Improved air-sea CO2 flux estimates from sailboat measurements

This repository contains files and MATLAB scripts used to generate figures for the paper titled "Improved air-sea CO2 flux estimates from sailboat measurements". The study shows through an observing system simulation study with different sampling schemes how integrating sailboat data improves global air-sea CO2 flux estimates. Overview We used fCO2 data sourced from HAMOCC (HAMburg Ocean Carbon Cycle model) model coupled to the ocean general circulation model MPIOM, contributing to GCB simulation at a monthly 1x1 degree resolution from 1982 until the end of 2021, along with other environmental variables. We used the model dataset as our starting point and applied different sailboat subsampling schemes to get pseudo-observations. The sailboat tracks are We used tracks from the IMOCA 60 sailboats “Seaexplorer-Yacht Club de Monaco” (until 2019 “Malizia”) and “Newrest - Art & Fenêtres” (now “Nexans - Art & Fenêtres”) during offshore sailing and training events from 2018 to 2021, available via SOCATv2022 (www.socat.info). We applied the 2-step neural network method SOM-FFN (Landschützer et al., 2013) to reconstruct air-sea CO2 flux estimates. Six sets of air-sea CO2 flux estimates were reconstructed, each with 40 ensemble members to account for random errors in the neural network. These are the datasets used in this study: Data X) - “truth”: Ground truth fCO2 values from the HAMOCC modelData A) - “existing sailboat”: Subsampled model fCO2 (X) generated using the SOCATv2022 sampling scheme, including sailboat data.Data B) - “without sailboat”: Subsampled model fCO2 (X) generated using the SOCATv2022 sampling scheme, excluding sailboat dataData C) - “3 circumnavigations”: Subsampled model fCO2 (X) generated using the SOCATv2022 sampling scheme including sailboat data and 3 Antarctic circumnavigations (Vendée Globe; 2012/13, 2016/17, 2020/21) (from which only the latter one is from the existing track and also included in A; the original track was used to subsample earlier years)Data D) - “2 different circumnavigations”: Subsampled model fCO2 (X) generated using the SOCATv2022 sampling scheme including sailboat data and 2 different round-the-world races (Vendée Globe 2020/21 and the preliminary The Ocean Race 2017/18) (for the latter one we used existing tracks from 2023)Data E) - “3 circumnav + meas. uncert.”: Data C, but with an added random measurement uncertainty ranging from +- 5µatmData F) - “3 circumnav + pos. offset”: Data C, but with an added + 5µatm measurement offset These air-sea CO2 flux density estimates are 4-dimensional, represented in mol C m-2 yr-1, with the following dimensions:- Ensemble size of flux reconstructions (40)- Time (months from 1982 to 2021, totaling 480 months)- Latitude (1-degree intervals, totaling 180 latitudes)- Longitude (1-degree intervals, totaling 360 longitudes) There are the following folders and MATLAB codes in this repository:- figures (contains codes for generating the figures in the manuscript)- functions (contains functions needed for generating the figures)- mat-files (contains MATLAB files needed for generating the figures)- netcdf (contains NetCDF files needed for generating the figures)- scripts (contains codes to generate the MATLAB and NetCDF files needed for generating the figures)