Drivers of population dynamics and juvenile mortality in northwest Atlantic harp seals
Data files
Nov 21, 2025 version files 330.71 MB
-
HS_code.zip
330.68 MB
-
HS_data.rdata
12.28 KB
-
README.md
16.41 KB
Abstract
Human-induced threats to terrestrial and marine wildlife are on the rise, and while some species may face a single major threat that is easily identifiable, others face multiple concurrent threats. Harp seals, an abundant pinniped in the north Atlantic that was historically depleted by human harvest, are one such species: while commercial and subsistence harvests remain a significant source of mortality, in recent decades their environment has undergone significant changes which could also impact population dynamics. Inferring the relative importance of various threats as drivers of population dynamics can be challenging, particularly for marine species where monitoring abundance is difficult: the use of integrated population models (IPM), which leverage multiple data sources to parameterize process-based models of population dynamics, provide one solution to this challenge. We developed a hierarchical Bayesian IPM with which to explore the shifting roles of anthropogenic and environmental factors in driving population trends. We used a competing hazards formulation for survival, enabling the partitioning of mortality into multiple discreet causes (hazards), and allowing us to assess how these different hazards varied over 7 decades (1952 – 2019). We fitted the model to available data on pup production, fecundity, age structure, human removals and environmental conditions, and used the fitted model to produce annual estimates of pup production and total abundance. We conducted a Bayesian life stage simulation analysis (LSA) to compare the relative contributions of various hazards to variation in population growth. We found that harvests of young of the year and adults were the primary contributors to variation in population trends from 1951-1982; however, after 1983 the relative importance of harvest mortality decreased while the impacts of natural mortality increased, especially for young of year (YOY), and since 2000 the impacts of YOY mortality from ice cover anomalies have become one of the strongest drivers of trends. Based on current climate models, which project warmer water and decreasing ice cover, we expect continued high levels of YOY mortality from environmental factors such as deteriorating ice conditions. These climate-related hazards are likely to become the dominant drivers of population dynamics in coming decades, which will in turn affect sustainable harvest levels for both Canada and Greenland. Our model will provide a useful tool for exploring future scenarios of climate impacts and management strategies.
Dataset DOI: 10.5061/dryad.2z34tmq0k
Description of the data and file structure
Data were used to develop and fit a hierarchical Bayesian IPM with which to explore the shifting roles of anthropogenic and environmental factors in driving population trends. We used a competing hazards formulation for survival, enabling the partitioning of mortality into multiple discreet causes (hazards), and allowing us to assess how these different hazards varied over 7 decades (1952 – 2019). We fitted the model to available data on pup production, fecundity, age structure, human removals, and environmental conditions, and used the fitted model to produce annual estimates of pup production and total abundance. We conducted a Bayesian life stage simulation analysis (LSA) to compare the relative contributions of various hazards to variation in population growth.
The model was developed in the R programming language, with Bayesian model fitting achieved using STAN software package. All the R and STAN code scripts necessary to set up and fit the model and analyze results are contained in the zip archive "HS_code.zip". All the raw data tables required for fitting are included as R data frames within the rdata file "HS_data.rdata". Details are provided below.
Files and variables
File: HS_code.zip
Description: Zip archive containing the all code (R and STAN scripts) required to set up and fit the model and analyze results, including generating figures
| File | Description |
|---|---|
| HS_model_shell.R | R code used to execute all analyses described in paper, and generate all results figures (see code annotation for details) |
| HSmodfit.stan | STAN code used to fit Integrated Population Model (see code annotation for details) |
| init_fun.R | Initialization values for estimated parameters for STAN (optional, to speed up processing) |
| HS_Results_fit.RDS | saved model fitting results, generated from included scripts (optional, to speed up processing) |
| cmdstan_sumstats.r | helper R script to generate model summaries and posterior samples after fitting model (see code annotation for details) |
| LSA_perturation_analysis.R | R script to run Bayesian LSA analysis (see code annotation for details) |
| LSA_rslts.RDS | saved results from Bayesian LSA analysis, generated from included scripts (optional, to speed up processing) |
File: HS_data.rdata
Description: Rdata file containing the 7 raw data tables (R data frames) required to fit the model. The metadata are summarized in the following Tables: the names of each data frame (df.*) and their descriptions are listed, and the rows below which contain details of each variable (column) contained in the data frame, including descriptions of how data were collected and the units of measure.
df.Age: Data table summarizing numbers sample by age class for each year of the study period
| Column | Description | Units |
|---|---|---|
| Age | Annual age class, values from 1 to 36 listed in rows | age of animal (years) |
| 1979-2019 | Column for each year, with values in rows indicating numbers of animals sampled for each age class | number animals |
df.HV: Data table summarizing human take data between 1951 and 2019, summarized by year. Data are categorized by age-class (pup vs adult) and source of human take.
| Column | Description | Units |
|---|---|---|
| YEAR | Year of sampling (values 1951-2019 in rows) | calendar year |
| Arctic_pup | Estimate of number pups harvested in Canadian Arctic subsistence hunt, by year | number animals |
| Byctch_pup | Estimate of number pups caught in fisheries by-catch, by year | number animals |
| Canad_pup | Estimate of number pups harvested in Canadian commercial hunt, by year | number animals |
| Greenlnd_pup | Estimate of number pups harvested in Greenland hunt, by year | number animals |
| Arctic_1plus | Estimate of number adults (age in years >0) harvested in Canadian Arctic subsistence hunt, by year | number animals |
| Byctch_1plus | Estimate of number adults (age in years >0) caught in fisheries by-catch, by year | number animals |
| Canad_1plus | Estimate of number adults (age in years >0) harvested in Canadian commercial hunt, by year | number animals |
| Greenlnd_1plus | Estimate of number adults (age in years >0) harvested in Greenland hunt, by year | number animals |
| Arctic_ppn_pup | Estimate of proportion of take pups in Canadian Arctic subsistence hunt, by year | proportion |
| Bycatc_ppn_pup | Estimate of proportion of take pups in fisheries by-catch, by year | proportion |
| Can_ppn_pup | Estimate of proportion of take pups in Canadian commercial hunt, by year | proportion |
| Grenlnd_ppn_pup | Estimate of proportion of take pups in Greenland hunt, by year | proportion |
| Arctic | Estimate of number animals (pups and adults) harvested in Canadian Arctic subsistence hunt, by year | number animals |
| Byctch | Estimate of number animals (pups and adults) caught in fisheries by-catch, by year | number animals |
| Canad | Estimate of number animals (pups and adults) harvested in Canadian commercial hunt, by year | number animals |
| Greenlnd | Estimate of number animals (pups and adults) harvested in Greenland hunt, by year | number animals |
| TOTAL | Estimate of total number of human takes, by year, by year | number animals |
| Arctic_ppn | Estimate of proportion of total take made up by Canadian Arctic subsistence hunt, by year | proportion |
| Byctch_ppn | Estimate of proportion of total take made up by fisheries by-catch, by year | proportion |
| Canad_ppn | Estimate of proportion of total take made up by Canadian commercial hunt, by year | proportion |
| Greenlnd_ppn | Estimate of proportion of total take made up by Greenland hunt, by year | proportion |
| PUPTOT | Estimate of total pup take, all sources, by year | number animals |
| ADLTOT | Estimate of total adult take (age in years >0), all sources, by year | number animals |
df.HV_uncert: Data table summarizing relative levels of precision ("certainty") associated with human take estimated (Table df.HV), summarized by year. Certainty levels are categorized as "low", "moderate" or "high". In the case of total catches, certainty levels can be interpreted in terms of the coefficient of variation (CV) associated with each reported estimate: low = CV of 0.2, moderate = CV of 0.1, high = CV of 0.05. In the case of proportions of take comprised of pups (young of year, or YOY columns), certainty levels can be interpreted in terms of the precision of a beta distribution: specifically, for a proportion of 0.5, low = sd of 0.1, moderate = sd of 0.05, high = sd of 0.01
| Column | Description | Units |
|---|---|---|
| Year | Year of sampling (values 1951-2019 in rows) | calendar year |
| Certainty_Arctic_HarvTot | Relative certainty associated with estimates of total take for Canadian Arctic subsistence hunt, by year | category: low, moderate or high |
| Certainty_Bycacth_HarvTot | Relative certainty associated with estimates of total take for fisheries by-catch, by year | category: low, moderate or high |
| Certainty_Canadian_HarvTot | Relative certainty associated with estimates of total take for Canadian commercial hunt, by year | category: low, moderate or high |
| Certainty_Greenland_HarvTot | Relative certainty associated with estimates of total take for Greenland hunt, by year | category: low, moderate or high |
| Certainty_Arctic_Harv_PYOY | Relative certainty associated with estimates of proportion young of year (pups) for Canadian Arctic subsistence hunt, by year | category: low, moderate or high |
| Certainty_Bycatch_Harv_PYOY | Relative certainty associated with estimates of proportion young of year (pups) for fisheries by-catch, by year | category: low, moderate or high |
| Certainty_Canadian_Harv_PYOY | Relative certainty associated with estimates of proportion young of year (pups) for Canadian commercial hunt, by year | category: low, moderate or high |
| Certainty_Greenland_Harv_PYOY | Relative certainty associated with estimates of proportion young of year (pups) for Greenland hunt, by year | category: low, moderate or high |
df.Ice: Table of ice anomaly index values (IC) for each year from 1959-2019 (rows) and for two breeding areas, Gulf and Front. Raw ice cover data, representing proporiton of area covered by first-year ice, were first extracted from the Canadian Ice Service of Environment Canada online repository (https://iceweb1.cis.ec.gc.ca/IceGraph/page1.xhtml?lang=en) for each of two periods: whelping (p = 1; defined for Gulf as week of 28 February, and for Front as week of 26 March) and post-weaning fast (p = 2; defined for Gulf as week of 5 March, and for Front as week of 9 April). IC values were then calculated as the deviation between proportional ice cover (averaged for the two periods) and the average proportional ice cover over reference years 1969-2000, re-scaled to unit variance.
| Column | Description | Units |
|---|---|---|
| Year | Year of study (values 1951-2019 in rows) | calendar year |
| Gulf_dwnld_nnw | The ice anomaly index value (IC) for the Gulf breeding area, by year | index value (real number) |
| Front_dwnld_nnw | The ice anomaly index value (IC) for the Front breeding area, by year | index value (real number) |
df.NLCI: Table of data on Newfoundland Climate Index (NLCI) for each year from 1959-2019 (rows). NLCI is a mosaic of 10 environmental components that provides a measure of the overall state of environmental conditions and ecosystem variability in the Northwest Atlantic between 1951-2019 (Cyr, F. and P.S. Galbraith. 2021. A climate index for the Newfoundland and Labrador shelf. Earth System Science Data 13: 1807-1828).
| Column | Description | Units |
|---|---|---|
| Year | Year of study (values 1951-2019 in rows) | calendar year |
| CI | NLCI values, by year | index value (real number) |
df.REP: Data table of reproductive data: numbers of females sampled each year, by age class, and the number of those females that was pregnant with a viable pup
| Column | Description | Units |
|---|---|---|
| Year | Year of sampling (values 1951-2019 in rows) | calendar year |
| Age | Age class of sampled females | age of female (years) |
| N | Number of females sampled by year and age class | number animals |
| Preg | Number of females sampled by year and age class that were pregnant | number animals |
| Prob | Proportion of sampled females that were pregnant, by year and age class | proportion |
df.PUP: Data table with survey estimates of pup counts (YOY) on ice whelping patches, by survey year (rows)
| Column | Description | Units |
|---|---|---|
| Year | Year of sampling (specific year values between 1951 and 2019, in rows) | calendar year |
| Sgulf_N | Survey estimates of pups in southern Gulf breeding area, by year | number animals |
| Sgulf_se | Standard error in estimates of pups in southern Gulf breeding area, by year | standard error in estimated number animals |
| Ngulf_N | Survey estimates of pups in northern Gulf breeding area, by year | number animals |
| N_gulf_se | Standard error in estimates of pups in northern Gulf breeding area, by year | standard error in estimated number animals |
| Front_N | Survey estimates of pups in Newfoundland Front breeding area, by year | number animals |
| Front_se | Standard error in estimates of pups in Newfoundland Front breeding area, by year | standard error in estimated number animals |
| Total_Npup | Survey estimates of pups, all breeding areas combined, by year | number animals |
| Total_se | Standard error in estimates of pups, all breeding areas combined, by year | standard error in estimated number animals |
Code/software
R programming environment (version 4.5.1)
R packages required:
'parallel','cmdstanr','posterior','gtools','stats','dplyr','reshape2','ggplot2','bayesplot','readxl','knitr','kableExtra', 'fitdistrplus','loo','gridExtra','cowplot','rstan', 'MASS','rsq','foreach','doParallel','doRNG','BSDA'
RStudio: integrated development environment (IDE) for R
STAN (cmdsan version 2.37.0): software for Bayesian data analysis
NOTE: Workflow is controlled by R script " HS_model_shell.R "
Access information
Other publicly accessible locations of the data:
- NA
Data was derived from the following sources:
- NA
