. Some SCR surveys do not require more than one occasion, because recaptures can occur between detectors rather than between occasions. This is not the case if detectors physically hold animals, as would mist nets or pitfall traps, for example, where more than one occasion is required in order to get any recaptures. With detectors like camera traps, which do not hold animals, it is possible to have only one occasion.

​

Session: In the SCR literature this means ``surveys, or parts of surveys, that are isolated from other surveys, or parts of surveys, in that no recaptures (or redetections) are possible between occasions''. Different sessions could be repeat surveys of the same area in space in such a way that recaptures of the same detection units between sessions is either impossible or if possible, not recorded. Different sessions could also be surveys of different parts of space in a larger survey region, such that distances between the parts of space surveyed by each session are large enough that no animal detected in one session could be detected in another session.

​

Study area: The surface area of the region within which you want to draw inference using your survey data (i.e., the surface area of the survey region). This should be defined before conducting the survey.

​

Study region: The region within which you want to draw inference using your survey data. This should be defined before conducting the survey.

​

​

​

​

​

banner 7.jpg

Continue with Chapter 4

bottom of page