DOI: 10.14466/CefasDataHub.138

Description

Environmental Niche Model (ENM) outputs for 49 commercial fish species under climate change until the decade of 2060 around northwestern Europe. A model ensemble of 5 ENMs was used (MaxEnt, Generalised Linear Models, Support Vector Machine, Random Forest and BIOCLIM ), and projections were made under three different emission scenarios: A1B, RCP4.5 and RCP 8.5. The data shows model agreement (normalised to 1) for presence/absence decadal projections from 2020 to 2060. Additionally we provide data on model performance, with the Area Under the Curve (AUC) scores of the Receiver Operator Characteristic (ROC) curve for each of the 5 ENMs trained for each combination of fish species and emission scenario. Only ENMs with an AUC score of at least 0.7 were considered. The data is based on the output of an ensemble of 5 Environmental Niche Modeling techniques: Maximum Entropy (MaxEnt), Generalised Linear Model (GLM), Support Vector Machine (SVM), Random Forest and BIOCLIM. Each model was trained to capture the environmental requirements of individual fish species, using the 20 year-average in the period 1997-2016 of the environmental variables: near-bed sea temperature, sea surface temperature, near-bed salinity, sea surface salinity, the difference in near-bed and surface salinity and temperature (an approximation of stratification), and depth.

Once the models were trained, they were used to study future habitat suitability for the 49 commercially-important fish species in 20-year averages centered in 2020, 2030, 2040, 2050 and 2060. To produce maps of the suitable habitat, the model projections were converted to binary presence/absence data using the thresholds that optimized the True Skill Statistics (TSS) of each model. The data shows the model agreement, normalised to 1 ("1" denotes all ENMs agree an area is suitable, "0" denotes all ENMs agree an area is not suitable). The number of ENMs considered for each combination of fish species and emission scenario depends on model performance. A ENM model was considered if its Area Under the Curve (AUC) score of the Receiver Operator Characteristic (ROC) curve was at least 0.7. Model performance is listed in the file "ModelperformanceAUC.xlsx".

Model input data: Climate change projection data from Met Office and Plymouth Marine Laboratory (PML) and species presence and abundance records from online databases (Ocean Biodiversity Information System (OBIS), Global Biodiversity Information System (GBIF) and Marine Scotland (Moriarty, M., Greenstreet, S.P.R. and Rasmussen, J. (2017) Derivation of Groundfish Survey Monitoring and Assessment Data Product for the Northeast Atlantic Area. Scottish Marine and Freshwater Science, 8, 240pp. DOI: 10.7489/1984-1). Models were trained using the R and Maxent softwares.

Study region: northeast Atlantic shelf. Latitudes between -17° and 9.25°; longitudes between 44° and 65°.

These datasets were produced for the study in the article "Climate change projections of commercial fish distribution and suitable habitat around northwestern Europe", by Bryony L. Townhill, Elena Couce, Jonathan Tinker, Susan Kay, John K. Pinnegar, in Fish and Fisheries (2023; http://doi.org/10.1111/faf.12773).

Contributors

Couce, Elena / Townhill, Bryony

Subject

Habitat / Model / Modelling / Fish / Fisheries /Climate

Start Date

01/01/2011

End Date

01/01/2070

Year Published

2023

Version

1

Citation

Couce and Townhill (2023). Modelled projections of habitat for commercial fish around North-western Europe under climate change, 2020 to 2060. Cefas, UK. V1. doi: https://doi.org/10.14466/CefasDataHub.138

Rights List

Open Government Licence

DOI

10.14466/CefasDataHub.138