Terrestrial carbon storage is dramatically decreased during glacial periods due to cold temperatures, increased aridity, and the presence of large ice sheets on land. Most of the carbon released by the terrestrial biosphere is stored in the ocean, where the light isotopic signature of terrestrial carbon is observed as a 0.32-0.7‰ depletion in benthic foraminiferal δ 13 C. The wide range in estimated δ 13 C change results from the use of different subsets of benthic δ 13 C data and different methods of weighting the mean δ 13 C by volume. We present a more precise estimate of glacial-interglacial δ 13 C change of marine dissolved inorganic carbon using benthic Cibicidoides spp. δ 13 C records from 480 core sites (more than 3 times as many sites as previous studies). We divide the ocean into eight regions to generate linear regressions of regional δ 13 C versus depth for the Late Holocene (0-6 ka) and Last Glacial Maximum (19-23 ka) and estimate a mean δ 13 C decrease of 0.38 ± 0.08‰ (2σ) for 0.5-5 km. Estimating large uncertainty ranges for δ 13 C change in the top 0.5 km, below 5 km, and in the Southern Ocean, we calculate a whole-ocean change of 0.34 ± 0.19‰. This implies a terrestrial carbon change that is consistent with recent vegetation model estimates of 330-694 Gt C. Additionally, we find that a well-constrained surface ocean δ 13 C change is essential for narrowing the uncertainty range of estimated whole-ocean δ 13 C change.

The carbon isotope composition (δ13C) of seawater provides valuable insight on ocean circulation, air‐sea exchange, the biological pump, and the global carbon cycle and is reflected by the δ13C of foraminifera tests. Here more than 1700 δ13C observations of the benthic foraminifera genus Cibicides from late Holocene sediments (δ13CCibnat) are compiled and compared with newly updated estimates of the natural (preindustrial) water column δ13C of dissolved inorganic carbon (δ13CDICnat) as part of the international Ocean Circulation and Carbon Cycling (OC3) project. Using selection criteria based on the spatial distance between samples, we find high correlation between δ13CCibnat and δ13CDICnat, confirming earlier work. Regression analyses indicate significant carbonate ion (−2.6 ± 0.4) × 10−3‰/(μmol kg−1) [CO32−] and pressure (−4.9 ± 1.7) × 10−5‰ m−1 (depth) effects, which we use to propose a new global calibration for predicting δ13CDICnat from δ13CCibnat. This calibration is shown to remove some systematic regional biases and decrease errors compared with the one‐to‐one relationship (δ13CDICnat = δ13CCibnat). However, these effects and the error reductions are relatively small, which suggests that most conclusions from previous studies using a one‐to‐one relationship remain robust. The remaining standard error of the regression is generally σ ≅ 0.25‰, with larger values found in the southeast Atlantic and Antarctic (σ ≅ 0.4‰) and for species other than Cibicides wuellerstorfi. Discussion of species effects and possible sources of the remaining errors may aid future attempts to improve the use of the benthic δ13C record.

A paleo-data compilation with 492 δ 13 C and δ 18 O observations provides the opportunity to better sample the Last Glacial Maximum (LGM) and infer its global properties, such as the mean δ 13 C of dissolved inorganic carbon. Here, the paleocompilation is used to reconstruct a steady-state water-mass distribution for theLGM, that in turn is used to map the data onto a 3D global grid. A global-mean marine δ 13 C value and a self-consistent uncertainty estimate are derived using the framework of state estimation (i.e., combining a numerical model and observations). The LGM global-mean δ 13 C is estimated to be 0.14 ± 0.20 at the two standard error level, giving a glacial-to-modern change of 0.32 ± 0.20 .The magnitude of the error bar is attributed to the uncertain glacial ocean circulation and the lack of observational constraints in the Pacific, Indian, and Southern Oceans. Observations in the Indian and Pacific Oceans generally have 10 times the weight of an Atlantic point in the computation of the global mean. To halve the error bar, roughly four times more observations are needed, although strategic sampling may reduce this number. If dynamical constraints can be used to better characterize the LGM circulation, the error bar can also be reduced to 0.05 to Preprint submitted to Quaternary Science Reviews August 5, 2015 0.1 , emphasizing that knowledge of the circulation is vital to accurately map δ 13 C DIC in three dimensions.

Abstract. We present a compilation of 127 time series δ 13 C records from Cibicides wuellerstorfi spanning the last deglaciation (20-6 ka) which is well-suited for reconstructing large-scale carbon cycle changes, especially for comparison with isotope-enabled carbon cycle models. The age models for the δ 13 C records are derived from regional planktic radiocarbon compilations . The δ 13 C records were stacked in nine different regions and then combined using volume-weighted averages to create intermediate, deep, and global δ 13 C stacks. These benthic δ 13 C stacks are used to reconstruct changes in the size of the terrestrial biosphere and deep ocean carbon storage. The timing of change in global mean δ 13 C is interpreted to indicate terrestrial biosphere expansion from 19-6 ka. The δ 13 C gradient between the intermediate and deep ocean, which we interpret as a proxy for deep ocean carbon storage, matches the pattern of atmospheric CO 2 change observed in ice core records. The presence of signals associated with the terrestrial biosphere and atmospheric CO 2 indicates that the compiled δ 13 C records have sufficient spatial coverage and time resolution to accurately reconstruct large-scale carbon cycle changes during the glacial termination.