Insights into processes of population decline using an integrated population model: The case of the St. Lawrence Estuary beluga (Delphinapterus leucas)

Abstract

a b s t r a c tIntegrated population models combine data from several sources into a single model to allow the simultaneous estimation of demographic parameters and the prediction of population trajectories. They are especially useful when survey data alone are insufficient to estimate precise vital rates and abundance, and to understand mechanisms of population growth and decline. The St. Lawrence Estuary (SLE) beluga population was depleted by intensive hunting over the past century, and had declined to 1000 individuals or less when it was afforded protection in 1979. Despite protective measures, the SLE population has shown no signs of recovery. Low abundance estimates and high calf mortalities observed in recent years have raised concerns as to its current status. An age-structured Bayesian model was used to describe population dynamics by integrating information from two different monitoring programs. The model included information on population size and proportion of young (<2 years-old) obtained from seven photographic aerial surveys flown between 1990 and 2009, and mortalities documented annually by a carcass monitoring program maintained from 1983 to 2012. Results suggest that the population was stable or slightly increasing from the end of the 1960s until the early 2000s when it numbered approximately 1000 belugas. The population then declined to 889 individuals (95%CI 672−1167) in 2012. Although neither dataset on its own could explain this decline, the integrated model was able to shed light on the internal processes involved. Results suggest substantial changes in population dynamics and age structure, moving from a stable period (1984−1998) characterized by a 3-year calving cycle and a population composed of 7.5% newborns and 42% immature individuals, to an unstable state (1999−2012) characterized by a 2-year calving cycle, high newborn mortality and a declining proportion of newborns and immatures (respectively, 6 and 33% in 2012). Independent indices of abundance, population age structure and calf production match model predictions, thus increasing our confidence in its conclusions. The lack of recovery, high adult mortality (6%) and highly variable newborn survival further increase concerns about this population.Crown

Testosterone and progesterone concentrations in blow samples are biologically relevant in belugas (Delphinapterus leucas)

“…Additionally, detecting maturity in belugas is difficult to perform through visual inspection alone; an improved understanding of the proportion of belugas that are mature through the use of blow sampling would aid in developing more accurate population models (Mosnier et al, 2015).…”

Site‐selection bias and apparent population declines in long‐term studies

“…Furthermore, populations often fluctuate randomly, exhibit temporally autocorrelated changes, or both, and such phenomena reflect complex underlying dynamics (e.g., Mosnier et al. ; McCain et al. ; Öst et al.…”

“…Studying population dynamics over time is labor intensive and expensive, and because total population censuses are rarely possible, biologists must deal with substantial measurement uncertainty (e.g., Reynolds et al 2011;d'Eon-Eggertson et al 2015;Rueda-Cediel et al 2018), especially for species with low or variable detection probabilities (Nichols et al 2000;Bailey et al 2004). Furthermore, populations often fluctuate randomly, exhibit temporally autocorrelated changes, or both, and such phenomena reflect complex underlying dynamics (e.g., Mosnier et al 2015;McCain et al 2016;Öst et al 2016). Incorrect conclusions about monitored populations can arise from short time series (Krebs 1991;White 2019); inconsistent methods (Hayward et al 2015); and nonrandom sampling (Yoccoz et al 2001), which is our focus here.…”

Modeling Quantitative Value of Habitats for Marine and Estuarine Populations

“…An additional classifier we did not consider was whether population models are formulated as deterministic or stochastic. Stochastic formulation implies that values used in simulations of some or all parameters are generated from probability distributions so that the values of those parameters differ through time within a simulation or differ among simulations (Higgins et al, 1997;Mosnier et al, 2015). All of our classes of models can be either deterministic or stochastic; the only difference within a category is whether one needs to know how habitat would affect not only the mean or best estimate of a parameter (deterministic case) but also potentially the variability in those parameter values for thorough adjustment of parameters in stochastic models.…”