Inversion of acoustic zooplankton measurement for adaptive physical-biological ocean forecast

The Poseidon project is aimed at bringing multi disciplinary oceanographic data together on an Information Technology backbone in real-time, for improved understanding and forecasting. In this framework zooplankton acoustic backscatter is needed for better biology understanding, and can in turn benefit from the input of physical and biological models. Zooplankton backscatter models are subdivided in three categories: fluid-like, hard elastic shells, and gas bladder animals. Zooplankton species neither dominant in number, size or biomass can overwhelm part of the acoustic target strength spectrum, implying a necessary species-specific measurement. Furthermore, the too few high frequencies sampled by available sonars leave the acoustic inversion widely underdetermined. Real data inversion from WHOI's BiomaperlI has provided plankton population estimation comparable to what nets data and counting had recorded. Multiple species acoustic inversion has been demonstrated with the fluid-like and the elastic-shelled model. Purely acoustic field data inversion would require unjustifiable assumptions and lead to unbearable levels of uncertainty, which have always been reduced by cameras and labor-intensive direct tows. While other methods remain necessary to validate large-scale acoustic data, the Poseidon project's adaptive modeling, sampling, and the automatic input of biological information as part of data assimilation could significantly reduce acoustic uncertainty. Other issues addressed are acoustic inversion simulation behavior with various target sizes, the inversion's probabilistic validation, multiple species inversion, bubbles detection, application on WHIIG's BiomaperlI data, sources of error and adaptive modeling.