As part of the continuing University of Washington engagement effort, and in preparation for the new National Science Foundation K12 education award focused on bringing OOI data into the classroom, Kelley collaborated with the Center for Environmental Visualization within the School of Oceanography to generate an earthquake exploration tool focused on seismic events within the global oceans from 1970 to present. We anticipate that one of the curriculum modules developed for the K12 program will be focused on geohazards, with an emphasis on the Cascadia Subduction Zone within the context of the “ring of fire.” A video of this animation is hosted on interactive oceans and a direct link to the developmental site is provided above. The animation will be used in a Queens College physical geology class this next year that has 150 students (Dr. Dax Soule). This effort is also in preparation for completing a similar visualization focused on Axial Seamount and Regional Cabled Array seismic data.
The oceanic bottom boundary layer (BBL) is the portion of the water column close to the seafloor where water motions and properties are influenced significantly by the seabed. This study (Reimers & Fogaren, 2021) reported in the Journal of Geophysical Research examines conditions in the BBL in winter on the Oregon shelf. Dynamic rates of sediment oxygen consumption (explicitly oxygen fluxes) are derived from high-frequency, near-seafloor measurements made at water depths of 30 and 80 meters. The strong back-and-forth motions of waves, which in winter form sand ripples, pump oxygen into surface sediments, and contribute to the generation of turbulence in the BBL, were found to have primed the seabed for higher oxygen uptake rates than observed previously in summer.
Since oxygen is used primarily in biological reactions that also consume organic matter, the winter rates of oxygen utilization indicate that sources of organic matter are retained in, or introduced to, the BBL throughout the year. These findings counter former descriptions of this ecosystem as one where organic matter is largely transported off the shelf during winter. This new understanding highlights the importance of adding variable rates of local seafloor oxygen consumption and organic carbon retention, with circulation and stratification conditions, into model predictions of the seasonal cycle of oxygen.
Supporting observations, which give environmental context for the benthic eddy covariance (EC) oxygen flux measurements, include data from instruments contained in OOI’s Endurance Array Benthic Experiment Package and Shelf Surface Moorings. Specifically, velocity profile time-series are drawn from records of a 300-kHz Velocity Profiler (Teledyne RDI-Workhorse Monitor), near-seabed water properties from CTD (SBE 16plusV2) and oxygen (Aanderaa-Optode 4831) sensors, winds from the surface buoy’s bulk meteorological package, and surface-wave data products from a directional wave sensor (AXYS Technologies) (see e.g., Fig 1 above).
Reimers, C. E., & Fogaren, K. E. (2021). Bottom boundary layer oxygen fluxes during winter on the Oregon shelf. Journal of Geophysical Research: Oceans, 126, e2020JC016828. https://doi.org/10.1029/2020JC016828
Among the detailed analyses undertaken in this two-part study was quantification of the impact of observations on the reduction of RMS error for estimates of the volume transport across an along-front transect (Fig. 1). Temperature and salinity data from moorings and gliders were impactful for the larger grids (G1, G2). As the grid resolution was increased (G3), submesoscale motions were resolved and velocity data from the moorings became more important for reduction of error variance. An analysis of the sensitivity of shelf-slope exchange indices (e.g. volume transport) to removal of an observation, compared to the direct impact of the observation, showed that the majority of observed variables (e.g., SST, SSH, T, S, U, V) were “synergistic” – providing value to the assimilation through their connection with other variables as represented in the model dynamics. For the highest resolution estimates (G3 grid), the Pioneer Array observing assets were more impactful than other observations (e.g., remote sensing, NDBC and IOOS buoys) in reducing uncertainty, with velocity data being the major contributor. This is not a complete surprise, since the Pioneer Array was “tuned” to these scales. Still, it is gratifying to see that the impact on model fidelity is quantifiable.
The two-part study undertaken by Levin et al. provides a wealth of additional information about the performance of assimilative models as well as the utility of in-situ observations for modeling and prediction. As the authors state, they have “just begun to scratch the surface” of approaches that can be applied to the assessment of model performance as well as the management of observing systems.
Levin J., H.G. Arango, B. Laughlin, E. Hunter, J. Wilkin, and A.M. Moore, 2020. Observation impacts on the Mid-Atlantic Bight front and cross-shelf transport in 4D-Var ocean state estimates: Part I – Multiplatform analysis,Ocean Modeling, 156, 101721, 1-17, doi 10.1016/j.ocemod.2020.101721.
Levin J., H.G. Arango, B. Laughlin, E. Hunter, J. Wilkin, and A.M. Moore, 2021. Observation impacts on the Mid-Atlantic Bight front and cross-shelf transport in 4D-Var ocean state estimates: Part II – The Pioneer Array, Ocean Modeling, 157, 101731, 1-17, doi 10.1016/j.ocemod.2020.101731.
[media-caption path="https://oceanobservatories.org/wp-content/uploads/2021/03/Screen-Shot-2021-03-30-at-5.51.41-PM.png" link="#"]Researchers used echosounder data from the Oregon Offshore site of the Coastal Endurance Array to develop a new methodology that makes it easier to extract dominant patterns and trends.[/media-caption]The ocean is like a underwater cocktail party. Imagine, as a researcher, trying to follow a story someone is telling while other loud conversations are in the background of a recording. This phenomenon, known as the “Cocktail Party Problem,” has been studied since the 1950s (Cherry, 1953; McDermott, 2009). Oceanographers face this challenge in sorting through ocean acoustics data, with its mixture of echoes from acoustic signals sent out to probe the ocean.
Oceanographer Wu-Jung Lee and data scientist Valentina Staneva, at the University of Washington, teamed up to tackle the challenge in a multidisciplinary approach to analyze the vast amounts of data generated by echosounders on Ocean Observatories Initiative (OOI) arrays. Their findings were published in The Journal of the Acoustical Society of America, where they proposed a new methodology that uses machine learning to parse out noisy outliers from rich echosounder datasets and to summarize large volumes of data in a compact and efficient way.
This new methodology will help researchers use data from long time series and extract dominant patterns and trends in sonar echoes to allow for better interpretation of what is happening in the water column.
The ocean is highly dynamic and complex at the Oregon Offshore site of the OOI Coastal Endurance Array, where echosounder data from a cabled sonar were used in this paper. At this site, zooplankton migrate on a diurnal basis from a few hundred meters to the surface, wind-stress curl and offshore eddies interact with the coastal circulation, and a subsurface undercurrent moves poleward. The echosounder data offer opportunities to better understand the animals’ response to immediate environmental conditions and long-term trends. During the total eclipse of the Sun in August 2017, for example, echosounders captured the zooplankton’s reaction to the suddenly dimmed sunlight by moving upwards as if it was dusk time for them to swim toward the surface to feed (Barth et al, 2018).
Open access of echosounder datasets from the OOI arrays offers researchers the potential to study trends that occur over extended stretches of time or space. But commonly these rich datasets are underused because they require significant processing to parse out what is important from what is not.
Echosounders work by sending out pulses of sound waves that bounce off objects. Based on how long it takes for the reflected echo to come back to the sensor, researchers can determine the distance of the object. That data can be visualized as an echogram, an image similar to an ultrasound image of an unborn baby.
But unlike an ultrasound of a baby, when an undersea acoustic sensor records a signal, it may be a combination of signals from different sources. For example, the signal might be echoes bouncing off zooplankton or schools of fish.[caption id="attachment_20566" align="alignleft" width="350"] (A) Data used in this work were collected by a three-frequency echosounder installed on a Regional Cabled Array Shallow Profiler mooring hosting an underwater platform (200 m water depth) and profiler science pod located at the Oregon Offshore site of the OOI Coastal Endurance Array (red triangle). The symbols indicate the locations of all OOI echosounders installed along the coast of Oregon and Washington. (B) The transducers are integrated into the mooring platform (from left to right: 120, 200, and 38 kHz). The platform also hosts an instrumented profiler that traverses the water column above the echosounder from ~ 200 m to ~ 5m beneath the ocean’s surface. (Image credit: UW/NSF-OOI/WHOI-V15).[/caption]
“When the scatterers are of different size, they will reflect the sound at different frequencies with different strengths,” said Lee. “So, by looking at how strong an echo is at different frequencies, you will get an idea of the range of sizes that you are seeing in your echogram.”
Current echogram analysis commonly requires human judgement and physics-based models to separate the sources and obtain useful summary statistics. But for large volumes of data that span months or even years, that analysis can be too much for a person or small group of researchers to handle. Lee and Staneva’s new methodology utilizes machine learning algorithms to do this inspection automatically.
“Instead of having millions of pixels that you don’t know how to interpret, machine learning reduces the dataset to a few patterns that are easier to analyze,” said Staneva.
Machine learning ensures that the analysis will be data-driven and standardized, thus reducing the human bias and replicability challenges inherently present in manual approaches.
“That’s the really powerful part of this type of methodology,” said Lee. “To be able to go from the data-driven direction and say, what can we learn from this dataset if we do not know what may have happened in a particular location or time period.”
Lee and Staneva hope that by making the echosounder data and analytical methods open access, it will improve the democratization of data and make it more usable for everybody, even those who do not live by the ocean.
In the future, they plan to continue working together and use their new methodology to analyze the over 1000 days of echosounder data from the OOI Endurance Coastal and Regional Cabled Array region.
Lee, W-J and Staneva, V (2021).Compact representation of temporal processes in echosounder time series via matrix decomposition. Special Issue on Machine Learning in Acoustics. The Journal of the Acoustical Society of America.
Barth JA, Fram JP, et al. (2018). Warm Blobs, Low-Oxygen Events, and an Eclipse: The Ocean Observatories Initiative Endurance Array Captures Them All.Oceanography, Vol 31.
McDermott, J (2009). The Cocktail Party Problem.Current Biology, Vol 19, Issue 22.
Cherry EC (1953). Some Experiments on the Recognition of Speech, with One and Two Ears.The Journal of the Acoustical Society of America. Vol. 25, No.5.
The Cabled Observatory Vent Imaging Sonar (COVIS) was installed on the OOI RCA in the ASHES hydrothermal field (Fig. 26 a-c) at the summit of Axial Seamount in 2018, resulting in the first long-term, quantitative monitoring of plume emissions (Xu et al., 2020). The sonar provides 3-dimensional backscatter images of buoyant plumes above the actively venting ‘Inferno’ and ‘Mushroom’ edifices, and two-dimensional maps of diffuse flow at temporal frequencies of 15 and 2 minutes, respectively. Sonar data coupled with in-situ thermal measurements document significant changes in plume variations (Fig. 26 d-f) and modeling results indicate a heat flux of 10 MW for the Inferno plume (Xu et al., 2020). COVIS will provide key data to the community investigating the impacts of eruptions on hydrothermal flow at this highly active volcano.
 Xu, G., Bemis, K., Jackson, D., and Ivakin, A., (2020) Acoustic and in-situ observations of deep seafloor hydrothermal discharge: OOI Cabled Array ASHES vent field case study. Earth and Space Science. Note: This project was funded by the National Science Foundation through an award to PI Dr. K. Bemis, Rutgers University – “Collaborative Research: Heat flow mapping and quantification at ASHES hydrothermal vent field using an observatory imaging sonar (#1736702). COVIS data are available through oceanobservatories.orgRead More
In February 2021 JGR Oceans article, Clare E. Reimers (Oregon State University) and Kristen Fogaren (Boston College) used data from the Endurance Array Oregon Shelf to advance understanding of how the benthic boundary layer on the Oregon Shelf in winter depends on surface-wave mixing and interactions with the seafloor.
The oceanic bottom boundary layer (BBL) is the portion of the water column close to the seafloor where water motions and properties are influenced significantly by the seabed. This study examines conditions in the BBL in winter on the Oregon shelf. Dynamic rates of sediment oxygen consumption (explicitly oxygen fluxes) are derived from high-frequency, near-seafloor measurements made at water depths of 30 and 80 m. The strong back-and-forth motions of waves, which in winter form sand ripples, pump oxygen into surface sediments, and contribute to the generation of turbulence in the BBL, were found to have primed the seabed for higher oxygen uptake rates than observed previously, in summer. Since oxygen is used primarily in biological reactions that also consume organic matter, the winter rates of oxygen utilization indicate that sources of organic matter are retained in, or introduced to, the BBL throughout the year. These findings counter former descriptions of this ecosystem as one where organic matter is largely transported off the shelf during winter. This new understanding highlights the importance of adding variable rates of local seafloor oxygen consumption and organic carbon retention, with circulation and stratification conditions, into model predictions of the seasonal cycle of oxygen.
The rest of the article can be accessed here.
In the summer of 2020 the Rutgers University Ocean Data Labs project worked with the Rutgers Research Internships in Ocean Science to support ten undergraduate students in a virtual Research Experiences for Undergraduates program. Two weeks of research methods training and Python coding instruction was followed by six weeks of independent study with a research mentor.
Dr. Rachel Eveleth (Oberlin College) was one of those mentors. Already using some of the Data Labs materials in her undergraduate oceanography course, she saw an opportunity to leverage the extensive OOI data holdings to engage students in cutting edge research on a limited budget during a time when her own field work was curtailed due to the COVID-19 pandemic. Dr. Eveleth advised Alison Thorson from Sarah Lawrence College (NY) and Brianna Velasco form Humboldt State University (CA) on the study of air-sea fluxes of CO2 on the US east and west coast, respectively.
Preliminary results were presented at the 2020 Fall AGU meeting. A poster authored by Thorson and Eveleth (ED037-0035) evaluated pCO2 data from the three Pioneer Array Surface Moorings during 2016 and 2017. They showed that the annual mean CO2 flux across all three sites for the two years was negative, meaning that the continental shelf acts as a sink of atmospheric carbon. The annual average flux was -0.85 to -1.6 mol C/(m2 yr), but the flux varied significantly between mooring sites and between years (Figure 23). Investigation of short-term variability in pCO2 concentration concurrent with satellite imagery of SST and Chlorophyll was consistent with temperature-driven, but biologically damped, changes.[media-caption path="https://oceanobservatories.org/wp-content/uploads/2021/02/Pioneer-for-Science-Highlights.png" link="#"]Figure 24. Hourly (dots) and monthly (lines) average air and water CO2 concentration observed at the Endurance Array Washington Offshore mooring during 2016 and 2017. From Velasco et al. (2020).[/media-caption]
A poster by Velasco, Eveleth and Thorson (ED004-0045) analyzed pCO2 data from the Endurance Array offshore mooring. Three years of nearly continuous data were available during 2016-2018. The seasonal cycle showed that the pCO2 concentration in water was relatively stable and near equilibrium with the air in winter, decreasing in late spring and summer (Figure 24). Short-term minima in summer were as low as 150 uatm. Like the east coast, the mean air-sea CO2 flux was consistently negative, meaning the coastal ocean acts as a carbon sink. The annual means at the Washington Offshore mooring for 2016, 2017 were -1.9 and -2.1 mol C/(m2 yr), respectively. The seasonal cycle appears to be strongly driven by non-thermal factors (on short time scales), presumably upwelling events and algal blooms.
These studies, although preliminary, are among the first to use multi-year records of in-situ CO2 flux from the OOI coastal arrays, and to our knowledge the first to compare such records between the east and west coast. Dr. Eveleth’s team intends to use the rich, complementary data set available from the OOI coastal arrays to investigate the mechanisms controlling variability and role of biological vs physical drivers.
In the summer of 2020, the Rutgers University Ocean Data Labs project worked with the Rutgers Research Internships in Ocean Science to support ten undergraduate students in a virtual Research Experiences for Undergraduates program. Rutgers led two weeks of research methods training and Python coding instruction. This was followed by six weeks of independent study with one of 13 research mentors.
Dr. Tom Connolly (Moss Landing Marine Labs, San Jose State University) advised Andrea Selkow from Austin College, Texas on her study of dissolved oxygen (DO) off the Washington and Oregon coasts using the OOI Endurance Array.
Selkow evaluated DO data from Endurance Array Surface Moorings during 2017 and 2018. She presented this work as a poster at the conclusion of her summer REU. Selkow focused on the question: Are there similarities in the dissolved oxygen concentrations off the coast of Oregon and Washington during a known low oxygen event? She also considered why there might exist differences based on the spatial variability of wind stress forcing, i.e., do the strong Oregon winds cause dissolved oxygen concentrations to be lower at the Oregon mooring compared to the Washington moorings. Finally, she reviewed the data and tried to answer whether the oxygen data were accurate or affected by biofouling.
She used datasets from the OR and WA Inshore Shelf Mooring time-series and WA Shelf Mooring time-series from Endurance Array. Her focus was on the seafloor data because that is where the lowest oxygen concentrations were expected to be observed.
Selkow focused her attention on low DO observed in the summer of 2017. While Barth et al. (2018) presented a report on these data for one event in July 2017, she expanded the analysis to include the Washington shelf and inshore moorings. She plotted time series data and used cruise data to validate these time series. While overall seasonal trends in DO were similar, she found dissolved oxygen is routinely more quickly depleted off the coast of Oregon than Washington during a low oxygen event (Figure 25). She also looked at the cross-shelf variability in DO time series and found dissolved oxygen is more quickly depleted at the shelf mooring than at the inshore shelf mooring. Upwelling is known to drive the low oxygen events and she inferred that the weaker southward winds over the Washington shelf may be why DO decreases at a slower rate off Washington than Oregon.
Barth, J.A., J.P. Fram, E.P. Dever, C.M. Risien, C.E. Wingard, R.W. Collier, and T.D. Kearney. 2018. Warm blobs, low-oxygen events, and an eclipse: The Ocean Observatories Initiative Endurance Array captures them all. Oceanography 31(1):90–97,
Selkow, A. and T. Connelly. Low Dissolved Oxygen off Washington and Oregon Coast Impacted by Upwelling in 2017, Accessed 13 Jan 2021.Read More
[media-caption path="/wp-content/uploads/2020/10/Screen-Shot-2020-10-29-at-1.22.57-PM.png" link="#"]Figure 18. Comparison of Sentinel-1 (S-1) SAR deep-learning predictions of significant wave height Hs and buoy measurements. (b) scatter plot of SAR Hs vs buoy Hs. (d) RMS error of SAR prediction vs. buoy measurements as a function of Hs; error bars show one standard deviation. From Quach et al. (2020).[/media-caption]
Synthetic Aperture Radar (SAR) sensors on satellites measure backscatter from the ocean surface and can be used to estimate wave height at very high spatial resolution (~10 m) relative to satellite altimetry. Two Sentinel-1 satellites of the European Space Agency (ESA) collected SAR measurements of the ocean surface from 2015-2018, together covering the entire globe every six days. Data-driven approaches to predicting significant wave height (Hs) from SAR have either used relatively limited in-situ data sets or used a wave model (e.g. WaveWatch-3) as the “training” data for a deep learning approach. Quach et al.(2020) improve on previous approaches to estimation of Hs from SAR by creating a comprehensive in-situ observational record. They compiled data from the US National Data Buoy Center and Coastal Data Information Program, Canadian Marine Environmental Data Services, the international OceanSITES project, and the OOI. Surface wave data sets from the OOI Irminger Sea, Argentine Basin and Southern Ocean surface buoys were used. The authors note the importance of the Southern Ocean Array, where “many of the largest wave heights are recorded… [from] an under sampled region of the ocean.”
The comprehensive in-situ data set is split into separate training and validation segments. When SAR Hs from training data are compared to altimeter Hs from the validation segment, the deep learning algorithm shows root-mean-square (RMS) error of 0.3 m, a 50% improvement relative to prior approaches. Comparison with the buoy validation segment (Fig. 18) shows RMS error of 0.5 m. The authors attribute the increased error to the larger number of extreme sea states in the observations and the relative paucity of extremes in the training data.
Observational sea state information is critical for understanding surface wave phenomena (generation, propagation and decay), predicting wave amplitudes, and estimating extreme sea states. Thus, the improvement in RMS error using the deep learning technique notable. The availability of in-situ data from extreme environments such as those sampled by the OOI Irminger Sea and Southern Ocean Arrays are key to validation of these new approaches.
The Cascadia Subduction Zone extends from northern California to British Columbia. It has experienced magnitude 9 megathrust events with a reoccurrence rate of every ~500 years over the past 10,000 years  and large earthquakes at intervals of ~ 200-1200 years . The last Cascadia megathrust rupture occurred on January 26, 1700 . When the next event occurs, it is estimated that financial losses would be ~ $60 billion USD with substantial loss of life. Hence, there is significant research focused on understanding seismic processes along this ~ 1100 km subduction zone, the generation of slow earthquakes, and causes of variation in seismicity along strike.
Understanding the factors that control seismic events was/is a major driver in the siting of OOI-RCA core geophysical instrumentation on the southern line of the Regional Cabled Array: the RCA is one of the few places in the world where seismic-focused instrumentation occurs on both the down-going tectonic plate and on the overlying margin. The offshore network is especially valuable in determining earthquake source depths that inform on interpolate dynamics . The central section of the Cascadia Margin is the only area that experiences repeat, measurable shallow crustal earthquakes [1-3]. RCA data flowing from the seismic network at Slope Base and Southern Hydrate Ridge, and from the Cascadia Initiative are providing new insights into factors controlling seismicity along this portion of the margin [1,4] (note because the RCA broadband seismometers are buried, they have lower noise levels at higher frequencies than the Cascadia Initiative instruments ).
Most recently, Morton et al.,  examined data from the Cascadia Initiative  and the RCA. Shallow earthquakes are focused in the area of a subducted seamount [1-3] and another cluster to the north (Fig. 1b and c). Based on earthquake locations, they suggest that subduction of the seamount produces stress heterogeneities, faulting, fracturing of the overriding Siletz terrane (old oceanic crust) (Fig 1b), and fluid movement promoting seismic swarms. Because this area is the most seismically active area along the Cascadia margin, it is an optimal area to examine the impacts of local earthquakes on, for example, gas hydrate deposits and fluid expulsion.
 Tréhu, A.M., Wilcock, W.S.D., Hilmo, R., Bodin, P., Connolly, J., Roland, E.C., and Braunmiller, R., (2018) The role of the Ocean Observatories Initiative in Monitoring the offshore earthquake activity of the Cascadia Subduction Zone. Oceanography, 31, 104-113.
 Tréhu, A.M., Blakely, R.J., and Williams, M., (2012) Subducted seamounts and recent earthquakes beneath the central Cascadia Forearc. Geology, 40, 103-106.
 Tréhu, A.M., Braunmiller, J., and Davis, E., (2015) Seismicity of the Central Cascadia Continental Margin near 44.5° N: a decadal view. Seismological Research Letters, 86, 819-829.
 Morton, Bilek, S.L., and Rowe, C.A. (2018) Newly detected earthquakes in the Cascadia subduction zone linked to seamount subduction and deformed upper plate. Geology, 46, 943-946.
 Satake, K.Shimazaki, K., Tsuji, Y., and Ueda, K., (1996) Time and size of a giant earthquake in Cascadia inferred from Japanese tsunami records of January 1700. Nature, 379, 246-249.
 Goldfinger, C., Nelson, C.H., Eriksson, E., et al., (2012) Turbidite event history: Methods and implications for Holocene paleoseismicity of the Cascadia Subduction Zone. US Geological Survey Professional Paper (1661-F), 184 pp.
 Toomey, D.R., Allen, R.M., Barclay, A.H., Bell, S.W., Bromirski, P.D. et al., (2014) The Cascadia Initiative: A sea change in seismological studies of subduction zones. Oceanography, 27, 138-150.