From land, sea,… and space: searching for whales in the vast ocean

By Solène Derville, Postdoc, OSU Department of Fisheries, Wildlife, and Conservation Science, Geospatial Ecology of Marine Megafauna Lab

The ocean is vast.

What I mean is that the vastness of the ocean is very hard to mentally visualize. When facing a conservation issue such as increased whale entanglement along the US West Coast (see OPAL project ), a tempting solution may  be to suggest « let’s go see where the whales are and report their location to the fishermen?! ». But, it only takes a little calculation to realize how impractical this idea is.

Let’s roll out the numbers. The US West Coast exclusive economic zone (EEZ) stretches from the coast out to 200 nautical miles offshore, as prescribed by the 1982 United Nations Convention on the Law of the Sea. It covers an area of 825,549 km² (Figure 1). Now, imagine that you wish to survey this area for marine mammals. Using a vessel such as the R/V Bell M. Shimada that is used for the Northern California Current Ecosystem surveys cruises (NCC cruises, see Dawn and Rachel’s last blog), we may detect whales at a distance of roughly 6 km (based on my preliminary results). This distance of detection depends on the height of the observer, hence the height of the flying bridge where she/he is standing (the observer’s height may also be accounted for, but unless she/he is a professional basket-ball player, I think it can be neglected here). The Shimada is quite a large ship and it’s flying bridge is 13 meters above the water. Two observers may survey the water on each side of the trackline.

Considering that the vessel is moving at 8 knots (~15 km/h), we may expect to be effectively surveying 180 km² per hour (6x2x15). That’s not too bad, right?

Again, perspective is the key. If we divide the West Coast EEZ surface by 180 km² we can estimate that it would take 2,752 hours to survey this entire region. With an average of 12 hours of daylight, this takes us to…

382 DAYS OF SURVEY, searching for marine mammals over the US West Coast. Considering that observations cannot be undertaken on days with bad weather (fog, heavy rain, strong winds…), it might take more than a year and a half to complete the survey! And what would the marine mammals have done in the meantime? Move…

This little math exercise proves that exhaustively searching for the needle in the haystack from a vessel is not the way to go if we are to describe whale distribution and help mitigate the risk of entanglement. And using another platform of observation is not necessarily the solution. The OPAL project has relied on a great collaboration with the United States Coast Guard to survey Oregon waters. The USCG helicopters travel fast compared to a vessel, about 90 knots (167 km/h). As a result, more ground is covered but the speed at which it is traveling prevents the observer from detecting whales that are very far away. Based on the last analysis I ran for the OPAL project, whales are usually detected up to 3 km from the helicopter (only 5 % of sightings exceed that distance). In addition, the helicopter generally only has capacity for one observer at a time.

If we replicate the survey time calculation from above for the USCG helicopter, we realize that even with a fast-moving aerial survey platform it would still take 137 days to cover the West Coast EEZ.

First, we can model and extrapolate. This approach is the path we are taking with the OPAL project: we survey Oregon waters in 4 different areas along the coast each month, then model observed whale densities as a function of topographic and oceanographic variables, and then predict whale probability of presence over the entire region. These predictions are based on the assumption that our survey design effectively sampled the variety of environmental conditions experienced by whales over the study region, which it certainly did considering that all sites are surveyed year-round.

An alternative approach that has been recently discussed in the GEMM Llab, is the use of satellite images to detect whales along the coast. A communication entitled « The Potential of Satellite Imagery for Surveying Whales » was published last month in the Sensors Journal (Höschle et al., 2021) and presents the opportunities offered by this relatively new technology. The WorldView-3 satellite, owned by the company Digitalglobe and launched in 2016, has made it possible to commercialize imagery with a resolution never reached before, of the order of 30 cm per pixel. These very high resolution (VHR) satellite images make it possible to identify several species of large whales (Cubaynes et al. al., 2019) and to estimate their density (Bamford et al., 2020). Furthermore, machine learning algorithms, such as Neural Networks, have proved quite efficient at automatically detecting whales in satellite images (Guirado et al., 2019, Figure 2). While several new ultra-high resolution imaging satellites are expected to be launched in 2021 (by Maxar Technologies and Airbus), this “remote” approach looks like a promising avenue to detect whales over vast regions while drinking a cup of coffee at the office.

But like any other data collection method, satellites have their drawbacks. We recently discovered that these VHR satellites are routinely switched off while passing above the ocean. Specific inquiries would need to be made to acquire data over our study areas, which would be at great expense. One of the cheapest provider I found is the Soar platform, that provides images at 50 cm resolution in partnership with the Chinese Aerospace Science and Technology Corporation. They advertise daily images anywhere on earth at \$10 USD per km². This might sound cheap at first glance, but circling back to our US West Coast EEZ area calculations, we estimate that surveying this region entirely with satellite imagery would cost more than \$8 million USD.

Yet, we have to look forward. The use of satellite imagery is likely to broaden and increase in the coming years, with a possible decrease in cost. Quoting Höschle et al. (2021) ‘To protect our world’s oceans, we need a global effort and we need to create opportunities for that to happen’.

Will satellites soon save whales?

References

Bamford, C. C. G. et al. A comparison of baleen whale density estimates derived from overlapping satellite imagery and a shipborne survey. Sci. Rep. 10, 1–12 (2020).

Cubaynes, H. C., Fretwell, P. T., Bamford, C., Gerrish, L. & Jackson, J. A. Whales from space: Four mysticete species described using new VHR satellite imagery. Mar. Mammal Sci. 35, 466–491 (2019).

Guirado, E., Tabik, S., Rivas, M. L., Alcaraz-Segura, D. & Herrera, F. Whale counting in satellite and aerial images with deep learning. Sci. Rep. 9, 1–12 (2019).

Höschle, C., Cubaynes, H. C., Clarke, P. J., Humphries, G. & Borowicz, A. The potential of satellite imagery for surveying whales. Sensors 21, 1–6 (2021).

The ups and downs of the ocean

By Solène Derville, Postdoc, OSU Department of Fisheries and Wildlife, Geospatial Ecology of Marine Megafauna Lab

As a GEMM lab post-doc working on the OPAL project, my main goal for 2021 will be to produce accurate predictive models of baleen whale distribution off the Oregon coast to reduce entanglement risk. For the past months, I have been compiling, cleaning, and processing about two years of data collected by Leigh Torres and Craig Hayslip during monthly repeat surveys conducted onboard United States Coast Guard (USCG) helicopters. These standardized surveys record where and when whales are observed off the Oregon coast. These presence and absence data may now be modeled in relation to habitat, while accounting for effort and detection (as several parameters, such as weather and sea state, can affect the capacity of observers to detect whales at the surface). Considering that several baleen whale species (namely, humpback, fin, blue and gray whales) are known to feed in the area, prey availability is expected to be a major driver of their distribution.

As prey distribution data are frequently the lacking component in the habitat model equation, whale ecologists often resort to using environmental proxies. Variables such as topography (e.g., the depth or slope of the seafloor), water physical and chemical characteristics (e.g., temperature, salinity, oxygen concentration) or ocean circulation (e.g., currents, turbulence) have proved to be good predictors for fish or krill distribution, and in turn potential predictors for whale suitable habitats. In my search for such environmental variables to be tested in our future OPAL models, I have been focusing my research on a fascinating ocean feature: sea height.

Sea height varies both temporally and spatially under the influence of multiple factors, from internal mass of the solid Earth to the orbital revolution of the moon. After reading this blog you will realize that the flatness of the horizon at sea is a deceiving perspective (Figure 1) …

Gravity and the geoid

We all know of Newton’s s discovery of gravity: the attraction force exerted by any object with a given mass on its surroundings. Yet, it is puzzling to think that the rate of acceleration of the apple falling on Newton’s head would have been different if Newton had been anywhere else on Earth.

Why is that and what does it have to do with sea height? On Earth, the standard gravity g is set at 9.80665 m/s2. This constant is called a “standard” because in fact, gravity varies at the surface of our planet, even if estimated at a fixed altitude. Indeed, as gravity is caused by mass, any change in relief or rock composition results in a change in gravity. For instance, magmatic activity in the upper mantle of the Earth and the crust causes a change in rock density and results in a change in gravity measured at the surface.

Gravity therefore is the first reason why the ocean surface is not flat. Gravity shapes an irregular surface called the “geoid”. This hypothetical ocean surface has equal gravitational potential anywhere on Earth and differs from the ellipsoid of reference by as much as 100 m! So to the question whether Earth is round or flat, I would say it is potato shaped (Figure 2)!

The geoid is an essential reference for understanding ocean currents and monitoring changes in sea-level. Hypothetically, if ocean water had equal density everywhere and at any depth, the sea surface should match with the geoid… but that’s not the case. Let’s see why.

Ocean dynamic topography

Not unlike the hills and valleys covering landscapes, the ocean surface also has its highs and lows. Except that in the ocean, the surface topography is ever changing. Sea surface height (SSH) measures the average height difference between the observed sea level and the ellipsoid of reference (Figure 3). SSH is mostly affected by ocean circulation and may vary by as much as ±1 m. Indeed, just like the rocks inside the Earth, the water in the ocean varies in density. The vertical and horizontal physical structuring of the ocean was extensively discussed by Dawn last November while she was preparing for her PhD Qualifying Exams. Temperature clearly is at the core of the processes. As thermal expansion increases the space between warming water particles, the volume of a given amount of liquid water increases with increasing temperature. Warmer waters therefore take up “more space” than cooler waters, resulting in an elevated SSH.

SSH may therefore be used as an indicator of oceanographic phenomena such as upwellings, where warm surface waters are replaced by deep, cooler, and nutrient-rich waters moving upwards. The California Current that moves southwards along the North American coast is known as one of the world’s major currents affiliated with strong upwelling zones, which often triggers increased biological productivity. Several studies conducted in the California Current system have found a link between the variations in SSH and whale abundance or foraging activity (Abrahms et al. 2019; Pardo et al. 2015; Becker et al. 2016; Hazen et al. 2016).⁠

SSH is measured by altimeter satellites and is made freely available by the European Space Agency and the US National Aeronautics and Space Administration. Lucky me! Numerous variables are derived from SSH, as shown in Figure 3. Among other things, I was able to download the daily maps of Sea Surface Height Anomaly (SSHa, also referred to as Sea Level Anomaly: SLA) over the Oregon coast from February 2019 to December 2020. SSHa is the difference between observed SSH at a specific time and place from the mean SSH field of reference calculated over a long period of time. Negative values of SSHa potentially suggest upwellings of cooler waters that could be associated with higher prey availability. Figure 4 shows an example of environmental data mining as I try to match SSHa with whale observations made during OPAL surveys. Figure 4B suggests increased whale occurrence where/when SSHa is lower.

Although encouraging, these preliminary insights are just the tip of the modeling iceberg. Many more testing and modeling steps will be required to determine confounding factors and relevant spatio-temporal scales at which these oceanographic variables may be influencing whale distribution off the Oregon coast. I am only at the start of a long road…

References

Abrahms, Briana, Heather Welch, Stephanie Brodie, Michael G. Jacox, Elizabeth A. Becker, Steven J. Bograd, Ladd M. Irvine, Daniel M. Palacios, Bruce R. Mate, and Elliott L. Hazen. 2019. “Dynamic Ensemble Models to Predict Distributions and Anthropogenic Risk Exposure for Highly Mobile Species.” Diversity and Distributions, no. December 2018: 1–12. https://doi.org/10.1111/ddi.12940.

Becker, Elizabeth, Karin Forney, Paul Fiedler, Jay Barlow, Susan Chivers, Christopher Edwards, Andrew Moore, and Jessica Redfern. 2016. “Moving Towards Dynamic Ocean Management: How Well Do Modeled Ocean Products Predict Species Distributions?” Remote Sensing 8 (2): 149. https://doi.org/10.3390/rs8020149.

Hazen, Elliott L, Daniel M Palacios, Karin A Forney, Evan A Howell, Elizabeth Becker, Aimee L Hoover, Ladd Irvine, et al. 2016. “WhaleWatch : A Dynamic Management Tool for Predicting Blue Whale Density in the California Current.” Journal of Applied Ecology 54 (5): 1415–28. https://doi.org/10.1111/1365-2664.12820.

Pardo, Mario A., Tim Gerrodette, Emilio Beier, Diane Gendron, Karin A. Forney, Susan J. Chivers, Jay Barlow, and Daniel M. Palacios. 2015. “Inferring Cetacean Population Densities from the Absolute Dynamic Topography of the Ocean in a Hierarchical Bayesian Framework.” PLOS One 10 (3): 1–23. https://doi.org/10.1371/journal.pone.0120727.

Remote Sensing Applications

By Leila Lemos, PhD candidate

Fisheries and Wildlife Department, OSU

I am finally starting my 3rd and last year of my PhD. Just a year left and yet so many things to do. As per department requirements, I still need to take some class credits, but what classes could I take? In this short amount of time it is important to focus on my research project and on what could help me better understand the many branches of the project and what could improve my analyses. Thinking of that, both my advisor (Dr. Leigh G. Torres) and I agreed that it would be useful for me to take a class on remote sensing. So, I could learn more about this field, as well as try to include some remote sensing analyses in my project, such as sea surface temperature (SST) and chlorophyll (i.e., as a productivity indicator) conditions over the years we have collected data on gray whales off the Oregon coast.

Our photogrammetry data indicates that whales gradually increased their body condition over the feeding seasons of 2016 and 2018, while 2017 is different. Whales were still looking skinny in the middle of the season, and we were not collecting many fecal samples up to that point (indicating not much feeding). These findings made us wonder if this was related to delayed seasonal upwelling events and consequently low prey availability. These questions are what motivated me the most to join this class so that we might be able to link environmental correlates with our observations of gray whale body condition.

If we stop to think about what remote sensing is, we have already been implementing this method in our project since the beginning, as my favorite definition for remote sensing is “the art of collecting information of objects or phenomenon without touching it”. So, yes, the drone is a type of sensor that remotely collects information of objects (in this case, whales).

However, satellites, all the way up in the space, are also remotely sensing the Earth and its objects and phenomena. Even from thousands of km above Earth, these sensors are capable of generating a great amount of detailed data that is easily and freely accessible (i.e., NASA, NOAA), and can be used for multiple applications in different fields of study. Satellites are also able to collect data from remote areas like the Antarctica and the Arctic, as well as other areas that are not easily reached by humans. One important application of the use of satellite imagery is wildlife monitoring.

For example, satellite data was used to detect variation in the abundance of Weddell seals (Leptonychotes weddellii) in Erebus Bay, Antarctica (LaRue et al., 2011). Because this is a well-studied seal population, the object of this study was to test if satellite imagery could produce reliable abundance estimates. The authors used high-resolution (0.6 m) satellite imagery (from satellites Quick-Bird-2 and WorldView-1) to compare counts from the ground with counts from satellite images in the same locations at the same time. This study demonstrated a reliable methodology for further studies to replicate.

Satellite imagery was also applied to estimate colony sizes of Adélie penguins in Antarctica (LaRue et al., 2014). High-resolution (0.6 m) satellite imagery combined with spectral analysiswas used to estimate the sizes of the penguin breeding colonies. Ground counts were also used in order to check the reliability of the applied method. The authors then created a model to predict the abundance of breeding pairs as a function of the habitat, which was identified terrain slope as an important component of nesting density.

The identification of whales using satellite imagery is also possible. Fretwell et al. (2014)pioneered this method by successfully identifing Southern Right Whales (Eubalaena australis) in the Golfo Nuevo, Península Valdés, in Argentina in satellite images. By using very high-resolution satellite imagery (50 cm resolution) and a water penetrating coastal band that was able to see deeper into the water column, the researchers were able to successfully identify and count the whales (Fig. 04). The importance of this study was very significant, since this species was extensively hunted from the 17ththrough to the 20thcentury. Since then, the species has shown a strong recovery, but population estimates are still at <15% of historical estimates. Thus, being able to use new tools to identify, count and monitor individuals in this recovering population is a great development, especially in remote and hard to reach areas.

Polar bears (Ursus maritimus) have also been studied in the Foxe Basin, in Nunavut and Quebec, Canada (LaRue et al., 2015). Researchers used high-resolution satellite imagery in an attempt to identify and count the bears, but spectral signature differences between bears and other objects were insufficient to yield useful results. Therefore, researchers developed an automated image differencing, also known as change detection, that identifies differences between remotely sensed images collected at different times and “subtract of one image from another”. This method correctly identified nearly 90% of the bears. The technique also generated false positives, but this problem can be corrected by a manual review.

Figure 05 shows the difference in resolution of two types of satellite imagery, the panchromatic (0.6 m resolution) and the multispectral (2.4 m resolution). LaRue et al. (2015)decided not to use the multispectral imagery due to resolution constraints.

A more recent study is being conducted by my fellow OSU Fisheries and Wildlife graduate student, Jane Dolliveron breeding colonies of three species of North Pacific albatrosses (Phoebastria immutabilis, Phoebastria nigripes, and Phoebastria albatrus)(Dolliver et al., 2017). Jane is using high-resolution multispectral satellite imagery (DigitalGlobe WorldView-2 and -3) and image processing techniques to enumerate the albatrosses. They are also using albatross species at multiple reference colonies in Hawaii and Japan (Fig. 06) to determine species identification accuracy and required correction factor(s). This will allow scientists to accurately count unknown populations on the Senkakus, which are uninhabited islands controlled by Japan in the East China Sea.

Using satellite imagery to count seals, penguins, whales, bears and albatrosses is just the start of this rapidly advancing technology. Techniques and resolutions are continuously improving. Methods can also be applied to many other endangered species, especially in remote areas, providing data on presence, abundance, annual productivity, population estimates and trends, changes in distribution, and breeding ground usage.

Other than directly monitoring wildlife, satellite images can also provide information on the environmental variables that can be related to wildlife presence, abundance, productivity and distribution.

Gentemann et al. (2017), for example, used satellite data from NASA to analyze SST variations along the west coast of the United States from 2002 to 2016. The NASA Jet Propulsion Laboratory produces global, daily, 1 km, multiscale ultra-high resolution, motion-compensated analysis of SST, and incorporates SSTs from eight different satellites. Researchers were able to identify warmer than usual SSTs (also called anomalies) along the Washington, Oregon, and California coasts from January 2014 to August 2016 (Fig.07) relative to previous years. This marine heat wave started in the Gulf of Alaska and ended in Southern California, where SST reached a maximum temperature anomaly of 6.2°C, causing major disturbances and substantial economic impacts.

Changes in SST and winds may alter events such as the coastal upwelling that supplies nutrients to sustain a whole food chain. A marine heat-wave event as described by Gentemann et al. (2017)could have significant impacts on the health of the marine ecosystem in the subsequent season (Gentemann et al., 2017).

These findings may even relate to our questions regarding the poor gray whale body condition we noticed in 2017: this marine heat wave that lasted until August 2016 along the US west coast could have impacted the ecosystem in the subsequent season. However, I must conduct a more detailed study to determine if this heat wave was related or if another oceanographic process was involved.

So, whether remotely sensed data is generated by satellites, drones, thermal imagery, robots (as I previously wrote about), or another type of technology, it can have important  and informative applications to monitor wildlife or environmental variables associated with their ecology and biology. We can take advantage of remotely sensed technology to aid wildlife conservation efforts.

References

Dolliver, J., et al., Multispectral processing of high resolution satellite imagery to determine the abundance of nesting albatross. Ecological Society of America, Portland, OR, United States., 2017.

Fretwell, P. T., et al., 2014. Whales from Space: Counting Southern Right Whales by Satellite. Plos One. 9,e88655.

Gentemann, C. L., et al., 2017. Satellite sea surface temperatures along the West Coast of the United States during the 2014–2016 northeast Pacific marine heat wave. Geophysical Research Letters. 44,312-319.

LaRue, M. A., et al., 2014. A method for estimating colony sizes of Adélie penguins using remote sensing imagery. Polar Biology. 37,507-517.

LaRue, M. A., et al., 2011. Satellite imagery can be used to detect variation in abundance of Weddell seals (Leptonychotes weddellii) in Erebus Bay, Antarctica. Polar Biology. 34,1727–1737.

LaRue, M. A., et al., 2015. Testing Methods for Using High-Resolution Satellite Imagery to Monitor Polar Bear Abundance and Distribution. Wildlife Society Bulletin. 39,772-779.

Scratching the Surface

By Dr. Leigh Torres, Assistant Professor, Oregon State University, Geospatial Ecology of Marine Megafauna Lab

I have been reminded of a lesson I learned long ago: Never turn your back on the sea – it’s always changing.

The blue whales weren’t where they were last time. I wrongly assumed oceanographic patterns would be similar to our last time out in 2014 and that the whales would be in the same area. But the ocean is dynamic – ever changing. I knew this. And I know it better now.

Below (Fig. 1) are two satellite images of sea surface temperature (SST) within the South Taranaki Bight and west coast region of New Zealand that we surveyed in Jan-Feb 2014 and again recently during Jan-Feb 2016. The plot on the left describes ocean surface conditions in 2014 and illustrates how SST primarily ranged between 15 and 18 ⁰C. By comparison, the panel on the right depicts the sea surface conditions we just encountered during the 2016 field season, and a huge difference is apparent: this year SST ranged between 18 and 23 ⁰C, barely overlapping with the 2014 field season conditions.

While whales can live in a wide range of water temperatures, their prey is much pickier. Krill, tiny zooplankton that blue whales seek and devour in large quantities, tend to aggregate in pockets of nutrient-rich, cool water in this region of New Zealand. During the 2014 field season, we encountered most blue whales in an area where SST was about 15 ⁰C (within the white circle in the left panel of Fig. 1). This year, there was no cool water anywhere and we mainly found the whales off the west coast of Kahurangi shoals in about 21 ⁰C water (within the white circle in the right panel of Fig. 1. NB: the cooler water in the Cook Strait in the southeast region of the right panel is a different water mass than preferred by blue whales and does not contain their prey.)

The hot water we found this year across the survey region can likely be attributed, at least in part, to the El Niño conditions that are occurring across the Pacific Ocean currently. El Niño has brought unusually settled conditions to New Zealand this summer, which means relatively few high wind events that normally churn up the ocean and mix the cool, nutrient rich deep water with the hot surface layer water. These are ideal conditions for Kiwi sun-bathers, but the ocean remains highly stratified with a stable layer of hot water on top. However, this stratification does not necessarily mean the ocean is un-productive – it only means that the SST satellite images are virtually useless for helping us to find whales this year.

Although SST data can be informative about ocean conditions, it only reflects what is happening in the thin, top slice of the ocean. Sub-surface conditions can be very different. Ocean conditions during our two survey periods in 2014 and 2016 could be more similar when compared underwater than when viewed from above. This is why sub-surface sensors and data collection is critical to marine studies. Ocean conditions in 2014 and 2016 could both potentially provide good habitat for the whales. In fact, where and when we encountered whales during both 2014 and 2016 we also detected high densities of krill through hydro-acoustics (Fig. 2). However, in 2014 we observed many surface swarms of krill that we rarely saw this recent field season, which could be due to elevated SST. But, we did capture cool drone footage this year of a brief sub-surface foraging event:

An overhead look of a blue whale foraging event as the animal approaches the surface. Note how the distended ventral (throat) grooves of the buccal cavity (mouth) are visible. This is a big gulp of prey (krill) and water. The video was captured using a DJI Phantom 3 drone in the South Taranaki Bight of New Zealand in on February 2, 2016 under a research permit from the New Zealand Department of Conservation (DOC) permit # 45780-MAR issued to Oregon State University.

Below are SST anomaly plots of January 2014 and January 2016 (Fig. 3). These anomaly plots show how different the SST was compared to the long-term average SST across the New Zealand region. As you can see, in 2014 (left panel) SST conditions in our study area were ~1 ⁰C below average, while in 2016 (right panel) SST conditions were ~1 ⁰C above average. So, what are normal conditions? What can we expect next year when we come back to survey again for blue whales across this region? These are challenging questions and illustrate why marine ecology studies like this one must be conducted over many years. One year is just a snap shot in the lifetime of the oceans.

Like all marine megafauna, blue whales move far and fast to adjust their distribution patterns according to ocean conditions. So, I can’t tell you what the ocean will be like in January 2017 or where the whales will be, but as we continue to study this marine ecosystem and its inhabitants our understanding of ocean patterns and whale ecology will improve. With every year of new data we will be able to better predict ocean and blue whale distribution patterns, providing managers with the tools they need to protect our marine environment. For now, we are just beginning to scratch the (sea) surface.