Barkley Canyon is a shelf-incising submarine canyon located ∼100 km west of Vancouver Island in the NE Pacific ( Figure 1 ). It possesses multiple head branches intersecting the continental shelf, all discharging into the main canyon axis at mid-slope depths of 400–600 m ( Figure 1 ). Barkley Canyon’s seafloor between 350 to 2000 m provides a range of benthic habitats with distinct biological communities (Campanyà-Llovet et al., 2018; Domke et al., 2017). At mid-canyon depths near ~1000 m, hydrate outcrops with methane-seep communities (vesicomyid clam beds and bacterial mats) occur near a plateau on the canyon’s northern flank (Thomsen et al., 2012; Riedel et al., 2022), with sparse cold-water coral communities occurring only a few hundred meters to the southeast (De Leo et al., 2024). The seeps are surrounded by bioturbated soft sediments, with megabenthic communities dominated by tanner crabs (Chionoecetes tanneri), buccinid gastropods (Neptunea pacifica), sablefish (Anoplopoma fimbria), and Pacific hagfish (Heptatretus stoutii) (Juniper et al., 2013; Seabrook et al., 2018, 2019).

Permanent hypoxic conditions, with dissolved oxygen concentrations below 0.5 ml.l^-1 (Diaz and Rosenberg, 2008; Vaquer-Sunyer and Duarte, 2008), occur between 870 and 970 m in Barkley Canyon (m=0.288; sd=0.017), with concentrations as low as 0.23 ml l^-1 (Ocean Networks Canada’s 12-yr time-series of observations). The low oxygen values coincide with the core of the oxygen minimum zone (OMZ) along this portion of the Cascadia Margin (De Leo et al., 2017), and appear to be an important driver of reduced megabenthic abundance and diversity in the canyon (Chauvet et al., 2018; Domke et al., 2017) and on the adjacent slope (De Leo et al., 2017).

Our study utilized Ocean Networks Canada’s seafloor cabled observatory infrastructure installed in the Barkley Canyon Wall (BCW) location at 870 m depth on soft sediments (De Leo et al., 2018), about 500 m from the nearest methane seep ( Figure 1 ). The observatory installations allowed real-time monitoring with a high-definition video camera system in combination with standard oceanographic sensors, which have been measuring core essential ocean variables (e.g., temperature, salinity, pressure, dissolved oxygen, turbidity, current speed and direction) at 1-Hz frequency since May 2009 (Barnes et al., 2010; De Leo et al., 2018).

On May 13 2014, three fresh-frozen and thawed adult humpback whale (Megaptera novaeangliae) cylindrical rib sections which had the bulk of flesh removed (each ~10-cm diameter and 30-cm long), one 20 x 20 x 10 cm block of untreated Douglas Fir wood (Pseudotsuga menziesii), and a ~30 x 30 x 30 cm block of authigenic carbonate (retrieved from a methane seep ~1000 m deep on the Costa Rica margin in 2009) were placed by a remotely operated vehicle (ROV) at 890 m depth near the northern wall of Barkley Canyon ( Figure 1 ). All substrate packages were mounted on 35.5 x 35.5 cm, high-density PVC plates 20 cm above the seafloor on fiberglass grating, and positioned ~1 m from each other and about 50 cm concentrically from a high-definition video camera connected to ONC’s cabled observatory ( Figure 1C ). Over a period of 8.3 months, 5-minute videos were recorded at 2-hr intervals. The camera system was mounted at a height of 65 cm above the seabed on a stainless-steel tripod frame with a ROS-485 Pan and Tilt (PT) unit. This allowed panning of the camera to each of the three substrates, and then holding position for at least 1 min during video recording. Two dimmable ROS LED lights (100 W, > 406 lm) provided illumination during video recording, and parallel laser beams 10-cm apart provided scaling. The 5-min recording/lights-on,1:55-h off schedule used is the standard for all video cameras installed on NEPTUNE’s Barkley Canyon observatory node, and is intended to limit behavioral artifacts in deep-sea organisms exposed to artificial lighting (Robert and Juniper, 2012). The standardized and automated routine of video recordings was started on May 16 at 00:00:00 hrs (UTC), 3 days after the experiment was deployed by the ROV. ONC’s online SeaTube Pro video player web interface (https://data.oceannetworks.ca/SeaTube) was used for performing video annotations. Animals were identified to the lowest possible taxonomic level and counts were performed from 30 s video segments from the 5-min videos at 2 different frequency intervals: 1) every 6 hours for the first 30 days, and 2) every 12 hours for the next 7.3 months. Only animals above the experiment plates were counted. Animals down to a length of ~1 cm (i.e., megafauna) could be identified. The video time-series was abruptly interrupted on January 12, 2015, ~8.3 months after the initial experiment deployment, when the Barkley Node subsea infrastructure was damaged by a bottom-trawling vessel. After this interruption, we used ROV video surveys to qualitatively assess multi-year trends in colonization by ecosystem engineers (i.e., Osedax sp., and Xylophaga sp.), bone/wood substrate degradation, and occurrence of bacterial mats. One of the three whale-bone rib sections was recovered on May 15 2016 (2 yr), a second one on August 1, 2018 (4 yr 3 mo), and the third and final along with the wood block were recovered on July 16 2023, after 9.2 years on the seafloor. Recovered bones and wood were frozen and examined under a compound microscope to assess the presence of Osedax/Xylophaga.

Bacterial-mat development on the whalebones was quantified from video frame grabs extracted at 72 h intervals for the entire 8.3 mo high-resolution observation period. The coverage of bacterial mats, identified as bone area covered by white filaments, was measured in square pixels using the masking tool in the post-processing image analysis software Image J (Schneider et al., 2012). No Bacterial mats were evident on the wood and carbonate substrates.

Co-located oceanographic instruments with sensors ~1.5 m above the seafloor, including a Sea-Bird Conductivity-Temperature-Depth-Dissolved Oxygen (CTD-O) sensor, a 2 MHz Nortek Aquadop Acoustic Doppler Profiler (ADCP), and a Wet Labs Eco FLNTU fluorometer/turbidity optical sensor, allowed the measurement at 1-Hz of the following environmental variables: bottom-water temperature (°C), salinity (PSU), dissolved oxygen concentration (ml.l^-1), horizontal current velocity (u, v, m.s^-1), acoustic backscatter (db), chl-a fluorescence (mg.l^-1) and turbidity (nephelometric turbidity units, NTU). All oceanographic data streams used in this paper are freely available for download at: https://data.oceannetworks.ca/DataSearch (search tree path: Pacific>Northeast Pacific>Barkley Canyon>MidWest and > MidEast).

Sea-surface environmental variability during the study period was assessed from long-term data available from the La Perouse Bank meteorological buoy station (C46206, latitude 48°50’24”N, longitude 126°00’00”), 60 km to the NE of the study site. The buoy is maintained by Environment and Climate Change Canada (ECCC), and data are accessible through the US National Oceanographic and Atmospheric Administration (NOAA) National Data Buoy Center (NDBC) (https://www.ndbc.noaa.gov/station_page.php?station=46206). Variables utilized were sea-surface temperature (°C), atmospheric pressure at sea level (dbar), significant wave height (m) and period (s), and wind speed (m/s) and direction. In addition, 8-day average ocean surface color data from the MODIS satellite, integrated over an area of 144 km² centered above our study site, were used as a proxy for surface primary productivity, following the same approach as Chauvet et al. (2018). Data were obtained from the NASA Earth Observations (NEO) Goddard Ocean Colour Group (https://modis.gsfc.nasa.gov/) and processed using SeaDAS data analysis software (Stramska, 2014).

A total of 10 faunal taxa were observed across all substrates ( Table 1 ), with a total of 4650, 5576 and 4877 observations of faunal individuals on whale bones, wood and carbonate, respectively, during the 8.3 mo (250 d) of high-frequency observations (see Supplementary Videos 1-3 for a quick overview of community succession over the 8.3 mo). Faunal colonists on all substrates over the 8.3 mo consisted of mobile species previously observed or collected (including by baited traps) in the surrounding community, with no evidence of bone/wood borers (i.e., decomposers) on bone/wood substrates.

Whale bones: Early colonization of whale bones was dominated by scavengers/omnivores. Multiple hagfish (Eptatretus deani), the low-oxygen-tolerant lysianassid amphipod Orchomenella obtusa, and shrimp (Heptacarpus sp.) were observed presumably feeding on the whale bones within the first five days ( Figure 2A ). Over the next 25-30 d, Orchomenella obtusa and Heptacarpus increased substantially in abundance, attaining peak levels of 16 and 93 individuals, respectively. Lysianassid amphipods remained the most abundant taxon on whale bones until ~45 d, and then declined rapidly as coverage by bacterial mats increased ( Figure 2A ). The abundance of Heptacarpus remained relatively stable from 30 d to 250 d, ranging up to 26 individuals. Colonization of whale bones by bacterial mats (dense white filaments) was first observed after three days, generally increased to a peak area near 140 days, and then declined to relatively stable lower levels from 170-250 d ( Figure 2A ; Supplementary Video 1 ). No evidence of bone-boring siboglinid worms in the genus Osedax (e.g., mucus sheaths, plumes, or bone borings) was observed on the whale bones over the entire 250-d period. Periodograms calculated from the 48-hr running averages of the two most abundant faunal species showed significant periods of 79 and 80 days only for Orchomenella obtusa ( Supplementary Figure 1A ; Table 2 ). Total faunal community abundance, also based on 48-h moving averages, peaked at ~80 individuals at ~35 days, while species richness peaked at 3 species after about ~12 d ( Figures 3A, B ). Alpha diversity on whalebones, as measured by Shannon H’ and Pielou’s J, showed initial peaks during the first~10 days, then decreased to about 40 d, and then increased to relatively constant values for the rest of the observation period ( Figures 3C, D ).

Approximately yearly ROV video monitoring of the whale bones following the 8.3 mo of high-frequency observations, as well as microscopic examination of recovered bones, indicated no colonization by Osedax of whale bones over a total period of 9.2 y ( Supplementary Figure 3 ). Bacterial mats persisted on the bones from 8.3 mo to 9.2 y, but areal coverage of mats on bone surfaces declined as sediment and brown hydrozoans accumulated to cover some areas of the bones ( Supplementary Figure 3 ). Heptacarpus sp. continued to be present at low abundances on the bones throughout the 9.2 y of the study.

Wood Substrate: Orchomenella obtusa and Heptacarpus sp. occurred in the first observation at 3 d. Orchomenella then increased rapidly to a secondary peak of ~50 individuals at 30 d, declined to zero abundance at ~48 d, then built up rapidly to a maximum of 57 at 59 d, and then gradually declined to essentially zero levels by 118 d ( Figure 2B ). Heptacarpus sp. very gradually increased from low abundance at ~30 d to a peak of 38 individuals at 243 d. Other species (Munidopsis quadrata, Chionoecetes tanneri) occurred as singletons or doubletons early (<51 d) and late (>150 d) in the 8.3 mo high-frequency observation period. No evidence of the wood-boring bivalve Xylophaga, such as siphons, borings,or feces, was observed at any time over the initial 8.3 mo. Periodograms showed significant periods for both O. obtusa between 60-75 and 110-120 days, and for Heptacarpus sp. between 34-36 and 45-46 days ( Supplementary Figure 1B ; Table 2 ). Community abundance based on 48-h moving averages showed secondary peaks of ~45 individuals at ~12 d and 215 d, with a primary maximum of ~60 inds. at ~60 d ( Figure 3A ). Species richness, also based on 48 h moving averages, peaked at 3 species around 25 d ( Figure 3B ). Shannon H’ and Pielous’s J were relatively high and constant during the first ~100 days, and then decreased to relatively low levels for the rest of the 8.3 mo observation period ( Figures 3C, D ).

Approximately yearly ROV video monitoring of the wood substrate following the 8.3-mo high-frequency observation period first revealed Xylophaga protruding from the wood block after 2 y ( Supplementary Figure 4 ). Nonetheless, the wood remained largely intact, with sharp corners and only small boreholes visible, at least until 4.3 y. By 7.3 y, the wood block appeared to be losing its sharp corners apparently from Xylophaga boring ( Supplementary Figure 4 ). Recovery of the wood substrate after 9.2 y revealed heavy riddling by Xylophaga borings, with very few live Xylophaga individuals remaining. From 4.3 y onward, the wood substrate became increasingly covered with brown hydrozoan colonies, achieving very heavy coverage with no visible gaps by 7.3 y ( Supplementary Figure 4 ). Heptacarpus persisted at low abundance on the wood substrate throughout the 9.2 y of observations.

Carbonate Substrate: The shrimp Heptacarpus sp. was present on the carbonate on day three with three individuals, and increased in abundance to 40 inds. at ~38 d ( Figure 2C ). Abundances remained relatively high throughout the 250 d of high-frequency observations. The amphipod O. obtusa reached abundances of 7- 9 individuals at 18 - ~58 d, then generally occurred at lower abundances (2-4 individuals) to ~120 d, and then intermittently as singletons ( Figure 2C ). Several other species, including the lithodid Paralomis verrilli, the squat lobster Munidopsis quadrata, the crab Chionoecetes tanneri, and the zoarcid fish Licenchelys sp., occurred sporadically as singletons or doubletons on the carbonate block. Periodogram analysis showed significant periods for Heptacarpus sp. between 35-40 and 70-76 days and for O. obtusa for 60, 70, 100,105,110 days ( Supplementary Figure 1C ; Table 2 ). Total community abundance (48 h moving average) exhibited multiple peaks of 38-40 individuals throughout the 250 d ( Figure 3A ). Species richness (48 h moving ave.) peaked at 3 at ~100 days ( Figure 3B ). Shannon H’ and Pielous’s J on the carbonate substrate increased over the first 12 days, declined until ~25 days, and then remained relatively low compared to whale-bone and wood substrates ( Figures 3C, D ). During approximately annual ROV monitoring from 8.3 mo to 9.2 y, the carbonate substrate became increasingly colonized with brown hydrozoans, with most surfaces covered by 7.3 y. Heptacarpus sp. persisted at low to moderate abundance on the carbonate out to 9.2 y ( Supplementary Figure 5 ).

Early patterns of faunal diversity and abundance differed between the bone/wood substrates and the carbonate block. Colonization of bone and wood substrates within the first 50-60 days was dominated by the scavenger O. obtusa, but bone substrates supported higher amphipod abundance. Amphipod abundance on bone substrates declined rapidly after ~45 days coincident with rapid growth of bacterial mats on whale-bone surfaces ( Figure 2A ; Supplementary Video 1 ); this rapid decline on the bones coincided with an abrupt increase in O. obtusa on the wood. However, overall amphipod abundance was significantly correlated across bone and wood substrates (τ = 0.528, p < 0.001). In contrast, O. obtusa abundance on the authigenic carbonate never attained high levels ( Figure 2C ), but was also significantly correlated with amphipod abundances on whale-bone (τ = 0.490, p < 0.001) and wood (τ = 0.607, p < 0.001) substrates. Furthermore, on the whale bones, community composition changed dramatically as bacterial mats became abundant at ~50 d ( Figure 4 ). Heterocarpus remained more abundant on the wood and carbonate substrates from ~100 d to 250 d than on the whale bones ( Figure 2 ).

Remarkable changes in community composition, indicated by beta diversity analyses, occurred over the first months of the experiment on all three substrates, in particular in the first ~25 days. The LCBD time series identified a period (day 54-113) of gradual community changes toward stabilization on wood and carbonate substrates, while communities established on the whale bones experienced substantial variations until ~175 d ( Figure 3E ). The species contribution to the beta diversity index (SCBD) ( Table 1 ) indicates the contribution of individual species to temporal variations in community composition. Among ten identified species, O. obtusa and Heptacarpus sp. made major contributions to temporal variations in community structure on whale-bone and wood substrates ( Table 1 ). On the carbonate substrate, only O. obtusa made major contributions to beta diversity, with much higher SCBD value changes compared to other species, indicating its importance in driving community heterogeneity ( Table 1 ). Total beta diversity (BD_TOT) values ( Table 1 ), suggest that communities on bone and wood varied more over time than on the carbonate substrate.

Community Trajectory Analysis (CTA) indicate largely directional dissimilarity changes in wood and carbonate communities ( Figure 4 - from left to right in the nMDS plots), with both communities converging at the end of the observation period (τ = -0.476, p = <0.001; Mann-Kendall test). Whale-bone communities exhibited directional change for the first ~120 days, but then lapsed into large, multidirectional changes, and diverged from wood (τ = 0.396, p = <0.001) and carbonate communities (τ = 0.140, p = <0.001). Trajectory lengths (L) were higher and directionality (DIR) was lower in the whale-bone substrate (L = 56.3, DIR = 0.393) compared to wood (L = 18.6, DIR = 0.533) and carbonate (L = 17.4, DIR = 0.501). Kendall rank correlation shows a positive trend (τ = 0.284, p = <0.001) between step lengths and bacterial-mat area on whale bones (i.e. when there was greater bacterial-mat coverage, the community was more dynamic).

Community responses clearly depended on the substrate (_adjR² = 0.597, p = 0.002) ( Figure 5 ). During the first 110 days, the communities on the wood and whale-bone differed substantially from that on the carbonate. After 110 days, communities on wood and carbonate converged, while those on the whale-bone remained quite different. Early differences between bone/wood and carbonate appeared to be driven by the abundance of O. obtusa, while later convergence between wood and carbonate appeared to be driven by relatively high abundances of Heptacarpus sp. ( Figures 2 , 5 ).

Environmental variables at ~1.5 m above the seafloor in Barkley Canyon are indicated at a 1-Hz frequency in Figure 6 . Throughout the 250 d of observation, dissolved oxygen concentrations were severely hypoxic, ranging between 0.23 - 0.29 ml.l^-1, with a mean of 0.25 ml.l^-1 (sd = 0.01) ( Figure 6A ). Bottom-water temperatures had a mean of 3.82 °C (max=4.26, min=3.4, sd=0.12). Turbidity averaged 0.24 NTU (max=24.2, min=0.05, sd=1.2), and chlorophyll-a concentration averaged 0.08 ug.l^-1 (max=0.46, min=0.07, sd=0.01) for the entire period ( Figure 6C ) with each of these variables exhibiting a few, isolated high peaks. Current velocities averaged 7.5 cm.s^-1 (max=33.5, min=0.05, sd=1.2), in a predominantly southerly direction ( Figure 6D ). Acoustic backscatter at 1.2-2.0 m above the seabed, which tracked suspended particle load, averaged 30.5 dB (max=49, min=27.1, sd=2.16) ( Figure 6E ).

At the sea surface, winds, wave height, atmospheric pressure, and SST exhibited the expected strong seasonal changes characteristic of the NE Pacific ( Figure 7 ). Wind speeds, which averaged 6 m.s^-1 (max=18.7, min=0, sd=3.5), significant wave height in meters (av=2.00, max=6.75, min=0.62, sd=1.10 m), and atmospheric pressure, in bars (av=1.01, max=1.04, min=0.98, sd=0.01), all exhibited enhanced variability during fall and winter months ( Figures 7A–C ). SST also followed a declining trend from summer to winter (av=13.3, max=17.7, min=9.5, sd=1.46). It is important to note that SST values were significantly higher, up to 3°C, than the 30-year average for the region, as our study period encompassed a severe and persistent marine heat wave, called ‘The Blob’ (Kintisch, 2015).

MODIS-Aqua chlorophyll, used as a proxy for sea-surface productivity, also showed strong temporal variability, ranging from 0.26 to 90 mg.m^-3, with highest values in spring and summer of 2013 ( Figure 7D ).

Communities established on all three substrates displayed significant linear trends. A total of 54 positive dbMEMs were calculated from the time distance matrix, which significantly explained the detrended community structure on all substrates (p = 0.001). Of those, between 14 and 18 dbMEMs per substrate were forward selected and used to model the faunal matrix ( Supplementary Figure 2 ). Most of the temporal structure in communities across all substrates was broad-scale (≥ 50 d periods), followed by medium scales (21-44 d) ( Table 3 ; Supplementary Figure 2 ).

On whale bones, broad and medium variations in community structure were significant and explained 28.4% and 15.7% of the total variance, respectively ( Table 3 ; Supplementary Figure 2 ). On the wood block, greater proportions of the variation in community structure were explained by broad and medium temporal scales: 46.2% and 21.5%, respectively. Broad scales also explained a high amount of variation in the community on the carbonate (21.3%), while medium (6.4%) and fine (2.2%) scales explained very small amounts.

Variations in community structure were also correlated with environmental variables at different time scales ( Table 3 , third line; Supplementary Figure 2 ). The variation explained by environmental variables never exceeded 0.35, indicating that biological processes (or unmeasured environmental variables), as opposed to the measured environmental variables, may play a bigger role in driving community patterns.

Significant explanatory variables show similarities and differences among the three substrate sub-models ( Table 3 lines 5 - 15; Supplementary Figure 2 ). For all three substrates, temperature and chlorophyll explained significant variation in medium-scale temporal patterns, whereas salinity and SST were related to broad-scale patterns. Oxygen was only significantly related to broad-scale patterns for Wood and Carbonate. There is also evidence that sea-surface variables, especially SST and Along-slope wind, were related to variations in the community structure on bone and wood substrates. The environmental and temporal variables explained 66.6% of the variation in the community on the whale-bone substrate ( Figure 8 ). The environmental variables explained 28.8% of the variation, and most is also explained by broad dbMEM and linear trend (20.3%). The variation explained by temporal variables independently of the environment is 40.0%, which may be associated with unmeasured environmental variables or biotic processes. Wood was the substrate with the most explained variance in the community (89.5%). More than half of the variance that is explained by the environmental variables is also explained by broad dbMEM and linear trend (56.1%), and 23.2% was explained by temporal variables independently of the environment. Similarly to the whale bones, the environmental and temporal variables explained 69.5% of the variation in the carbonate. However, the environmental variables explained most of the variation (52.3%) and 20.2% was explained by temporal variables independently of the environment, similar compared to the wood.

Questions 1 and 2: Does community structure vary over time on whale-bone, wood, and carbonate, and, if so, are faunal-community temporal patterns similar across substrate types?

Faunal abundances, species diversity, and community structure varied substantially over the 250 d of high-frequency observations on each substrate, with markedly different patterns across substrates. Temporal changes on bones included (a) very high abundances of O. obtusa during the first ~50 d, and (b) abundant bacterial mats after this period. In contrast, wood sustained moderate abundances, and the carbonate low abundances, of these amphipods during the first 60 days, and both failed to develop visible bacterial mats ( Figure 2 ). The abundance of O. obtusa on whale bones uniquely exhibited two significant periods of ~80 days; Early dominance by the mobile scavenging amphipod O. obtusa with a period of ~80 d likely reflects availability of soft tissue on the fresh whale bones. The later dominance by bacterial mats on the bones indicates the presence of sulfur-oxidizing bacteria utilizing sulfide produced by anaerobic bacterial decomposition of labile organic material on/in the whale bones (Smith and Baco, 2003; Smith et al., 2015). These changes suggest a rapid transition from a mobile scavenger stage to a sulfophilic stage of succession on the whale bones without an intervening enrichment opportunist stage characteristic of communities on intact whale carcasses (Smith and Baco, 2003; Smith et al., 2015). The absence of an enrichment opportunist stage likely results both from very limited availability of whale soft tissue on the whale bones at time of deployment and the absence of the bone-degrading worm Osedax over the 8.3 mo high-frequency observation period.

The wood community was also initially dominated by O. obtusa, but abundances were substantially lower than on the whale bones for the first ~60 days, yet still much higher than on carbonate ( Figure 2 ). This pattern likely reflects intermediate levels of food availability for the scavenging amphipod on wood, i.e., much lower than on whale bones, but much higher than on the carbonate. Because the wood showed no evidence of wood-degrading Xylophaga bivalves or sulfur-oxidizing microbial mats during the 8.3 mo of high-frequency observations, an intermediate level of food availability on wood over this interval likely results from growth of cellulose-degrading microbes on the wood surfaces (Palacios et al., 2006, 2009; Fagervold et al., 2014; Hampel et al., 2022). Over the interval 50-75 days, as O. obtusa declined on whale bones, this amphipod abruptly increased in abundance on the wood to levels 2-3 fold higher than on the bones, and then gradually declined; we interpret this as a bone mass effect (sensu Leibold et al., 2004; Young et al., 2022) in which growing bacterial mats excluded large numbers of amphipods from the whale-bone surfaces, yielding immigration to the wood block. Interestingly, the abundance of O. obtusa remained low on the carbonate throughout these 8.3 mo, exhibiting no substantial mass effect from the whale bones; we hypothesize that the carbonate provided little food for the scavenging amphipod and was much less desirable as a roosting location for amphipods feeding on the more organic-rich substrates of whale bones and wood (Smith, 1985).

Wood and carbonate communities also differed from those on whale bones by sustaining similar, higher abundances of Heptacarpus sp. for most of the 8.3 mo high-frequency observations ( Figure 2 ). Species evenness declined and stabilized over time on wood and carbonate as these communities became increasingly dominated by Heptacarpus sp. ( Figure 3 ). The shrimp appears to have been attracted to the physical structure of the wood and carbonate, and likely was excluded from bone surfaces by high concentrations of amphipods and microbial mats.

Communities on all three substrates showed directional changes in nMDS plots ( Figure 4 ) over the first ~100 d, with wood and carbonate communities converging to a relatively stable, similar structure after 200 d, while the bone community exhibited substantial variations in community structure after 150 d as abundances of O. obtusa, Heptacarpus sp. and bacterial mats continued to vary. The greater trajectory length and lower directionality on bones compared to wood and carbonate likely reflects the initially higher food supply on bones (e.g, remnants of soft tissue), which attracted higher abundances of scavengers such as O. obtusa and Eptatretus. Bone-community divergence from wood and carbonate in latter stages likely is driven by the development of extensive microbial mats, and sulfide efflux, at bone surfaces.

Our qualitative ROV observations between 8.3 mo and 9.2 y ( Supplementary Figures 3, 4 ) provide some insights into longer-term successional processes on our substrates. The persistence of microbial mats on our whale bones between 8.3 mo and 9.2 y indicates the presence of the sulphophilic stage on parts of the bones for 9.2 y, although colonization of hydroids in some areas suggest the bones are gradually transitioning to the final reef stage (Smith et al., 2015). The wood parcel appeared to be transitioning from the intermediate opportunist stage to the last senescence stage between 7.3 and 9.2 y (Pop Ristova et al., 2017).

There are few studies of co-located bone and wood at the deep-sea floor and these have focused on macrofauna. Young et al. (2022) also found that co-located whale-bone, wood, and rock substrates at the deep-sea floor at bathyal depths in the NE Pacific supported distinct macrofaunal communities. However, their study compared substrates after 15 mo at the seafloor and the bone/wood substrates were colonized by the ecosystem engineering taxa Osedax/Xylophaga. Similarly, Pereira et al. (2022) found different macrofaunal communities on Xylophaga-bored wood, cow bones (uncolonized by Osedax) and carbonate placed away from seepage for 7.4 y on the bathyal on the Costa Rica margin. Our results indicate that megafaunal community divergence can occur between bone and wood substrates even in the absence of Osedax and Xylophaga. Early divergence in our study was caused mainly by changes in species proportions and likely resulted from higher availability of labile organic matter on whale bones. Greater divergence between our bone and wood communities occurred with the development of microbial mats on the bones, indicative of the sulfophilic stage. We suggest that sulfate reducing bacteria within our whale bones, and within other whale falls (Smith et al., 2015; Alfaro-Lucas et al., 2017) also act as ecosystem engineers by anaerobically degrading bone lipids to provide a chemical energy source (sulfide) for mats of sulfur-oxidizing bacteria on the bones, producing a new reducing habit. This sulfide efflux may then exclude other biota such as scavenging amphipods and facilitate development of the sulfophilic stage of whale-fall community succession.

Question 3: Are faunal community patterns correlated with key near-bottom and/or sea-surface environmental variables?

Communities on all three substrates showed significant periods of temporal variability on broad (≥ 54 d) and medium (21- 29 d) time scales, and carbonate showed additional significant variability on fine (6-7 d) time scales, with ~ 30 - 35% of variability explained by measured environmental variables ( Table 3 ). This suggests that biological processes, e.g., growth of microbial mats (or unmeasured environment variables), as opposed to the measured environmental variables, explain 65-70% of the observed community patterns in our study. The significant relationship between oxygen and broad-scale patterns in wood and carbonate communities, with ~15% of variability explained, and the lack of such a relationship for whale bones, indicates greater independence from oxygen variations in the whale-bone associated community. Lower sensitivity to reduced oxygen might be expected in whale-bone communities that persist in organic-rich whale-fall habitats (Smith et al., 2015). Orchomenella obtusa, in particular, has been shown to tolerate very low oxygen conditions to exploit food-rich habitats (De Robertis et al., 2001). In addition, chlorophyll-a, an index of phytodetritus availability (Thomsen et al., 2017), explained 8 to ~12% of medium-scale variability across all treatments. This suggests that phytodetritus may be a food source for many of the species observed in the study.

The amount of temporal variability explained in our study by environmental variables is comparable to levels of variability explained by environmental variables in similar time-series studies of benthic community dynamics on the bathyal NE Pacific margin (Matabos et al., 2014; Chauvet et al., 2018), but the partitioning of the variability was different. Only eight percent of the community variability at Barkley Canyon Axis (Chauvet et al., 2018) was explained by the combined effects of environmental parameters and temporal structure, and most of the variation was explained by the temporal model only (30%). For our wood and carbonate substrates, most of the variation was explained by the combined effects of environmental and temporal variables in the variation partitioning analysis, suggesting an important role of the environmental variables in these communities. In contrast, for the whale bones, much of the variation (40.3%) was explained by the temporal variables only, compared to 26.7% for environmental and temporal variables combined. This suggests that biological processes (e.g., growth of microbial mats and/or interspecific relationships such as completion and predation) may be more important than physical drivers generally in controlling whale-bone megabenthic community structure over time scales of weeks to months.

It is noteworthy that we observed no colonization of the bone/wood substrates, or bone/wood degradation, by the ecosystem engineers Osedax/Xylophaga after 1.4 y, and no Osedax colonization or degradation of the bones even after 9.2 y. ROV images indicated some colonization of the wood by Xylophaga after 2 y, but any burrow-holes and siphons were very small and barely visible ( Supplementary Figure 4 ). These rates of bone/wood colonization and degradation by Osedax/Xylophaga contrast sharply with rates in other bathyal settings with higher oxygen concentrations along the NE Pacific margin. For example, Lundsten et al. (2010a) found substantial Osedax colonization after six months on whale bones at 633 and 1018 m in Monterey Canyon (dissolved concentrations of 0.476 - 0.568 ml/l), and Smith et al. (2014) observed dense Osedax colonies on whale bones after 18 mo at oxygen levels of ~ 0.60 ml/l in Santa Cruz Basin (Juniper et al., 2021). Grupe (2014) observed 3 species of Osedax in whale bone after 1 year at Hydrate Ridge North off Oregon at ~600 m (0.34-0.52 ml/l O₂). Osedax was also reported on a whale skeleton at 1288 m on Clayoquot Slope (Lundsten et al., 2010b) where oxygen ranged annually from 0.33 - 0.57 ml/l over 10.8 years of observation from Sep 2013 to June 2024 (ONC Data Archive: https://data.oceannetworks.ca/PlottingUtility?refLink=Mjg0NjN8MTE4NDcHEQ).

For wood falls, Voight (2007) documented massive numbers of Xylophaga (including adults) and substantial decomposition in Douglas-fir blocks after 10 months on the Oregon margin at ~1500-2600 m depths with dissolved oxygen values of ~ 0.7 - 1.6 ml/l, and massive colonization was observed in Douglas-fir parcels after 6 mo in Santa Catalina Basin at 1240 m (shown in Glover et al., 2013) at 0.5 ml l^-1. of oxygen. The slow colonization of the wood by Xylophaga in our study apparently retarded community successional, since the development of the opportunistic stage depends on wood degradation by Xylophaga (Pop Ristova et al., 2017).

Slow to no colonization of Osedax/Xylophaga at our site could result from low-oxygen stress and/or from a lack of nearby larval sources. Reduced colonization rates seem unlikely to result from a lack of larval sources for Xylophaga since the west coast of Vancouver Island is heavily forested, yielding substantial wood inputs (Voight, 2007, 2009; Wohl and Iskin, 2021), and submarine canyons such as Barkley are known to accumulate wood debris and have been argued to be the preferred habitat for Xylophaga (Romano et al., 2013). For Osedax, a well developed whale-fall community with Osedax has been described ~90 km away on the Clayoquot Slope (Lundsten et al., 2010b). Furthermore, humpback and gray-whale migration routes and feeding grounds lie along the west coast of Vancouver Island (Calambokidis and Barlow, 2020; Carretta et al., 2020), yielding an abundant source of whales falls and potential larval sources for Osedax in the region. Given larval sources in the area, the along-axis currents in Barkley Canyon should facilitate larval transport across our experimental substrates (De Leo et al., 2018).

Thus, we hypothesize that no/slow colonization of Osedax/Xylophaga on our bone/wood substrates resulted from low-oxygen stress. Since oxygen levels fell to 0.22 ml/l during the 8.3 mo of our high-frequency observations, and Osedax has been observed at concentrations of 0.33 - 0.5 ml/l, this suggests an oxygen threshold for Osedax species in the region between 0.22 and 0.33 ml/l. Similarly, wood colonization and degradation by Xylophaga appears to be slowed by an oxygen threshold between 0.22 and 0.5 ml/l. These thresholds are similar to those of many macrofaunal and megafaunal taxa occurring near oxygen minimum zones (Levin, 2003; Chu and Tunnicliffe, 2015; Chu et al., 2018). Based on our results, we posit that OMZ expansion due to climate change, yielding increasing bathyal areas with oxygen concentrations between 0.22 and 0.33 ml/l (Levin, 2018; Breitburg et al., 2018; Sampaio et al., 2021) is likely to reduce habitat availability for bathyal Osedax and Xylophaga, and lower the decomposition rates of whale bones and wood on the NE Pacific margin. Other species that rely on these two engineering taxa to modify the substrate before they settle could also be affected.

Total megafaunal species richness on our whale bones (a maximum of four species at any time point) also is low compared to whale skeletons studied with similar video techniques in better oxygenated bathyal sites (9 - 45 taxa at any time point; Lundsten et al., 2010a, 2010b; Juniper et al., 2021). We think this also is partly related to oxygen stress; however, it may also be a function of the smaller size of the whale-bone habitat in our experiment compared to studies of entire whale skeletons (e.g., Lundsten et al., 2010b) because the areas of island habitats have been widely documented to be positively related to species richness (Ziegler et al., 2017).

Finally, it is also noteworthy that megafaunal community structure on our whale-bone, wood, and control treatments after 8.3 months was much more similar than observed for macrofauna on co-located whale-bone, wood and control substrates after 15 months in better oxygenated waters on the nearby Oregon-Washington margin (Young et al., 2022). Although there may be alternative explanations related to different faunal size classes and time intervals, we hypothesize that the lack of the ecosystem engineers Osedax and Xylophaga, in concert with oxygen stress, stymied the development of distinct whale-bone and wood-fall communities in Barkley Canyon compared to sites with higher oxygen availability (e.g., Bernardino et al., 2010; Lundsten et al., 2010a, 2010b; Young et al., 2022). Similarly, where active methane seepage apparently prevented colonization of Osedax and Xylophaga, the macrofaunal colonists of bone, wood and carbonate were very similar (Pereira et al., 2022). We thus posit that OMZ expansion will reduce the contribution of whale falls and wood falls to beta diversity on the NE Pacific margin.