Comparing the performance of three common species distribution modelling frameworks for freshwater environments through application to eel species in New Zealand DOI
Anthony R. Charsley, Nokuthaba Sibanda, Simon Hoyle

et al.

Canadian Journal of Fisheries and Aquatic Sciences, Journal Year: 2022, Volume and Issue: 80(3), P. 533 - 548

Published: Nov. 22, 2022

Globally, many freshwater species are depleting and require population-level assessments. Many distribution modelling frameworks available for such assessments, but comparisons needed to understand their predictive performance under different settings. K-fold cross-validation techniques were employed compare the of three commonly used frameworks: machine learning, spatiotemporal modelling, Gaussian process (GP) modelling. Through application New Zealand populations longfin eel ( Anguilla dieffenbachii) shortfin australis), area receiver operating characteristic curve (AUC) true skill statistic (TSS) model metrics estimated. All produced approximately consistent maps differed in performance. AUC TSS results indicated that predictions from framework most accurate, followed by GP However, all performed similarly when training test data spatially independent. In addition having best performance, showed greatest promise advancement assessment is therefore recommended. The useful ecologists resource managers make informed decisions on appropriateness a research objective.

Language: Английский

SSN2: The next generation of spatial stream network modeling in R DOI Creative Commons
Michael Dumelle, Erin E. Peterson, Jay M. Ver Hoef

et al.

The Journal of Open Source Software, Journal Year: 2024, Volume and Issue: 9(99), P. 6389 - 6389

Published: July 26, 2024

The SSN2 R package provides tools for spatial statistical modeling, parameter estimation, and prediction on stream (river) networks.SSN2 is the successor to SSN (Ver Hoef, Peterson, Clifford, & Shah, 2014), which was archived alongside broader changes in

Language: Английский

Citations

5

Impacts of different types of data integration on the predictions of spatio-temporal models: A fishery application and simulation experiment DOI
Arnaud Grüss, Richard L. O’Driscoll, James T. Thorson

et al.

Fisheries Research, Journal Year: 2025, Volume and Issue: 284, P. 107321 - 107321

Published: March 11, 2025

Language: Английский

Citations

0

tinyVAST: Multivariate Spatio-Temporal Models using Structural Equations DOI Open Access
James T. Thorson, Sean C. Anderson

Published: March 13, 2025

Language: Английский

Citations

0

tinyVAST: R Package With an Expressive Interface to Specify Lagged and Simultaneous Effects in Multivariate Spatio‐Temporal Models DOI Creative Commons
James T. Thorson, Sean C. Anderson, Pamela Goddard

et al.

Global Ecology and Biogeography, Journal Year: 2025, Volume and Issue: 34(4)

Published: April 1, 2025

ABSTRACT Aim Multivariate spatio‐temporal models are widely applicable, but specifying their structure is complicated and may inhibit wider use. We introduce the R package tinyVAST from two viewpoints: software user statistician. Innovation From viewpoint, adapts a used formula interface to specify generalised additive combines this with arguments spatial interactions among variables. These specified using arrow notation (from structural equation models) or an extended arrow‐and‐lag that allows simultaneous, lagged recursive dependencies variables over time. The also specifies domain for areal (gridded), continuous (point‐count) stream‐network data. statistician constructs sparse precision matrices representing multivariate variation, parameters estimated by linear mixed model (GLMM). This expressive encompasses vector autoregressive, empirical orthogonal functions, factor analysis ARIMA models. Main Conclusion To demonstrate, we fit data survey platforms sampling corals, sponges, rockfishes flatfishes in Gulf of Alaska Aleutian Islands. then compare eight alternative structures different assumptions about habitat drivers detectability. Model selection suggests towed‐camera bottom trawl gears have variation detectability sample same underlying density positively associated sponges while negatively corals. conclude can be test hypotheses research real‐world policy evaluation.

Language: Английский

Citations

0

Spatially varying catchability for integrating research survey data with other data sources: case studies involving observer samples, industry-cooperative surveys, and predators as samplers DOI
Arnaud Grüss, James T. Thorson, Owen F. Anderson

et al.

Canadian Journal of Fisheries and Aquatic Sciences, Journal Year: 2023, Volume and Issue: unknown

Published: June 15, 2023

Spatio-temporal models are widely applied to standardise research survey data and increasingly used generate density maps indices from other sources. We developed a spatio-temporal modelling framework that integrates (treated as “reference dataset”) sources (“non-reference datasets”) while estimating spatially varying catchability for the non-reference datasets. demonstrated it using two case studies. The first involved bottom trawl observer spiny dogfish ( Squalus acanthias) on Chatham Rise, New Zealand. second cod predators samplers of juvenile snow crab Chionoecetes opilio) abundance, integrated with industry-cooperative surveys in eastern Bering Sea. Our leveraged strengths individual (the quality reference dataset quantity data), downweighting influence datasets via estimated catchabilities. They allowed generation annual longer time-period provision one single index rather than multiple each covering shorter time-period.

Language: Английский

Citations

9

Integrating survey and observer data improves the predictions of New Zealand spatio-temporal models DOI Creative Commons
Arnaud Grüss, Anthony R. Charsley, James T. Thorson

et al.

ICES Journal of Marine Science, Journal Year: 2023, Volume and Issue: 80(7), P. 1991 - 2007

Published: Aug. 22, 2023

Abstract In many situations, species distribution models need to make use of multiple data sources address their objectives. We developed a spatio-temporal modelling framework that integrates research survey and collected by observers onboard fishing vessels while accounting for physical barriers (islands, convoluted coastlines). demonstrated our two bycatch in New Zealand deepwater fisheries: spiny dogfish (Squalus acanthias) javelinfish (Lepidorhynchus denticulatus). Results indicated employing observer-only or integrated is necessary map fish biomass at the scale exclusive economic zone, interpolate local indices (e.g., east coast South Island) years with no but available observer data. also showed that, if enough are available, fisheries analysts should: (1) develop both an model relying on survey-only data; (2) given geographic area, ultimately choose index produced based reliability interannual variability index. conducted simulation experiment, which predictions virtually insensitive consideration barriers.

Language: Английский

Citations

7

Understanding the dynamics of fish spawning phenology and habitat in a changing ecosystem using a long-term ichthyoplankton monitoring dataset DOI
Noah Hunt, Katrina Rokosz, Yong Chen

et al.

Canadian Journal of Fisheries and Aquatic Sciences, Journal Year: 2024, Volume and Issue: 81(12), P. 1728 - 1739

Published: Aug. 9, 2024

As ecosystems change, understanding the consequences for fish population dynamics and habitat use is essential resource management. Using white perch ( Morone americana) survey data on early life stages collected during a long-term ichthyoplankton monitoring program in Hudson River (New York, USA), an ecosystem under immense pressure from climate ecological shifts, anthropogenic activities, we evaluated drivers of changes egg abundance spawning between 1980 2017. Results indicated that associated nonlinearly with temperature, conductivity, discharge, depth, location, week year. We also found has declined within river over time. Additionally, shifts hotspots activity were identified, including evidence lower extent moved upriver since 1980. This study indicates histories are changing. It highlights utility broadening our ecology age big changing ecosystems.

Language: Английский

Citations

2

Connecting population functionality with distribution model predictions to support freshwater and marine management of diadromous fish species DOI Creative Commons
Chloé Dambrine, Patrick Lambert, Sophie Elliott

et al.

Biological Conservation, Journal Year: 2023, Volume and Issue: 287, P. 110324 - 110324

Published: Oct. 9, 2023

Diadromous fish species have a complex life cycle during which they migrate between marine and freshwater habitats. They experience multiple human-induced pressures in both environments, likely exacerbated by climate change, leading to dramatic population declines across their distribution ranges. Currently Species Distribution Models (SDMs) been applied separately continental habitats improve our understanding of lifecycles help with management. Integrating the freshwater-sea continuum into decisions would now be step further improving With this objective, we developed decision tree that links SDM outputs current observations functionality suggested management guidance options for viability these species. Potential effects change were included through future projections guide integrative long-term Several criteria proposed assess validity considering main sources uncertainties local expert knowledge on habitat status. The framework was approximately one hundred catchments from southern Portugal Scandinavia four diadromous At European level, differed two anadromous catadromous Platichthys flesus Chelon ramada populations seemed better state than those Alosa alosa A. fallax. Finally, national experts, focused distributed along latitudinal gradient test methodology demonstrate challenges terms continuity.

Language: Английский

Citations

1

Comparing the performance of three common species distribution modelling frameworks for freshwater environments through application to eel species in New Zealand DOI
Anthony R. Charsley, Nokuthaba Sibanda, Simon Hoyle

et al.

Canadian Journal of Fisheries and Aquatic Sciences, Journal Year: 2022, Volume and Issue: 80(3), P. 533 - 548

Published: Nov. 22, 2022

Globally, many freshwater species are depleting and require population-level assessments. Many distribution modelling frameworks available for such assessments, but comparisons needed to understand their predictive performance under different settings. K-fold cross-validation techniques were employed compare the of three commonly used frameworks: machine learning, spatiotemporal modelling, Gaussian process (GP) modelling. Through application New Zealand populations longfin eel ( Anguilla dieffenbachii) shortfin australis), area receiver operating characteristic curve (AUC) true skill statistic (TSS) model metrics estimated. All produced approximately consistent maps differed in performance. AUC TSS results indicated that predictions from framework most accurate, followed by GP However, all performed similarly when training test data spatially independent. In addition having best performance, showed greatest promise advancement assessment is therefore recommended. The useful ecologists resource managers make informed decisions on appropriateness a research objective.

Language: Английский

Citations

1