Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

A Changing Foraminiferal Landscape in the Öresund: Cataloguing the Potentially Invasive Nonionella sp. T1 and Assessing Deep Learning for Species Identification

Pålsson, Malin LU (2026) In Dissertations in Geology at Lund University GEOR11 20261
Department of Earth and Environmental Sciences (MGeo)
Abstract
Benthic foraminifera are valuable environmental indicators, as species assemblages can vary depending on their surrounding marine conditions. Previous studies have reported the potentially invasive Nonionella sp. T1 along the Swedish west coast, a non-indigenous species (NIS) with abilities to denitrify and adapt to varying environments. In this thesis, with the aim of studying foraminiferal species assemblages, the southernmost finding of Nonionella sp. T1 in the region was observed in sediment collected south of Ven in the Öresund. There was a rapid increase of Nonionella sp. T1 in the shallower levels of the studied sediment core, alongside shifting abundances of the general species assemblages. In addition to Nonionella sp. T1, low... (More)
Benthic foraminifera are valuable environmental indicators, as species assemblages can vary depending on their surrounding marine conditions. Previous studies have reported the potentially invasive Nonionella sp. T1 along the Swedish west coast, a non-indigenous species (NIS) with abilities to denitrify and adapt to varying environments. In this thesis, with the aim of studying foraminiferal species assemblages, the southernmost finding of Nonionella sp. T1 in the region was observed in sediment collected south of Ven in the Öresund. There was a rapid increase of Nonionella sp. T1 in the shallower levels of the studied sediment core, alongside shifting abundances of the general species assemblages. In addition to Nonionella sp. T1, low abundances of two other non-indigenous species were observed, Ammonia confertitesta and Trochammina hadai. The presence of three non-indigenous species of benthic foraminifera further emphasized the need for close monitoring of marine conditions and marine sediments in the Öresund.
Manual identification of foraminifera is both time-demanding and labour-intensive, hence automation by machine learning methods could potentially improve efficiency. Application of accurate deep learning models for species identification would enable quicker and more extensive analyses of sediment material and the foraminifera content therein. Without any additional training, an already existing deep learning model trained to identify 29 species of benthic foraminifera was assessed on species identification on foraminifera from the sediment samples collected in the Öresund. The performance of the You Only Look Once (YOLO) v7x model was tested on both foraminifera and sediment particles in varied imaging conditions. The accuracy of the model’s output data varied depending on magnification, brightness, focus, and pixel resolution. There was no universal optimal imaging setup, which made direct application on species identification of unpicked samples challenging. However, the YOLOv7x model showed great potential as a first step in automated identification. (Less)
Please use this url to cite or link to this publication:
author
Pålsson, Malin LU
supervisor
organization
course
GEOR11 20261
year
type
H2 - Master's Degree (Two Years)
subject
keywords
benthic foraminifera, non-indigenous species, Nonionella sp. T1, species identification, machine learning, deep learning, marine sediments
publication/series
Dissertations in Geology at Lund University
report number
735
language
English
id
9247052
date added to LUP
2026-07-29 14:18:45
date last changed
2026-07-29 14:18:45
@misc{9247052,
  abstract     = {{Benthic foraminifera are valuable environmental indicators, as species assemblages can vary depending on their surrounding marine conditions. Previous studies have reported the potentially invasive Nonionella sp. T1 along the Swedish west coast, a non-indigenous species (NIS) with abilities to denitrify and adapt to varying environments. In this thesis, with the aim of studying foraminiferal species assemblages, the southernmost finding of Nonionella sp. T1 in the region was observed in sediment collected south of Ven in the Öresund. There was a rapid increase of Nonionella sp. T1 in the shallower levels of the studied sediment core, alongside shifting abundances of the general species assemblages. In addition to Nonionella sp. T1, low abundances of two other non-indigenous species were observed, Ammonia confertitesta and Trochammina hadai. The presence of three non-indigenous species of benthic foraminifera further emphasized the need for close monitoring of marine conditions and marine sediments in the Öresund.
Manual identification of foraminifera is both time-demanding and labour-intensive, hence automation by machine learning methods could potentially improve efficiency. Application of accurate deep learning models for species identification would enable quicker and more extensive analyses of sediment material and the foraminifera content therein. Without any additional training, an already existing deep learning model trained to identify 29 species of benthic foraminifera was assessed on species identification on foraminifera from the sediment samples collected in the Öresund. The performance of the You Only Look Once (YOLO) v7x model was tested on both foraminifera and sediment particles in varied imaging conditions. The accuracy of the model’s output data varied depending on magnification, brightness, focus, and pixel resolution. There was no universal optimal imaging setup, which made direct application on species identification of unpicked samples challenging. However, the YOLOv7x model showed great potential as a first step in automated identification.}},
  author       = {{Pålsson, Malin}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Dissertations in Geology at Lund University}},
  title        = {{A Changing Foraminiferal Landscape in the Öresund: Cataloguing the Potentially Invasive Nonionella sp. T1 and Assessing Deep Learning for Species Identification}},
  year         = {{2026}},
}