Results and Discussion

Estimates of core-top PPa based on ML and ANN at site ODP 1019 (475 ± 104 and 345 ± 155 gC/m²/y, respectively) are consistent with the modern observed values (426 gC/m²/y) (21). The youngest (mid-Holocene) PPa reconstructions for core W8709-13PC (564 ± 104 and 454 ± 155 gC/m²/y for ML and ANN, respectively) are slightly higher than the modern atlas value of 351 gC/m²/y. Both methods estimate PPa during glacial times about double that of the late Holocene; the ML estimates display higher variance than the ANN estimates at both core sites during the glacial interval (Fig. 2 A and B). PPa decreases abruptly between ∼17 and 16 ka and again between ∼13 and ∼11 ka, which is coincident with times of rapidly increasing atmospheric CO₂ (22).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Diatom-based paleoenvironmental indices across the last glacial termination in the Northeastern Pacific. For derivation, see Dataset S2. PPa reconstructions for (A) ODP 1019 and MD02-2499 (ML in black and ANN in gray) and (B) W8709-13PC (ML in black and ANN in gray); thin lines indicate the uncertainty envelope of each method. (C) Upwelling index based on Q-mode factor analysis for MD02-2499 (black) and W8709-13PC (gray). (D) Abundance of freshwater diatoms as a percentage of total diatoms (a proxy for river discharge) at ODP 1019 and MD02-2499 (black) and W8709-13PC (gray). (E) Percentage of large diatoms [relative to total marine diatoms; a proxy for export efficiency at ODP 1019 and MD02-2499 (black) and W8709-13PC (gray)]. Age model's details are in SI Materials and Methods.

To assess oceanographic mechanisms driving productivity change, we calculated a diatom-based upwelling index (Fig. 2C and SI Materials and Methods) in a manner similar to ref. 23, but excluding freshwater diatoms as these track river input (15) (Fig. 2D). A diatom size index (Fig. 2E) traces the efficiency of carbon export from the sea surface. Larger diatoms today are associated with high-export coastal upwelling systems replete with iron, whereas small diatoms are associated with low-export open ocean systems that are iron limited (24). As expected, this index is generally higher at our coastal upwelling site (ODP 1019 and MD02-2499) than offshore (W8709-13PC). The offshore Oregon system today is under mild iron stress, such that iron availability regulates the magnitude of phytoplankton blooms (25). Although some of the peaks in PPa are coincident with peaks in the upwelling index, in general, the upwelling index is lower during the LGM, consistent with a southward shift of the subpolar gyre (14). Substantially higher freshwater input documented during glacial time, especially during flood events (Fig. 2D) (15), may also contribute nutrients including silica and iron that are expected drivers of productivity. Overall, high glacial primary productivity tracks long-term increases in apparent river inputs, but in detail, reconstructed PPa highs do not match the individual peaks in freshwater diatom abundance, so the connection is likely indirect. Although not constrained here, such an indirect connection may reflect the tendency of river-borne iron in colloids to flocculate and sink quickly on encountering salt water (26); suboxic remobilization of iron from accumulation sites in shelf sediments makes it available to the biota during the upwelling season (27). Relatively low flooded shelf area during the glacial maximum may have limited this mechanism for iron fertilization despite strong regional river input. Thus, high primary productivity in the region may reflect nitrate, phosphate, and silicic acid inputs from strong river fluxes, whereas relatively low export may have reflected iron limitation.

To assess temporal relationships in carbon fluxes and burial rates, we average all productivity and burial parameters within nonoverlapping consecutive 30-cm increments (SI Materials and Methods) and scale the diatom size index linearly to export efficiency (i.e., the ratio export/primary productivity) such that it spans a range of 20–50%. This is similar to the modern range of this ratio from the subtropical North Pacific to the high-latitude upwelling systems (28). Although the exact calibration of the export index is uncertain, our results are relatively insensitive to reasonable variations in the scaling. The results indicate that, although the reconstructed PPa in core W8709-13PC was highest during glacial time (Last Glacial Maximum), the inferred export productivity peaked during the deglaciation (including the Bølling-Allerød, Younger Dryas, and Heinrich event 1 intervals) and the observed carbon burial was highest during interglacial (Holocene) time (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Variations in components of the biological pump across the glacial termination at site W8709-13PC, calculated in 30-cm averaged intervals. (A) Primary productivity (PPa), export productivity (EP), and organic carbon burial (OCB) averaged in 30-cm nonoverlapping consecutive intervals with uncertainty estimates (Dataset S3). (B) Export efficiency (EP/PPa), burial efficiency (OCB/PPa), and seafloor preservation of organic carbon (OCB/EP) expressed as percentages. Uncertainties are ±1σ (Dataset S4).

These results indicate that even under an extreme scenario of changes in export productivity, carbon burial across the last termination does not track productivity at our study site; this is also the case for preservation-corrected productivity indices based on an empirical combination of organic carbon contents and sedimentation rate (29), as well as for other common geochemical productivity indices such as opal percentage and accumulation rate and barium concentration and accumulation rate (SI Materials and Methods and Figs. S4 and S5). North of our study area, mass accumulation rates of marine organic carbon are controlled primarily by sedimention rates (especially in late glacial time), but also rise during interglacial time when sedimentation rates were lower (30). It is possible that organic carbon-based estimates of export productivity may work in some regions when confounding influences of preservation are relatively small. Thus, it is plausible that organic carbon-based estimates of export productivity may work in some regions where confounding influences are relatively small.

Carbon burial at our Gorda Ridge site is likely dominated by preservation effects, expressed either as sea floor preservation (i.e., OCB/EP) or carbon burial efficiency (i.e., OCB/PPa). Both parameters increased during the deglacial transition when both primary productivity and the inferred export productivity were low (Fig. 3B). Across the deglacial transition, carbon burial decreases with increasing primary productivity (Fig. 4A) and is unrelated to export productivity (Fig. 4B), but shows a strong positive correlation to sedimentation rate (with one outlier, Fig. 4C). These findings support hypotheses that link the process of organic carbon preservation either to sorption on mineral surfaces (10, 11) or to integrated oxygen exposure time (9, 31), both of which are influenced directly by sediment accumulation rate. There is no significant correlation between our estimates of PPa from diatom species assemblages and sedimentation rate (r = −0.3, P = 0.18), implying no obvious artifacts in PPa estimates related to this process.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Determinants of carbon burial across the glacial termination at site W8709-13PC. (A) Carbon burial decreases with increasing reconstructed PPa (r = 0.8, P = 0.001). (B) Carbon burial is not significantly correlated to estimated export productivity (r = 0.3, P = 0.18). (C) Carbon burial increases proportional to sedimentation rate (r = 0.5, P = 0.05), suggesting that preservation effects linked to sedimentation rate, rather than productivity, control carbon burial. (D) δ¹⁵N (per mil) decreases with increasing reconstructed PPa (r = 0.9, P < 0.001). (E) δ¹⁵N (per mil) is not significantly correlated to estimated export productivity (r = 0.2, P = 0.28). (F) δ¹⁵N (per mil) does not increase proportionally to sedimentation rate (r = 0.1, P = 0.38), suggesting that sedimentation rate does not control the δ¹⁵N (per mil) record. All parameters are plotted as means and 1σ uncertainties for nonoverlapping consecutive 30-cm intervals (Dataset S3).

The apparent changes in export efficiency, which decoupled primary and export production through time, must have plausible ecological controls. In the North Pacific, this likely reflects the transition from a subpolar ecosystem during glacial time (14), similar to that of the modern Alaska Gyre where carbon export is relatively low and primary production is mostly recycled within the water column, to a coastal upwelling system in interglacial time dominated by large diatoms that sink rapidly and are more effective in exporting organic carbon to the deep sea (10, 32). Our PPa estimates are significantly correlated (r = −0.91, P < 0.001) with δ¹⁵N (33). The association of high PPa with low δ¹⁵N (Fig. 4D) is also consistent with a glacial ecosystem similar to that of the Alaska Gyre, with relatively low fractional nitrate utilization and export efficiency.

Export efficiency peaked during the last deglacial transition, during a time of sea level rise. Such an effect is consistent with hypothesized sources of both particulate and dissolved iron from the continental shelf during sea level rise (16) and from growing hypoxia on the shelf and upper slope during deglacial warming (34, 35). Hypoxia in parts of the North Pacific during the last deglaciation may have been sustained in part by iron feedback that produced an ecological shift to a more efficient exporting ecosystem (36).