Before proceeding with a detailed discussion of age and cohort structure in CMR analysis, we need to be clear on what we mean by "age" and "cohort", because these terms tend to be used in somewhat interchangeable and confusing way in some of the CMR literature.
By age we mean the actual chronological age (e.g., in years) of an organism at various times during the study. Age is sometimes known accurately, either because individuals are followed from birth through time, or because age can be reliably assessed from field data (e.g., from dentition). In many cases age can only be roughly assessed as to being in broad intervals, e.g., birds can often be determined based on plumage to be hatching year (HY) or later (AHY), or sometimes HY, second year (SY), and after second year (ASY), unless they were originally captured as HY and are now being recaptured x years later (in which case we know they are exactly x years old upon recapture).
Cohort in the context of CMR means any group of animals released with marks at the same time. An age cohort of animals would mean a release cohort of the same age (e.g., birth year), so for instance 100 fawns tagged in 2013, or 50 1-year-old deer tagged in 2012. The key idea is that the potential recaptures of cohorts involves 2 issues, and these may or may not be separable, depending on the data structure.
Time elapsed since release of the cohort, which includes an "aging" process
Chronological time of the release period, and subsequent survival intervals
To illustrate, suppose we have 2 release cohorts of fawns, 1 in 2013 and one in 2014, and we conduct a 5-year recapture study. The ages (0 -5) are entered in the corresponding recapture years (rows are cohorts, columns recapture periods)
Notice that the animals of the same age (take 2 year olds) survive to the next age (3) over 2 different calendar intervals: 2015-2016 for the 2013 cohort, and 2016-2017 for the 2014 cohort. We could try to interpret the rate at which each cohort survives to each subsequent year as being age-specific, but that ignores the fact that survival (for all animals, or for specific ages) may also vary over time, that is, be (as it was for the original CJS model) time-specific. In fact, it is this very situation that gives rise to the "all different PIM" CJS model, in which every combination of cohort and occasion has a unique survival index. It is possible, in fact, to estimate parameters under somewhat simplified (constrained) versions of the "all different" model, as we will see the problem is typically complicated by the fact that the release cohorts are not (or may not be) of a single, identifiable age.
Next: Cohort based analysis