Correction to “A joint atmosphere‐ocean inversion for surface fluxes of carbon dioxide: 1. Methods and global‐scale fluxes”

Estimates have been corrected so that river carbon flow is manifested as a land sink and a preindustrial ocean source of 0. 45 PgC yr 1 (see auxiliary material). Contemporary air-sea fluxes (F contemp = F preindust + F anthro) have been scaled to the 1992 -1996 period by assuming the anthropogenic component is proportional to the atmospheric concentration perturbation, but air-land fluxes are not scaled. Original estimates, uncorrected for river carbon and unscaled in time, are given in parentheses. ''IPCC-TAR 90s'' is the estimate of Prentice et al. 2001 for the 1990s; ''T3L1'' and ''T3L2'' are the TransCom3 control inversions for level 1 Gurney et al. , 2002 and level 2 Gurney et al. , 2004, respectively, both for the period 1992 -1996. ''Bopp O 2 /N 2 '' and ''Keeling O 2 /N 2 '' represent the oxygen analyses of Bopp et al. 2002 for the period 1990 -1996 and Keeling and Garcia 2002 for the 1990s, respectively. ''Joint'' is the current joint inversion for the period 1992 -1996. ''Gloor inverse'' is the previous ocean inversion of Gloor et al. 2003 scaled to 1992 -1996, a result nearly identical to that of McNeil et al. 2003 from CFC analysis. ''Matsumoto OCMIP'' is the summary of forward ocean carbon cycle simulations Orr et al. , 2001, as reported by Matsumoto et al. 2004, also scaled to 1992 -1996. ''Tak99'' and ''Tak02'' represent estimates based on the DpCO 2 analyses of Takahashi et al. 1999, 2002 respectively, using quadratic (''k u 2 '' Wanninkhof, 1992) and cubic (''k u 3 '' Wanninkhof and McGillis, 1999) gas transfer velocity parameterizations. ''MOM3 Forward'' are the 1992 -1996 fluxes from OCMIP2 biotic simulations for the five models of the MOM3 suite used in the present study (see Table