- This event has passed.
Neural networks and the seawater CO2 system. From the global ocean to the Ría de Vigo
Daniel Broullón, from the Oceanic Processes in Global Change Group at IIM-CSIC, will defend his doctoral thesis at the University of Vigo (DOMAR Programme), in person, in the new graduation hall of the Faculty of Marine Sciences at the University of Vigo.
The thesis was co-supervised by Fiz Fernández and Rosa Reboreda, from the Institute of Marine Research of the CSIC.
ABSTRACT
Earth’s climate has changed many times throughout geological history, driven by interactions among the atmosphere, geosphere, hydrosphere, cryosphere and biosphere. Today, the concentration of greenhouse gases such as carbon dioxide, methane and nitrous oxide has increased sharply since the Industrial Revolution due to the burning of fossil fuels. One immediate consequence is the warming of the planet, but about a quarter of the carbon dioxide emitted to the atmosphere dissolves in the ocean, altering seawater chemistry and causing ocean acidification.
Understanding ocean acidification requires knowledge of the seawater CO2 system, including dissolved inorganic carbon, total alkalinity, partial pressure of CO2 and pH. These variables are essential to quantify the ocean carbon cycle, the exchange of CO2 between atmosphere and ocean, the buffering capacity of seawater and the vulnerability of marine ecosystems. International datasets compiled over recent decades make it possible to produce global climatologies and assess changes from the open ocean to coastal systems.
This doctoral thesis contributes to a better understanding of acidification at both global and local scales. At the global level, it develops climatologies for key variables of the seawater CO2 system, such as total alkalinity and total dissolved inorganic carbon, using neural networks trained with international oceanographic databases. These products improve previous spatial and temporal resolution and provide robust tools for biogeochemical modelling and other studies.
The first part of the thesis designs a global climatology of total alkalinity and a neural network able to estimate this variable anywhere in the ocean with low error. The second part develops an equivalent climatology for total dissolved inorganic carbon, combining GLODAP and Lamont-Doherty datasets and validating the model against independent observations and ocean time-series stations. Together, these products provide monthly fields that help describe the marine carbon cycle and support estimates of other CO2 system variables.
The regional part of the thesis focuses on the Ría de Vigo, a coastal environment of high ecological and economic importance. Neural networks were trained to reconstruct weekly time series of pH and total alkalinity from 1992 to 2019 at several stations and depth ranges. The results show strong short-term variability, clear spatial differences between the inner and outer ría, and long-term pH trends indicating acidification rates higher than those typically observed in the open ocean.
The final part of the thesis analyses hydrogen ion concentration and aragonite saturation state in the Ría de Vigo to identify the processes driving variability in the local CO2 system. Seasonal and long-term changes are controlled by photosynthesis and respiration, carbonate precipitation and dissolution, freshwater inputs, solar radiation, anthropogenic CO2 uptake and remineralisation of organic matter. The magnitude of the observed trends highlights the need for action to protect the ecological value and economic viability of the Ría de Vigo.