Koster historical biodiversity assessment

Occurrence Observation
Latest version published by Wildlife.ai on Sep 5, 2025 Wildlife.ai
Home:
http://www.subsim.se
Publication date:
5 September 2025
Published by:
Wildlife.ai
License:
CC-BY 4.0

Download the latest version of this resource data as a Darwin Core Archive (DwC-A) or the resource metadata as EML or RTF:

Data as a DwC-A file download 72,369 records in English (847 KB) - Update frequency: as needed
Metadata as an EML file download in English (21 KB)
Metadata as an RTF file download in English (12 KB)

Description

!Important! - Download this dataset as a Darwin-Core Archive file to ensure both individualCount and organismQuantity are included. As of now, data downloads as GBIF Annotated Archives do not include organismQuantity. Dataset of species records extracted from video footage recorded by remotely operated vehicles (ROVs) in the marine protected area Kosterhavets nationalpark on the Swedish west-coast. The original movies were collected during 1997-2023. This data set is based on videos of 70 transects across slopes and rock walls in the National Park at depths between 7-105 m. Species records were extracted from the movies using Yolov8 model, while depth information was extracted with the easyOCR python package from the ROV video overlays. Original videos are archived and accessible at Tjärnö Marine Laboratory’s (Univerisity of Gothenburg). The analysis was performed using the Swedish platform for subsea image analysis (www.subsim.se). We acknowledge the support of the technical officers and ROV pilots at Tjärnö Marine Laboratory, in particular Tomas Lundälv, Roger Johannesson, and Joel White.

Data Records

The data in this occurrence resource has been published as a Darwin Core Archive (DwC-A), which is a standardized format for sharing biodiversity data as a set of one or more data tables. The core data table contains 73,098 records.

This IPT archives the data and thus serves as the data repository. The data and resource metadata are available for download in the downloads section. The versions table lists other versions of the resource that have been made publicly available and allows tracking changes made to the resource over time.

Versions

The table below shows only published versions of the resource that are publicly accessible.

How to cite

Researchers should cite this work as follows:

Nilsson C, Anton V, Burman E, Germishuys J, Obst M (2025). Koster historical biodiversity assessment. Version 1.10. Wildlife.ai. Occurrence dataset. https://doi.org/10.15468/rzhmef

Rights

Researchers should respect the following rights statement:

The publisher and rights holder of this work is Wildlife.ai. This work is licensed under a Creative Commons Attribution (CC-BY 4.0) License.

GBIF Registration

This resource has been registered with GBIF, and assigned the following GBIF UUID: 51d0bd32-e215-45ea-a04d-47a474336125.  Wildlife.ai publishes this resource, and is itself registered in GBIF as a data publisher endorsed by GBIF New Zealand.

Keywords

Occurrence; Machine Observations; YOLOv8; Benthic Invertebrates; Koster; Hard Substrate; Marine Biology; Marine Biodiversity; Marine Ecology; Subtidal Zone; Observation; Machine Observations; YOLOv8; Benthic Invertebrates; Koster; Hard Substrate; Marine Biology; Marine Biodiversity; Marine Ecology; Subtidal Zone

Contacts

Christian Nilsson
  • Metadata Provider ●
  • Originator ●
  • Point Of Contact ●
  • Principal Investigator
  • Researcher
University of Gothenburg
  • Anders Zornsgatan 34B
412 72 Gothenburg
Västra Götaland
SE
  • 0730682795
Victor Anton
  • Publisher
Emil Burman
  • Programmer
Jurie Germishuys
  • Programmer
Matthias Obst
  • Owner
Joel White
  • Point Of Contact
  • Research Engineer
University of Gothenburg, Tjärnö Marine Laboratory
  • Laboratorievägen 10
452 96 Strömstad
Västra Götaland
SE
  • +46 31-786 96 03
Victor Anton
  • Publisher
  • GM
Wildlife.ai
NZ
Emil Burman
  • Programmer
  • Researcher
University of Gothenburg
  • Medicinaregatan 7B
413 90 Gothenburg
Västra Götaland
SE
Jurie Germishuys
  • Programmer
  • Data Scientist
Combine Control Systems AB
  • Västra Hamngatan 8
411 17 Gothenburg
Västra Götaland
SE
Matthias Obst
  • Owner
  • Researcher
University of Gothenburg
  • Medicinaregatan 7B
413 90 Gothenburg
Västra Götaland
SE
  • +4676-618 38 27

Geographic Coverage

Data collected from the area west of the island of Yttre Vattenholmen. For additional information, feel free to contact authors.

Bounding Coordinates South West [58.9, 11.1], North East [58.9, 11.1]

Temporal Coverage

Start Date / End Date 1997-08-27 / 2023-10-09

Project Data

Biodiversity assessment of distribution size and relative abundance of 17 unique benthic invertebrate taxa. Assessment made by applying a YOLOv8 model trained on image data from ROV footage of the study site to 70 ROV transects from 1997-2023.

Title Depth Learning - Using Deep-Learning Object Detection Software to Investigate Spatiotemporal Vertical Ecological Trends on a Submarine Canyon Wall in Northern Skagerrak
Study Area Description Rock walls and slopes from 7-105m in the area west of the island of Yttre Vattenholmen.

The personnel involved in the project:

Christian Nilsson
  • Principal Investigator
Victor Anton
  • Publisher
Emil Burman
  • Programmer
Jurie Germishuys
  • Programmer
Matthias Obst
  • Owner

Sampling Methods

Sampling was performed by ROV from Tjärnö Marine Laboratory. Transects were taken for various purposes and are non-standardized. Thus, time spent at each depth and distance to substrate may vary. Transects are defined as consecutive filming of the study site until departure and may have depths removed if the ROV was not filming the habitat of interest for this study (hard substrate) at these depths. Additionally, inconsistency may exist between substrate depths and ROV depths where the ROV was not filming perpendicular to the seafloor. For further information, contact Joel White or Christian Nilsson.

Study Extent Archived ROV footage from the area west of Yttre Vattenholmen was utilized. All footage available with sufficient depth information was utilized. Footage was collected for various purposes, therefore sampling frequency & temporal resolution varies.
Quality Control Average maxcount per depth (organismQuantity) was added to provide insight regarding false positives. If individualCount for an observation is significantly higher than organismQuantity a false positive may be possible. For further information, contact Christian Nilsson.

Method step description:

  1. ROV depth for each image frame of videos was extracted from the video overlay using the EasyOCR python package. Depth was connected to YOLOv8 model observations through frame number in R, after which maximum and mean individual count was summarized from each transect for each taxon.

Bibliographic Citations

  1. Nilsson C. L., Germishuys J., Burman E., Anton V., White J., and Obst M. (2024). Koster historical invertebrate model - SUBSIM 17tx. Zenodo. https://doi.org/10.5281/zenodo.13589902 https://doi.org/10.5281/zenodo.13589902
  2. Nilsson C. L., Faurby S., Burman E., Germishuys J., and Obst. M. 2025. “ Applying Deep Learning to Quantify Drivers of Long-Term Ecological Change in a Swedish Marine Protected Area.” Ecology and Evolution 15, no. 9: e72091. https://doi.org/10.1002/ece3.72091. https://doi.org/10.1002/ece3.72091
  3. Nilsson C. L. (2025). ShrimpFather7/Koster-Deep-Learning-Ecology. Zenodo. https://doi.org/10.5281/zenodo.15249144 https://doi.org/10.5281/zenodo.15249144

Additional Metadata

Acknowledgements
Alternative Identifiers 51d0bd32-e215-45ea-a04d-47a474336125
https://ipt.gbif.org.nz/resource?r=koster_historical_assessment