Addressing temporal trends in survivorship from cross-sectional sampling designs: A modelling framework with applications for megafauna conservation

Studying survival in megafauna populations is a challenge. Survival can vary over time and can be altered by increasing pressures from human activities. Considering time variations in inter-annual survival or cumulative survival is necessary to evaluate conservation status and anticipate detrimental demographic changes before large declines in abundance materialize. Estimating survival is straightforward with mark-recapture methods but remains a challenge if individuals cannot be easily identified or recaptured. This is often the case for marine mammals because natural marks are insufficient to identify individual or because tagging raises insuperable logistical difficulties. Here, we propose to fill this methodological gap using age-at-death data with a flexible Bayesian linear modelling framework. We designed a simulation study with several scenarios of decline in survival and size of the data set. We also applied the approach to a real dataset on harbour porpoise (Phocoena phocoena). We evaluated the ability of a model that includes both a trend and yearly random effects to estimate time variations in survivorship, age-specific survival, and hazard rates. We compared the performance of the model to that of simpler models. Our results show that the model is very effective in detecting a trend in survivorship, but can yield biased estimates of age-specific survival rates. We provide a pragmatic modelling framework to address time variations in a megafauna species survivorship using age-at-death data. This framework will enhance our ability to study populations that cannot be individually identified and to design proactive conservation management measures.