Using a Mathematical Model of Artificial Neural Networks to Analysis CO2 Exchange across the Sea Surface in the Arctic Ocean

The research hypothesis was to determine that the air-sea CO₂ fluxes depend largely on the difference in partial pressure of CO₂ (pCO₂) in seawater and air, than on gas transfer velocity. For this we used Artifical Neural Networks (ANN's), a powerful non-linear modeling tool for mapping performance, and tested those methods for so spare and difficult area-which is the Arctic Ocean. The ANN's were first used 20-30 years ago (Kohonen, 2001) and since then there has been a significant increase in their application to solving environmental problems. They are now commonly used in atmospheric, oceanography, and meteorology sciences.

While atmospheric pCO₂ shows relative homogeneity, marine pCO₂ varies strongly both temporarily and spatially, wherefore proving the hypothesis required achievement of the objective that was: estimate of marine pCO₂ distribution dependent on the sea surface temperature, sea surface salinity, biological activity, gas transfer velocity, and/or wind speed. The analysis was based on data from 2013-2017 year collected from two sources: publicity available satellite observation (like SOCAT, ESA/GlobWave Altimeret, NCEP/NCAR reanalysis, GlobColour GSM), and archives of Air-Sea Interaction Laboratory IO PAS (in-situ observation using LI-COR 7550 obtained during 5 Arctic cruises operated by Air-Sea Interaction Laboratory IO PAS, from r/v Oceania).

High uncertainty about the size of the sink/source of CO₂ in the Arctic Ocean reflects the storage of coordinated in situ measurements and difficulties in providing logistic support in a hostile environmental, wherefor the coverage of in sity pCO₂ measurements in the North Atlantic, still remains unevenly distributed in time and space. A potential alternative solution is to provide Earth Observation data. The Arctic Ocean is one of the strongest sinks for CO₂ in the world’s ocean throughout the year with seasonal variability, and so plays an important role in the carbon cycle. Despite of that, there are a lot of speculations over future trend of the carbon dioxide. Estimates of the current CO₂ flux, for this region, still are burden with high uncertainty, given that there are competitive processes to decreasing of increasing the rate of exchange. The magnitude of the ocean sink can be determine using air-sea flux estimates based on satellite observation as well as in situ measurements.

This work was financed by the National Science Centre grant ID number 2016/21/N/ST10/00387 and co-financed by Centre of Polar Studies "POLAR-KNOW" (a project of the Polish Ministry of Science).