Auxiliary Material for
Deglacial whole-ocean d13C change estimated from 480 benthic foraminiferal records
Carlye Peterson, Lorraine Lisiecki, and Joseph Stern
(University of California Santa Barbara, Santa Barbara, California)
Paleoceanography, 2014

Introduction
This data set contains benthic Cibicidoides spp. d13C records from 480 marine core sites from the Atlantic, Pacific, and Indian Oceans. Table "ts01.txt" contains core location in degrees latitude and longitude, water depth (meters), Holocene d13C, LGM d13C, data type (time series = 1; time slice = 2; PE = phytodetritus effect), and literary references. Table "ts02.txt" contains latitudinal and longitudinal boundaries of ocean regions. Figure "fs01.eps" shows a map of the Holocene d13C data and modern d13C from shallow depths (0.5-2 km). Figure "fs02.eps" shows a map of the Holocene d13C data and modern d13C from deep depths (2-5 km). Figure "fs03.eps" shows the regional profiles of Holocene d13C data from all sites and sites with complementary LGM data. Figure "fs04.eps" shows the regional profiles of LGM d13C from all sites and sites with complementary Holocene data. Figure "fs05.eps" shows (a) the modern and Holocene d13C data from the South Atlantic and (b) the modern d13C data from the southern portions (south of -55) of the Pacific, Indian, and Atlantic oceans as well as Holocene d13C data from south of -55 in the Pacific and Atlantic Oceans. Figure "fs06.eps" shows how the deglacial d13C change of the Northwest Atlantic region changes with increasing order of polynomial line fit. Figure "fs07.eps" shows how the deglacial d13C change of the Indian Ocean region changes with increasing order of polynomial line fit.

1. ts01.txt Locations of core records of Holocene  d13C and LGM d13C
1.1 Column "data type (time series = 1; time slice = 2; PE = phytodetritus effect)", 1 denotes time series d13C record, 2 denotes time slice d13C record, PE denotes suspected phytodetritus effect on the benthic d13C record
1.2 Column "Latitude (S = -)", degrees, latitude of core locations
1.3 Column "Longitude (W = -)", degrees, longitude of core locations
1.4 Column "Depth (meters)", meters, depth of cores beneath mean sea level.
1.5 Column "Holocene d13C", d13C per mil, carbon isotope value of Holocene.
1.6 Column "LGM d13C", d13C per mil, carbon isotope value of LGM.
1.7 Column "References", references for d13C records.

2. ts02.txt, table of latitudinal and longitudinal boundaries of regions of the ocean used for regional d13C estimates in this study.
2.1 Column "Ocean region"
2.2 Column "N boundary lat", degrees, latitude of Northern boundary of region
2.3 Column "S boundary lat", degrees, latitude of Southern boundary of region
2.4 Column "E boundary long", degrees, longitude of Eastern boundary of region
2.5 Column "W boundary long", degrees, longitude of Western boundary of region

3.  fs01.eps, Map of shallow sites (0.5-2 km) with Modern d13C (diamonds) and Holocene d13C (stars) data. Symbol colors correspond to d13C values scaled to the modern and benthic d13C data sets. Low spatial coverage is obvious in the Northwest, Southeast, South Atlantic regions, and the Atlantic sector of the Southern Ocean as well as the interior of the North Pacific, and South Indian Ocean 40°-50° south. This makes modern vs. Holocene d13C comparisons difficult without a more sophisticated extrapolation technique (e.g. a model, flow paths).
4.  fs02.eps, Map of deep sites (2-5 km) with Modern d13C (diamonds) and Holocene d13C (large circles) data. Symbol colors correspond to d13C values scaled to the modern and benthic d13C data sets. Again, low spatial coverage in the Northwest Atlantic, the Atlantic sector of the Southern Ocean, especially in the Southeastern Atlantic, most of the interior of the North Pacific, and the Southern Indian between 40°-50° south. Although modern d13C spatial coverage is sparse in the North Pacific, the available deep Holocene and modern d13C data seem to agree well zonally. Similarly, South Pacific Holocene d13C and modern d13C data agree well zonally, although this relationship is more clear because modern d13C is more densely sampled in this region.
5.  fs03.eps, Profiles of regional Holocene d13C, sites with only Holocene d13C data are in blue, sites that have an LGM pair are in red, and the line fit for all Holocene d13C data is plotted in blue. The "unpaired" d13C data fit well with the rest of the paired d13C data. Using a d13C dataset of only paired Holocene and LGM sites gives the same mean d13C change as the entire 480 d13C record compilation.
6.  fs04.eps, Profiles of regional LGM d13C data, sites with only LGM d13C data are in black, sites that have both LGM and Holocene d13C are in pink, and the line fit for all LGM d13C data is plotted in black. The "unpaired" d13C data fit well with the rest of the paired d13C data. Using a d13C dataset of only paired Holocene and LGM sites gives the same mean d13C change as the entire 480 d13C record compilation.
7.  fs05.eps, Assumptions for d13C value of Southern Ocean. (a) Modern DIC d13C data (red stars) from the South Atlantic region showing good agreement with South Atlantic benthic Holocene d13C (pink diamonds) where the data overlap. (b) Comparison of South Atlantic d13C data (from a) with the Pacific sector of the Southern Ocean (south of -55°) modern DIC d13C (black dots) and benthic Holocene d13C (black squares), and the Indian sector (south of -55°) modern DIC d13C as blue dots. There are no cores in the Indian sector south of -55°. Holocene d13C and modern DIC d13C agree well from depths 0-3.5 km.
8. fs06.eps, Northwest Atlantic regional trends of increasing polynomial order. Volume-weighted regional d13C change derived from each trend is provided in each plot. In the Northwest Atlantic, higher-order polynomial fits generate estimates of mean d13C change that are within 0.005‰ of our linear regression-based estimate.
9. fs07.eps, Indian regional trends of increasing polynomial order. Volume-weighted regional d13C change derived from each trend is provided in each plot. In the Indian, higher-order polynomial fits up to 4th order generate estimates of mean d13C change that are within 0.01‰ of our linear regression-based estimate (-0.417‰) while an 8th order polynomial gives a mean d13C change of +0.478‰, an example of how polynomial line fits can produce unrealistic estimates.
10. text01.txt, supplemental text about age models and sensitivity to polynomial fits.
11. text02.txt, list of complete references for cores in table "ts01.txt"