This web application uses cookies and other tracking technologies to ensure you get the best experience.
SC/69B/SM/01
Download this file5.8 MB |
Download |
View directly in browser5.8 MB |
View in browser |
Resource ID
22162
Access
Open
Document Number
SC/69B/SM/01
Full Title
Spatially-explicit models of density improve estimates of Eastern Bering Sea beluga (Delphinapterus leucas) abundance and distribution from line-transect surveys
Author
M.C. Ferguson, P.B. Conn, And J.T. Thorson
Authors Summary
SM## investigated spatially explicit models and ensemble modeling techniques for estimating animal abundance from line-transect survey data. Spatially explicit models are expected to be statistically more efficient, resulting in more precise abundance estimates, than design-based abundance estimators that rely heavily on assumptions about survey design. Ensemble models allow model selection uncertainty to propagate to the abundance estimator. The authors developed density surface models using stochastic partial differential equations and basis-penalty smoothers for a case study, aerial line-transect survey data for belugas (Delphinapterus leucas) from the Eastern Bering Sea (EBS) stock. The authors compared EBS beluga abundance estimates for 2017 and 2022 that were derived using post-stratified, design-based abundance estimators with analogous estimates derived using spatially explicit and ensemble modeling methods. Although the design-based estimators were less precise than individual spatially-explicit models (with one exception), precision (CV) was essentially equivalent between the design-based and ensemble model-averaged abundance estimators. Because spatial models identify spatial patterns in animal density (the number of individuals per unit area) at finer resolutions than design-based models, the authors argued that ensembles of spatially-explicit density models provide a reasonable path forward for estimating cetacean abundance and distribution in a way that is useful to management and conservation efforts.
Publisher
IWC
Publication Year
2024
Abstract
We investigate spatially explicit models and ensemble modeling techniques for estimating animal abundance from line-transect survey data. Spatially explicit models are expected to be statistically more efficient, resulting in more precise abundance estimates, than design-based abundance estimators that rely heavily on assumptions about survey design. Ensemble models allow model selection uncertainty to propagate to the abundance estimator. We develop density surface models using stochastic partial differential equations and basis-penalty smoothers for a case study, belugas (Delphinapterus leucas) from the Eastern Bering Sea (EBS) stock. EBS belugas are upper trophic level predators in a rapidly changing ecosystem and are a vital nutritional and cultural resource for Alaska Natives. Effective management of this stock requires regular monitoring to derive accurate and unbiased estimates of abundance. Since 1992, aerial line-transect surveys have been the primary means of surveying and estimating abundance of EBS belugas in the region. We compared EBS beluga abundance estimates for 2017 and 2022 that were derived using post-stratified, design-based abundance estimators with analogous estimates derived using spatially explicit and ensemble modeling methods. Although the design-based estimators were less precise than individual spatially-explicit models (with one exception), precision (CV) was essentially equivalent between the design-based and ensemble model-averaged abundance estimators. The design-based models estimated that there were 12,269 belugas in 2017 (CV = 0.12) and 20,635 belugas in 2022 (CV = 0.31; the study area was larger in 2022). The ensemble spatial models estimated that there were 11,597 belugas in 2017 (CV = 0.12) and 17,197 belugas in 2022 (CV = 0.33). Among the individual spatially-explicit models, abundance estimates ranged from 11,242 to 11,962 (CV = 0.11 to 0.12) in 2017 and 12,593 to 21,508 (CV = 0.18 to 0.29) in 2022. Because spatial models identify spatial patterns in beluga density (the number of belugas per unit area) at finer resolutions than design-based models, we argue that ensembles of spatially-explicit density models provide a reasonable path forward for estimating EBS beluga abundance and distribution in a way that is useful to management and conservation efforts.