1 Text S1 2 In an effort to understand our observations regarding the PETM at Mead and Dee 3 Streams, we have built a simple box model for dissolved O2 concentrations in the ocean. 4 It is comprised of four oceanic reservoirs with a cycling time of 2 kyrs between them. 5 Reservoirs include a low-latitude shallow ocean (0-500 mbsl), a high-latitude shallow 6 ocean (0-500 mbsl), a global intermediate water ocean (500-2000 mbsl), and a global 7 deep water ocean (2000-5000 mbsl). We assume that the high-latitude surface ocean 8 served as the source of dissolved O2 entering intermediate and deep ocean water masses, 9 similar to the modern ocean. Dissolved O2 in intermediate ocean are thus impacted by 10 water flow from the high-latitude shallow ocean, by oxidation of sinking organic carbon 11 and, during carbon injection of the PETM, potentially by aerobic oxidation of methane. 12 Deep ocean dissolved O2 is treated as a mixture of high-latitude surface ocean and 13 intermediate dissolved O2, and the low-latitude surface ocean is calculated independent of 14 the other reservoirs. Here we focus on variations in dissolved O2 in the intermediate 15 ocean because these results relate to the bioturbation records we have generated for Mead 16 and Dee Streams. 17 We calculate dissolved O2 in the surface ocean reservoirs as a function of 18 temperature based on solubility relationships presented by Garcia and Gordon [1992]. To 19 understand changes in dissolved O2 during the PETM (Figure S1), we first set the high- 20 latitude surface ocean temperature to 9C. This value comes from benthic foraminifera 21 18O records [Zachos et al., 2001], which should be representative of both deep ocean 22 temperatures and the temperature at the location those waters formed. We then increase 23 surface ocean temperatures by 5.5 C over the duration of the PETM (Figure S1). This is 24 generally consistent with temperature variations records from both surface and deep 25 ocean reservoirs [e.g., Kennett and Stott, 1991; Zachos et al., 2001, 2003, Sluijs et al., 26 2006]. The temperature increase relates to a decrease in dissolved O2 of ~0.03 mols/m3 in 27 the high-latitude surface ocean during the warmest point of the PETM (Figure S1). 28 The drop in dissolved O2 that creates the modern OMZ within intermediate waters 29 is driven by oxidation of organic carbon sinking from the surface ocean. We assume in 30 our model that a similar mass of organic carbon was oxidized in the early Paleogene 31 OMZ. Therefore, we model the effect of organic carbon oxidation on the OMZ dissolved 32 O2 content by decreasing the number of O2 moles in our intermediate water reservoir by a 33 value necessary to reduce the modern dissolved O2 concentration of sinking surface 34 waters to the modern mean OMZ dissolved O2 concentration, a drop of ~0.23 mols/m3. 35 Our model suggests that the rise in temperature would have driven average dissolved O2 36 concentrations of intermediate waters to a minimum of ~0.016 mols/m3 during the 37 warmest portion of the PETM (Figure S1). 38 The amount and location of carbon added during the PETM remain the source of 39 debate. Nonetheless, several lines of evidence support the idea that it was methane 40 released from the seafloor when gas hydrate dissociated in marine sediment [e.g., 41 Dickens et al., 1995, 1997; Sluijs et al., 2007; Zeebe et al., 2009]. This is an especially 42 appealing mechanism to drop dissolved O2 concentrations in intermediate waters because 43 the methane would necessarily enter these waters directly [Dickens et al., 1995; Dickens, 44 2000]. To get a theoretical methane input, we assume a 13C composition of –60 ‰ and 45 then estimate the injection rate and duration by taking the first derivative of an 46 interpolated curve fit to the bulk carbonate 13C record at Site 690 [Bains et al., 1999], an 47 approach similar to that taken by Dickens [2001]. The methane carbon input can be 48 modeled as a ~2.4 x 1018 g (~2.0 x 1017 mole) injection of reduced carbon into the upper 49 2000 m of the ocean within ~65 kyrs (Figure S1). This overall carbon input is somewhat 50 similar to that suggested by Zeebe et al. [2009], which was based on global records of 51 deep-sea carbonate dissolution. 52 Aerobic methane oxidation removes 2 moles of dissolved O2 for every one mole 53 of C. If all methane in the above postulated release oxidized to CO2 within intermediate 54 waters, average dissolved O2 concentrations within the OMZ would have fallen to 0 55 mols/m3 for ~10 kyrs at the onset of the PETM, and below ~0.01 mols/m3 for another 56 ~50 kyrs (Figure S1). This may be consistent with the observed bioturbation records at 57 Mead and Dee Stream, if they are representative of large regions of intermediate water 58 depth. 59 To assess the relationship between our regional South Pacific records and the modeled 60 global mean results, we have fit modern global OMZ dissolved O2 data [Boyer et al., 61 2006] to a normal cumulative distribution function. We then shifted this function to our 62 modeled mean OMZ values with a standard deviation that reduces as a function of the 63 mean. This approach maintains the shape of the modern dissolved O2 distribution at any 64 given mean level and can be used to calculate a percent of intermediate waters exposed to 65 hypoxic conditions. Results indicate that ~6% of the latest Paleocene OMZ was 66 characterized by hypoxic or lower dissolved O2 levels, a value roughly double the amount 67 of the modern ocean because of warmer surface ocean temperatures (Figure S2). 68 Additional warming during the PETM would have risen the amount of hypoxic waters at 69 intermediate depths to ~23%. The combined effect of warming and methane oxidation, as 70 modeled above, would further this to ~59% (Figure S2), a result which certainly warrants 71 a more sophisticated modeling approach but nonetheless offers an intriguing link to our 72 NZ records. The 13C and bioturbation records generated at Mead and Dee Streams 73 suggest a direct relationship between carbon injection and oxygen depletion. If the 74 dissolved oxygen distributions we have used are in fact applicable, these records indicate 75 that the New Zealand margin would have been in the ~36% of the intermediate ocean our 76 model suggests would have only been exposed to hypoxic waters as a result of both 77 warming and methane oxidation. 78 79 Auxiliary Figure Captions 80 Figure S1. Dissolved O2 Box Model Constraints and Results. (a) Idealized PETM high- 81 latitude surface ocean temperature curve (see Appendix 1 for details) and the resultant high- 82 latitude surface ocean dissolved O2 content. (b) The ODP Site 690 bulk carbonate 13C record 83 (purple diamonds and line) [Bains et al., 1999], an interpolated curve fit to this 13C data (blue 84 circles and line), and the injection rate of –60 ‰ C necessary to drive the interpolated 13C curve 85 (green circles and line). (c) The modeled oxygen minimum zone (OMZ) mean dissolved O2 86 response to the PETM increase in high-latitude surface temperature (grey circles and line) and 87 the combined effect of this warming and the oxidation of the methane injection illustrated in 88 pane “b” (orange diamonds). Distribution points (i.e. D1, D2, D3) represent the mean dissolved 89 O2 values used to constrain distribution functions illustrated in Figure S2. 90 91 Figure S2. OMZ Dissolved O2 Cumulative Distributions. The OMZ normal dissolved 92 O2 cumulative distributions for the modern ocean (grey circles and line; data from Boyer 93 et al., 2006), the latest Paleocene (blue diamonds and line; representing point D1 in 94 Figure S1c), and two PETM realizations, one produced from PETM warming alone 95 (orange circles and line; representing point D2 in Figure S1c), and one produced by the 96 combined effect of PETM warming and methane oxidation (red diamonds and line; 97 representing point D3 in Figure S1c). 98 99 Auxiliary Table Captions 100 Table S1. Mead Stream and Dee Stream Geochemical and B% Results. *Sample depths 101 reported in previous studies have been shifted slightly to account for our updated 102 lithostratigraphic logs. See noted references for original heights by Field Number. 103 104 Table S2. Sample-sets used for B% error evaluation. 105 106 Table S3. Mead Stream and Dee Stream Age Models. *The Dee Stream outcrop 107 section between the PETM and H1/ELMO/ETM2 is complicated by a combination of 108 faults, heavily vegetated rock face, and a broad swing in the river that has created 109 separate but internally continuous intervals of obviously usable outcrop; one containing 110 the PETM and one the early Eocene H1-I2 CIEs (Nicolo et al., 2007). Therefore, because 111 this section has not yet been logged in sufficient detail to fully resolve the true 112 stratigraphic height of the HI CIE onset relative to the PETM, here we report heights 113 consist with prior work [Hancock et al., 2003; Nicolo et al., 2007] but rely on the pre- 114 PETM sedimentation rate to calculate Dee Stream sample ages for the sections both 115 below and above the PETM CIE (Figure 7 and Figure 8). Age references relate to the 116 following superscript lettering: A = Westerhold et al., 2007; B = Farley and Eltgroth, 117 2003; C = Röhl et al., 2000; D = the mean age of B and C; E = Hancock et al., 2003.