Speaker
Description
Recruitment variability and population connectivity are central to the assessment and management of fish stocks, yet the environmental variability effect on the eggs and larvae dispersion and survival remains largely unresolved as it is not directly observable. In this contribution, we present our recent developments towards a Digital Twin of the Ocean (DTO) targeting the simulation of Early Life Stages (ELS) of different fish species, especially small pelagic fish, integrating observations, numerical model forcing and simulation outputs in a service layer in which FAIR principles are applied end-to-end.
The Digital Twin starts from annual egg-production surveys providing data on the spatial distribution of egg concentration during the peak spawning period. These observations are stored in a PostGIS database and are catalogued and served through a GeoNetwork. Numerical model output describing the physics and the biogeochemistry of the ocean are available from our IEO hydrodynamical ROMS model configuration coupled to a biogeochemical model configuration but also from other model products, namely MeteoGalicia and Copernicus Marine Service. Model results are stored following CF conventions in a THREDDS server and are also catalogued in the GeoNetwork, so that forcing sources are plug-and-play. Egg-production data is used to initialise Lagrangian model simulations and numerical models are used to force the Lagrangian particles through all the simulation period. Particle trajectories are computed with the offline trajectory model OpenDrift, which we have extended with a generic Individual Based Model (IBM) module that allows the simulation of physical processes and parameterised, configurable biological behaviour (temperature-dependent development, vertical migration, mortality) of the ELS of different fish species, though it is currently applied to species of commercial interest in Galicia (sardine, anchovy, hake and octopus). OpenDrift outputs are also CF compliant NetCDF which are stored and catalogued in the mentioned service-based infrastructure, so that they are findable, accessible and reusable as the observations that generated them.
The objective of the digital twin is to simulate several years of the evolution of early life stages of fish to shed light on the effect that the inter-annual variability of the environmental variables has on the recruitment and connectivity, which is relevant to stock assessment and management. Even though these simulations can run on local workstations, the use of the CESGA HPC infrastructure allows parallel simulations (independent runs covering several release years, parameter settings and different sets of particles released at the same time that are executed concurrently). In this presentation, we will show how this DTO can be used to illustrate inter-annual connectivity estimates and recruitment indicators, and to discuss how this workflow allows reproducible simulations that are transferable to other species and regions, and how what-if scenarios could be performed.