North Atlantic temperature control on deoxygenation in the northern tropical Pacific

Ocean oxygen content is decreasing with global change. A major challenge for modelling future declines in oxygen concentration is our lack of knowledge of the natural variability associated with marine oxygen inventory on interannual and multidecadal timescales. Here, we present 10 annually resolved 200 year-long records of denitrification, a marker of deoxygenation, from a varved sedimentary archive in the North Pacific oxygen minimum zone covering key periods over the last glacial–interglacial cycle. Spectral analyses on these records reveal strong signals at periodicities typical of today’s Atlantic multidecadal oscillation. Modern subsurface circulation reanalyses regressed on the positive Atlantic and Pacific Climatic Oscillation indices further confirm that North Atlantic temperature patterns are the main control on the subsurface zonal circulation and therefore the most likely dominant driver of oxygen variability in the tropical Pacific. With currently increasing temperatures in the Northern Hemisphere high latitudes and North Atlantic, we suggest deoxygenation will intensify in the region.

Cold SSTs excursions in the GoC correspond to active upwelling periods whilst warm SSTs are symptomatic of surface stratification and tropical surface water invading the GoC (Supplementary Figure 3a, ref (6)).These excursions are in phase with the ENSO 3.4 index and the NPGO index, a climate fluctuation reflected by changes in the intensity of the central and eastern branches of the North Pacific gyre circulations and the properties of the California Current (7).Time series analyses in the frequency domain using spectral analyses with various windowing settings (Blackman-Tuckey -not shown-and Max Entropy) show a good match between the ENSO3.4index and the SST record spectra, with corresponding periodicities and power at sub-decadal scale (Supplementary Figure 3b).The SST spectrum shows an additional peak at about 12-year periodicity that corresponds to the cyclicity of the NPGO climate variability.
Temporal changes in upwelling dynamics reflected in the GoC SST record control the biogenic and detrital export fluxes to the sediment.In Supplementary Figure 3c, we show 3 monthaverage contributions (% of total weight accumulation per varve couplets) of biogenic opal (diatoms) calculated from sediment trap data (3,6).During El Niño events such as in 1991-92, ocean stratification and wet climate on land drive marine productivity down and increases detrital (riverine) inputs to the sea floor causing the relative annual biogenic silica contribution to be low (20%).This pattern is repeated during the 1998 El Nino event although the sequence is incomplete.During normal years, upwelling of cold and nutrient subsurface waters promotes marine productivity and dryness on land resulting in high biogenic silica export and decreased dilution of the biogenic fluxes by riverine inputs (Supplementary Figure 3c).This results in high relative annual biogenic silica contribution.We reconstruct annual biogenic silica contribution in our varved sections as an indicator of upwelling related biological production.

Supplementary note 3: Spectral analyses and cross-correlations
Annually resolved Nitrogen isotopes records in 10 sections of our Guaymas basin sediment core were analysed in the frequency domain using spectral analyses with various windowing settings (Blackman-Tuckey -not shown-and Max Entropy, see method section).The main recurring periodicities were identified in the 10 records, namely around 80, 35, 25, 10 and 5-8 years.These periodicities are similar to today's AMO, PDO, NPGO and ENSO observed periodicities (8)(9)(10).The data in table S1 show a predominance of the multidecadal periodicities compared to the shorter periodicities in all records and the absence of clear AMOlike periodicities in the glacial and stadial sections #26 and #17, respectively.Supplementary Table 1: Relative power (in %) of the various periodicities identified in each denitrification (δ 15 N) record.
We examined whether there was significant and consistent relationship between the upwelling/biological productivity indicator (Biogenic Silica) and the denitrification indicator (δ 15 N) in the 10 selected records.Supplementary Figure 4 shows the cross-correlation plots for each couple of annually resolved records (red line compared to the cross-correlation plot of identical records in black, for reference).Section 10, 45, 46 show some degree of correlation, albeit not statistically significant between Biogenic silica and denitrification, with a correlation coefficient of around 0.4 at 0 year lag, while the other 7 records show no positive correlation or even slightly negative relationship between δ 15 N and biogenic silica.Detrended records, i.e. filtered to remove variability at periodicity greater than 75 years, show even lower degrees of correlation than the non-detrended records which are shown in Supplementary Figure 4.This indicate that at multi-to sub-decadal timescales, there is no correlation between biogenic export and hypoxia-induced denitrification.There is no systematic lead-lag relationship between the 2 parameters.Coherence spectra between biogenic silica and nitrogen isotope time series analyses are shown in Supplementary Figure 5 and corroborate our findings that there is no significant relationship between productivity and denitrification changes in the GoC.The spectra are below the 95% confidence limit for most of the frequencies and there are no frequencies at which the spectra show systematically good coherence across all sections.
The cross-correlation plot for Section 26 (non-detrended) stands out and exhibits a negative correlation between Biogenic silica and δ 15 N suggesting that denitrification decreases with increasing biogenic export.It is worth noting that this negative correlation is not seen between the detrended records (Supplementary Figure 4).Section 26 was deposited just before the last glacial maximum (26 ka BP).It has been shown that denitrification intensity was dramatically reduced during the last glacial maximum in the ETNP (11)(12)(13) and therefore, it is not excluded that the nitrogen isotopic signal recorded in this section is only partly influenced by denitrification (in response to hypoxia) and responded to other biogeochemical processes such as nitrate utilisation by the biota (14).Under strong upwelling conditions and in absence of, or reduced hypoxia/denitrification, biological productivity and δ 15 N are expected to be negatively correlated.Therefore, we do not include section 26 in for the spectral analyses of multi-to sub-decadal deoxygenation changes (section 1 of the paper).