Advanced fish biomass estimation using acoustic data and machine Learning

09.01.2025

Sustainability in the fish mapping industry involves using advanced technologies and methods to ensure that fishing practices are environmentally responsible and do not deplete fish populations.
For sustainability initiatives to succeed, they must be competitive in the market like any other business interest. By doing so, sustainability efforts will be less reliant on government subsidies and the support of non-governmental organisations (NGOs).

Close up of monitors for fishmapping.

Sustainable fish mapping and biomass estimation

Sustainovate is a European consultancy firm that promotes sustainable marine business development and innovation by bridging the gap between science, industry, and government. It advocates data-driven technologies for industrial fisheries, offering tools for autonomous data collection, processing, analysis, and collaboration using cloud computing, IoT, and AI. The firm specialises in managing complex data types, such as hydroacoustic and visual imagery data, for this purpose.

Technical and scientific challenges

The main challenge here is demonstrating the added value of a fish-mapping model in estimating fish biomass based on acoustic recordings during commercial fisheries on North Sea herring.

Complex tools are needed to analyse 5+TB of continuously collected acoustic data in real-time. We investigate the feasibility of realizing the complex analysis chains using a workflow engine, which can facilitate human interaction and collaborations with third parties.

Methodology

We attempted to explain the fish biomass using machine learning methods. We chose to estimate the biomass of herring in the limited area of longitudes within [-3, 3] degrees and latitudes within [56, 62] degrees north from July 1st to August 30th of each year during a period of 7 years. We compare two spatiotemporal datasets stemming from:

  • acoustic measurements from the trawlers where the measurements cover our area of interest, and they are quite homogeneously distributed in space.
  • acoustic measurements from the scientific surveys where the measurements are biased towards the fishing grounds and do not sample equally the area of interest.

Scientific impact

The study develops a first glance at the idea of giving biomass estimates using statistical methods. Thus, we provide a comparison in terms of total estimated biomass for each year between scientific and commercial vessel observations. The timespan corresponds to fishing seasons (July-August) for 7 years in the northern North Sea. A Python-based framework has been put in place to easily integrate more upcoming data.

This project will improve Sustainovate’s scientific toolbox and expertise in the field of provision of annual fish biomass indices. It will increase the company’s competitiveness in that sector and will contribute significantly to sustainable fisheries.

Key benefits

  • Framework comparing Biomass information from Fishing fleet and Scientific survey
  • Reusability
  • Improved data analysis
  • Gain of expertise in Biomass estimation in the context of Marine science
Bilde
Sustainovate logo.

Established in 2008 to address the need for business- and market-oriented marine research in the Dutch fishing industry, Sustainovate has operated as a limited company since 2012, registered in Oslo and active in both Norway and the Netherlands.