| Summary: | The ocean regulates the climate by annually absorbing roughly 25 % of anthropogenic CO 2 emissions from the atmosphere. In order to quantify the capacity of the ocean carbon sink from observations, measurements of the sea surface partial pressure of CO 2 (pCO 2 ) are essential. Building on the existing observational networks, we can utilize neural networks and other statistical methods, to interpolate data gaps in time and space creating homogeneous pCO 2 maps to estimate the exchange of CO 2 through the air-sea interface. However, uncertainties in these neural network interpolations are still substantial, particularly in less frequently monitored ocean regions such as the Southern Ocean. Trying to close existing data gaps, MPI is working with a novel, cost efficient and environmentally friendly fleet: sailboats. Sailboat pCO 2 has been regularly collected since 2018, however, their added value has not yet been quantified. Here, we quantify the added value and rate of improvement of underway pCO 2 data from such racing events by creating a twin of all available SOCAT observations, excluding data from sailboat races. We apply the SOM-FFN technique on all pCO 2 observations in SOCAT as well as the twin dataset and calculated the sea surface pCO 2 and subsequently the air-sea CO 2 exchange. By comparing the reconstructive air-sea CO 2 fluxes, we were able to quantify the difference, representing the added value of sailboat racing events. Our results show that the reconstructions on SOCAT and the twin dataset significantly differ in the air-sea CO 2 flux density on regional scales by up to 1.26 mol m -2 yr- 1 . 99 % of the significant differences fall below 0.40 mol m⁻² yr⁻¹.While differences are within the noise in many regions, significant differences can be detected in the less frequently monitored Southern Ocean, where pCO 2 data from single events, such as the Vendée Globe are added, as well as in the North Atlantic, where the majority of racing events took place. While the results after 5 years of data ...
|
|---|