Periodic Reporting for period 2 - ANTICS (Advancing Novel imaging Technologies and data analyses in order to understand Interior ocean Carbon Storage)
Okres sprawozdawczy: 2023-04-01 do 2024-09-30
Photosynthesis in the ocean converts approximately 100 Gt of carbon dioxide (CO2) into organic matter every year, of which 5-15% sinks to the deep ocean. The depth to which this organic matter sinks is important in controlling the magnitude of ocean carbon storage, as changes in this flux attenuation depth drive variations in atmospheric pCO2 of up to 200 ppm. Efforts to produce global maps of flux attenuation have yielded starkly contrasting global patterns, blocking our understanding of ocean carbon storage and our ability to predict it. The bottleneck is our ignorance of the spatiotemporal variability of the processes that control flux attenuation.
ANTICS will directly address this knowledge gap by using an innovative synthesis of cutting-edge in situ imaging, machine learning and novel data analyses to mechanistically understand ocean carbon storage. Using state-of-the-art imaging technologies, I will collect data on the size, distribution and composition of organic matter particles and measure their sinking velocity in the upper 600 m across the Atlantic. I will design a neural network model that allows the conversion of in situ images into carbon fluxes, and develop analysis routines of particle size spectra that quantify the processes causing flux attenuation: remineralisation, physical aggregation/disaggregation, fragmentation/repackaging by zooplankton. By statistically linking these outputs to seasonality, depth, primary production and temperature, I will be able to determine which processes dominate under specific environmental conditions. This step change in our understanding will allow ANTICS to resolve flux attenuation spatially and temporally. I will use this pioneering knowledge to validate and inform the parametrization of the marine biogeochemical component of the UK’s Earth system model used for carbon cycle forecasting in the next IPCC assessments.
ANTICS will directly address this knowledge gap by using an innovative synthesis of cutting-edge in situ imaging, machine learning and novel data analyses to mechanistically understand ocean carbon storage. Using state-of-the-art imaging technologies, I will collect data on the size, distribution and composition of organic matter particles and measure their sinking velocity in the upper 600 m across the Atlantic. I will design a neural network model that allows the conversion of in situ images into carbon fluxes, and develop analysis routines of particle size spectra that quantify the processes causing flux attenuation: remineralisation, physical aggregation/disaggregation, fragmentation/repackaging by zooplankton. By statistically linking these outputs to seasonality, depth, primary production and temperature, I will be able to determine which processes dominate under specific environmental conditions. This step change in our understanding will allow ANTICS to resolve flux attenuation spatially and temporally. I will use this pioneering knowledge to validate and inform the parametrization of the marine biogeochemical component of the UK’s Earth system model used for carbon cycle forecasting in the next IPCC assessments.
During three research cruises across the Atlantic, including a transect from the Falkland Islands to the UK, we collected particle size spectra, particle and plankton images, particle flux measurements, and turbulence data. Additionally, we identified supplementary datasets for inclusion into our analyses. A new Red Camera Frame and data logger system was designed, tested, and successfully deployed on these cruises. A detailed manual for operating the Red Camera Frame and data logger, including step-by-step instructions and troubleshooting tips, was also developed to ensure consistent data collection and ease of use by other researchers. We intercalibrated the three imaging systems on the Red Camera Frame to compile continuous size spectra, with this work also being readied for submission to Frontiers in Marine Science. We also developed and scientifically deployed a new Marine Snow Catcher, named 'Yuki,' and this work is being prepared for submission to a peer-reviewed journal. We conducted an in-depth analysis and published findings on the current state of knowledge regarding the relationship between sinking velocities and particle size. A camera system to measure in situ sinking velocities of particles was designed, manufactured and integrated into a complex mooring for deployment. An image processing workflow for holographic particle images was developed, resulting in two published conference proceedings and two manuscripts currently under preparation for submission to peer-reviewed journals. This effort has fostered collaboration with the University of Aberdeen and a business partner. A community survey on the use of in situ imaging of plankton for large-scale ecosystem assessment and policy decision-making was conducted, analyzed, and published in a peer-reviewed journal. We developed an unsupervised image classification algorithm for plankton and particle images, which is currently under revision for a peer-reviewed journal. Our analysis of flux profiles from the Southern Ocean indicates that diatoms are inefficient vectors of organic carbon to the deep ocean, contrary to current best knowledge; this work has been submitted for peer review. An analysis of high-resolution image profiles, which is currently prepared for submission, further shows that species-resolved taxonomic information is critical for understanding particle flux dynamics. Finally, our analysis of particle size spectra and turbulence profiles suggests that stronger aggregates form under turbulence, and this work is currently under review in a peer-reviewed journal.
So far, we have achieved significant progress beyond the state of the art through the development of novel instrumentation, advanced data workflows, and new scientific insights. We introduced an integration platform and data logger designed to synchronize the data streams of multiple in situ camera systems, enhancing the efficiency and use of underwater imaging. We developed a new Marine Snow Catcher design, and we designed and manufactured the “FluxCam”, a camera system specifically for measuring in situ sinking velocities of particles.
To complement these innovations, we established new data workflows that significantly enhance image processing and analysis. Our machine-learning-based approach for processing holographic particle images has improved the speed and accuracy of the analysis of holographic particle images. Furthermore, we developed an unsupervised image classification algorithm that identifies and categorizes various particle types without prior labeling, streamlining the data analysis process. We also created methods for compiling continuous size spectra from different in situ camera systems, allowing for comprehensive and coherent analyses of particle size distributions.
Our research has yielded novel scientific insights that challenge existing assumptions and contribute to a deeper understanding of ocean dynamics. We discovered that stronger turbulence results in flatter slopes of aggregate size distributions due to the rapid formation of large aggregates, with a shift in the size spectra slopes implying that strongly bonded small aggregates are formed with increasing turbulence.
Moreover, we found that diatoms in the Southern Ocean are not as significant for carbon transfer through the twilight zone as previously thought based on the ballast hypothesis. Our data show that diatom fluxes attenuate faster than organic carbon fluxes. By integrating ecological perspectives and high-resolution chemical measurements, we demonstrated that grazer repackaging and diatom buoyancy regulation decouple silica and carbon fluxes. This decoupling suggests that the presence of diatoms does not directly correlate with carbon sinking efficiency, addressing a key uncertainty in the modeling of ocean carbon storage.
We expect to further advance our understanding of ocean carbon storage through sinking particles by applying our new data analysis workflows on the data collected during the research cruises spanning the entire Atlantic Ocean.
To complement these innovations, we established new data workflows that significantly enhance image processing and analysis. Our machine-learning-based approach for processing holographic particle images has improved the speed and accuracy of the analysis of holographic particle images. Furthermore, we developed an unsupervised image classification algorithm that identifies and categorizes various particle types without prior labeling, streamlining the data analysis process. We also created methods for compiling continuous size spectra from different in situ camera systems, allowing for comprehensive and coherent analyses of particle size distributions.
Our research has yielded novel scientific insights that challenge existing assumptions and contribute to a deeper understanding of ocean dynamics. We discovered that stronger turbulence results in flatter slopes of aggregate size distributions due to the rapid formation of large aggregates, with a shift in the size spectra slopes implying that strongly bonded small aggregates are formed with increasing turbulence.
Moreover, we found that diatoms in the Southern Ocean are not as significant for carbon transfer through the twilight zone as previously thought based on the ballast hypothesis. Our data show that diatom fluxes attenuate faster than organic carbon fluxes. By integrating ecological perspectives and high-resolution chemical measurements, we demonstrated that grazer repackaging and diatom buoyancy regulation decouple silica and carbon fluxes. This decoupling suggests that the presence of diatoms does not directly correlate with carbon sinking efficiency, addressing a key uncertainty in the modeling of ocean carbon storage.
We expect to further advance our understanding of ocean carbon storage through sinking particles by applying our new data analysis workflows on the data collected during the research cruises spanning the entire Atlantic Ocean.