A spatial hierarchical model for abundance of three ice-associated seal species in the eastern Bering Sea

Abstract Estimating the abundance of seals inhabiting sea ice is complicated because the areas are large, the ice area and distribution may change rapidly, and it is impractical to detect and count a (typically unknown) portion of the population that is in the water, rather than hauled out on the ice. We propose a method for resolving these issues by using a series of daily estimates that are imprecise by themselves, but yield an acceptable estimate when they are combined. Population surveys of spotted seals, ribbon seals and bearded seals were conducted over 279,880 km 2 of the Bering Sea between 13 April and 26 May 2007. During that period, the sea-ice conditions and spatial distribution of seals changed dramatically. We surveyed 2748 km 2 using line transect methods from a helicopter deployed from the US Coast Guard icebreaker Healy . Corrections for incomplete availability of seals used a generalized linear mixed model for seal haul-out probability using sensors on seals with data transmitted by satellite. We accounted for incomplete detection through standard distance-sampling methods along with a double-observer model. The availability and detection models were combined in a data model for local abundance in space and time. To accommodate the shifting ice and seal movements, we then developed a hierarchical spatially-autocorrelated regression model using remotely sensed sea ice concentration data to predict abundance at each survey date. While abundance estimation was very imprecise for each date, we were able to combine them to obtain good estimates of overall population abundance even though the population was spatially dynamic. The proposed hierarchical model combined submodels and accounted for their sources of uncertainty. Spotted seals were most abundant within the study area (233,700, 95% CI 137,300–793,100), followed by bearded seals (61,800, 95% CI 34,900–171,600) and ribbon seals (61,100, 95% CI 35,200–189,300).

[1]  A. Goldberger,et al.  On the Exact Covariance of Products of Random Variables , 1969 .

[2]  D L Borchers,et al.  Accommodating Unmodeled Heterogeneity in Double‐Observer Distance Sampling Surveys , 2006, Biometrics.

[3]  R. Littell.  SAS System for Mixed Models , 1996 .

[4]  D. Borchers,et al.  Mark-Recapture Models for Line Transect Surveys , 1998 .

[5]  Galin L. Jones,et al.  Fixed-Width Output Analysis for Markov Chain Monte Carlo , 2006, math/0601446.

[6]  L. Mark Berliner,et al.  Hierarchical Bayesian Time Series Models , 1996 .

[7]  J. Besag,et al.  Bayesian image restoration, with two applications in spatial statistics , 1991 .

[8]  Jay M. Ver Hoef,et al.  Fast computing of some generalized linear mixed pseudo-models with temporal autocorrelation , 2010, Comput. Stat..

[9]  Stuart Geman,et al.  Bayesian image restoration , 1986 .

[10]  K. Pollock.  A Family of Density Estimators for Line-Transect Sampling , 1978 .

[11]  Catherine A Calder,et al.  Accounting for uncertainty in ecological analysis: the strengths and limitations of hierarchical statistical modeling. , 2009, Ecological applications : a publication of the Ecological Society of America.

[12]  David Barber,et al.  Bayesian Time Series Models , 2011 .

[13]  L. A. Goodman.  On the Exact Variance of Products , 1960 .

[14]  R Core Team,et al.  R: A language and environment for statistical computing. , 2014 .

[15]  J. Overland,et al.  Status review of the ribbon seal (Histriophoca fasciata) , 2008 .

[16]  E. F. Becker,et al.  Combining Line Transect and Double Count Sampling Techniques for Aerial Surveys , 1997 .

[17]  J. Hoef.  Who Invented the Delta Method , 2012 .

[18]  Sw. Banerjee,et al.  Hierarchical Modeling and Analysis for Spatial Data , 2003 .

[19]  David R. Anderson,et al.  Advanced Distance Sampling: Estimating Abundance of Biological Populations , 2004 .

[20]  R. Suydam,et al.  Habitat use and habitat selection by spotted seals (Phoca largha) in the Bering Sea , 2000 .