Evaluating the robustness of generalized additive models as a tool for threshold detection in variable environments DOI Creative Commons
A. Raine Detmer, Eric J. Ward, Mary E. Hunsicker

et al.

Ecosphere, Journal Year: 2025, Volume and Issue: 16(3)

Published: March 1, 2025

Abstract As global climate change and anthropogenic activities amplify widespread environmental variability, there is a strong need for management strategies that incorporate relationships between ecosystem components. This especially apparent when changes in drivers cause threshold responses (abrupt, nonlinear changes) ecosystems. Such ecological thresholds can provide useful reference points decisions. However, methods detecting empirical datasets may fail to find an existing threshold, one does not exist, or be biased their estimates of locations. These types misspecifications result high conservation socioeconomic costs. Simulation studies mitigate these risks by providing information about method performance across different scenarios. Here, we constructed series simulations evaluate the robustness detection with generalized additive models (GAMs) exposed variety common, real‐world data characteristics. GAMs generally performed best time were long, observation error was low, crossed fairly frequently, covariates accounted for. Over realistic ranges values, frequency crossing had stronger effects on detectability than length. Importantly, found depend both shape relationship statistical definition location. case study, applied this dataset relating ocean temperature spatial distribution Pacific hake ( Merluccius productus ), largest volume fishery US West Coast. While suggest no evidence relationship, our indicated approximately equal chances true false given currently available data. Our results general guidelines where likely robust are context indicator development ecosystem‐based variable world.

Language: Английский

Evaluating the robustness of generalized additive models as a tool for threshold detection in variable environments DOI Creative Commons
A. Raine Detmer, Eric J. Ward, Mary E. Hunsicker

et al.

Ecosphere, Journal Year: 2025, Volume and Issue: 16(3)

Published: March 1, 2025

Abstract As global climate change and anthropogenic activities amplify widespread environmental variability, there is a strong need for management strategies that incorporate relationships between ecosystem components. This especially apparent when changes in drivers cause threshold responses (abrupt, nonlinear changes) ecosystems. Such ecological thresholds can provide useful reference points decisions. However, methods detecting empirical datasets may fail to find an existing threshold, one does not exist, or be biased their estimates of locations. These types misspecifications result high conservation socioeconomic costs. Simulation studies mitigate these risks by providing information about method performance across different scenarios. Here, we constructed series simulations evaluate the robustness detection with generalized additive models (GAMs) exposed variety common, real‐world data characteristics. GAMs generally performed best time were long, observation error was low, crossed fairly frequently, covariates accounted for. Over realistic ranges values, frequency crossing had stronger effects on detectability than length. Importantly, found depend both shape relationship statistical definition location. case study, applied this dataset relating ocean temperature spatial distribution Pacific hake ( Merluccius productus ), largest volume fishery US West Coast. While suggest no evidence relationship, our indicated approximately equal chances true false given currently available data. Our results general guidelines where likely robust are context indicator development ecosystem‐based variable world.

Language: Английский

Citations

0