Analysis of the spatio-temporal variability of pCO2 and air-sea flux of CO2 over the global coastal ocean during the last two decades: a satellite approach – CO2COAST

Coastal zones are under strong natural and anthropogenic pressures including climate change, which undoubtedly affect the quality of coastal waters essential for biogeochemical, ecological, and economic reasons. Economic, social and cultural impact could arise through more accurate and targeted blue carbon (i. e. the one captured by the world’s ocean) accounting and by raising awareness of the role of coastal waters in global warming. However, due to the relatively limited spatio-temporal coverage of in situ data, uncertainties in coastal carbon fluxes are such, that the CO2 flux in the coastal margins remains a poorly constrained term in global carbon budgets. Satellite remote sensing allows for the collection of various physical and biological parameters at global scales at different temporal resolutions not accessible from other observation methods. In this context, the objectives of CO2COAST are to estimate, for the first time, the global coastal waters CO2 flux from satellite observations at high spatial resolution, and assess and analyse its seasonal, inter-annual, and the trend over the 25 last years. At last, this unique data set will allow to scrupulously analyze the controversial respective contribution of estuaries and coastal shelf.

The existing approaches dedicated to the estimation of pCO2 from space, although convenient to develop, present a regional character that prevent applications to global scale, the main objective of CO2COAST. Only the recent model of Laruelle et al. (2017) deals with global coastal waters using Self Organizing Map (SOM) allowing to interpolate experimental data. However, their validation exercise reveals strong differences in the performance according to the considered coastal provinces, with large errors ranging from 20 to 53 µatm (typical error in open ocean are around 17 µatm, Landschützer et al., 2014). According to the authors, “This is likely because these coastal regions have complex biogeochemical dynamics and high frequency variability that cannot be fully captured with the current generation of data interpolation techniques using the limited available predictor data”. Moreover, and very importantly, this latter method, while global, only provides pCO2 maps at low spatial resolution (1/4 ° ~ 27 km) and is dependent on the sea surface salinity values which are still not well resolved in coastal waters from satellite (35-50 km of resolution). These products, provided at low spatial resolution, do not allow to properly account for the large heterogeneity of the coastal domain. For instance, it is common along eastern and western boundary currents to find continental shelves as narrow as 10–20 km. Additionally, biogeochemical fronts, very productive, associated with river plumes, coastal currents and upwelling areas are characterized by spatial scales of the order of tens of kilometers or even smaller. Therefore, high spatial resolution (about 1 km) has to be considered to properly account for the contribution of shelves and estuaries in the CO2 cycle. At last, the pCO2 data provided by Laruelle et al. (2017) is partly based on standard Chl satellite products which are subject to large uncertainties (up to 200%) in such complex bio-optical environments (Loisel et al., 2013), necessarily impacting the pCO2 retrieval.

The innovative aspect of CO2COAST mainly relies on its global character, its high spatial resolution, and on the use of a novel approach (Optical Water Class-based algorithms) to consider the physical and biogeochemical complexity and variety of coastal waters aiming at i) improving the retrieval of pCO2, and ii) estimating pCO2 in under-sampled coastal regions. Preliminary tests performed very recently at LOG strongly support the proposed methodology.