Carbon fluxes in subduction zones can be better constrained by including new estimates of carbon concentration in subducting mantle peridotites, consideration of carbonate solubility in aqueous fluid along subduction geotherms, and diapirism of carbonbearing metasediments. Whereas previous studies concluded that about half the subducting carbon is returned to the convecting mantle, we find that relatively little carbon may be recycled. If so, input from subduction zones into the overlying plate is larger than output from arc volcanoes plus diffuse venting, and substantial quantities of carbon are stored in the mantle lithosphere and crust. Also, if the subduction zone carbon cycle is nearly closed on time scales of 5-10 Ma, then the carbon content of the mantle lithosphere + crust + ocean + atmosphere must be increasing. Such an increase is consistent with inferences from noble gas data. Carbon in diamonds, which may have been recycled into the convecting mantle, is a small fraction of the global carbon inventory. Key carbon reservoirs and transport mechanisms can now be better quantified. These include carbon concentration ([C]) in altered mantle lithologies (this paper plus refs. 9 and 10), carbon solubility in aqueous fluids at subduction zone conditions (this paper plus refs. 11 and 12), the volume of altered peridotites in subducting oceanic plates (especially ref. 13), the volume of altered peridotite in the mantle wedge, and the nature of metasedimentary diapirs rising from subducting crust (14).In this paper, we reevaluate carbon fluxes in several tectonic settings. We start with carbon uptake during hydrothermal alteration near midocean ridges, followed by an estimate of carbon addition during alteration of shallow mantle peridotite at the "outer rise," where subducting oceanic plates bend before subduction. We then consider carbon transfer in fluids and melts derived from the subducting plate. Finally, we review carbon outputs from arc volcanoes and via diffuse venting.Carbon Uptake During Hydrothermal Alteration of Basaltic Oceanic Crust: 22-29 Mt C/y Following Alt and Teagle (15), we compiled data on [C] in altered oceanic crust (Dataset S1; also see SI Text).[C] is now thought to be higher in volcanic rocks and lower in gabbros. These differences offset each other, so our estimate of 500-600 ppm carbon in oceanic crust agrees with Alt and Teagle. Oceanic plates are consumed at an average of ∼0.05 m/y along the ∼44,500-km length of global subduction zones (4). These values and [C] in altered oceanic crust yield a carbon flux of 22-29 Mt C/y (Dataset S2).Seismic data and seafloor outcrops imply that 5-15% of oceanic crust that formed at slow-and ultraslow-spreading ridges is composed of altered mantle peridotite, with the extent of serpentinization varying from ∼100% at the seafloor to ∼0% at ∼7 km depth (16,17). Because slow-and ultraslow-spreading crust is formed at ∼30% of midocean ridges (18,19), crust composed of altered peridotite is 1-4% of the total, and not a significant part of the global ca...