Rapid deep ocean deoxygenation and acidification threaten life on Northeast Pacific seamounts

Abstract Anthropogenic climate change is causing our oceans to lose oxygen and become more acidic at an unprecedented rate, threatening marine ecosystems and their associated animals. In deep‐sea environments, where conditions have typically changed over geological timescales, the associated animals, adapted to these stable conditions, are expected to be highly vulnerable to any change or direct human impact. Our study coalesces one of the longest deep‐sea observational oceanographic time series, reaching back to the 1960s, with a modern visual survey that characterizes almost two vertical kilometers of benthic seamount ecosystems. Based on our new and rigorous analysis of the Line P oceanographic monitoring data, the upper 3,000 m of the Northeast Pacific (NEP) has lost 15% of its oxygen in the last 60 years. Over that time, the oxygen minimum zone (OMZ), ranging between approximately 480 and 1,700 m, has expanded at a rate of 3.0 ± 0.7 m/year (due to deepening at the bottom). Additionally, carbonate saturation horizons above the OMZ have been shoaling at a rate of 1–2 m/year since the 1980s. Based on our visual surveys of four NEP seamounts, these deep‐sea features support ecologically important taxa typified by long life spans, slow growth rates, and limited mobility, including habitat‐forming cold water corals and sponges, echinoderms, and fish. By examining the changing conditions within the narrow realized bathymetric niches for a subset of vulnerable populations, we resolve chemical trends that are rapid in comparison to the life span of the taxa and detrimental to their survival. If these trends continue as they have over the last three to six decades, they threaten to diminish regional seamount ecosystem diversity and cause local extinctions. This study highlights the importance of mitigating direct human impacts as species continue to suffer environmental changes beyond our immediate control.

Seamounts can be viewed as a microcosm for the impact of climate change on benthic ecosystems. They occur worldwide (Yesson et al., 2020), are often isolated, consist of relatively uniform volcanic substrates that, due to natural gradients in water properties, span a huge range of environmental conditions and habitats, and are home to many species that are fixed in place. Additionally, they are oases of life in the deep ocean, as they enhance primary productivity, concentrate local productivity, and provide refugia for some continental slope species (Rowden, Dower, Schlacher, Consalvey, & Clark, 2010).
Thus, they are considered essential ecosystems in need of protection (IUCN, 2017). It has been long established that mountains are vulnerable to climate change (Huber, Bugmann, & Reasoner, 2006), changing faster than other terrestrial habitats (Nogués-Bravo, Araújo, Errea, & Martinez-Rica, 2007). Although understudied, as underwater mountains, the same might be expected for seamounts.
There is a large cluster of seamounts (undersea mountain >1,000 m tall) in the Northeast Pacific (NEP). These seamounts, with varying summit depths and geomorphologies, form one of the densest undersea mountain ranges in the world. This region is also home to a large oxygen minimum zone (OMZ), between ~480 and 1,700 m depth, which contains some of the lowest oxygen levels in the global ocean (Paulmier & Ruiz-Pino, 2009). This water is "old," that is, has long been away from the surface (~1,000 years, Gebbie & Huybers, 2012), and has experienced significant biological consumption (Talley, 2011).
Many of the seamounts intersect the OMZ, so their summits and flanks pass through a wide range of environmental conditions. The remaining seamounts contend with the OMZ as a chemical ceiling. Benthic organisms are generally found in stratified bands encircling the seamounts (Du Preez, Curtis, & Clarke, 2016), the depths of which are determined by both environmental (e.g., temperature, oxygen; Wishner, Levin, Gowing, & Mullineaux, 1990) and biological (e.g., competition; Victorero, Robert, Robinson, Taylor, & Huvenne, 2018) factors.
In addition to the already low oxygen levels, this region has been experiencing loss of oxygen, that is, deoxygenation, faster than the global average (Cummins & Ross, 2020;Schmidtko et al., 2017). Whitney, Freeland, and Robert (2007) reported significant shoaling of the 1.5 m/L oxygen horizon in the NEP. Changes in oxygen levels are expected to alter seamount ecosystems and their associated communities (Clark, Watling, Rowden, Guinotte, & Smith, 2011;Levin, 2003;Wishner et al., 1990).
Concurrent with deoxygenation, total carbon dioxide (DIC) is increasing in the NEP, because carbon and oxygen are so closely linked through respiration (Anderson & Sarmiento, 1994) and due to uptake of anthropogenic carbon from the atmosphere (Gruber et al., 2019;Sabine et al., 2004). Increased carbon causes significant changes in ocean carbon chemistry, including increased acidity (reduction in pH; termed "ocean acidification," OA, Caldeira & Wickett, 2003;IPCC, 2011), and may negatively impact marine organisms in multiple ways (Haigh et al., 2015). The calcium carbonate saturation state (Ω) provides a chemical measure of the energetic cost for calcification, specific to polymorph (Supporting Information, Section S1). It places a constraint on calcifying organisms (e.g., Andersson, Mackenzie, & Bates, 2008;Spalding, Finnegan, & Fischer, 2017) because they must manipulate local chemistry to build and maintain their structures (Adkins, Boyle, Curry, & Lutringer, 2003;Weiner & Dove, 2003). OA increases these energetic costs by reducing Ω (Spalding et al., 2017). The present-day surface ocean is usually supersaturated (Ω > 1), becoming less saturated with depth, driven mainly by increases in pressure and DIC (Jiang et al., 2015;Li, Takahashi, & Broecker, 1969). Calcium carbonate polymorphs in direct contact with seawater below their associated saturation horizons (Ω = 1) will dissolve. Saturation horizons are already remarkably shallow in the old, carbon-rich NEP relative to other ocean basins  and are getting shallower (Feely et al., 2012).
Here, we focus on an NEP seamounts study region (Figure 1) that contains a dense cluster of 47 seamounts, 32 of which intersect the OMZ (oxygen < 1 ml/L), and is bisected by the long-term oceanographic monitoring section, Line P. We calculate trends in oxygen from 1960 to present in the deep ocean bathing these seamounts F I G U R E 1 Map of our Northeast Pacific seamounts study region. Line P oceanographic monitoring stations are indicated by red squares; major stations (P12, P16, and P26) are highlighted in yellow (main map) and black (inset). The triangle symbols indicate seamount summits in our study region (DFO, 2019), the four in black were surveyed during the 2017 benthic imaging survey cruise (Table 1). The proposed Offshore Pacific marine protected area is outlined in blue. Bathymetry is from GEBCO 2014 Grid, version 20141103, http://www.gebco.net and examine the mechanisms driving these trends. Similarly, we calculate deep Ω (aragonite and calcite) trends from 1987 to present.
We explicitly resolve the evolution of the OMZ and saturation horizons in time, while critically investigating uncertainties in the data and derived trends. Modern benthic seamount survey data from four of the seamounts in the region were analyzed to identify key indicators, including cold water corals, sponges, echinoderms, and rockfish, based on their depth distributions relative to the chemical horizons. We identify potential implications of oxygen and Ω trends to these key species considering their life histories and mineral compositions. Finally, we discuss these trends in the context of ecosystem diversity and marine protected area (MPA) planning, which is timely as this NEP seamount cluster is a significant component of a proposed MPA (DFO, 2019).

| Oceanographic data
The oceanographic data were collected as part of the Line P monitoring program (Figure 1), starting in 1959 and sampled at least three times a year (Cummins & Masson, 2012;Freeland, 2007;Mackas & Galbraith, 2012;Whitney et al., 2007). Discrete bottle chemistry data were collected throughout the time series along with electronic sensor measurements (i.e., CTD observations) starting in 1969 for temperature and salinity and in 2001 for oxygen (Seabird 43 dissolved oxygen sensor). For all data (oceanographic and ecological), we refer to depth rather than pressure (i.e., 1 dbar = 1 m) for convenience, even though at 3,000 dbar, this simplified conversion is off by ~1.5%. Bottle samples have been analyzed for oxygen (present: automated Winkler Photometric System) and macronutrient (present: Technicon Autoanalyzer II) concentrations since 1959, and are collected from 24 discrete depths at each major station ( Figure 1). Since 1987, a subset of the bottle samples have been analyzed for dissolved inorganic carbon (DIC) concentration and total alkalinity (TA) following the standard sampling and analysis protocol (Dickson & Goyet, 1994;Dickson, Sabine, & Christian, 2007; S1).
TA becomes more accurate later in the record when the open-cell alkalinity system (Dickson et al., 2007) was adopted. Like salinity, dilution (precipitation and evaporation) exerts the primary control over variations in ocean TA (Millero, Lee, & Roche, 1998), allowing us to use the recent TA observations to extend the record back in time to cover the full DIC time series using empirical relationships (S2).
We use Ω to indicate the biological impact of increasing DIC because Ω directly impacts calcifying marine organisms like corals (Kleypas et al., 1999) and echinoderms (Dupont, Ortega-Martinez, & Thorndyke, 2010) and because Ω and acidity (pH-which also affects animal physiology; Claiborne, Edwards, & Morrison-Shetlar, 2002) covary at the depths and salinity ranges in our study zone.

| Data analysis
Monthly, 5 m gridded temperature, salinity, density, oxygen and DIC data were prepared for each station (Figure 1) following Cummins and Ross (2020). Given the longer time series and more regular sampling of oxygen, the monthly oxygen data were averaged by year for subsequent analysis. The uncertainties calculated for the time series in each depth bin at each station reflected mainly the standard deviation between profiles for oxygen (intra-annual variability) and the uncertainties in the observations and constants used to determine the carbon system for Ω (S3).
Next, the data for the stations were averaged spatially to represent the changes over our NEP seamounts study region. Across all stations, oxygen below about 300 m behaved similarly with depth; thus, we use an uncertainty-weighted average across the stations P06-P16 (within the seamount study region) but also included TA B L E 1 Summary of 2017 seamount benthic imaging expedition (DFO-Pac2017-036)  Figure 1) to compute the area-average Ω profiles. The uncertainties in oxygen and Ω were propagated and combined with the spatial standard deviations (S3).
Multiple definitions of "hypoxic" appear in the literature (from 0.5 to 1.5 ml/L; e.g., Breitburg et al., 2018;Levin, 2003;Paulmier & Ruiz-Pino, 2009;Whitney et al., 2007). We define the OMZ as water with less than 1 ml/L oxygen (Clark et al., 2011) and label these waters as hypoxic and those with less than 0.5 ml/L as severely hypoxic.
Depths of the top and bottom of the OMZ and carbonate saturation state horizons (Ω Ca = 1 and Ω Ar = 1, 0.7) were determined for each oxygen/Ω profile and averaged over time and then space just like the study region averaged profiles.
Finally, all trends were calculated using a weighted least squares linear fit and the uncertainty on the trend estimated using bootstrapping (Efron & Gong, 1983). Trends in oxygen at fixed density were calculated by creating a regular density grid and using the coincident density profile to map oxygen onto that grid at each time step, then extracting data from a constant density bin. To average over the depth distributions of the indicator taxa (see Section 2.2 for definition), the depth distribution of the oxygen or Ω was averaged at each time step, weighting by the abundance of the animals in each depth bin. For all of these trends, the trend uncertainty was taken to be the 68% confidence intervals (equivalent to a standard deviation to combine it with uncertainty propagated from the temporal and spatial averages) as determined by bootstrapping; that is, randomly resampling (with replacement) the data to make 1,000 estimates of the weighted least-squares linear fit and looking at the 16th and 84th percentiles.

| Ecological data
Benthic ecology data were collected on the 2017 DFO expedition to Union, Dellwood, Unnamed (UN) 16, and UN 18 seamounts ( Figure 1; Table 1) aboard the CCGS John P. Tully. Visual surveys of the benthic ecosystems were conducted from 1 m above the seafloor using a DFO drop camera system. BOOTS (Bathyal Ocean Observation and Televideo System) has a forward-facing HD Minizeus video camera with 10 cm scaling lasers (Gale et al., 2017).
Potential transects were randomly generated for each seamount "side," with the final transect selection dependent on ocean and weather conditions (Table 1). A dive started with BOOTS at ~2,150 m (maximum equipment depth) in the morning, traveling slowly upslope (0.3-0.5 knots) to reach the summit by day's end (5-11 hr dives).
We present depth (bathymetric) distributions of a subset of indicator taxa, derived from video annotation of abundance for all observed taxa as a function of depth. From the 10 BOOTS dives, 75.5 hr of video was annotated (protocol in Curtis et al., 2015), resulting in 265,135 1 s records, from which individual living organisms were resolved to the lowest taxonomic level possible with confidence, then georeferenced. In total, 27 taxonomic groupings of cold water corals and sponges were resolved, with an additional 78 taxa of other invertebrates and fish, between approximately 300 and 2,150 m depth. These data were merged into a single dataset, which was deemed suitable because taxa showed similar depth distribution independent of dive (S5). For each of these 105 taxa, the depth distribution was quantified by calculating the mean (µ), standard deviation (σ; a proxy for depth range size), and upper and lower limits.

| Indicator taxa
The large number of observed taxa was parsed down into nine indicators by focusing on taxonomic groups with narrow depth distributions in three key depth zones (defined below). The idea is that a narrow depth range on the steep seamount flanks represents a narrow ecological niche, since many marine environmental variables are correlated with depth. Specializing within a narrow depth range increases vulnerability to environmental change, potentially leading to extinction or extirpation (Gallagher, Hammerschlag, Cooke, Costa, & Irschick, 2015). In addition, the smaller the inhabitable isolated area, the less likely an immigration or recolonization rescue (MacArthur & Wilson, 2001). In comparison, a generalist species that can thrive within a large depth range is likely more versatile and resilient to change.
Because corals and sponges are often large, conspicuous, sessile animals, they are candidates as proxy indicators for ocean health and monitoring environmental changes (Girard & Fisher, 2018).
Additionally, some of the same criteria that qualify them as defining features of Vulnerable Marine Ecosystems make them important in this study: they create unique/rare habitats; are of special importance for many species; are vulnerable and slow to recover; are biologically productive and diverse; and are structurally complex (

| Trends in oxygen, carbonate saturation state, and boundaries
The ocean bathing the NEP seamounts has been losing oxygen rapidly over the past 60 years, reducing the oxygen inventory in the upper 3,000 m by 15% ( Figure 2). While showing considerable interannual variability, the upper bound of the OMZ has remained constant over the 60 year time series (mean ± standard deviation: 480 ± 50 m). In contrast, the bottom of the OMZ (mean: 1,700 m) has much smaller variability and shows a significant deepening trend (3.1 ± 0.5 m/year, Figure 2).
The seamounts are almost entirely surrounded by water un-

| Seamount indicator taxa, their depth ranges, and the chemical changes therein
The exposed rocky flanks of the seamounts support dense forests of sessile, habitat-forming, filter-feeders such as soft corals F I G U R E 2 False color plots of dissolved oxygen (a) and aragonite saturation state (Ω Ar , b) over the Line P time series, averaged over our NEP seamounts study region ( Figure 1). Overlaid on (a) are the calcite saturation depth (yellow), and the upper (red) and lower (blue) boundaries of the oxygen minimum zone. The legend indicates the trends associated with each of these boundaries. In (b), the Ω Ar = 0.7 (red) horizon is shown. Also overlaid are the isopycnal depths for the 26.   (Figure 1). The red background shows the mean oxygen concentration over the full time series, highlighting the OMZ (i.e., white is outside the OMZ). The dashed white lines delineate the region between 800 and 1,200 m which corresponds to oxygen <0.5 ml/L and dotted blue line indicates the mean depth of the calcite saturation horizon. A range of metrics on the species' depth ranges are plotted (mean, horizontal black lines; ±1 standard deviation, dark gray boxes; maximum and minimum depth observed, thin dark gray lines). *The sea lily has a bimodal depth distribution (gap between 1,150 and 700 m); the black coral also has a bimodal distribution (dip between 900 and 850 m); the rest have roughly bell-shaped unimodal distributions (except for the rockfish, which is truncated at the top) (S5). All taxa were found on all seamounts with habitat (seafloor) in their depth range (mean ± 1 standard deviation) with the exception of the bugle sponge which was not observed on UN 16 (S5) TA B L E 2 Each indicator taxon, the zone it belongs to, its common and scientific name, number of individuals observed in the 2017 visual surveys, the mean (µ) and standard deviation (σ) of its depth range (2017), its vertical distribution-weighted mean oxygen level and trend (1960-2019); its vertical distribution-weighted mean calcite and aragonite saturation states (Ω

TA B L E 2 (Continued)
the three shallowest seamounts, while the undulated glass sponge inhabited all four (S5).
These eight taxa had roughly unimodal depth distributions narrower than over two-thirds of all other annotated organisms (except black coral; S5). For example, these depth specialists inhabited depth ranges three to nine times smaller than the ubiquitous Florometra serratissima (2σ 952 m; Figure 4h). This sea lily was included in the analysis to represent widely distributed taxa, as dense fields or individuals were observed on all four seamounts, across ~2,150 to 550 m, almost the entire depth range surveyed, with a gap between 1,150 and 700 m (bimodal distribution; S5).
The oxygen and saturation states integrated over the depth bands of these taxa show that current chemical conditions may already cause some organisms stress (Table 2). Furthermore, the oxygen in the water bathing the corals and sponges is declining so rapidly that if trends continue, in the worst-case scenario, conditions may become fatal within many of their typical life spans (Table 2).
The absence of this compounding effect may allow for sponge taxa to persist under deoxygenation longer than the corals. The mobility of the sea lily, as well as its broad depth range, may help it endure the changing oxygen conditions. Still, within several generations, if trends continue, oxygen within the entire modern depth range of the sea lily will disappear.
While there is no significant long-term oxygen trend within the rockfish depth range, the movement of the OMZ upper bound likely has and will continue to limit the small habitable area of this hypoxia-sensitive fish. In 2017, it appears a threshold existed near the OMZ upper bound (µ−σ = 435 m), which restricted the species to the shallowest part of its preferred depth range (800-200 m, down to 2,830 m; Table 2). Carbonate saturation states are also declining at and above these depthsand may cause direct or indirect (i.e., food web) impacts (Haigh et al., 2015;Hamilton et al., 2017). Brittle stars are vulnerable to OA (Table 2), and the mean carbonate saturation state bathing the ~200 vertical meters of dense brittle star mats, which play a significant role in seamount energy transfer and represent a large proportion of the local benthic productivity and biomass, may be declining.

| Climate change impacts on oxygen and calcite in the intermediate and deep NEP
The waters bathing the seamounts are some of the oldest in the global oceans. The majority were last at the surface (i.e., ventilated) allowing oxygen uptake from the atmosphere (and CO 2 expulsion) ~1,000 years ago (Gebbie & Huybers, 2012), prior to the Anthropocene. Thus, the significant chemical changes that we observe on timescales of decades in the deep water are not expected.
Only the tallest summits (Union and Dellwood, Table 1)  The oxygen trends in this zone (~120-700 m depth in the NEP) are consistent with similar observations on common density surfaces near the ventilation region (Nakanowatari, Ohshima, & Wakatsuchi, 2007;Sasano et al., 2015) and previous local analyses (Crawford & Peña, 2016;Whitney et al., 2007). These are driven at least partly by a reduction in ventilation associated with climate change (Deutsch, Emerson, & Thompson, 2006). Changes in circulation have forced the downward migration of isopycnals (Figures 2 and 3). It is this migration that brings oxygen to the top of the OMZ and keeps its upper boundary from shoaling, even though oxygen is decreasing on each isopycnal (Figure 3a).
This observation is most likely due to circulation changes and possibly flow convergence further offshore in the NEP between ~100 and 500 m (Cummins & Ross, 2020), which may result from changes in wind forcing or distant overturning circulation (Ohshima, Nakanowatari, Riser, Volkov, & Wakatsuchi, 2014).
Slowing down of flow may also be responsible for the changes in oxygen on deep isopycnals. While the deep water is too old to have experienced ventilation during the Anthropocene (mean age 1,000-1,500 years, Gebbie & Huybers, 2012;Matsumoto, 2007), anthropogenically driven changes in circulation (Deutsch, Emerson, & Thompson, 2005;Deutsch et al., 2006;Masuda et al., 2010;Talley et al., 2016) could mean that it is progressively older and has experienced more oxygen consumption (e.g., Lauvset et al., 2020) when it arrives at the seamounts. Trends were weaker on the deeper isopycnals, but they were always negative. Changes at these depths could also result from increased organic matter export (causing oxygen consumption).

| Potential loss of ecosystem diversity
The changing chemical environment in the deep NEP will likely cause significant loss of seamount ecosystem diversity. This loss can be investigated using ecosystem classification systems based on environmental variables. For instance, Clark et al. (2011) identified five "biologically meaningful" environmental variables as part of a seamount classification system (biogeographic province, export productivity, summit depth, dissolved oxygen, and proximity) and showed that it gives biologically realistic groupings in comparison with biological data. At present, this scheme shows four classes among the seamounts in our study region (DFO, 2019;S6); Union is the single member of one class (oxic; summit <800 m), Dellwood is in another (hypoxic; summit <800 m), and the UN seamounts are in a third (hypoxic; summit >800 m). Seamount summit oxygen level is a key variable in the scheme, highlighting the potential impact of climate-driven trends. Projected changes in oxygen at the summits in our study area-from oxic to hypoxic-will cause a loss of the second largest seamount class (oxic; summit >800 m) as it transforms into the hypoxic class at a rate of approximately 1 seamount every 8 years. If the OMZ continues to expand at its present rate, an entire seamount class-representing an ecosystem type-will be lost by 2240, homogenizing the seamount cluster.
Additionally, while taxa were found at similar depth bands across the different seamounts sampled (Figure 5), their absolute and relative abundances were not the same, that is, community structure differed (S5  Figures 2 and 5).
Union supports a unique offshore shallow-water community (isolated from the continental slope by 100s of km; Figure 1; Table 1).
Additionally, it is home to a number of corals ( Figure 4) that already live in low Ω waters and will experience increasing stress in coming years.

| Potential impact of chemical trends on NEP seamount communities
While classification schemes are useful for simplifying the description of seamount ecosystems, their thresholds are only meaningful to the bulk assemblage of species (Clark et al., 2011). Individual, or groups of similar, species are influenced by their local environment-even if they are not near an identified threshold. While temperature change drives important changes in Atlantic seamounts (Puerta et al., 2020), temperature trends in the NEP are much weaker (<0.5°C/century below 500 m, Cummins & Ross, 2020), and thus, we have focused on chemical trends. With strong chemical gradients at many depths, the narrow depth distributions of the indicator taxonomic groups identified here (Figures 4 and 5) suggest that their ranges are already limited by their chemical environment Wishner et al., 1990). Therefore, they may be particularly sensitive to oxygen and Ω trends, especially if they are long-lived-as are many of the indicator taxa (see Table 2)-or run out of suitable habitat (e.g., if their ideal chemical zone, e.g., Table 2, passes the summit of the seamount). While variability that overlies the observed trends (Figures 2 and 3) could also play a role in triggering distribution changes, we focus on potential impacts of the long-term trends (Table 2). This variability is accounted for in the trend uncertainties (S3); if a trend is significant, it has "emerged" from the background variability (in  Table 2). Similarly, on Union seamount, these rockfish were observed inhabiting primarily the upper half of their preferred depth range (Table 2), seemingly to avoid the OMZ. Additionally, the size, weight, and abundance of rockfish are expected to decrease with lower dissolved oxygen levels (Cross, 1987;Keller et al., 2017;Koslow, Goericke, Lara-Lopez, & Watson, 2011;Vetter & Lynn, 1997). There may also be negative impacts associated with OA (Hamilton et al., 2017).
pertusum reefs in the Southeast Atlantic, an extensive food supply may be increasing our corals capability to adapt to extreme hypoxic conditions (e.g., Hebbeln et al., 2020). Both the bamboo coral and bubblegum coral were found in waters with low oxygen (Isidella spp.:  (Table 2). Chu et al. (2019) show that black coral's habitat influences do not appear to include inorganic carbon or oxygen.
Like black coral, the glass sponges, both bugle (Pinulasma n. sp.) and undulated glass (cf Tretodictyum n. sp.), are not composed of carbonate, and therefore, Ω is not expected to limit their niches (Table 2). However, oxygen levels below 1 ml/L may pose physiological limitations for glass sponges, and they are already rare when oxygen <1.4-2.1 ml/L and silicic acid, the key building block for glass sponges, is abundant (Leys et al., 2004;Whitney et al., 2005). Glass sponge reefs were present over 500 million years ago (Gehling & Rigby, 1996), implying enormous resiliency to extreme conditions in both oxygen and Ω (Bindoff et al., 2019), which they will need in the coming centuries.
Models predict that North Atlantic seamounts may act as temporary refugia from deep OA for cold water corals (Tittensor, Baco, Hall-Spencer, Orr, & Rogers, 2010), but that climate-induced changes will ultimately reduce suitable habitat for corals (up to 100% reduction) and fishes, by 2100 . While Ban, Alidina, Okey, Gregg, and Ban (2016)

| Mortalities associated with low oxygen
Although some organisms are adapted to low oxygen conditions, all need oxygen for respiration and if deprived for too long, mortality will result, even for the resilient glass sponges (Leys et al., 2004).
Our data show visual evidence that bamboo coral, living near the upper bound of the OMZ and so generally tolerant of hypoxia (Etnoyer, 2008) and oxygen variability (Figure 2), likely experienced one or more episodic mortality events. Starting at 650-700 m, there is a sharp decline in living bamboo coral abundance, with very few found within the severely hypoxic OMZ core and none below (S5). Thus, the offshore population of this species of conservation concern and commercial value (DFO, 2012) is already under threat. As one of the longest lived fish species (up to 205 years old; Table 2), rougheye rockfish population recovery could take decades to centuries.

| Difficulty in forming structural elements under ocean acidification
A first-order effect of OA is the increased energetic cost to form carbonate body parts (Spalding et al., 2017), such as sclerites in corals. Many organisms are successful in low Ω waters (Figures 2   and 5), but energetic costs will increase as Ω decreases under OA, leaving less energy for other functions like reproduction (Kroeker et al., 2013). For example, some echinoderm species have shown slowed arm regrowth (Donachy & Watabe, 1986) and muscle wastage (Wood, Spicer, & Widdicombe, 2008) under low Ω. Brittle stars contribute significantly to local seamount productivity and provide community structure. Their mean depth is just below the depth range of significant Ω trends; some portion is likely experiencing decreasing Ω (Table 2; S5). Brittle stars have strong but highly soluble high-magnesium-calcite structures, so they already require significant energy to calcify (Table 2). Arm regrowth is frequent (Donachy & Watabe, 1986); thus, the stress of declining Ω threatens what may already be a delicate balance for this taxon. Sea lily structures are more soluble than those of brittle stars (Iglikowska, Najorka, Voronkov, Chełchowski, & Kukliński, 2017); however, sea lilies calcify less frequently and are more fleshy (P. Lambert, personal communication, November 11, 2019), with organic coatings isolating body parts from seawater.
Organic coverings are likely key to the success of several of our indicator taxa. As with the brittle stars, parts of other echinoderms (McClintock et al., 2011 and references within) and bubblegum and bamboo corals (Bayer & Macintyre, 2001;Bostock et al., 2015;Haigh et al., 2015;Le Goff et al., 2017;Thresher et al., 2011) are composed of high-magnesium-calcite. Bubblegum corals (Pargagorgia spp.) in Antarctica tolerate low Ω (Ω Ar ~ 0.7) due in large part to the fact that the minerals are covered with organic material and thus not directly exposed to seawater (Bostock et al., 2015).  (Table 2; Jantzen & Schmidt, 2013). Survival in the low Ω waters on NEP seamounts must come at a significant energetic cost (Carreiro-Silva et al., 2014;Gómez, Wickes, Deegan, Etnoyer, & Cordes, 2018).
Even for species that contain no or little carbonate, if they are living (and breathing) animals, they must maintain ion transport over membranes (which can be impacted by pH) and must rid themselves of CO 2, so there will likely be OA impacts on chemical regulation in all animals (Claiborne et al., 2002). Ion transport difficulty may be a factor in the observation that high CO 2 levels (1,125 µatm) induced anxiety relative to ambient CO 2 levels (483 µatm) in rockfish, as observed through scototaxic (light/dark preference) testing (Hamilton, Holcombe, & Tresguerres, 2014). There is also some evidence suggesting that OA impacts will be more severe in larval stages (e.g., Hurst et al., 2019).

| Ocean acidification impacts on larvae and dispersal
OA impacts many aspects of reproduction; spawning, larval development, and juvenile mortality. For example, Primnoa pacifica (same suborder as the bamboo coral; inhabits the study seamounts; data not shown) experienced reduced spawning in low pH (pH = 7.55; Rossin, 2018 (Nilsson & Sköld, 1996). Behavioral change under low oxygen has been shown for glass sponges; they reduce (energetically expensive) filtering (i.e., feeding) rates (Leys & Kahn, 2018).
However, they can likely survive long periods of low oxygen if currents are strong enough because they take advantage of currents for feeding (Leys et al., 2011). OA may also impact membrane pumps of glass sponges, reducing feeding efficiency (Conway, Whitney, Leys, Barrie, & Krautter, 2017).
The biggest effect of deoxygenation and OA may come through food supply. As the OMZ expands-particularly the upper boundary-the depth to which zooplankton and myctophids migrate daily may shoal as they avoid hypoxic waters (Bianchi, Galbraith, Carozza, Mislan, & Stock, 2013). This daily migration directly and indirectly delivers organic carbon to depth (Steinberg et al., 2000). Hypoxic conditions have been shown to affect the daytime zooplankton depth in many locations (Devol, 1981;Netburn & Koslow, 2015). This avoidance of a rising OMZ boundary would have cascading effects on the already food-limited systems below (compounding any coincident increase in energy demand). The resident detritus-based food web on seamounts is already considered fragile, with many organisms near the margin of energetic sustainability, and has also taken a long time to develop (Pitcher & Bulman, 2007). Additionally, OA may increase primary production slightly (Eberlein et al., 2017;Haigh et al., 2015) but decrease its nutritive value causing cascading negative impacts to secondary producers (Riebesell et al., 2007;Rossoll et al., 2012).

| Implications for seamount management
States and organizations are moving to protect and monitor an unprecedented amount of the world's oceans. Many seamounts, including those in and around our study area, are identified as marine conservation targets and are within proposed marine protected areas (MPAs), whose objectives are to safeguard the wealth of marine animals living on the seamounts from harmful human activities (e.g., the proposed Offshore Pacific MPA, DFO, 2019; Figure 1). At the same time, the area of OMZs is expanding worldwide (Breitburg et al., 2018). Our findings show that benthic ecosystems and species in our NEP seamounts region are experiencing changes in their chemical environment. These deep-water chemical stressors have a remote source and cannot be prevented by any spatial management plan. However, impacts from these large-scale stressors would be compounded by others that the MPA can mitigateeliminating the controllable impacts may allow these ecosystems to withstand climate-forcing changes better or for longer (Puerta et al., 2020). With seamounts, seamount-like features (Yesson et al., 2020), expanding OMZs (Breitburg et al., 2018), and other climate-forcing changes (e.g., Sweetman et al., 2017) distributed throughout the world's oceans, this study highlights the global importance of mitigating direct human impacts as species continue to suffer environmental changes that are beyond our control in the immediate future, and the need for climate-informed ecosystem-based management, monitoring, and protection, including built-in resilience to uncertainties or projected changes in ocean conditions over multiple timescales (e.g., Gehlen et al., 2015).

| SUMMARY
Benthic communities on NEP seamounts are experiencing significant long-term changes in their chemical environment; the OMZ is expanding, deep oxygen levels are decreasing (15% since 1960), and waters above the OMZ are becoming more corrosive. These changes are caused by global reduction in ventilation, anthropogenic carbon accumulation, and, likely, a slowdown in deep largescale circulation. Many organisms already occupy narrow depth zones at or beyond their optimal oxygen limits. If these chemical trends continue as they have over the last three to six decades, the seamount ecosystems and their associated animals will be severely impacted. The study taxa are a subset of indicators that represent four phyla and have varying depth distributions, ecological roles, and life-history traits. Despite the biological variety and distribution across almost two vertical kilometers, evidence shows all of these taxa are vulnerable to the observed chemical trends-results that are likely applicable to many seamount taxa, even the generalists. Some populations are likely already experiencing impacts, while others may be more resilient (potentially glass sponges and sea lilies). Regardless, we predict that within the next century, the indicator taxa, many cold water corals, will suffer the compounding effects of, among other things, impacts to their distribution, respiration, metabolism, growth, dissolution, feeding, behavior, reproduction, and, ultimately, their survival or extirpation in our changing oceans. These trends and ecological implications warrant moving beyond the precautionary management and protection approaches toward plans that support and build resilience for future large-scale declines in ocean ecosystem health.

ACK N OWLED G EM ENTS
We acknowledge the contributions of Tammy Norgard throughout this work. We are indebted to all the scientists and crew members who have made the Line P monitoring program possible for over 60 years, especially Marie Robert and Frank Whitney; current and previous program leads, respectively. We thank Chelsea Stanley Canada funding and the detailed reviews provided by anonymous reviewers.

DATA AVA I L A B I L I T Y S TAT E M E N T
The chemical and physical data that support the findings of this study are openly available from the Line P data archive at https://