Explaining the divergence of population trajectories for two interacting waterfowl species

Gibson, Daniel 1 ; Arnold, Todd1; Buderman, Frances2 ; Koons, David 3

Published Nov 19, 2024 on Dryad. https://doi.org/10.5061/dryad.hqbzkh1n9

Data files

Nov 19, 2024 version files 115.43 MB

crop.rdata
3.34 KB
estimated_vital_rates.rdata
115.22 MB
model_files.rdata
47.86 KB
pond.rdata
2.23 KB
README.md
10.95 KB
winter.rdata
143.32 KB

Abstract

Identifying the specific environmental features and associated density-dependent processes that limit population growth is central to both ecology and conservation. Comparative assessments of sympatric species allow for inference into how ecologically similar species differentially respond to their shared environment, which can be used to inform community-level conservation strategies. Comparative assessments can nevertheless be complicated by interactions and feedback loops among the species in question. We developed an integrated population model based on sixty-one years of ecological data describing the demographic histories of Canvasbacks (Aythya valisineria) and Redheads (Aythya americana), two species of migratory diving ducks that utilize similar breeding habitats and affect each other’s demography through interspecific nest parasitism. We combined this model with a transient life table response experiment to determine the extent that demographic rates, and their contributions to population growth, were similar between these two species. We found that demographic rates and, to a lesser extent, their contributions to population growth covaried between Canvasbacks and Redheads, but the trajectories of population abundances widely diverged between the two species during the end of the 20^th century due to inherent differences between the species life-histories and sensitivities to both environmental variation and harvest pressure. We found that annual survival of both species increased during years of restrictive harvest regulations; however, recent harvest pressure on female Canvasbacks may be contributing to population declines. Despite periodic, and often dramatic, increases in breeding abundance during wet years, the number of breeding Canvasbacks declined by 13% whereas the number of breeding Redheads has increased by 37% since 1961. Reductions in harvest pressure and improvements in submerged aquatic vegetation throughout the wintering grounds have mediated the extent to which populations of both species contracted during dry years in the Prairie Pothole Region. However, continued degradation of breeding habitats through climate-related shifts in wetland hydrology and agricultural conversion of surrounding grassland habitats may have exceeded the capacity for demographic compensation during the non-breeding season.

Journal Name: Ecological Monographs (submitted)

Title: Explaining the divergence of population trajectories for two interacting waterfowl species.

Author(s):

Gibson, D.(1,2a), T.W. Arnold (2), F.E. Buderman (3) D.N. Koons (1),

1 Department of Fisheries, Wildlife, and Conservation Biology, University of Minnesota, St. Paul, MN 55455

2 Department of Ecosystem Science and Management, Pennsylvania State University, University Park, PA 16802

3 Department of Fish, Wildlife, and Conservation Biology & Graduate Degree Program in Ecology, Colorado State University, Fort Collins, Colorado 80523 USA

a Corresponding Author: gibso678@umn.edu

Decomposing the drivers of Canvasback and Redhead population change: Code and data to develop explantory variables, build a population model, and perform a transient life table response experiment

We have provided the raw agricultural (crop.rdata), wetland abundance (ponds.rdata), and winter habitat (winter.rdata) data, and Bayesian model code, model constraints, and initial values needed to
run the environmental covariate sub-models (generate_covariates.R) that, inconjunction with data describing variation in waterfowl survival, fecundity, and population sizes (model_files.rdata)
were used to inform the integrated population model (IPM_code.R) for Canvasback and Redhead ducks in North America from 1961 - 2021. The environmental and population sub-models were parameterized using the Nimble
library platform in R. The posterior distributions from these models (estimated_vital_rates.rdata) were used to perform a transient lifetable response experiment (LTRE_Code.R)

Data Descriptions

crop.rdata

crop.dat : The log-scaled amount of active cropland (in thousands of hectares) within the Prairie Pothole Region, which was calculated and provided by Canada and United States Agriculture Census data. Census data from each country occurred approximately every 5 years. Values during years in which a census was not completed are represented with NAs.
crop.constants : The model constraints required to perform the agricultural conditions sub-model in Nimble. See generate_covariates.R
crop.inits : The initial values for parameters of interest required to perform the agricultural conditions sub-model in Nimble. See generate_covariates.R

winter.rdata

winter.dat : Data file containing informating describing the amount of sub-aquatic vegetation or salinity in the Cheasapeake Bay or Laguna Madre, as well as potential environmental correlates.
- sav : Sub-aquatic Vegetation (log-scaled hectares) in the Cheasapeake Bay, USA
- salinity.lgm : Measured salinity (log-scaled) throughout the Laguna Madre of Texas, USA
- salinity.chb : Measured salinity (log-scaled) in the Cheasapeake Bay, USA
- covs.lgm : Environmental covariates fit to describe salinity in the Laguna Madre of Texas, USA
  - ACE : Accumulated Cyclone Energy
  - RESIDS : Residual Sea Surface Temperaturesn the Laguna Madre
  - PREC6 : Precipitation in the Laguna Madre during the spring and summer
  - FLOW6 : Freshwater flow in the Laguna Madre during the spring and summer
- covs.chb :
  - ACE : Accumulated Cyclone Energy
  - RESIDS : Residual Sea Surface Temperatures in the Cheasapeake Bay
  - PREC : Precipitation in the Laguna Madre during the spring and summer in the Cheasapeake Bay
winter.constants : The model constraints required to perform the winter habitat sub-model in Nimble. See generate_covariates.R
winter.inits : The initial values for parameters of interest required to perform the winter habitat sub-model in Nimble. See generate_covariates.R

pond.rdata

pond.dat : The estimated number of ponds (in thousands) in the prairies ($prairies) and the associated standard error ($sig_ponds) that were generated from the annual Breeding Population Survey for Waterfowl and their Habitats. The breeding population survey was not conducted in 2020 and 2021 due to COVID restrictions and values during this time frame are represented by NAs
pond.constants : The model constraints required to perform the breeding pond abundance sub-model in Nimble. See generate_covariates.R
pond.inits : The initial values for parameters of interest required to perform the breeding pond abundance sub-model in Nimble. See generate_covariates.R

model_files.rdata

dat : The data file necessary to run the Redhead and Canvasback Integrated Population Model in Nimble. This file contains the annual estimates in each modeled covariate and their associated uncertainity. Likewise, this file
also includes the demographic data necessary estimate waterfowl fecundity (age ratios at harvest), cause-specific mortality (band recoveries), and abundance (bpop surveys and total harvest).
Due to the complexity of these data, please contact the lead author for questions regarding these data, their use, and the modeling procedure.
constants : The model constraints required to perform the integrated population model for Canvasbacks and Redheads in Nimble. See IPM_code.R
inits : The initial values for parameters of interest required to perform the integrated population model for Canvasbacks and Redheads in Nimble. See IPM_code.R

estimated_vital_rates.rdata

File components correspond with the estimates generated from the Integrated Population Model (IPM_code.R) that were required to perform the Life Table Response Experiment (LTRE_Code.R)

Reproductive Success
- fecundity: individual draws (dim 3 from) posterior distributions for annual per-capita fecundity rates for Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2) from 1961:2020 (dim 2)
- bp individual draws (dim 2 from) posterior distributions for the discrepancy between HY and AHY female reproducive success in Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2)
Abundance:
- naf: individual draws (dim 3 from) posterior distributions for the breeding abudances of AHY female Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2) from 1961:2020 (dim 2)
- nam: individual draws (dim 3 from) posterior distributions for the breeding abudances of AHY male Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2) from 1961:2020 (dim 2)
- njf: individual draws (dim 3 from) posterior distributions for the breeding abudances of HY female Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2) from 1961:2020 (dim 2)
- njm: individual draws (dim 3 from) posterior distributions for the breeding abudances of HY male Canvasbacks (dim 1: level: 1) and Redheads (dim 1: level: 2) from 1961:2020 (dim 2)
Harvest Mortality
- kappa_ahy_c_f: individual draws (dim 2 from) posterior distributions for the harvest probabilities AHY female Canvasbacks from 1961:2020 (dim 1)
- kappa_ahy_c_m: individual draws (dim 2 from) posterior distributions for the harvest probabilities AHY male Canvasbacks from 1961:2020 (dim 1)
- kappa_ahy_r_f: individual draws (dim 2 from) posterior distributions for the harvest probabilities AHY female Redheads from 1961:2020 (dim 1)
- kappa_ahy_r_m: individual draws (dim 2 from) posterior distributions for the harvest probabilities AHY male Redheads from 1961:2020 (dim 1)
- kappa_hy_c_f: individual draws (dim 2 from) posterior distributions for the harvest probabilities HY female Canvasbacks from 1961:2020 (dim 1)
- kappa_hy_c_m: individual draws (dim 2 from) posterior distributions for the harvest probabilities HY male Canvasbacks from 1961:2020 (dim 1)
- kappa_hy_r_f: individual draws (dim 2 from) posterior distributions for the harvest probabilities HY female Redheads from 1961:2020 (dim 1)
- kappa_hy_r_m: individual draws (dim 2 from) posterior distributions for the harvest probabilities HY male Redheads from 1961:2020 (dim 1)
Winter Mortality
- s.winter_ahy_c_f: individual draws (dim 2 from) posterior distributions for the natural winter survival probabilities AHY female Canvasbacks from 1961:2020 (dim 1)
- s.winter_ahy_c_m: individual draws (dim 2 from) posterior distributions for the natural winter survival probabilities AHY male Canvasbacks from 1961:2020 (dim 1)
- s.winter_ahy_r_f: individual draws (dim 2 from) posterior distributions for the natural winter survival probabilities AHY female Redheads from 1961:2020 (dim 1)
- s.winter_ahy_r_m: individual draws (dim 2 from) posterior distributions for the natural winter survival probabilities AHY male Redheads from 1961:2020 (dim 1)
- s.winter_hy_c_f: individual draws (dim 2 from) posterior distributions for the natural winter survival HY female Canvasbacks from 1961:2020 (dim 1)
- s.winter_hy_c_m: individual draws (dim 2 from) posterior distributions for the natural winter survival HY male Canvasbacks from 1961:2020 (dim 1)
- s.winter_hy_r_f: individual draws (dim 2 from) posterior distributions for the natural winter survival HY female Redheads from 1961:2020 (dim 1)
- s.winter_hy_r_m: individual draws (dim 2 from) posterior distributions for the natural winter survival HY male Redheads from 1961:2020 (dim 1)
Summer Mortality
- s.summer_c_f: individual draws (dim 2 from) posterior distributions for the natural summer survival probabilities female Canvasbacks from 1961:2020 (dim 1)
- s.summer_c_m: individual draws (dim 2 from) posterior distributions for the natural summer survival probabilities male Canvasbacks from 1961:2020 (dim 1)
- s.summer_r_f: individual draws (dim 2 from) posterior distributions for the natural summer survival probabilities female Redheads from 1961:2020 (dim 1)
- s.summer_r_m: individual draws (dim 2 from) posterior distributions for the natural summer survival probabilities male Redheads from 1961:2020 (dim 1)

Model Code Descriptions

generate_covariates.R is an R object that contains the code to run each of the environmental sub-models. Results from these models are already embedded in the model_files.rdata file.
- Data prerequisite: crops.rdata, ponds.rdata, winter.rdata
IPM_code.R is an R object that contains the Nimble code necessary to perform the multispecies intergrated population model.
- Data prerequisite: model_files.rdata
LTRE_Code.R is an R object that contains the code necessary to perform a transient life table response experiment based on vital rates estimated from the multispecies IPM.
- Data prerequisite: estimated_vital_rates.rdata

Code/Software

R is required to load each .rdata file and run each .R script; the script was created using version 4.4.0.
The Nimble library package in R in addition to its dependencies are required to perform the Bayesian analyses presented in generate_covariates.R and IPM_code.R

DATA COLLECTION

We combined a series of long-term data sets into a single integrated population model that provided insights into how variation in seasonal survival (band releases and recoveries) and offspring production (harvest age-ratios) contributed to fluctuations in population growth (breeding survey, harvest estimates) for Canvasbacks and Redheads from 1961–2021.

Banding Data – Information regarding the banding and subsequent harvest of ducks was acquired from the GameBirds Database CD (Bird Banding Lab, USGS Patuxent Wildlife Research Center, Laurel MD, USA, version August 2022). Male and female Canvasbacks and Redheads were captured following breeding but prior to the hunting season (Pre-Hunting) as ducklings (Local) or hatch year (HY; fledged juvenile) individuals as well as after hatch year (AHY; adult) individuals or following the hunting season (Post-Hunting) as an undifferentiated mixture of second year (SY) and after second year (ASY) individuals captured and released across North America from 1961–2022. We limited the pre-harvest banding data for both species to include all individuals banded and released alive in areas within the Canadian provinces of Alberta, Manitoba, Saskatchewan, as well as the states of Minnesota, Montana, North Dakota, and South Dakota within the USA (Fig. 1). For the pre-hunting banding group, we retained individuals captured between 1961–2021 during the late summer (Jul 15^th – Sep 15^th) with a known sex (M or F) and age-class (local, HY or AHY) that were released without any additional markers considered to meaningfully affect survival of an individual (e.g., nasal saddles or dual banding were permissible but telemetered individuals were excluded; Lameris & Kleyheeg, 2017). For post-hunting banding, we limited the spatial boundary of banding efforts to only consider individuals released from the Atlantic, Central, or Mississippi Flyways (Fig. 1). We followed the same data selection procedures, but limited releases to occur between Jan. 1^st – March 15^th from 1962–2022. Because too few banders differentiated SY from ASY at time of banding, we treated all post-hunting samples as AHY adults. Individuals banded during this period that were reported to be harvested during the winter they were originally banded were censored from the analysis, as the underlying model assumption was that this cohort of individuals had already survived the current hunting season.

For both seasonal banding efforts, we only included recoveries of hunter-shot individuals harvested between September and February in which a known year-of-death could be ascertained. In addition to self-reported recoveries (i.e., reported by the hunter), we included hunter-harvested individuals that were instead reported by federal, state, or provincial entities (e.g., outcomes of hunter check stations or other forms of solicitation). We limited the dataset to only include recoveries of hunter-harvested individuals killed within 15 years of initial banding, which represented > 99% of pre-hunting and post-hunting recoveries. This cut-off was arbitrarily selected but did not meaningfully bias parameter estimation while vastly improving computational efficiency by bypassing the estimation of hundreds of zero-equivalent cell probabilities (personal communication S. Bonner).

Harvest Intensity – We used the average number of Canvasbacks or Redheads allowed to be harvested per day (i.e., bag limit; (Appendix S1: Tables S1a-b) across the U.S. portions of the Atlantic, Mississippi, Central, and Pacific flyways during each year of the study as an index of harvest regulatory pressure. Annual harvest restrictions were acquired from the published literature (Péron et al., 2012), the annual release of the Late-Season Migratory Bird Hunting Regulations (e.g., USFWS 2022), and direct requests to the U.S. Fish and Wildlife Service. For these species, liberal harvest regulations were bag limits of two (Canvasbacks) and two to four (Redheads) allowable harvest per day, whereas conservative harvest regulations were either a bag limit of one individual per day or total closure.

Harvest Composition – Data describing the age and sex structure of the harvested Canvasback and Redhead populations were derived from the annual Parts Collection surveys conducted by the U.S. Fish and Wildlife Service (USFWS) where a subset of hunters submit a wing from every duck they harvested (Pearse et al. 2014). These data were acquired through a direct request to the U.S. Fish and Wildlife Service. Additionally, estimates of the total number of Canvasbacks and Redheads harvested in the United States and Canada were derived from the Harvest Information Program (Steeg et al., 2002) and Canadian National Harvest Survey (Smith et al., 2022), respectively.

Breeding Duck and Pond Densities – The relative number of breeding Canvasbacks and Redheads, as well as the relative amount of their breeding habitat (i.e., flooded ponds) within the Prairies were calculated using count data from the USFWS Waterfowl Breeding Population and Habitat Survey (hereafter BPOP; Smith, 1995), which has conducted an annual survey of breeding waterfowl and their habitats throughout the core part of these species’ breeding ranges (i.e., central Canada and the north-central United States) during the spring from 1961 through 2022 (U.S. Fish and Wildlife Service, 2022). However, BPOP surveys did not occur during 2020 and 2021. For the purposes of this study, we limited the spatial extent of BPOP survey to only include transects flown within Alberta, Manitoba, Saskatchewan, Montana, North Dakota, and South Dakota.

Agriculture Development – The amounts of active cropland in the Prairies during each year of the study were estimated from Canada and United States Agriculture Census data (see Buderman et al., 2020). Annual estimates of active cropland acreages were summarized to represent an index of agricultural development during 1961–2021. Although agricultural development is predicted to have greater impact on upland-nesting dabbling ducks (Duncan and Devries 2018), it also impacts the wetland habitats in which Canvasbacks and Redheads forage and nest, as well as the predator communities that can access overwater nesting pochards (Sargeant et al. 1993, Bartzen et al. 2010).

Winter Habitat – Winter habitat conditions were assumed to be related to submerged aquatic vegetation (SAV) within the Chesapeake for Canvasbacks and environmental salinity (TDS; total dissolved solids) in the Laguna Madre for Redheads. Although Redheads likely respond to variation in SAV, time series data describing SAV were not available for the Laguna Madre. Therefore, we assumed that annual fluctuations in salinity were an informative proxy of both SAV conditions and osmotic constraints (Quammen and Onuf 1993, Moore 2009), which in turn was representative of winter habitat conditions that simultaneously influenced Redhead food availability and harvest risk (Ballard et al. 2021)..

Climate Data – We used the average Pacific/North American (PNA; Leathers et al., 1991) teleconnection pattern from April–July as an index of drought severity or environmental stress during the breeding season throughout the Prairies, and average sea-surface temperatures (SST) from September–March in the Chesapeake and Laguna Madre as an index of winter severity for Canvasbacks and Redheads, respectively (see Data Availability statement).