km) (3, 4 and 5, 14, 15 and 16). This distribution has been attributed to a shoaling of high 13C/low Cd North Atlantic source waters compensated by a greater influx of low 13C/high Cd Antarctic bottom waters. This replacement of some NADW by glacial Antarctic bottom water (AABW) notwithstanding, mid-depth northern Atlantic waters (˜3 km) have higher 13C and lower Cd than do waters of the eastern tropical Pacific. The structure and temporal variability are complex in detail (17). Whether the LGM NADW conveyor belt was global in scope as it is today has been debated considerably (see ref. 18 for my recent review of this debate). The matter may have been settled recently by studies of 231Pa/230Th by Yu et al. (19). 231Pa generated from decay of 235U in the Atlantic is “missing” from Atlantic sediments. Yu et al. make a convincing case that this deficiency is due to NADW-borne transport of 231Pa out of the Atlantic into the Antarctic Circumpolar Current, where 231Pa is trapped into sediments at levels exceeding its regional production rate. Because this deficiency persists during the LGM, the conveyor must have continued to move 231Pa out of the Atlantic.
Northern Indian Ocean. Kallel et al. (20) reported that upper waters of the northern Indian Ocean had higher levels of 13C than those at present. Naqvi et al. (21) questioned this interpretation because it was based upon a Geochemical Ocean Sections/LGM comparison rather than a core top/LGM comparison (because there are significant differences between Geochemical Ocean Sections d13C and core top C. wuellerstorfi d13C). More recently, foraminiferal Cd evidence from several species of aragonitic and calcitic benthic foraminifera was presented (22); this evidence shows that upper northern Indian Ocean nutrient depletion (strongest in the Arabian Sea) could be reconciled with the evidence of Naqvi et al. (21).
The major problem following this observation is the difficulty in accounting for the cause of 13C enrichment/Cd depletion. The Red Sea cannot be the source because its sill was too shallow during the LGM (due to sea level depression). Other sources (Indonesian basins, Antarctic sources) cannot be ruled out entirely but seem unlikely. In the “Global Picture” below, I suggest that upper glacial North Atlantic intermediate/deep water (GNAI/DW) is the source of this low SCO2 water.
Antarctic. The Antarctic has been a major sticking point for deep water paleoceanography. Most published LGM d13C data show values that are much lower than those of today (in some cases lower than anywhere else in the ocean) (4, 23, 24 and 25). Cd evidence is also self-consistent but contradictory to d13C evidence: Cd is either the same as it is today or slightly lower (7, 23, 24). So d13C data indicate that Southern Ocean deep water was high in SCO2, whereas Cd data indicate that it was moderate or lower in SCO2. The d13C evidence also is difficult to reconcile with lowered glacial atmospheric pCO2 (25). Attempts have been made to resolve this conundrum. These explanations can account for part of the discrepancy but not all (26). I have reviewed this situation recently (24) and argue that the Southern Ocean “Mackensen Effect,” whereby d13C of C. wuellerstorfi is observed to be low under waters of high productivity (27), is stronger during the LGM, hence 13C data are too low. However, this solution is not universally agreed to by stable isotope paleoceanographers.
Northwest Pacific. The northwest Pacific has been another difficult area; however, in this case the problem includes internal inconsistency for each tracer as well as tracer-to-tracer discrepancies. Some evidence is consistent with formation of a low-S,CO2 LGM deep water [e.g., Cd is consistently lower in LGM benthic foraminifera at all depths compared with levels in the eastern tropical Pacific (7), but the few available core top Cd measurements in this region are inexplicably low (7, 28)]. Some northwest Pacific sites show d13C similar to that found in the eastern tropical Pacific (29), but other sites show values that are enriched in 13C (4).
Global Picture. On the broad scale considered here, LGM deep water paleochemical studies have three hits and two misses: Cd and d13C data are consistent and informative on LGM chemical distributions in the eastern tropical Pacific, North and equatorial Atlantic, and northern Indian Ocean, but disagree substantially in the Southern Ocean and northwest Pacific. If the Southern Ocean disagreement is attributed to a productivity-related artifact in d13C (and Cd evidence is accepted), and if the northwest Pacific is left as an open question, a synthesis of LGM paleochemical data can be offered as a testable hypothesis (Fig. 1). In Fig. 1 , major flows are represented in “sections” along western boundaries of major ocean basins, with flow divided into upper deep and lower deep sections (approximately 1.5–2.5 and 2.5–5 km). Areas where bottom water is possibly formed, formation regions that are uncertain as to exact location but must have occurred within some broad region, and formation regions that are possible but