Atlantic salmon habitat-abundance modeling using machine learning methods - research and evaluation of ML models

Aug 2, 2026

Upload by Maeve

Climate change and anthropogenic activities degrade fish habitat suitability, increasing the need for accurate species abundance models to support conservation decisions. This study applies machine learning to characterize the habitat-abundance relationship of juvenile Atlantic salmon in the Teno catchment across Finland and Norway. Support Vector Regression, Random Forest, Gradient Boosting, and Support Vector Classification are compared to capture nonlinear complexity. Models using substrate, shade, and vegetation inputs show improved performance. Regression and classification results are contrasted to highlight modeling challenges, limitations, and the need to address temporal variation for better precision.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Motivation from climate change and human impacts

Role of habitat quality indicators in salmon abundance

Machine learning in ecological abundance and community modeling

Study context and overview

This is an electronic reprint of the original article. This reprint may differ from the original in pagination and typographic detail. Author(s): Bähar Jelovica, Jaakko Erkinaro, Panu Orell, Bjørn Kløve, Ali Torabi Haghighi, Hannu Marttila Title: Atlantic salmon habitat-abundance modeling using machine learning methods Year: 2024 Version: Published version Copyright: The Author(s) 2024 Rights: CC BY 4.0 Rights url: [https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/) [Please cite the original version:](Please cite the original version:) Bähar Jelovica, Jaakko Erkinaro, Panu Orell, Bjørn Kløve, Ali Torabi Haghighi, Hannu Marttila, Atlantic salmon habitat-abundance modeling using machine learning methods, Ecological Indicators, Volume 160, 2024, 111832, ISSN 1470-160X, [https://doi.org/10.1016/j.ecolind.2024.111832](https://doi.org/10.1016/j.ecolind.2024.111832) . Ecological Indicators 160 (2024) 111832 Contents lists available at ScienceDirect Ecological Indicators journal [homepage: www.elsevier.com/locate/ecolind](homepage: www.elsevier.com/locate/ecolind) | Atlantic salmon habitat-abundance learning methods | | | modeling using machine | | | --- | --- | --- | --- | --- | | Ba¨har Jelovicaa, *, Jaakko Erkinarob, Panu Orell b, Bjørn Kløve a, Ali Torabi Haghighia, Hannu Marttilaa
a Water, Energy and Environmental Engineering Research Unit, University of Oulu, Finland b Natural Resource Institute Finland (LUKE), Finland | | | | | | A R T I C L E I N F O | | A B S T R A C T | | | | Keywords:
Atlantic salmon Abundance Machine Learning Modelling Habitat-Abundance Relationship Arctic | | Climate change and anthropogenic activities have impacts on fish habitat suitability, demanding more accurate modeling of species abundance for effective conservation and management. In this study, we applied Machine Learning techniques to model the habitat-abundance relationship of juvenile Atlantic salmon (Salmo salar) in the Teno catchment in Finland and Norway. To capture the complexity and nonlinearity of the habitat-abundance relationship, we employed Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting (GB), and Support Vector Classification (SVC) and compared their performances. Among the regression models considered, those incorporating input variables such as substrate, shade, and vegetation demonstrate higher performance. Support Vector Regression yields the highest mean cross-validation score (R2 = 0.58), and Gradient Boosting produces the highest test score (R2 = 0.6) among the regression techniques. The mean cross-validation and test scores obtained for the classification models are notably higher compared to the regression models across all scenarios. A comparison between regression and classification results highlights the challenges of accurately modeling the habitat-abundance relationship. This study provides insights into the challenges and potential of machine learning techniques for juvenile Atlantic salmon habitat-abundance modeling in complex riverine habitat environments. The findings emphasize the importance of considering the limitations of machine learning models, particularly in ecological contexts, and the need for further research to address temporal variations and improve the precision of habitat-abundance modeling. | | | 1. Introduction Climate change and anthropogenic activities influence fish habitat suitability by changing the river temperature (Isaak et al., 2012), vegetation distribution (Ackerly et al., 2015), food availability (Cameron et al., 2019), water quality (Ritson et al., 2014), and hydrological regimes (Jelovica et al., 2022). Since distribution and abundance of fluvial fish are strongly impacted by the habitat (Armstrong et al., 2003), it is essential to measure the indicators which are reflecting the habitat quality and could support river conservation and improvements (Giorgio et al., 2016).

FAQs

What problem does the study address for Atlantic salmon conservation?

It addresses how climate change and human activities alter habitat suitability, making it difficult to model juvenile Atlantic salmon abundance accurately for conservation and management.

Which machine learning methods are compared in the study?

Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting (GB), and Support Vector Classification (SVC) are used and compared by performance.

What habitat variables improve modeling performance?

Inputs including substrate, shade, and vegetation are reported to produce higher performance, especially within the regression techniques.

Aug 2, 2026

Upload by Maeve