CoastalFutures - Future Scenarious for Sustainable Marine Ecosystems of Contested Marine Areas

Thünen young researchers presented their results at two international conferences. Tahereh Nakisa gave a talk about an online model library of larval fish at the 46th Larval Fish Conference that took place in Lisbon. Tahereh developed this library as a part of her PhD thesis. The library includes freely-available model code and data on larval stages of various fish species. Dr. Dario Fiorentino presented his findings about fish habitat modelling at the 5th International Symposium on “Effects of Climate Change on the World’s Ocean” in Bergen. This symposium brought together experts from around the world to better understand climate effects on ocean ecosystem and to elaborate possible adaptation and mitigation measures to ensure a sustainable future of our oceans. Abstract of both presentations are shown below.

“Online Model Library of larval fish: a tool to better understand larval survival in marine ecosystems”. T. Nakisa, A. Akimova, U. Daewel, I. Nunez-Riboni, C. Schrum

The stability and resilience of fish populations in the face of exploitation depends strongly on the annual recruitment of young-of-the-year. Understanding this phenomenon is crucial for the sustainable management of fish stocks. To achieve this understanding, previous studies have used individual-based models (IBMs), which simulate the behavior and interactions of individual fish and their environment. However, these models are often dispersed across various publications and are not easily accessible to the wider scientific community. To address this issue, we developed an online library of fish early life stages (ELSs), which brings together IBMs from various fish species, including Atlantic cod (Gadus morhua), Atlantic herring (Clupea harengus), European and sprat (Sprattus sprattus), among others. Our library provides comprehensive access to these models, including their summaries, codes, and quality control measures to to ensure their accuracy. Our Online Library of Fish ELSs not only features models but also includes current various data for model calibration and validation. One of goals is to develop a generic approach for larval fish IBMs suitable for various species. To achieve this, we review and compare the foraging behavior of different species to identify similarities and differences. Overall, our Online Library of Fish ELSs aims to provide a one-stop resource for researchers interested in larval IBMs as a powerful tool to investigate drivers of fish recruitment. By consolidating these models and data in one place, we hope to facilitate future research in this critical field and contribute to the sustainable management of fish populations.

“Improving selection of fish habitat models for climate studies”. D. Fiorentino, I. Núñez-Riboni, D. Oesterwind, M. E. Pierce, A. Akimova

Climate-induced shifts in fish distribution have been documented to trigger changes in marine ecosystems and their services. Species distribution modelling is widely used to study past and predict future species distribution as required by climate smart marine spatial planning to promote sustainable use of marine ecosystems. Traditional species distribution models rely on cross-validation (CV) techniques to evaluate model performance but rarely consider 1) model transferability, i.e. the ability to extrapolate to future conditions or area with no observations, and 2) the ability of CV to effectively choose ecological meaningful models.

To address these issues, we tested different types of abundance-environment relationships in a framework of generalized modeling applied to 12 commercial fish species. We designed three CV routines: 1) random-split CV, i.e. the traditional approach 2) space-blocking -CV, to control for sampling design and 3) niche-blocking CV, model fitting with optimal environmental conditions. Finally, we validated partial effects of temperature between models.

Our results reveal that the most widely used CV routine (random-split CV) often selects a species distribution model that lacks transferability (less suitable for extrapolation) as the best model. Therefore, to prevent erroneous predictions, we recommend to additionally verify the functional form of temperature partial effect to ensure model adherence with the principle of ecological significance and parsimony principles. We show that this step is critical for generating accurate projections of species distribution patterns in the face of future climate change scenarios.