CMR involves the capture, marking, and re-encounter ("capturing") of a sample of animals. As discussed more in a bit, we will generalize a bit the terms "marking" and "capturing" of animals, since broadly speaking CMR (and the related topic of mark-resighting) have many variations on the standard definition of a physical capture and tag or other artificial mark.
Regardless of the exact nature of the CMR process, it typically has one or more goals, all related to providing some type of inference, or probabilistic statement, about population parameters. Population parameters include abundance (N) and density (D=N/area) for closed populations, as well as immigration, emigration, and mortality (survival) for open populations.
For each of these types of parameters, we may also be interested in variation over space (for spatially stratified populations), or in relation to habitat or other attributes, and over time if the population is considered to be open. Therefore, an additional and very important type of inference beyond just estimating parameters is modeling variation.
In addition to these goals, we have to deal with the reality that detection (capture probability) is by definition incomplete-- we do not capture all animals at each sample. Not only that, capture probability (p) is typically thought to vary over sample occasions, or in response to previous capture, or even among individual animals for other reasons. Failure to correctly account for this incomplete and heterogeneous detection can and does lead to serious estimates in parameters we are interested in estimating, and so needs to be taken into account in our statistical models.
Next: Types of CMR data