| Summary: | Self organizing maps (SOMs) have been used to estimate seawater fugacity of CO2 (fCO2) distribution in the Nordic Seas and the northern North Atlantic Ocean for 2005-2007. Four maps were produced for each region using various combinations of the parameters MLD, SST, CHL and SSS to train the SOM. The combination of MLD+SST gave the most realistic fCO2 map for all the basins, which was surprising as one would expect that more input parameters would generate more realistic maps. SOMs were estimated in the northern North Atlantic Ocean and the Nordic Seas, the northern North Atlantic Ocean, and the Nordic Seas. The SOMs for the Nordic Seas were also merged with the SOMs for the northern North Atlantic Ocean, which generated better results than the SOM for the whole study area. The merged SOMs had minimal issues with discontinuity. All the SOMs generated incorrect fCO2 values for the deepest mixed layer depths in the Greenland Sea. A theory is that this may be the result of an artifact in the training dataset. This illustrate the importance of having a broad data coverage when using the SOM technique. The most optimal SOM for the Nordic Seas provided realistic estimates of fCO2 distribution. This SOM had an uncertainty of 12.5 uatm, which is close to the LSCOP target of 10.8 uatm for estimation of net fCO2 flux for the Northern North Atlantic. Master i Meteorologi og oseanografi MAMN-GEOF GEOF399
|
|---|