Short communication Estimating age-specific survival rates from historical ringing data ROBERT A. ROBINSON* British Trust for Ornithology, Thetford, Norfolk IP24 2PU, UK Keywords: mark–recovery, Sandwich Tern, Sterna sandvicensis . In most bird species, survival varies with age, young and immature birds generally having a lower probability of survival than older birds. Quantifying such variation is necessary if the population dynamics of the species is to be properly understood. A key aim of national ringing schemes is to monitor birds through time changes in sur- vival rates, which can be estimated from the number of marked individuals found dead and reported to the scheme (e.g. Baillie 2001, Williams et al. 2002). Although data currently collected may be computerized, data received historically on paper have often not been. Thus while the total number of birds ringed each year is often known, the number ringed in each age-cohort, or over a particular time period within a year, may not be. This complicates the analysis of survival in species, such as seabirds, where survival and reporting probabilities may differ markedly with age. Here I demonstrate how recovery data in combination with annual ringing totals may be used to generate age-specific survival rates by estimating the proportion of birds marked in each age- class from more recent computerized data. As a case study, I estimate survival rates in Sandwich Terns Sterna sandvicensis ringed in Britain and Ireland. METHODS In the years prior to 2000, records of birds ringed in Brit- ain and Ireland were submitted on paper forms and these have not been computerized. The number of juveniles (those in their first calendar year of life) and adults (those in their second and subsequent calendar years) marked cannot easily be tabulated from these paper records, as data on the more than 700 000 birds ringed each year are listed by ring number rather than by species. As each paper schedule details only up to 50 birds, relevant records for a species may be spread over hundreds of sheets each year. Thus, the number of full-grown (total- ling 924 between 1990 and 2000) Sandwich Terns ringed each year is known, but not the proportion of birds marked in each age-class; 16 628 individuals were ringed as pulli in the same period. Although a relatively small number of birds were ringed as adults, these need to be included to enable age-specific survival and recovery probabilities to be separated (Anderson et al. 1985). Ringing details of all birds subsequently recovered were, however, routinely computerized retrospectively when the report was received, so age at ringing for these indi- viduals is known. Since 2000, ringers have been supplied with a com- puter package that allows them to enter and store their ringing data, and in the years 2000–2007, ringing details of all Sandwich Terns were computerized by the ringers. In this period, 19 505 pulli and 488 full-grown Sand- wich Terns were ringed. From these records, it is possible to calculate the relative proportion of full-grown birds ringed as juveniles and adults in each year (with the exception of 2003, when the only birds caught were in a single cannon-net catch, providing a biased age-sample). On average, 38% of birds caught as full-grown individu- als during this period were juvenile birds of the year (range 25–47%). I thus estimated survival rates for birds ringed from 1990 to 2006, assuming a constant proportion (38%) of full-grown birds were juvenile birds of the year, with the actual numbers ringed in each age-class used for the period 2000–2006, to which were added the individuals ringed as pulli in each year. Note that this assumes an unchanging age-structure amongst marked birds rather than the population at large. In addition, I assume mor- tality during the marking period is minimal, which is unlikely to be strictly true, resulting in a small down- ward bias in survival rates. On the other hand, consider- ing full-grown juveniles and chicks in the same age-class will result in a small upward bias in survival rates, although there are insufficient recoveries of Sandwich Terns ringed as chicks to permit separation of these birds in this case. All years ran from June of one year to May the following year, with most birds ringed in the first 3 months of each period (i.e. June–August), thus satisfy- ing the assumption for mark–recovery periods that the length of ringing period (3 months) is short relative to the length of the recovery period (12 months). I included all birds found dead by 31 May 2007. To test the effect of the assumption regarding the proportion of each age-class marked, I also ran the analyses assuming 28% and 48% of full-grown birds were ringed as juve- niles in the years 1990–99. To model age structure, I initially considered a model with four age-classes and estimated survival and report- ing probabilities for birds in their first, second and third calendar years and those four or more years old. This is *Email: rob.robinson@bto.org ª 2010 The Author Journal compilation ª 2010 British Ornithologists’ Union Ibis (2010), 152, 651–653