6). This counters the amplification of the sink regions just to the north. MERRA forcing produces the smallest sink in the North Pacific and North Atlantic basins (Fig. 5). The weaker sink in the North Pacific can be attributed to a source region see more east of the Sea of Okhostk

(Fig. 6), and the North Atlantic to a local source in the Labrador Sea. MERRA-estimated fluxes in these two basins is about 0.15 mol C m−2 y−1 (39%) lower in the North Pacific than the strongest sink and 0.33 mol C m−2 y−1 (21%) lower in the North Atlantic. The strongest sink in both cases is produced by NCEP2. In the tropical basins, the estimates of air–sea carbon fluxes by NCEP2 produce the strongest source in 3 of the 4 major basins (Fig. 5). Sometime this is closer to the in situ estimates relative to the other forcings, as in the Equatorial Atlantic, and sometimes it is a larger departure, as in the Equatorial Indian. The large source represented by NCEP2 forcing in the Equatorial Pacific is derived from a very strong local flux along the Peru coast (Fig. 6). Although a smaller manifestation appears in NCEP1 and ECMWF forcing, it does not appear in MERRA-forcing, PLX4032 which leads to its representation of

the smallest Equatorial Pacific source. ECMWF departs strongly from the other forcings in the North Indian, and is nearly 3 times the fluxes estimated by the lowest reanalysis (NCEP1), but is closer to the in situ estimates (Fig. 5). This stronger source can be attributed

to local intensification offshore of Somalia (Fig. 6), which feature is either much smaller in the other forcings (NCEP1) or non-existent (MERRA and NCEP2). Estimates of FCO2 in Reverse transcriptase the sub-polar basins are more similar among the forcings than the high latitudes and tropics (Fig. 5), exhibiting the lowest ranges of estimates of all the basins. ECMWF is the strongest sink in 4 of the 5 basins, while MERRA forcing is the lowest in 2 basins (North Central Pacific and Atlantic). All the forcings indicate a much stronger sink estimate in the South Atlantic and Pacific than the in situ estimates. Global area-weighted mean partial pressures show similar relationships among the four reanalysis forcings and with the data (Fig. 7). The deviations from data are much smaller than the flux estimates: all are within 1% of data global means, with ECMWF the outlier at 0.6%. NCEP1 pCO2 is closest to the data, with a difference < 1 μatm, or −0.1%. All forcings also show positive and statistically significant correlations across basins, with values similar to the fluxes. On basin scales the pCO2 mean differences between the forcings and data are smaller, and more consistent with one another than for the basin fluxes (Fig. 7). The South Atlantic is a notable exception, which exhibits a departure from the data for all forcings similar to the fluxes. NCEP2 forcing is noticeably closer to the data pCO2 but it is still low by 26 μatm (about 7%).