The relevant time scales in estimating the air–sea CO2 exchange in a mid-latitude region

A 1D biogeochemical model simulating the nitrogen and carbon cycles is validated, using continuous observations obtained with the Carioca buoy (sea surface temperature (SST), pCO2, fluorescence) at the DYFAMED station (NW Mediterranean Sea) in 1995–1997, as well as other in situ data. Although the average surface pCO2 is generally slightly over-saturated (8±8 μatm), the NW Mediterranean Sea is a weak sink for atmospheric CO2 (0.15±0.07 mol C m−2 yr−1), because the highest fluxes occur in winter and spring during the period of under-saturation when the winds are strong. The seasonal cycle is modulated by synoptic events (Mistral bursts), which can have a strong impact on the behaviour of the system and on the fluxes in winter and spring. While the atmospheric forcing constrains the annual balances and fluxes of tracers and CO2, the winter pre-conditioning of the bloom plays a major role in driving the interannual variability. Sensitivity analyses indicate that atmospheric forcing in this highly variable region should be averaged over periods of no longer than one week to simulate the biological and air–sea CO2 fluxes correctly. Also, because the model used is rather simple, this time scale is an upper limit. The sampling period must be no greater than a few days in order to estimate the CO2 fluxes with an error smaller than 20%. In the NW Mediterranean Sea, it is difficult to use “satellite proxies” (SST, sea colour) to estimate annual air–sea CO2 fluxes because SST does not vary much in winter and during the beginning of the bloom, while chlorophyll can either remain low (winter) or change rapidly (beginning of the bloom). At least several direct observations of pCO2 are needed during winter to define “initial conditions”.