About my research
My PhD is part of the Asia-Africa BlueTech Superhighway (AABS) project, which is coordinated by WorldFish and funded by UK International Development under the UK’s Climate and Ocean Adaptation and Sustainable Transition (COAST) program of the Blue Planet Fund (BPF). Kenya, Tanzania, Mozambique, among other countries, are at the focus of the first phase (Phase 1) of the project. During this phase, “Work Package 1: Digital Coasts” focusses on creating digital information tools to collect data on Small-Scale Fisheries (SSF) and ocean conditions for assessment, monitoring, and prediction purposes. My PhD is integrated into this work package, and I will be contributing to (i) the establishment of an ocean model (using SINMOD from SINTEF) for the Western Indian Ocean (WIO) covering Kenya, Tanzania, and Mozambique, and to (ii) the development AI-powered tools for SSF of the above-mentioned African countries.
At the core of the fisheries part of the project lies the Peskas framework developed by WorldFish, which has been validated for SSF of Timor-Leste and is currently being scaled up in the regions of interest for the AABS project. Peskas consists in solar-powered GPS tracking devices (developed and produced by Pelagic Data Systems - PDS), that are deployed on artisanal fishing vessels, and of “enumerators” (local trained and untrained personnel) that collect information of the landings when the vessels return to shore. These data are collected using Kobo forms on tablets and typically include photographs of the catches. This system produces vast amounts of data, but this data can be unprecise, inconsistent, or incomplete due to the complex nature of these SSF and to regional differences in gear types, species names, language, and other factors. Consequently, there is a need to develop innovative tools to automate data collection processes, improve accuracy and efficiency, analyse the data collected, and provide useful predictions for management purposes. Machine learning (ML) algorithms, particularly Deep Learning (DL), are powerful and continuously improving tools for time-series analysis and classification tasks, and their applications for fisheries monitoring and management are growing rapidly. During this PhD I aim to develop models to identify fishing activity and gear utilization from GPS tracks, as well as identifying fish species from photographs of the catches.
Simultaneously to the fisheries-related projects, SINTEF Ocean will establish a numerical ocean model based on the SINMOD architecture for the WIO. This model includes both physical and biological components and can produce predictions of multiple physical and biological metrics for up to a few days ahead. The ocean model serves different purposes within the AABS project. Firstly, it is meant as a tool that fishers and stakeholders can access and utilize freely for their activities. For example, they might be interested in current patterns, tides, and Sea Surface Temperature (SST), which could be relevant in their decisions of when and where to fish. Secondly, its output can be used to extract and test forecasters of Catch Per Unit Effort (CPUE). Thirdly, the high-resolution oceanographic data produced by the model will be used (in collaboration with Macquarie University) to study the physical and biological processes connecting different areas and fishing grounds in the WIO. These processes, collectively known as “connectivity”, are crucial for fish populations and marine communities and must therefore be considered in fisheries management to promote sustainability. After completing the operational setup of the model, SINTEF will integrate data-assimilation capabilities to utilize observational data for fine-tuning and correction of biases.
The overarching goal of the PhD project, and an important milestone of the AABS project, is to integrate the information generated by the Peskas fisheries monitoring framework and the ocean model predictions into a digital toolbox that local fishers and managers can access to look at fisheries data and optimize fishing activities, while ensuring sustainable practices. This toolbox will be automated through the implementation of AI and feedback algorithms. As a result, this system will be dynamic and constantly improving based on its performance and the feedback provided. Fishers will be able to use this system to plan their fishing activities, while managers will be able to monitor fishing activities, weigh the impact of management measures (limiting effort, restricting access to given areas or gears, etc.), and use this information for marine spatial planning.
Participating organizations: WorldFish, Wildlife Conservation Society, NTNU, Pelagic Data Systems.
Affiliated organizations: KMFRI, KeFS, ZAFIRI, TAFIRI, InOM, IDEPA.
