The deep and intermediate waters of the subarctic North Pacific Ocean contain some of the highest nutrient levels in the world . Accordingly productivity peaks over glacial–interglacial cycles, including those covered in this study, have been linked to changes in the regional halocline and water column stratification which would alter the advection of nutrient- and carbon-rich sub-surface waters into the photic zone . A key difference between the two productivity events in MIS 5e and MIS 5b/c is the response of the biological community to raised photic zone nutrient availability. Although productivity is high during MIS 5e, values are low for δ30Sidiatom at ODP Site 882 (<1.0 ‰, n=3) and for δ15Ndiatom (<6 ‰) at a nearby site (49∘72′ N, 168∘30′ E) (Fig. 3). In contrast during MIS 5b/c the productivity peak is concordant with an increase in δ30Sidiatom and δ15Ndiatom to ca. 1.2–1.3 ‰ and >6 ‰ respectively (Fig. ).

Changes in δ30Sidiatom may reflect either increased biological uptake of silicic acid (consumption) and/or changes in the supply of silicic acid to the photic zone. The modern-day regional stratified water column is best represented by a closed system model in which a finite amount of silicic acid exists for biomineralisation . In contrast an unstratified water column would be reflected by an open system model with continual supply of silicic acid. By assuming that the two productivity peaks reflect a weakening in the stratification, an open system model can be used to investigate the controls on δ30Sidiatom: δ30Sidiatom=δ30Si(OH)4+ϵ⋅f, where δ30Si(OH)4 is the isotopic composition of dissolved silicic acid supplied to the photic zone, ϵ is the enrichment factor between diatoms and dissolved silicic acid and f is the fraction of utilised Si(OH)4 remaining in the water. Existing work from the North Pacific Ocean has estimated δ30Si(OH)4 at 1.23 ‰ and ϵ as 1.0 . Using changes in Si(OH)4 consumption (Eq. ) and siliceous productivity (opal), the supply of Si(OH)4 into the photic zone can be constrained relative to mean conditions during MIS 5e as Si(OH)4supply=Opalsample/OpalMIS 5e1-fconsumedsample1-fconsumedMIS 5e. Estimates of Si(OH)4 consumption and supply from Eqs. () and () are only applicable for intervals when the water column represents an open system (e.g. the productivity peaks in MIS 5e and MIS 5b/c) and are dependant on modern-day estimates of δ30Si(OH)4 being representative of past conditions. This assumption is based on evidence that δ30Si(OH)4 is relatively resilient to change, outside of seasonal biological fluxes, over timescales similar to this study except in extreme circumstances linked to major reductions in the flux of riverine silicon into the ocean . The results show that the productivity peaks in MIS 5e and MIS 5b/c are both closely correlated with elevated levels of Si(OH)4 being supplied to the photic zone (Fig. ), supporting the suggestion that these intervals are linked to a reduction in water column stratification and an increase in the vertical flux of nutrients bearing sub-surface waters into the photic zone . However, whilst the increase in Si(OH)4 supply in MIS 5b/c is matched by a corresponding increase in biological consumption of Si(OH)4, increasing the ratio of regenerated to preformed nutrients in the ocean interior, the opposite occurs during MIS 5e when the rates of Si(OH)4 consumption are at their lowest over the analysed interval. (Fig. ). Whilst reduced Si(OH)4 consumption during MIS 5e could be linked to iron limitation, records indicate that aeolian dust deposition was equally low during both the MIS 5e and the MIS 5b/c productivity peaks. However, as before we are unable to account for changes in iron supply from non-aeolian sources.

Understanding the mechanisms that regulate changes in atmospheric concentrations of CO2 (pCO2) remains a key objective in palaeoclimatology. Previous research has demonstrated that the Southern Ocean and low-latitude oceans act as the dominant source/sink of atmospheric CO2 over glacial–interglacial cycles . Whilst the North Pacific Ocean does not need to be invoked to explain the full amplitude of glacial–interglacial changes, recent work has advocated a potential role for the region in regulating atmospheric pCO2 over the last termination .

Today the net annual ocean–atmosphere exchange of CO2 in the subarctic north-west Pacific Ocean is close to zero, but alters from being a sink of atmosphere CO2 in spring to a source in winter (Fig. a). This seasonal variability can be attributed to changes in the biological pump and in SST which affects the solubility of CO2 . A weakening of the halocline stratification in MIS 5e and MIS 5b/c would have increased the advection of nutrient- and carbon-rich waters from the ocean interior, raising photic zone pCO2(aq) and the potential for CO2 to be ventilated into the atmosphere due to an air–sea disequilibrium in CO2. This, however, is dependant on the response and relative efficiency of the biological pump in taking advantage of the increased nutrient supply and altering the ratio of regenerated : performed nutrients to re-export carbon into the deep ocean . Whilst changes in the temperature/thermocline may also have been important, the only SST record for the region does not contain the temporal resolution to investigate this further and does not provide a surface–subsurface depth temperature transect.

During MIS 5e a scenario of both higher pCO2(aq) and incomplete/low rates of nutrient utilisation suggests the regional ocean could have ventilated CO2 into the atmosphere faster than the soft-tissue biological pump reabsorbed and sequestered CO2 into the deep ocean (Fig. b) despite evidence for higher organic carbon export (BioBa) in this period (Fig. ). The culmination of this interval at the end of MIS 5e would have resulted in the system returning to a stratified state, perhaps similar to the modern-day water column (Fig. a) with minimal air–sea fluxes of CO2. In contrast the opal productivity peak in MIS 5b/c is marked by similar levels of photic zone Si(OH)4 supply as in MIS 5e but with ca. 20 % higher rates of Si(OH)4 consumption (Figs.  and c). The combination of high siliceous productivity (opal/δ13Cdiatom) and a highly efficient biological pump (δ30Sidiatom/δ15Ndiatom) during this interval suggests that the net flux of ocean–atmosphere CO2 exchanges arising from the sea–air disequilibrium could have remained close to zero if photosynthetic carbon demand were similar to the rate of sub-surface carbon flux to the photic zone. Whilst a highly efficient soft-tissue biological pump raises the possibility for the region to have acted as a net sink of atmospheric CO2, the potential and significance for this is limited by the relatively low proportion of surface waters which reach the deep ocean interior and low BioBa in this interval (Fig. ).

Although the data suggest that changes in the regional photic zone may have contributed to variations in atmospheric pCO2 during MIS 5, both via the soft-tissue biological pump and associated changes in ocean alkalinity, it is not possible to quantify the magnitude of any fluxes or access whether they were accompanied by a change in diatom silicification and cellular Si : C ratios. Firstly, insufficient purified diatom material remains to measure diatom silicon concentrations. Secondly, although diatom elemental carbon measurements obtained during the analysis of δ13Cdiatom increase from <0.3 wt% in MIS 5e to ca. 0.4 wt% in MIS 5b–d (see Supplement Table S1), the analytical reproducibility for Cdiatom is relatively high at 0.1 % (1σ) and measurements are derived from the cell wall material and not the bulk cellular matter formed duringphotosynthesis. Furthermore, whilst other cores from the region show a similar double peak in opal productivity during MIS 5 records at other sites suggest that the second peak is restricted to MIS 5b with no increase in MIS 5c . Such discrepancies either suggest poor stratigraphic controls on the age model for either core, or the potential for significant spatial variability across the region and reiterates that the magnitude of any ocean–atmosphere fluxes of CO2 would be low compared to those occurring elsewhere in the marine system such as the Southern Ocean and low-latitude oceans.

Records show that the decline in siliceous productivity for both intervals culminates with large decreases in δ18Odiatom of ca. 3–5 ‰ from ca. 113 kaBP and 85 kaBP (Fig. , red shading). The magnitude of change is too large to be driven by reductions in deep water upwelling or shifts in ocean water masses from both higher and lower latitudes, which would only alter δ18Owater by ca. 1 ‰ . Instead the drop in δ18Odiatom suggests an input of isotopically depleted freshwater that may be similar in origin to events documented at the same site during the late Pliocene/early Quaternary .

Although the modern-day regional halocline is maintained by high precipitation and low evaporation , it is difficult to envisage a sufficient increase in precipitation to initiate a 3–5 ‰ decrease in δ18Odiatom. This is reiterated by evidence that monsoonal activity was largely stable during MIS 5b–e . At the same time the potential for a glacial source is questioned by evidence indicating a restricted glaciation in north-east Russia, closest to ODP Site 882, at the Last Glacial Maximum (LGM) , although other work suggests these ice sheets may have been considerably larger prior to the LGM . Recent work has shown that both of the major decreases in δ18Odiatom coincide with increases in IRD accumulation in some, but not all, cores from the Okhotsk Sea . It has also been argued that the regional water column was regulated by significant inputs of meltwater from the North American ice sheets during the last deglaciation . In either case, the decrease in δ18Odiatom at the end of each siliceous productivity peak suggests that inputs of freshwater helped re-establish/strengthen the halocline, limiting the upwelling of nutrient-/CO2-rich sub-surface waters and biological activity. However, with the decrease in δ18Odiatom only occurring after the initial decline in productivity, freshwater can only be acting as a secondary control in re-establishing the halocline.

Previous work has suggested a link between changes in the subarctic north-west Pacific Ocean and the Southern Ocean . The most viable mechanisms for synchronous changes between polar regions are temperature- and salinity-driven variations in water column density . For example a cooling of polar SST would reduce the rate of sub-surface upwelling into the photic zone , lowering nutrient availability and potentially triggering the initial decline in siliceous productivity. At the same time, a decrease in SST would increase the sensitivity of the water column to subsequent changes in salinity, making the region highly vulnerable to inputs of freshwater which would strengthen the water column and inhibit productivity . Additional reductions in siliceous productivity may then arise from lower North Atlantic overturning and associated deep-water incursions and upwelling in the North Pacific . Support for a series of events similar to this at ODP Site 882 lies with the concordant decreases at ODP Site 882 between supplied Si(OH)4/δ30Sidiatom/opal and Antarctic (δD)/NGRIP (δ18Oice) ice-core records at the start of each productivity decline from ca. 118 and 89 kaBP respectively (Fig. ). The final switch to a low productivity system then coincides with the later decreases in δ18Odiatom at 113 and 85 kaBP, suggesting that the climatic deterioration associated with lower δD/δ18Oice may have fuelled the increase in precipitation and/or an advancement of regional glaciers around the North Pacific Basin that triggered the increase in freshwater input. Whilst it remains unclear what initiated either siliceous productivity peak, it can be speculated that reductions in freshwater after 100 kaBP could have weakened the halocline and created the conditions for the second productivity bloom to eventually develop later in MIS 5b/c.