Do you lose SLEAP over video analysis of gray whale behavior? Not us in the GEMM Lab! 

Celest Sorrentino, University of California, Santa Barbara, Department of Ecological, Evolution, and Marine Biology, GEMM Lab NSF REU intern

Are you thinking “Did anyone proofread this blog beforehand? Don’t they know how to spell SLEEP?”  I completely understand this concern, but not to fear: the spelling of SLEAP is intentional! We’ll address that clickbait in just a moment. 

My name is Celest Sorrentino, a first-generation Latina undergrad who leaped at the opportunity to depart from the beaches of Santa Barbara, California to misty Newport, Oregon to learn and grow as a scientist under the influential guidance of Clara Bird, Dr. Leigh Torres and the powerhouse otherwise known as the GEMM lab. As a recent NSF REU (Research Experience for Undergraduates) intern in the GEMM Lab at Oregon State University, I am thrilled to have the chance to finally let you in on the project Clara, Leigh and I have been working on all summer. Ready for this?

Our project uses a deep-learning platform called SLEAP A.I. ( that can predict and track multiple animals in video to track gray whale mother calf pairs in drone footage. We also took this project a step further and explored how the distance between a gray whale mother and her calf, a proxy for calf independence, varied throughout the season and by different calf characteristics. 

In this story, we’ve got a little bit for everyone: the dynamic duo of computer vision and machine learning for my data scientist friends, and ecological inquest for my cetacean researcher friends. 

About the Author

Before we begin, I’d like to share that I am not a data scientist. I’ve only ever taken one coding class. I also do not have years of gray whale expertise under my belt (not yet at least!). I’m entering my 5th year at University of California, Santa Barbara as a double major in Ecology and Evolution (BS) as well as Italian Studies (BA). I am sharing this information to convey the feasibility of learning how to use machine-learning as a solution to streamline the laborious task of video analysis, which would permit more time towards answering your own ecological question, as we did here.

Essential Background

Hundreds of Hours of Drone footage

Since 2016, the GEMM Lab has been collecting drone footage of gray whales off the Oregon Coast to observe gray whale behavior in more detail (Torres et al. 2018). Drones have been shown to increase observational time of gray whales by a three-fold (Torres et al. 2018), including the opportunity to revisit the video with fresh eyes at any time one pleases. The GEMM Lab has flow over 500 flights in the past 6 years, including limited footage of gray whale mother-calf pairs. Little is known about gray whale mother-calf dynamics and even less about factors that influence calf development. As we cannot interview hundreds of gray-whale mother-calf pairs to develop a baseline for this information, we explore potential proxies for calf development instead (similar to how developmental benchmarks are used for human growth). 

Distance and Development

During our own life journey, each of us became less and less dependent on our parents to survive on our own. Formulating our first words so that we can talk for ourselves, cracking an egg for our parents so that we can one day cook for ourselves, or even letting go of their hand when crossing the street. For humans, we spend many years with our kin preparing for these moments, but gray whale mother-calf pairs only have a few months after birth until they separate. Gray whale calves are born on their wintering grounds in Baja California, Mexico (around February), migrate north with their mothers to the foraging grounds, and are then weaned during the foraging season (we think around August). This short time with their mother means that they have to become independent pretty quickly (about 6 months!).

Distance between mother and calf can be considered a measure of independence because we would expect increased distance between the pair as calf independence increases. In a study by Nielson et al (2019), distance between Southern Right Whale mother-calf pairs was found to increase as the calf grew, indicating that it can serve as a good proxy for independence. The moment a mother-calf pair separates has not been documented, but the GEMM lab has footage of calves during the foraging season pre-weaning that can be used to investigate this process.  However, video analysis is no easy feat: video analysis can range from post-processing, diligent evaluation, and video documentation (Torres et al. 2018). Although the use of UAS has become a popular method for many researchers, the extensive time required for video analysis is a limitation. As mentioned in Clara’s blog, the choice to pursue different avenues to streamline this process, such as automation through machine learning, is highly dependent on the purpose and the kind of questions a project intends to answer.


 In a world where modern technology is constantly evolving to cater towards making everyday tasks easier, machine learning leads the charge with its ability for a machine to perform human tasks. Deep learning is a subset of machine learning in which the model learns and adapts the ability to perform a task given a dataset. SLEAP (Social LEAP Estimation of Animal Poses) A.I. is an open-source deep-learning framework created to be able to track multiple subjects, specifically animals, throughout a variety of environmental conditions and social dynamics. In previous cases, SLEAP has tracked animals with distinct morphologies and conditions such as mice interactions, fruit flies engaging in courtship, and bee behavior in a petri dish (Pereira 2020). While these studies show that SLEAP could help make video analysis more efficient, these experiments were all conducted on small animals and in controlled environments. However, large megafauna, such as gray whales, cannot be cultivated and observed in a controlled Petri dish. Could SLEAP learn and adapt to predict and track gray whales in an uncontrolled environment, where conditions are never the same (ocean visibility, sunlight, obstructions)? 


In order to establish a model within SLEAP, we split our mother-calf drone video dataset into training (n=9) and unseen/testing (n=3) videos. Training involves teaching the model to recognize gray whales, and necessitated me to label every four frames using the following labels (anatomical features): rostrum, blowhole, dorsal, dorsal-knuckle, and tail (Fig. 1). Once SLEAP was trained and able to successfully detect gray whales, we ran the model on unseen video. The purpose of using unseen video was to evaluate whether the model could adapt and perform on video it had never seen before, eliminating the need for a labeler to retrain it. 

We then extracted the pixel coordinates for the mom and calf, calculated the distance between their respective dorsal knuckles, and converted the distance to meters using photogrammetry (see KC’s blog  for a great explanation of these methods).  The distance between each pair was then summarized on a daily scale as the average distance and the standard deviation. Standard deviation was explored to understand how variable the distance between mother-calf pair was throughout the day. We then looked at how distance and the standard deviation of distance varied by day of year, calf Total Length (TL), and calf Body Area Index (BAI; a measure of body condition). We hypothesized that these three metrics may be drivers of calf independence (i.e., as the calf gets longer or fatter it becomes more independent from its mother).  

Fig 1. Example of a labelled frame from SLEAP, highlighting labels: rostrum, blowhole, dorsal, dorsal-knuckle, and tail. 


SLEAP A.I. was able to successfully detect and track gray whale mother-calf pairs across all videos (that’s a total of 1318 frames!). When evaluating how the average distance changed across Day of Year, calf Total length, and calf BAI, the plots did not demonstrate the positive relationship we anticipated (Fig 2A). However, when evaluating the standard deviation of distance across Day of Year, calf Total Length, and calf BAI, we did notice that there does appear to be an increase in variability of distance with an increase in Day of Year and calf Total length (Fig 2B)

Fig 2A: Relationship between average distance(m) between each mother and calf (colors represent different named calves) relative to Day of Year, Total length of calf, and BAI of calf. 
Fig 2B: Relationship between standard deviation of  distance(m) between each mother and calf (colors represent different named calves) relative to Day of Year, Total length of calf, and BAI of calf.

Concluding thoughts

These results are monumental! We demonstrated the feasibility to use AI to create a model that can track gray whale pairs in drone footage, which is a fantastic tool that can be applied to updated datasets in the future. As more footage of gray whale mother-calf pairs are collected, this video can be quickly uploaded to SLEAP for model evaluation, predictions can be exported, and results subsequently included in the distance analysis to update our plots and increase our understanding. Our data currently provide a preliminary understanding of how the distance between mother-calf pairs changes with Day of Year, Total length, and BAI, but we are now able to continue updating our dataset as we collect more drone footage. 

I suppose you can say I did mislead you a bit with my title, as I have lost some SLEEP recently. But, not over video analysis per say but rather in the form of inspiration. Inspiration toward expanding my understanding of machine learning so that it can be applied toward answering pressing ecological questions. This project has only propelled me to dig my heels in and investigate further the potential of machine learning to analyze dense datasets for huge knowledge gains.

Fig 3A: Snapshot of Celest working in SLEAP GUI.


This project was made possible in partnership by the continuous support by Clara Bird, Dr. Leigh Torres, KC Bierlich, and the entire GEMM Lab!


Nielsen, M., Sprogis, K., Bejder, L., Madsen, P., & Christiansen, F. (2019). Behavioural development in southern right whale calves. Marine Ecology Progress Series629, 219–234.

Pereira, Talmo D., Nathaniel Tabris, Junyu Li, Shruthi Ravindranath, Eleni S. Papadoyannis, Z. Yan Wang, David M. Turner, et al. “SLEAP: Multi-Animal Pose Tracking.” Preprint. Animal Behavior and Cognition, September 2, 2020.

Torres, Nieukirk, S. L., Lemos, L., & Chandler, T. E. (2018). Drone Up! Quantifying Whale Behavior From a New Perspective Improves Observational Capacity. Frontiers in Marine Science, 5.

New publication by GEMM Lab reveals sub-population health differences in gray whales 

Dr. KC Bierlich, Postdoctoral Scholar, OSU Department of Fisheries, Wildlife, & Conservation Sciences, Geospatial Ecology of Marine Megafauna (GEMM) Lab

In a previous blog, I discussed the importance of incorporating measurement uncertainty in drone-based photogrammetry, as drones with different sensors, focal length lenses, and altimeters will have varying levels of measurement accuracy. In my last blog, I discussed how to incorporate photogrammetric uncertainty when combining multiple measurements to estimate body condition of baleen whales. In this blog, I will highlight our recent publication in Frontiers in Marine Science ( led by GEMM Lab’s Dr. Leigh TorresClara Bird, and myself that used these methods in a collaborative study using imagery from four different drones to compare gray whale body condition on their breeding and feeding grounds (Torres et al., 2022).

Most Eastern North Pacific (ENP) gray whales migrate to their summer foraging grounds in Alaska and the Arctic, where they target benthic amphipods as prey. A subgroup of gray whales (~230 individuals) called the Pacific Coast Feeding Group (PCFG), instead truncates their migration and forages along the coastal habitats between Northern California and British Columbia, Canada (Fig. 1). Evidence from a recent study lead by GEMM Lab’s Lisa Hildebrand (see this blog) found that the caloric content of prey in the PCFG range is of equal or higher value than the main amphipod prey in the Arctic/sub-Arctic regions (Hildebrand et al., 2021). This implies that greater prey density and/or lower energetic costs of foraging in the Arctic/sub-Arctic may explain the greater number of whales foraging in that region compared to the PCFG range. Both groups of gray whales spend the winter months on their breeding and calving grounds in Baja California, Mexico. 

Figure 1. The GEMM Lab field team following a Pacific Coast Feeding Group (PCFG) gray whale swimming in a kelp bed along the Oregon Coast during the summer foraging season. 

In January 2019 an Unusual Mortality Event (UME) was declared for gray whales due to the elevated numbers of stranded gray whales between Mexico and the Arctic regions of Alaska. Most of the stranded whales were emaciated, indicating that reduced nutrition and starvation may have been the causal factor of death. It is estimated that the population dropped from ~27,000 individuals in 2016 to ~21,000 in 2020 (Stewart & Weller, 2021).

During this UME period, between 2017-2019, the GEMM Lab was using drones to monitor the body condition of PCFG gray whales on their Oregon coastal feeding grounds (Fig. 1), while Christiansen and colleagues (2020) was using drones to monitor gray whales on their breeding grounds in San Ignacio Lagoon (SIL) in Baja California, Mexico. We teamed up with Christiansen and colleagues to compare the body condition of gray whales in these two different areas leading up to the UME. Comparing the body condition between these two populations could help inform which population was most effected by the UME.

The combined datasets consisted of four different drones used, thus different levels of photogrammetric uncertainty to consider. The GEMM Lab collected data using a DJI Phantom 3 Pro, DJI Phantom 4, and DJI Phantom 4 Pro, while Christiansen et al., (2020) used a DJI Inspire 1 Pro. By using the methodological approach described in my previous blog (here, also see Bierlich et al., 2021a for more details), we quantified photogrammetric uncertainty specific to each drone, allowing cross-comparison between these datasets. We also used Body Area Index (BAI), which is a standardized relative measure of body condition developed by the GEMM Lab (Burnett et al., 2018) that has low uncertainty with high precision, making it easier to detect smaller changes between individuals (see blog here, Bierlich et al., 2021b). 

While both PCFG and ENP gray whales visit San Ignacio Lagoon in the winter, we assume that the photogrammetry data collected in the lagoon is mostly of ENP whales based on their considerably higher population abundance. We also assume that gray whales incur low energetic cost during migration, as gray whale oxygen consumption rates and derived metabolic rates are much lower during migration than on foraging grounds (Sumich, 1983). 

Interestingly, we found that gray whale body condition on their wintering grounds in San Ignacio Lagoon deteriorated across the study years leading up to the UME (2017-2019), while the body condition of PCFG whales on their foraging grounds in Oregon concurrently increased. These contrasting trajectories in body condition between ENP and PCFG whales implies that dynamic oceanographic processes may be contributing to temporal variability of prey available in the Arctic/sub-Arctic and PCFG range. In other words, environmental conditions that control prey availability for gray whales are different in the two areas. For the ENP population, this declining nutritive gain may be associated with environmental changes in the Arctic/sub-Arctic region that impacted the predictability and availability of prey. For the PCFG population, the increase in body condition across years may reflect recovery of the NE Pacific Ocean from the marine heatwave event in 2014-2016 (referred to as “The Blob”) that resulted with a period of low prey availability. These findings also indicate that the ENP population was primarily impacted in the die-off from the UME. 

Surprisingly, the body condition of PCFG gray whales in Oregon was regularly and significantly lower than whales in San Ignacio Lagoon (Fig. 2). To further investigate this potential intrinsic difference in body condition between PCFG and ENP whales, we compared opportunistic photographs of gray whales feeding in the Northeastern Chukchi Sea (NCS) in the Arctic collected from airplane surveys. We found that the body condition of PCFG gray whales was significantly lower than whales in the NCS, further supporting our finding that PCFG whales overall have lower body condition than ENP whales that feed in the Arctic (Fig. 3). 

Figure 2. Boxplots showing the distribution of Body Area Index (BAI) values for gray whales imaged by drones in San Ignacio Lagoon (SIL), Mexico and Oregon, USA. The data is grouped by phenology group: End of summer feeding season (departure Oregon vs. arrival SIL) and End of wintering season (arrival Oregon vs. departure SIL). The group median (horizontal line), interquartile range (IQR, box), maximum and minimum 1.5*IQR (vertical lines), and outliers (dots) are depicted in the boxplots. The overlaid points represent the mean of the posterior predictive distribution for BAI of an individual and the bars represents the uncertainty (upper and lower bounds of the 95% HPD interval). Note how PCFG whales at then end of the feeding season (dark green) typically have lower body condition (as BAI) compared to ENP whales at the end of the feeding season when they arrive to SIL after migration (light brown).
Figure 3. Boxplots showing the distribution of Body Area Index (BAI) values of gray whales from opportunistic images collected from a plane in Northeaster Chukchi Sea (NCS) and from drones collected by the GEMM Lab in Oregon. The boxplots display the group median (horizontal line), interquartile range (IQR box), maximum and minimum 1.5*IQR (vertical lines), and outlies (dots). The overlaid points are the BAI values from each image. Note the significantly lower BAI of PCFG whales on Oregon feeding grounds compared to whales feeding in the Arctic region of the NCS.

This difference in body condition between PCFG and ENP gray whales raises some really interesting and prudent questions. Does the lower body condition of PCFG whales make them less resilient to changes in prey availability compared to ENP whales, and thus more vulnerable to climate change? If so, could this influence the reproductive capacity of PCFG whales? Or, are whales that recruit into the PCFG adapted to a smaller morphology, perhaps due to their specialized foraging tactics, which may be genetically inherited and enables them to survive with reduced energy stores?

These questions are on our minds here at the GEMM Lab as we prepare for our seventh consecutive field season using drones to collect data on PCFG gray whale body condition. As discussed in a previous blog by Dr. Alejandro Fernandez Ajo, we are combining our sightings history of individual whales, fecal hormone analyses, and photogrammetry-based body condition to better understand gray whales’ reproductive biology and help determine what the consequences are for these PCFG whales with lower body condition.

Did you enjoy this blog? Want to learn more about marine life, research, and conservation? Subscribe to our blog and get a weekly message when we post a new blog. Just add your name and email into the subscribe box below.



Bierlich, K. C., Hewitt, J., Bird, C. N., Schick, R. S., Friedlaender, A., Torres, L. G., … & Johnston, D. W. (2021). Comparing Uncertainty Associated With 1-, 2-, and 3D Aerial Photogrammetry-Based Body Condition Measurements of Baleen Whales. Frontiers in Marine Science, 1729.

Bierlich, K. C., Schick, R. S., Hewitt, J., Dale, J., Goldbogen, J. A., Friedlaender, A.S., et al. (2021b). Bayesian Approach for Predicting Photogrammetric Uncertainty in Morphometric Measurements Derived From Drones. Mar. Ecol. Prog. Ser. 673, 193–210. doi: 10.3354/meps13814

Burnett, J. D., Lemos, L., Barlow, D., Wing, M. G., Chandler, T., & Torres, L. G. (2018). Estimating morphometric attributes of baleen whales with photogrammetry from small UASs: A case study with blue and gray whales. Marine Mammal Science35(1), 108–139.

Christiansen, F., Rodrı́guez-González, F., Martı́nez-Aguilar, S., Urbán, J., Swartz, S., Warick, H., et al. (2021). Poor Body Condition Associated With an Unusual Mortality Event in Gray Whales. Mar. Ecol. Prog. Ser. 658, 237–252. doi:10.3354/meps13585

Hildebrand, L., Bernard, K. S., and Torres, L. G. (2021). Do Gray Whales Count Calories? Comparing Energetic Values of Gray Whale Prey Across Two Different Feeding Grounds in the Eastern North Pacific. Front. Mar. Sci. 8. doi: 10.3389/fmars.2021.683634

Stewart, J. D., and Weller, D. (2021). Abundance of Eastern North Pacific Gray Whales 2019/2020 (San Diego, CA: NOAA/NMFS)

Sumich, J. L. (1983). Swimming Velocities, Breathing Patterns, and Estimated Costs of Locomotion in Migrating Gray Whales, Eschrichtius Robustus. Can. J. Zoology. 61, 647–652. doi: 10.1139/z83-086

Torres, L.G., Bird, C., Rodrigues-Gonzáles, F., Christiansen F., Bejder, L., Lemos, L., Urbán Ramírez, J., Swartz, S., Willoughby, A., Hewitt., J., Bierlich, K.C. (2022). Range-wide comparison of gray whale body condition reveals contrasting sub-population health characteristics and vulnerability to environmental change. Frontiers in Marine Science. 9:867258.

The many dimensions of a fat whale: Using drones to measure the body condition of baleen whales 

Dr. KC Bierlich, Postdoctoral Scholar, OSU Department of Fisheries, Wildlife, & Conservation Sciences, Geospatial Ecology of Marine Megafauna (GEMM) Lab

In my last blog, I discussed how to obtain morphological measurements from drone-based imagery of whales and the importance of calculating and considering uncertainty, as different drone platforms have varying levels of measurement uncertainty. But how does uncertainty scale and propagate when multiple measurements are combined, such as when measuring body condition of the whole animal? In this blog, I will discuss the different methods used for measuring body condition of baleen whales from drone-based imagery and how uncertainty differs between these metrics.

Body condition is defined as the energy stored in the body as a result of feeding and is assumed to indicate an animal’s overall health, as it reflects the balance between energy intake and investment toward growth, maintenance and reproduction (Peig and Green, 2009). Thus, body condition reflects the foraging success of an individual, as well as the potential for reproductive output and the quality of habitat. For example, female North American brown bears (Ursus arctos) in high quality habitats were in better body condition, produced larger litter sizes, and lived in greater population densities compared to females in lower quality habitats (Hilderbrand et al., 1999). As Dawn Barlow and Will Kennerley discussed in their recent blog, baleen whales are top predators and serve as ecosystem sentinels that shed light not only on the health of their population, but on the health of their ecosystem. As ocean climate conditions continue to change, monitoring the body condition of baleen whales is important to provide insight on how their population and ecosystem is responding. 

As discussed in a previous blog, drones serve as a valuable tool for obtaining morphological measurements of baleen whales to estimate their body condition. Images are imported into photogrammetry software, such as MorphoMetriX (Torres and Bierlich, 2020), to measure the total length of an individual and that is then divided into perpendicular width segments (i.e., in 5 or 10% increments) down the body (Fig. 1). These total length and width measurements are then used to estimate body condition in either 1-, 2-, or 3-dimensions: a single width (1D), a projected dorsal surface area (2D), or a body volume measure (3D). These 1D, 2D, and 3D measurements of body condition can then be standardized by total length to produce a relative measure of an individual’s body condition to compare among individuals and populations. 

Figure 1. An example of a Pacific Coast Feeding Group (PCFG) gray whale measured in MorphoMetriX (Torres & Bierlich, 2020).

While several different studies have used each of these dimensions to assess whale body condition, it is unclear how these measurements compare amongst each other. Importantly, it is also unclear how measurement uncertainty scales across these multiple dimensions and influences inference, which can lead to misinterpretation of data. For example, the surface area and volume of two geometrically similar bodies of different sizes are not related to their linear dimensions in the same ratio, but rather to the second and third power, respectively (i.e., x2 vs. x3).  Similarly, uncertainty should not be expected to scale linearly across 1D, 2D, and 3D body condition measurements. 

The second chapter of my dissertation, which was recently published in Frontiers in Marine Science and includes Clara Bird and Leigh Torres as co-authors, compared the uncertainty associated with 1D, 2D, and 3D drone-based body condition measurements in three baleen whale species with different ranges in body sizes: blue, humpback, and Antarctic minke whales (Figure 2) (Bierlich et al., 2021). We used the same Bayesian model discussed in my last blog, to incorporate uncertainty associated with each 1D, 2D, and 3D estimate of body condition. 

Figure 2. An example of total length and perpendicular width (in 5% increments of total length) measurements of an individual blue, humpback and Antarctic minke whale. Each image measured using MorphoMetriX (Torres and Bierlich, 2020). 

We found that uncertainty does not scale linearly across multi-dimensional measurements, with 2D and 3D uncertainty increasing by a factor of 1.45 and 1.76 compared to 1D, respectively. This result means that there is an added cost of increased uncertainty when utilizing a multidimensional body condition measurement. Our finding is important to help researchers decide which body condition measurement best suits their scientific question,  particularly when using a drone platform that is susceptible to greater error – as discussed in my previous blog. However, a 1D measurement only relies on a single width measurement, which may be excluding other regions of an individual’s body condition that is important for energy storage. In these situations, a 2D or 3D measure may be more appropriate.

We found that when comparing relative measures of body condition (standardized by total length of the individual), each standardized metric was highly correlated with one another. This finding suggests that 1D, 2D, and 3D metrics will draw similar relative predictions of body condition for individuals, allowing researchers to be confident they will draw similar conclusions relating to the body condition of individuals, regardless of which standardized metric they use. However, when comparing the precision of each of these metrics, the body area index (BAI) – a 2D standardized metric – displayed the highest level of precision. This result highlights how BAI can advantageously detect small changes in body condition, which is useful for comparing individuals or even tracking the same individual over time.

BAI was developed by the GEMM Lab (Burnett et al., 2018) and was designed to be similar to body mass index (BMI) in humans [BMI = mass (kg)/(height (m))2], where BAI uses the calculated surface area as a surrogate for body mass. In humans, a healthy BMI range is generally considered 18.5–24.9, below 18.5 is considered underweight, above 24.9 is considered overweight, and above 30 is considered obese (Flegal et al., 2012). Identifying a healthy range in BAI for baleen whales is challenged by a limited knowledge of what a “healthy” body condition range is for a whale. We found strong evidence that a healthy range of BAI is species-specific, as each species displayed a distinctive range in BAI: blue whales: 11–16; AMW: 17–24; humpback whales: 23–32; humpback whale calves: 23–28 (Fig. 3). These differences in BAI ranges likely reflect differences in the body shape of each species (Fig. 4). For example, humpbacks have the widest range of BAI compared to these other two species, which was also reflected in their larger variation in perpendicular widths (Figs. 2-4). Thus, it seems that BAI offers conditionally “scalefree” comparisons between species, yet it is unreasonable to set a single, all-whale BAI threshold to determine “healthy” versus “unhealthy” body condition.  Collecting a large sample of body condition measurements across many individuals and demographic units over space and time with information on vital rates (e.g., reproductive capacity) will help elucidate a healthy BAI range for each species.

Figure 3. Body area index (BAI) for each species. AMW = Antarctic minke whale.  Figure from Bierlich et al. (2021).
Figure 4. A) Absolute widths (m) and B) relative widths, standardized by total length (TL) to help elucidate the different body shapes of Antarctic minke whales (AMW; n = 40), blue whales (n = 32), humpback whales (n = 40), and humpback whale calves (n = 15). Note how the peak in body width occurs at a different percent body width between species, demonstrating the natural variation in body shape between baleen whales. Figure from Bierlich et al. (2021).

Over the past six years, the GEMM Lab has been collecting drone images of Pacific Coast Feeding Group (PCFG) gray whales off the coast of Oregon to measure their BAI (see GRANITE Project blog). Many of the individuals we encounter are seen across years and throughout the foraging season, providing an opportunity to evaluate how an individual’s BAI is influenced by environmental variation, stress levels, maturity, and reproduction. These data will in turn help determine what the healthy range in BAI for gray whales is. For example, linking BAI to pregnancy – whether a whale is currently pregnant or becomes pregnant the following season – will help determine what BAI is needed to support calf production. We are currently analyzing hundreds of body condition measurements from 2016 – 2021, so stay tuned for upcoming results!


Bierlich, K. C., Hewitt, J., Bird, C. N., Schick, R. S., Friedlaender, A., Torres, L. G., … & Johnston, D. W. (2021). Comparing Uncertainty Associated With 1-, 2-, and 3D Aerial Photogrammetry-Based Body Condition Measurements of Baleen Whales. Frontiers in Marine Science, 1729.

Burnett, J. D., Lemos, L., Barlow, D., Wing, M. G., Chandler, T., & Torres, L. G. (2018). Estimating morphometric attributes of baleen whales with photogrammetry from small UASs: A case study with blue and gray whales. Marine Mammal Science35(1), 108–139.

Flegal, K. M., Carroll, M. D., Kit, B. K., & Ogden, C. L. (2012). Prevalence of Obesity and Trends in the Distribution of Body Mass Index Among US Adults, 1999-2010. JAMA307(5), 491.

Hilderbrand, G. V, Schwartz, C. C., Robbins, C. T., Jacoby, M. E., Hanley, T. A., Arthur, S. M., & Servheen, C. (1999). The importance of meat, particularly salmon, to body size, population productivity, and conservation of North American brown bears. Canadian Journal of Zoology77(1), 132–138.

Peig, J., & Green, A. J. (2009). New perspectives for estimating body condition from mass/length data: the scaled mass index as an alternative method. Oikos118(12), 1883–1891.

Torres, W., & Bierlich, K. C. (2020). MorphoMetriX: a photogrammetric measurement GUI for morphometric analysis of megafauna. Journal of Open Source Software5(45), 1825–1826.

Drones with lasers: almost as cool as “sharks with laser beams attached to their heads”

Dr. KC Bierlich, Postdoctoral Scholar, OSU Department of Fisheries, Wildlife, & Conservation Sciences, Geospatial Ecology of Marine Megafauna (GEMM) Lab

The recent advancement in drones (or unoccupied aircraft systems, UAS) has greatly enhanced opportunities for scientists across a broad range of disciplines to collect high-resolution aerial imagery. Wildlife researchers in particular have utilized this technology to study large elusive animals, such as whales, to observe their behavior (see Clara Bird’s blog) and obtain morphological measurements via photogrammetry (see previous blog for a brief history on photogrammetry and drones). However, obtaining useful measurement data is not as easy as flying the drone and pressing record. For this blog, I will provide a brief overview on the basics of using photogrammetry to extract morphological measurements from images collected with drones, as well as the associated uncertainty from using different drone platforms. 

During my PhD at Duke University, I co-developed an open-source photogrammetry software called MorphoMetriX to measure whales in images I collected using drones (Fig. 1) (Torres and Bierlich, 2020) (see this blog for some fieldwork memoirs!). The software is designed to be flexible, simple to use, and customizable without knowledge of scripting languages. Using MorphoMetriX, measurements are made in pixels and then multiplied by the ground sampling distance (GSD) to convert to standard units (e.g., meters) (Fig. 2A). GSD represents the distance on the ground each pixel represents (i.e., the linear size of the pixel) and therefore sets the scale of the image (i.e., cm per pixel). Figure 2A describes how GSD is dependent on the camera sensor, focal length lens, and altitude. Thus, drones equipped with different cameras and focal length lenses will have inherent differences in GSD as altitude increases (Fig. 2B). A larger GSD increases the length each pixel represents in a photo and results in a lower resolution image, potentially obscuring important features in the photo and introducing measurement error.

Figure 1. An example of a Pacific Coast Feeding Group gray whale measured in MorphoMetriX (Torres & Bierlich, 2020).
Figure 2: Overview of photogrammetry methods and calculating ground sampling distance (GSD). A) Photogrammetry methods for how each image is scaled to convert measurements in pixels to standard units (e.g., meters). Altitude is the distance between the camera lens and whale (usually at the surface of the water). Figure from Torres and Bierlich (2020). B) The exact (not accounting for distortion or altitude error) ground sampling distance (GSD) for two drone platforms commonly used to obtain morphological measurements of cetaceans. The difference in GSD between the P4Pro and Inspire 1 is due to the difference in sensor width and focal lengths of the cameras used. Figure from Bierlich et al. (2021).

Obtaining accurate altitude information is a key component in obtaining accurate measurements. All drones are equipped with a barometer, which measures altitude from changes in pressure. In general, barometers usually yield low accuracy in the altitude recorded, particularly for low-cost sensors commonly found on small, off-the-shelf drones (Wei et al., 2016). Dawson et al. (2017) added a laser altimeter (i.e., LightWare SF11/C, to a drone, which yields higher accuracy in the altitude recorded. Since then, several studies have adopted use of a laser altimeter to study different species of baleen whales (i.e., Gough et al., 2019; Christiansen et al., 2018).

The first chapter of my dissertation, which was published last year in Marine Ecology Progress Series, compared the accuracy of several drones equipped with different camera sensors, focal length lenses, and a barometer vs. laser altimeter (Bierlich et al., 2021). We flew each drone over a known sized object floating at the surface and collected images at various altitudes (between 10 – 120 m). We used the known size of the floating object to determine the percent error of each measurement at each altitude. We found that 1) there is a lot of variation in measurement error across the different drones when using a barometer to measure altitude and 2) using a laser altimeter dramatically reduces measurement error for each drone (Fig. 3).

Figure 3. The % error for measurements from different drones. Black dashed line represents 0% error (true length = 1.48 m). The gray dashed lines represent under- and over-estimation of the true length by 5% (1.41 and 1.55 m, respectively).

These findings are important because if a study is analyzing measurements that are from more than one drone, the uncertainty associated with those measurements must be taken into account to know if measurements are reliable and comparable. For instance, let’s say we are comparing the body length of two different populations and found that population A is significantly longer than population B. From looking at Figure 3, that significant difference in length between population A and B could be unreliable as the difference may be due to the bias introduced by the type of drone, camera sensor, focal length lens, and whether a barometer or laser altimeter was used for recording altitude. In other words, without incorporating uncertainty associated with each measurement, how do you trust your measurement? 

Hence, the National Institute of Standards and Technology (NIST) states that a measurement is complete only when accompanied by a quantitative statement of its uncertainty (Taylor & Kuyatt, 1994). In our Bierlich et al. (2021) study, we develop a Bayesian statistical model where we use the measurements of the known-sized object floating at the surface (what was used for Fig. 3) as training data to predict the lengths of unknown-sized whales. This Bayesian approach views data and the underlying parameters that generated the data (such as the mean or standard deviation) as random, and thus can be described by a statistical distribution. Using Bayes’ Theorem, a model of the observed data (called the likelihood function), is combined with prior knowledge pertaining to the underlying parameters (called the prior probability distribution) to form the posterior probability distribution, which serves as updated knowledge about the underlying parameter. For example, if someone told me they saw a 75 ft blue whale, I would not be phased. But if someone told me they saw a 150 ft blue whale, I would be skeptical – I’m using prior knowledge to determine the probability of this statement being true. 

The posterior probability distribution produced by this Bayesian approach can also serve as new prior information for subsequent analyses. Following this framework, we used the known-sized objects to first estimate the posterior probability distribution for error for each drone. We then used that posterior probability distribution for error specific to each drone platform as prior information to form a posterior predictive distribution for length of unknown-sized whales. The length of an individual whale can then be described by the mean of this second posterior predictive distribution, and its uncertainty defined as the variance or an interval around the mean (Fig. 4). 

Figure 4. An example of a posterior predictive distribution for total length of an individual blue whale. The black bars represent the uncertainty around the mean value (the black dot) – the longer black bars represent the 95% highest posterior density (HPD) interval, and the shorter black bars represent the 65% HPD interval. 

For over half a decade, the GEMM Lab has been collecting drone images of Pacific Coast Feeding Group (PCFG) of gray whales off the coast of Oregon to measure their morphology and body condition (see GRANITE Project Blog). We have been using several different types of drones equipped with different cameras, focal length lenses, barometers, and laser altimeters. These measurements from different drones will inherently have different levels of error associated with them, so adapting these methods for incorporating uncertainty will be key to ensure our measurements are comparable and analyses are robust. To do this, we fly over a known-sized board (1 m) at the start of each flight to use as training data to generate a posterior predictive distribution for length of the an unknown-sized PCFG gray whale that we fly over (Fig. 5). Likewise, we are working closely with several other collaborators who are also using different drones. Incorporating measurement uncertainty from drones used across research labs and in different environments will help ensure robust analyses and provide great opportunity for some interesting comparisons – such as differences in gray whale body condition on their feeding grounds in Oregon vs. their breeding grounds in Baja, Mexico, and morphological comparisons with other baleen whale species, such as blue and humpback whales. We are currently wrapping up measurement from thousands of boards (Fig. 5) and whales (Fig. 1) from 2016 – 2021, so stay tuned for the results!

Figure 5. An example of a known-sized object (1 m long board) used as training data to assess measurement uncertainty. 


Bierlich, K.C., Schick, R.S., Hewitt, J., Dale, J., Goldbogen, J.A., Friedlaender, A.S., Johnston D.J. (2021). A Bayesian approach for predicting photogrammetric uncertainty in morphometric measurements derived from UAS. Marine Ecology Progress Series. DOI:

Christiansen F, Vivier F, Charlton C, Ward R, Amerson A, Burnell S, Bejder L (2018) Maternal body size and condition determine calf growth rates in southern right whales. Mar Ecol Prog Ser 592: 267−281

Dawson SM, Bowman MH, Leunissen E, Sirguey P (2017) Inexpensive aerial photogrammetry for studies of whales and large marine animals. Front Mar Sci 4: 366

Gough, W.T., Segre, P.S., Bierlich, K.C., Cade, D.E., Potvin, J., Fish, F. E., Dale, J., di Clemente, J., Friedlaender, A.S., Johnston, D.W., Kahane-Rapport, S.R., Kennedy, J., Long, J.H., Oudejans, M., Penry, G., Savoca, M.S., Simon, M., Videsen, S.K.A., Visser, F., Wiley, D.N., Goldbogen, J.A. (2019). Scaling of swimming performance in baleen whales. Journal of Experimental Biology222(20).  

Taylor, B. N., and Kuyatt, C. E. (1994). Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results. Washington, DC: National Institute of Standards and Technology. 1–25.

Torres, W.I., & Bierlich, K.C. (2020). MorphoMetriX: a photogrammetric measurement GUI for morphometric analysis of megafauna. Journal of Open Source Software5(45), 1825.  

Wei S, Dan G, Chen H (2016) Altitude data fusion utilizing differential measurement and complementary filter. IET Sci Meas Technol (Singap) 10: 874−879

The costs and benefits of automated behavior classification

Clara Bird, PhD Student, OSU Department of Fisheries, Wildlife, and Conservation Sciences, Geospatial Ecology of Marine Megafauna Lab

“Why don’t you just automate it?” This is a question I am frequently asked when I tell someone about my work. My thesis involves watching many hours of drone footage of gray whales and meticulously coding behaviors, and there are plenty of days when I have asked myself that very same question. Streamlining my process is certainly appealing and given how wide-spread and effective machine learning methods have become, it is a tempting option to pursue. That said, machine learning is only appropriate for certain research questions and scales, and it’s important to consider these before investing in using a new tool.

The application of machine learning methods to behavioral ecology is called computational ethology (Anderson & Perona, 2014). To identify behaviors from videos, the model tracks individuals across video frames and identifies patterns of movement that form a behavior. This concept is similar to the way we identify a whale as traveling if it’s moving in a straight line and as foraging if it’s swimming in circles within a small area (Mayo & Marx, 1990, check out this blog to learn more). The level of behavioral detail that the model is able to track  depends on the chosen method (Figure 1, Pereira et al., 2020). These methods range from tracking each animal as a simple single point (called a centroid) to tracking the animal’s body positioning in 3D (this method is called pose estimation), which range from providing less detailed to more detailed behavior definitions. For example, tracking an individual as a centroid could be used to classify traveling and foraging behaviors, while pose estimation could identify specific foraging tactics. 

Figure 1. Figure from Pereira et al. (2020) illustrating the different methods of animal behavior tracking that are possible using machine learning.

Pose estimation involves training the machine learning algorithm to track individual anatomical features of an individual (e.g., the head, legs, and tail of a rat), meaning that it can define behaviors in great detail. A behavior state could be defined as a combination of the angle between the tail and the head, and the stride length. 

For example, Mearns et al. (2020) used pose estimation to study how zebrafish larvae in a lab captured their prey. They tracked the tail movements of individual larvae when presented with prey and classified these movements into separate behaviors that allowed them to associate specific behaviors with prey capture (Figure 2). The authors found that these behaviors occurred in a specific sequence, that the behaviors kept the prey within the larvae’s line of sight, and that the sequence was triggered by visual cues.  In fact, when they removed the visual cue of the prey, the larvae terminated the behavior sequence, meaning that the larvae are continually choosing to do each behavior in the sequence, rather than the sequence being one long behavior event that is triggered only by the initial visual cue. This study is a good example of the applicability of machine learning models for questions aimed at kinematics and fine-scale movements. Pose estimation has also been used to study the role of facial expression and body language in rat social communication (Ebbesen & Froemke, 2021). 

Figure 2. Excerpt from figure 1 of Mearns et al. (2020) illustrating (A) the camera set up for their experiment, (B) how the model tracked the eye angles and tail of the larvae fish, (C) the kinematics extracted from the footage. In panel (C) the top plot shows how the eyes converged on the same object (the prey) during prey capture event, the middle plot shows when the tail was curved to the left or the right, and the bottom plot shows the angle of the tail tip relative to the body.

While previous machine learning methods to track animal movements required individuals to be physically marked, the current methods can perform markerless tracking (Pereira et al., 2020). This improvement has broadened the kinds of studies that are possible. For example, Bozek et al., (2021) developed a model that tracked individuals throughout an entire honeybee colony and showed that certain individual behaviors were spatially distributed within the colony (Figure 3). Machine learning enabled the researchers to track over 1000 individual bees over several months, a task that would be infeasible for someone to do by hand. 

Figure 3. Excerpt from figure 1 of Bozek et al., (2021) showing how individual bees and their trajectories were tracked.

These studies highlight that the potential benefits of using machine learning when studying fine scale behaviors (like kinematics) or when tracking large groups of individuals. Furthermore, once it’s trained, the model can process large quantities of data in a standardized way to free up time for the scientists to focus on other tasks.

While machine learning is an exciting and enticing tool, automating behavior detection via machine learning could be its own PhD dissertation. Like most things in life, there are costs and benefits to using this technique. It is a technically difficult tool, and while applications exist to make it more accessible, knowledge of the computer science behind it is necessary to apply it effectively and correctly. Secondly, it can be tedious and time consuming to create a training dataset for the model to “learn” what each behavior looks like, as this step involves manually labeling examples for the model to use. 

As I’ve mentioned in a previous blog, I came quite close to trying to study the kinematics of gray whale foraging behaviors but ultimately decided that counting fluke beats wasn’t necessary to answer my behavioral research questions. It was important to consider the scale of my questions (as described in Allison’s blog) and I think that diving into more fine-scale kinematics questions could be a fascinating follow-up to the questions I’m asking in my PhD. 

For instance, it would be interesting to quantify how gray whales use their flukes for different behavior tactics. Do gray whales in better body condition beat their flukes more frequently while headstanding? Does the size of the fluke affect how efficiently they can perform certain tactics? While these analyses would help quantify the energetic costs of different behaviors in better detail, they aren’t necessary for my broad scale questions. Consequently, taking the time to develop and train a pose estimation machine learning model is not the best use of my time.

That being said, I am interested in applying machine learning methods to a specific subset of my dataset. In social behavior, it is not only useful to quantify the behaviors exhibited by each individual but also the distance between them. For example, the distance between a mom and her calf can be indicative of the calves’ dependence on its mom (Nielsen et al., 2019). However, continuously measuring the distance between two individuals throughout a video is tedious and time intensive, so training a machine learning model could be an effective use of time. I plan to work with an intern this summer to develop a machine learning model to track the distance between pairs of gray whales in our drone footage and then relate this distance data with the manually coded behaviors to examine patterns in social behavior (Figure 4).  Stay tuned to learn more about our progress!

Figure 4. A mom and calf pair surfacing together. Image collected under NOAA/NMFS permit #21678

Did you enjoy this blog? Want to learn more about marine life, research, and conservation? Subscribe to our blog and get a weekly alert when we make a new post! Just add your name into the subscribe box on the left panel.  


Anderson, D. J., & Perona, P. (2014). Toward a Science of Computational Ethology. Neuron84(1), 18–31.

Bozek, K., Hebert, L., Portugal, Y., Mikheyev, A. S., & Stephens, G. J. (2021). Markerless tracking of an entire honey bee colony. Nature Communications12(1), 1733.

Ebbesen, C. L., & Froemke, R. C. (2021). Body language signals for rodent social communication. Current Opinion in Neurobiology68, 91–106.

Mayo, C. A., & Marx, M. K. (1990). Surface foraging behaviour of the North Atlantic right whale, Eubalaena glacialis , and associated zooplankton characteristics. Canadian Journal of Zoology68(10), 2214–2220.

Mearns, D. S., Donovan, J. C., Fernandes, A. M., Semmelhack, J. L., & Baier, H. (2020). Deconstructing Hunting Behavior Reveals a Tightly Coupled Stimulus-Response Loop. Current Biology30(1), 54-69.e9.

Nielsen, M., Sprogis, K., Bejder, L., Madsen, P., & Christiansen, F. (2019). Behavioural development in southern right whale calves. Marine Ecology Progress Series629, 219–234.

Pereira, T. D., Shaevitz, J. W., & Murthy, M. (2020). Quantifying behavior to understand the brain. Nature Neuroscience23(12), 1537–1549.

Of snakes and whales: How food availability and body condition affect reproduction

Clara Bird, PhD Student, OSU Department of Fisheries, Wildlife, and Conservation Sciences, Geospatial Ecology of Marine Megafauna Lab

Over six field seasons the GEMM lab team has conducted nearly 500 drone flights over gray whales, equaling over 100 hours of footage. These hours of footage are the central dataset for my PhD dissertation, so it’s up to me to process them all. This process can be challenging, tedious, and daunting, but it is also quite fun and a privilege to be the one person who gets to watch all the footage. It’s fascinating to get to know the whales and their behaviors and pick up on patterns. It motivates me to get through this video processing step and start doing the data analysis. Recently, it’s been especially fun to notice patterns that I’ve seen mentioned in the literature. One example is adult social behavior. 

There are two categories of social behavior that I’m interested in studying: maternal behavior, defined as interactions between a mom and its calf, and general social behaviors, defined as social interactions between non-mom/calf pairs. In this blog I’ll focus on general social behaviors, but if you’re interested in maternal behavior check out this blog. General social behavior, which I’ll refer to as social behavior moving forward, includes tactile interactions and promiscuous behaviors (Torres et al. 2018; Clip 1). While gray whales in the PCFG range are primarily foraging, researchers have observed increases in social behavior towards the end of the foraging season (Stelle et al., 2008; Torres et al., 2018). We think that this indicates that the whales are starting to focus less on feeding and more on breeding. This tradeoff of foraging vs. socializing time is interesting because it comes at an energetic cost.

Clip 1. Example of social interaction between a male and female gray whale off the coast of Oregon, USA. Collected under NOAA/NMFS permit #21678

Broadly, animals need to balance the energetic demands of survival with those of reproduction. They need to reproduce to pass on their genes, but reproduction is energetically demanding, and animals also need to survive and grow to be able to reproduce. The decision to reproduce is costly because reproduction requires energetic investment and time investment since animals do not forage (gaining energy) when they are socializing. Consequently, only animals with sufficient energy reserves (i.e., body condition) to invest in reproduction actually engage in reproduction. Given these costs associated with reproduction, we expect to see a relationship between social behavior and body condition (Green, 2001) with mainly animals in good body condition engaging in social behavior because these animals have sufficient reserves to sustain the cost. Furthermore, since body condition is an indicator of foraging success and prey availability, environmental conditions can also affect social behavior and reproduction through this pathway. 

Rahman et al. (2014) used a lab experiment to study the relationship between nutritional stress and male guppy courtship behavior (Figure 1). In their experiment they tested for the effects of both decreased diet quantity and quality on the frequency of male courtship behaviors. Rahman et al (2014) found that individuals in the low-quantity group were significantly smaller than those in the high-quality group and that diet quantity had a significant effect on the frequency of courtship behaviors. Males fed a low-quantity diet performed fewer courtship behaviors. Interestingly, there was no significant effect of diet quality on courtships behavior, although there was some evidence of an interaction effect, which suggests that within the low-quantity group, males fed with high-quality food performed more courtship behaviors that those fed with low-quality food. This study is interesting because it shows how foraging success (diet quantity and quality) can affect courting behavior. 

Figure 1. A guppy (Rahman et al., 2013)

However, guppies are not the ideal species for comparison to gray whales because gray whales and guppies have quite different life history traits. A more fitting comparison would be with an example species with more in common with gray whales, such as viviparous capital breeders. Viviparous animals develop the embryo inside the body and give live birth. Capital breeders forage to build energy reserves and then rely on those energy reserves during reproduction. Surprisingly, I found asp vipers to be a good example species for comparison to gray whales.

Asp vipers (Figure 2) are viviparous snakes who are considered capital breeders because they forage prior to hibernation, and then begin reproduction immediately following hibernation without additional foraging. Naulleau & Bonnet (1996) conducted a field study on female asp vipers to determine if there was a difference in body condition at the start of the breeding season between females who reproduced or not during that season. To do this they marked individuals and measured their body condition at the start of the breeding season and then recaptured those individuals at the end of the breeding season and recorded whether the individual had reproduced. Interestingly, they found that there was a strongly significant difference in body condition between females that did and did not reproduce. In fact, they discovered that no female below a certain body condition value reproduced, meaning that they found a body condition threshold for reproduction. 

Figure 2. An asp viper

Additionally, a study on water pythons found that their body condition threshold for reproduction shifted over time in response to prey availability (Madsen & Shine, 1999). These authors found that females lowered their threshold after several consecutive years of poor prey availability. These studies are really exciting to me because they address questions that the GRANITE project team is interested in tackling.

Understanding the relationship between body condition and reproduction in gray whales is an important puzzle piece for our work. The aim of the GRANITE project is to understand how the effects of stressors on individual whales scales up to population level impacts (read Lisa’s blog to learn more). Reproduction rates play a big role in population dynamics, so it is important to understand what factors affect reproduction. Since we’re studying these whales on their foraging grounds, assessing body condition provides an important link between foraging behavior and reproduction. 

For example, if an individual’s response to a stressor is to forage less, that may lead to poorer body condition, meaning that they may be less likely to reproduce. While reduced reproduction in one individual may not have a big effect on the population, the same response from multiple individuals could impact the population’s dynamics (i.e., increasing or decreasing abundance). Understanding these different relationships between behavior, body condition, and reproduction rates is a big undertaking, but it’s exciting to be a member of the GRANITE team as this strong group of scientists works to bring together different data streams to work on this big picture question. We’re all deep into data processing right now so stay tuned over the next few years to learn more about gray whale social behavior and to find out if fat whales are more social than skinny whales. 

Did you enjoy this blog? Want to learn more about marine life, research, and conservation? Subscribe to our blog and get weekly updates and more! Just add your name into the subscribe box on the left panel.  


Green, A. J. (2001). Mass/Length Residuals: Measures of Body Condition or Generators of Spurious Results? Ecology82(5), 1473–1483.[1473:MLRMOB]2.0.CO;2

Madsen, T., & Shine, R. (1999). The adjustment of reproductive threshold to prey abundance in a capital breeder. Journal of Animal Ecology68(3), 571–580.

Naulleau, G., & Bonnet, X. (1996). Body Condition Threshold for Breeding in a Viviparous Snake. Oecologia107(3), 301–306.

Rahman, M. M., Kelley, J. L., & Evans, J. P. (2013). Condition-dependent expression of pre- and postcopulatory sexual traits in guppies. Ecology and Evolution3(7), 2197–2213.

Rahman, M. M., Turchini, G. M., Gasparini, C., Norambuena, F., & Evans, J. P. (2014). The Expression of Pre- and Postcopulatory Sexually Selected Traits Reflects Levels of Dietary Stress in Guppies. PLOS ONE9(8), e105856.

Stelle, L. L., Megill, W. M., & Kinzel, M. R. (2008). Activity budget and diving behavior of gray whales (Eschrichtius robustus) in feeding grounds off coastal British Columbia. Marine Mammal Science24(3), 462–478.

Torres, L. G., Nieukirk, S. L., Lemos, L., & Chandler, T. E. (2018). Drone up! Quantifying whale behavior from a new perspective improves observational capacity. Frontiers in Marine Science5(SEP).

Memoirs from above: drone observations of blue, humpback, Antarctic minke, and gray whales

By KC Bierlich, Postdoctoral Scholar, OSU Department of Fisheries, Wildlife, & Conservation Sciences, Geospatial Ecology of Marine Megafauna Lab

With the GRANITE field season officially over, we are now processing all of the data we collected this summer. For me, I am starting to go through all the drone videos to take snapshots of each whale to measure their body condition. As I go through these videos, I am reflecting on the different experiences I am fortunate enough to have with flying different drones, in different environments, over different species of baleen whales: blue, humpback, Antarctic minke, and now gray whales. Each of these species have a different morphological design and body shape (Woodward et al., 2006), which leads to different behaviors that are noticeable from the drone. Drones create immense opportunity to learn how whales thrive in their natural environments [see previous blog for a quick history], and below are some of my memories from above. 

I first learned how drones could be used to study the morphology and behavior of large marine mammals during my master’s degree at Duke University, and was inspired by the early works of John Durban (Durban et al., 2015, 2016) Fredrick Christiansen (Christiansen et al., 2016) and Leigh Torres (Torres et al., 2018). I immediately recognized the value and utility of this technology as a new tool to better monitor the health of marine mammals. This revelation led me to pursue a PhD with the Duke University Marine Robotics and Remote Sensing (MaRRS) Lab led by Dr. Dave Johnston where I helped further develop tools and methods for collecting drone-based imagery on a range of species in different habitats. 

When flying drones over whales, there are a lot of moving parts; you’re on a boat that is moving, flying something that is moving, following something that is moving. These moving elements are a lot to think about, so I trained hard, so I did not have to think about each step and flying felt intuitive and natural. I did not grow up playing video games, so reaching this level of comfort with the controls took a lot of practice. I practiced for hours over the course of months before my first field excursion and received some excellent mentorship and training from Julian Dale, the lead engineer in the MaRRS Lab. Working with Julian and the many hours of training helped me establish a solid foundation in my piloting skills and feel confident working in various environments on different species. 

Blue whales offshore of Monterey, California. 

In 2017 and 2018 I was involved in collaborative project with the MaRRS Lab and Goldbogen Lab at Stanford University, where we tagged and flew drones over blue whales offshore of Monterey, California. We traveled about an hour offshore and reliably found groups of blue whales actively feeding. Working offshore typically brought a large swell, which can often make landing the drone back into your field partner’s hands tricky as everything is bobbing up and down with the oscillations of the swell. Fortunately, we worked from a larger research vessel (~56 ft) and quickly learned that landing the drone in the stern helped dampen the effects of bobbing up and down. The blue whales we encountered often dove to a depth of around 200 m for about 20-minute intervals, then come to the surface for only a few minutes. This short surface period provided only a brief window to locate the whale once it surfaced and quickly fly over it to collect the imagery needed before it repeated its dive cycle. We learned to be patient and get a sense of the animal’s dive cycle before launch in order to time our flights so the drone would be in the air a couple of minutes before the whale surfaced. 

Once over the whales, the streamlined body of the blue whales was noticeable, with their small, high aspect ratio flippers and fluke that make them so well adapted for fast swimming in the open ocean (Fig. 1) (Woodward et al., 2006). I also noticed that because these whales are so large (often 21 – 24 m), I often flew at higher altitudes to be able fit them within the field of view of the camera. It was also always shocking to see how small the tagging boat (~8 m) looked when next to Earth’s largest creatures. 

Figure 1. Two blue whales surface after a deep dive offshore of Monterey, Ca. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03)

Antarctic minke whales and humpback whales along the Western Antarctic PeninsulaA lot of the data included in my dissertation came from work along the Western Antarctic Peninsula (WAP), which had a huge range of weather conditions, from warm and sunny days to cold and snowy/foggy/rainy/windy/icy days. A big focus was often trying to keep my hands warm, as it was often easier to fly without gloves in order to better feel the controls. One of the coldest days I remember was late in the season in mid-June (almost winter!) in Wilhemina Bay where ice completely covered the bay in just a couple hours, pushing the whales out into the Gerlache Strait; I suspect this was the last ice-free day of the season. Surprisingly though, the WAP also brought some of the best conditions I have ever flown in. Humpback and Antarctic minke whales are often found deep within the bays along the peninsula, which provided protection from the wind. So, there were times where it would be blowing 40 mph in the Gerlache Strait, but calm and still in the bays, such as Andvord Bay, which allowed for some incredible conditions for flying. Working from small zodiacs (~7 m) allowed us more maneuverability for navigating around or through the ice deep in the bays (Fig. 2) 

Figure 2. Navigating through ice-flows along the Western Antarctic Peninsula. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03 and ACA permits 2015-011 and 2020-016.)

Flying over Antarctic minke whale was always rewarding, as they are very sneaky and can quickly disappear under ice flows or in the deep, dark water. Flying over them often felt like a high-speed chase, as their small streamlined bodies makes them incredibly quick and maneuverable, doing barrel rolls, quick banked turns, and swimming under and around ice flows (Fig. 3). There would often be a group between 3-7 individuals and it felt like they were playing tag with each other – or perhaps with me!  

Figure 3. Two Antarctic minke whales swimming together along the Western Antarctic Peninsula. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03 and ACA permits 2015-011 and 2020-016.)

Humpbacks displayed a wide range of behaviors along the WAP. Early in the season they continuously fed throughout the entire day, often bubble net feeding in groups typically of 2-5 animals (Fig. 4). For as large as they are, it was truly amazing to see how they use their pectoral fins to perform quick accelerations and high-speed maneuvering for tight synchronized turns to form bubble nets, which corral and trap their krill, their main food source (Fig. 4) (Woodward et al., 2006). Later in the season, humpbacks switched to more resting behavior in the day and mostly fed at night, taking advantage of the diel vertical migration of krill. This behavior meant we often found humpbacks snoozing at the surface after a short dive, as if they were in a food coma. They also seemed to be more curious and playful with each other and with us later in the season (Fig. 5).

We also encountered a lot of mom and calf pairs along the WAP. Moms were noticeably skinny compared to their plump calf in the beginning of the season due to the high energetic cost of lactation (Fig. 6). It is important for moms to regain this lost energy throughout the feeding season and begin to wean their calves. I often saw moms refusing to give milk to their nudging calf and instead led teaching lessons for feeding on their own.

Figure 4. Two humpback whales bubble-net feeding early in the feeding season (December) along the Western Antarctic Peninsula. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03 and ACA permits 2015-011 and 2020-016.)
Figure 5. A curious humpback whale dives behind our Zodiac along the Western Antarctic Peninsula. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03 and ACA permits 2015-011 and 2020-016.)
Figure 6. A mom and her calf rest at the surface along the Western Antarctic Peninsula. Note how the mom looks skinnier compared to her plump calf, as lactation is the most energetically costly phase of the reproductive cycle. (Image credit: Duke University Marine Robotics and Remote Sensing under NOAA permit 14809-03 and ACA permits 2015-011 and 2020-016.)

Gray whales off Newport, Oregon

All of these past experiences helped me quickly get up to speed and jump into action with the GRANITE field team when I officially joined the GEMM Lab this year in June. I had never flown a DJI Inspire quadcopter before (the drone used by the GEMM Lab), but with my foundation piloting different drones, some excellent guidance from Todd and Clara, and several hours of practice to get comfortable with the new setup, I was flying over my first gray whale by day three of the job. 

The Oregon coast brings all sorts of weather, and some days I strangely found myself wearing a similar number of layers as I did in Antarctica. Fog, wind, and swell could all change within the hour, so I learned to make the most of weather breaks when they came. I was most surprised by how noticeably different gray whales behave compared to the blue, Antarctic minke, and humpback whales I had grown familiar with watching from above. For one, it is absolutely incredible to see how these huge whales use their low-aspect ratio flippers and flukes (Woodward et al., 2006) to perform low-speed, highly dynamic maneuvers to swim in very shallow water (5-10 m) so close to shore (<1m sometimes!) and through kelp forest or surf zones close to the beach. They have amazing proprioception, or the body’s ability to sense its movement, action, and position, as gray whales often use their pectoral fins and fluke to stay in a head standing position (see Clara Bird’s blog) to feed in the bottom sediment layer, all while staying in the same position and resisting the surge of waves that could smash them against the rocks (Video 1) . It is also remarkable how the GEMM Lab knows each individual whale based on natural skin marks, and I started to get a better sense of each whale’s behavior, including where certain individuals typically like to feed, or what their dive cycle might be depending on their feeding behavior. 

Video 1. Two Pacific Coast Feeding Group (PCFG) gray whales “head-standing” in shallow waters off the coast of Newport, Oregon. NOAA/NMFS permit #21678

I feel very fortunate to be a part of the GRANITE field team and to contribute to data collection efforts. I look forward to the data analysis phase to see what we learn about how the morphology and behavior of these gray whales to help them thrive in their environment. 


Christiansen, F., Dujon, A. M., Sprogis, K. R., Arnould, J. P. Y., and Bejder, L. (2016).Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales. Ecosphere 7, e01468–18.

Durban, J. W., Fearnbach, H., Barrett-Lennard, L. G., Perryman, W. L., & Leroi, D. J. (2015). Photogrammetry of killer whales using a small hexacopter launched at sea. Journal of Unmanned Vehicle Systems3(3), 131-135.

Durban, J. W., Moore, M. J., Chiang, G., Hickmott, L. S., Bocconcelli, A., Howes, G., et al.(2016). Photogrammetry of blue whales with an unmanned hexacopter. Mar. Mammal Sci. 32, 1510–1515.

Torres, L. G., Nieukirk, S. L., Lemos, L., & Chandler, T. E. (2018). Drone up! Quantifying whale behavior from a new perspective improves observational capacity. Frontiers in Marine Science5, 319.

Woodward, B. L., Winn, J. P., and Fish, F. E. (2006). Morphological specializations of baleen whales associated with hydrodynamic performance and ecological niche. J. Morphol. 267, 1284–1294.

Little whale, big whale, swimming in the water: A quick history on how aerial photogrammetry has revolutionized the ability to obtain non-invasive measurements of whales

Dr. KC Bierlich, Postdoctoral Scholar, OSU Department of Fisheries, Wildlife, & Conservation Sciences, Geospatial Ecology of Marine Megafauna Lab

The morphology and body size of an animal is one of the most fundamental factors for understanding a species ecology. For instance, fish body size and fin shape can influence its habitat use, foraging behavior, prey type, physiological performance, and predator avoidance strategies (Fig 1). Morphology and body size can thus reflect details of an individual’s current health, likelihood of survival, and potential reproductive success, which directly influences a species life history patterns, such as reproductive status, growth rate, and energetic requirements. Collecting accurate morphological measurements of individuals is often essential for monitoring populations, and recent studies have demonstrated how animal morphology has profound implications for conservation and management decisions, especially for populations inhabiting anthropogenically-altered environments (Miles 2020) (Fig. 1). For example, in a study on the critically endangered European eel, De Meyer et al. (2020) found that different skulls sizes were associated with different ecomorphs (a local variety of a species whose appearance is determined by its ecological environment), which predicted different diet types and resulted with some ecomorphs having a greater exposure to pollutants and toxins than others. However, obtaining manual measurements of wild animal populations can be logistically challenging, limited by accessibility, cost, danger, and animal disturbance. These challenges are especially true for large elusive animals, such as whales that are often in remote locations, spend little time at the surface of the water, and their large size can preclude safe capture and live handling.

Figure 1. Top) A pathway framework depicting how the morphology of an animal influences its habitat use, behavior, foraging, physiology, and performance. These traits all affect how successful an animal is in its environment and can reflect an individual’s current health, likelihood of survival, and potential reproductive success. This individual success can then be scaled up to assess overall population health, which in turn can have direct implications for conservation. Bottom) an example of morphological differences in fish body size and fin shape from Walker et al. (2013). Fineness ratio (f) = length of body ­÷ max body width. 

Photogrammetry is a non-invasive method for obtaining accurate morphological measurements of animals from photographs. The two main types of photogrammetry methods used in wildlife biology are 1) single camera photogrammetry, where a known scale factor is applied to a single image to measure 2D distances and angles and 2) stereo-photogrammetry, where two or more images (from a single or multiple cameras) are used to recreate 3D models. These techniques have been used on domestic animals to measure body condition and estimate weight of dairy cows and lactating Mediterranean buffaloes (Negretti et al., 2008; Gaudioso et al., 2014) and on wild animals to measure sexual dimorphism in Western gorillas (Breuer et al., 2007), shoulder heights of elephants (Schrader et al., 2006), nutritional status of Japanese macaques (Kurita et al., 2012), and the body condition of brown bears (Shirane et al., 2020). Over 70 years ago, Leedy (1948) encouraged wildlife biologists to use aerial photogrammetry from aircraft for censusing wild animal populations and their habitats, where photographs can be collected at nadir (straight down) or an oblique angle, and the scale can be calculated by dividing the focal length of the camera by the altitude or by using a ratio of selected points in an image of a known size. Indeed, aerial photogrammetry has been wildly adopted by wildlife biologists and has proven useful in obtaining measurements in large vertebrates, such as elephants and whales.

Whitehead & Payne (1978) first demonstrated the utility of using aerial photogrammetry from airplanes and helicopters as a non-invasive technique for estimating the body length of southern right whales. Prior to this technique, measurements of whales were traditionally limited to assessing carcasses collected from scientific whaling operations, or opportunistically from commercial whaling, subsistence hunting, stranding events, and bycatch. Importantly, aerial photogrammetry provides a method to collect measurements of whales without killing them. This approach has been widely adopted to obtain body length measurements on a variety of whale and dolphin species, including bowhead whales (Cubbage & Calambokidis, 1987), southern right whales (Best & Rüther, 1992), fin whales (Ratnaswamy and Wynn, 1993), common dolphins (Perryman and Lynn, 1993), spinner dolphins (Perryman & Westlake 1998), and killer whales (Fearnbach et al. 2012). Aerial photogrammetry has also been used to measure body widths to estimate nutritive condition related to reproduction in gray whales (Perryman and Lynn, 2002) and northern and southern right whales (Miller et al., 2012). However, these studies collected photographs from airplanes and helicopters, which can be costly, limited by weather and infrastructure to support aircraft research efforts and, importantly, presents a potential risk to wildlife biologists (Sasse 2003). 

The recent advancement and commercialization of unoccupied aircraft systems (UAS, or drones) has revolutionized the ability to obtain morphological measurements from high resolution aerial photogrammetry across a variety of ecosystems (Fig. 2). Drones ultimately bring five transformative qualities to conservation science compared to airplanes and helicopters: affordability, immediacy, quality, efficiency, and safety of data collection. Durban et al. (2015) first demonstrated the utility of using drones for non-invasively obtaining morphological measurements of killer whales in remote environments. Since then, drone-based morphological measurements have been applied to a wide range of studies that have increased our understanding on different whale populations. For example, Leslie et al. (2020) used drone-based measurements of the skull to distinguish a unique sub-species of blue whales off the coast of Chile. Groskreutz et al. (2019) demonstrated how long-term nutritional stress has limited body growth in northern resident killer whales, while Stewart et al. (2021) found a decrease in body length of North Atlantic Right whales since 1981 that was associated with entanglements from fishing gear and may be a contributing factor to the decrease in reproductive success for this endangered population. 

Drone imagery is commonly used to estimate the body condition of baleen whales by measuring the body length and width of individuals. Recently, the GEMM Lab used body length and width measurements to quantify intra- and inter-seasonal changes in body condition across individual gray whales (Lemos et al., 2020). Drones have also been used to measure body condition loss in humpback whales during the breeding season (Christiansen et al., 2016) and to compare the healthy southern right whales to the skinnier, endangered North Atlantic right whales (Christiansen et al., 2020). Drone-based assessments of body condition have even been used to measure how calf growth rate is directly related to maternal loss during suckling (Christiansen et al., 2018), and even estimate body mass (Christiansen et al., 2019). 

Drone-based morphological measurements can also be combined with whale-borne inertial sensing tag data to study the functional morphology across several different baleen whale species. Kahane-Rapport et al. (2020) used drone measurements of tagged whales to analyze the biomechanics of how larger whales require longer times for filtering the water through their baleen when feeding. Gough et al. (2019) used size measurements from drones and swimming speeds from tags to determine that a whale’s “walking speed” is 2 meters per second – whether the largest of the whales, a blue whale, or the smallest of the baleen whales, an Antarctic minke whale. Size measurements and tag data were combined by Segre et al. (2019) to quantify the energetic costs of different sized whales when breaching. 

Taken together, drones have revolutionized our ability to obtain morphological measurements of whales, greatly increasing our capacity to better understand how these animals function and perform in their environments. These advancements in marine science are particularly important as these methods provide greater opportunity to monitor the health of populations, especially as they face increased threats from anthropogenic stressors (such as vessel traffic, ocean noise, pollution, fishing entanglement, etc.) and climate change. 

Drone-based photogrammetry is one of the main focuses of the GEMM Lab’s project on Gray whale Response to Ambient Noise Informed by Technology and Ecology (GRANITE). This summer we have been collecting drone videos to measure the body condition of gray whales feeding off the coast of Newport, Oregon (Fig. 2). As we try to understand the physiological stress response of gray whales to noise and other potential stressors, we have to account for the impacts of overall nutritional state of each individual whale’s physiology, which we infer from these body condition estimates. 

Figure 2. Drones can help collect images of whales to obtain morphological measurements using photogrammetry and help us fill knowledge gaps for how these animals interact in their environment and to assess their current health. Bottom photo is an image collected by the GEMM Lab of a gray whale being measured in MorphoMetriX software to estimate its body condition. 


Best, P. B., & Rüther, H. (1992). Aerial photogrammetry of southern right whales, Eubalaena australis. Journal of Zoology228(4), 595-614.

Breuer, T., Robbins, M. M., & Boesch, C. (2007). Using photogrammetry and color scoring to assess sexual dimorphism in wild western gorillas (Gorilla gorilla). American Journal of Physical Anthropology134(3), 369–382. 

Christiansen, F., Vivier, F., Charlton, C., Ward, R., Amerson, A., Burnell, S., & Bejder, L. (2018). Maternal body size and condition determine calf growth rates in southern right whales. Marine Ecology Progress Series592, 267–281. 

Christiansen, F. (2020). A population comparison of right whale body condition reveals poor state of North Atlantic right whale, 1–43. 

Christiansen, F., Dujon, A. M., Sprogis, K. R., Arnould, J. P. Y., & Bejder, L. (2016). Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales. Ecosphere7(10), e01468–18. 

Christiansen, F., Sironi, M., Moore, M. J., Di Martino, M., Ricciardi, M., Warick, H. A., … Uhart, M. M. (2019). Estimating body mass of free-living whales using aerial photogrammetry and 3D volumetrics. Methods in Ecology and Evolution10(12), 2034–2044. 

Cubbage, J. C., & Calambokidis, J. (1987). Size-class segregation of bowhead whales discerned through aerial stereo-photogrammetry. Marine Mammal Science3(2), 179–185. 

De Meyer, J., Verhelst, P., & Adriaens, D. (2020). Saving the European Eel: How Morphological Research Can Help in Effective Conservation Management. Integrative and Comparative Biology23, 347–349. 

Gaudioso, V., Sanz-Ablanedo, E., Lomillos, J. M., Alonso, M. E., Javares-Morillo, L., & Rodr\’\iguez, P. (2014). “Photozoometer”: A new photogrammetric system for obtaining morphometric measurements of elusive animals, 1–10.

Gough, W. T., Segre, P. S., Bierlich, K. C., Cade, D. E., Potvin, J., Fish, F. E., … Goldbogen, J. A. (2019). Scaling of swimming performance in baleen whales. Journal of Experimental Biology222(20), jeb204172–11. 

Groskreutz, M. J., Durban, J. W., Fearnbach, H., Barrett-Lennard, L. G., Towers, J. R., & Ford, J. K. B. (2019). Decadal changes in adult size of salmon-eating killer whales in the eastern North Pacific. Endangered Species Research40, 1 

Kahane-Rapport, S. R., Savoca, M. S., Cade, D. E., Segre, P. S., Bierlich, K. C., Calambokidis, J., … Goldbogen, J. A. (2020). Lunge filter feeding biomechanics constrain rorqual foraging ecology across scale. Journal of Experimental Biology223(20), jeb224196–8. 

Leedy, D. L. (1948). Aerial Photographs, Their Interpretation and Suggested Uses in Wildlife Management. The Journal of Wildlife Management12(2), 191. 

Lemos, L. S., Burnett, J. D., Chandler, T. E., Sumich, J. L., and Torres, L. G. (2020). Intra- and inter-annual variation in gray whale body condition on a foraging ground. Ecosphere 11.

Leslie, M. S., Perkins-Taylor, C. M., Durban, J. W., Moore, M. J., Miller, C. A., Chanarat, P., … Apprill, A. (2020). Body size data collected non-invasively from drone images indicate a morphologically distinct Chilean blue whale (Balaenoptera musculus) taxon. Endangered Species Research43, 291–304. 

Miles, D. B. (2020). Can Morphology Predict the Conservation Status of Iguanian Lizards? Integrative and Comparative Biology

Miller, C. A., Best, P. B., Perryman, W. L., Baumgartner, M. F., & Moore, M. J. (2012). Body shape changes associated with reproductive status, nutritive condition and growth in right whales Eubalaena glacialis and E. australis. Marine Ecology Progress Series459, 135–156. 

Negretti, P., Bianconi, G., Bartocci, S., Terramoccia, S., & Verna, M. (2008). Determination of live weight and body condition score in lactating Mediterranean buffalo by Visual Image Analysis. Livestock Science113(1), 1–7. 

Ratnaswamy, M. J., & Winn, H. E. (1993). Photogrammetric Estimates of Allometry and Calf Production in Fin Whales, \emph{Balaenoptera physalus}. American Society of Mammalogists74, 323–330. 

Perryman, W. L., & Lynn, M. S. (1993). Idendification of geographic forms of common dolphin(\emph{Delphinus Delphis}) from aerial photogrammetry. Marine Mammal Science9(2), 119–137. 

Perryman, W. L., & Lynn, M. S. (2002). Evaluation of nutritive condition and reproductive status of migrating gray whales (\emph{Eschrichtius robustus}) based on analysisof photogrammetric data. Journal Cetacean Research and Management4(2), 155–164. 

Perryman, W. L., & Westlake, R. L. (1998). A new geographic form of the spinner dolphin, stenella longirostris, detected with aerial photogrammetry. Marine Mammal Science14(1), 38–50. 

Sasse, B. (2003). Job-Related Mortality of Wildlife Workers in the United States, 1937- 2000, 1015–1020. 

Segre, P. S., Potvin, J., Cade, D. E., Calambokidis, J., Di Clemente, J., Fish, F. E., … & Goldbogen, J. A. (2020). Energetic and physical limitations on the breaching performance of large whales. Elife9, e51760.

Shirane, Y., Mori, F., Yamanaka, M., Nakanishi, M., Ishinazaka, T., Mano, T., … Shimozuru, M. (2020). Development of a noninvasive photograph-based method for the evaluation of body condition in free-ranging brown bears. PeerJ8, e9982. 

Shrader, A. M., M, F. S., & Van Aarde, R. J. (2006). Digital photogrammetry and laser rangefinder techniques to measure African elephants, 1–7. 

Stewart, J. D., Durban, J. W., Knowlton, A. R., Lynn, M. S., Fearnbach, H., Barbaro, J., … & Moore, M. J. (2021). Decreasing body lengths in North Atlantic right whales. Current Biology.

Walker, J. A., Alfaro, M. E., Noble, M. M., & Fulton, C. J. (2013). Body fineness ratio as a predictor of maximum prolonged-swimming speed in coral reef fishes. PloS one8(10), e75422.

The learning curve never stops as the GRANITE project begins its seventh field season

Clara Bird, PhD Student, OSU Department of Fisheries, Wildlife, and Conservation Sciences, Geospatial Ecology of Marine Megafauna Lab

When I thought about what doing fieldwork would be like, before having done it myself, I imagined that it would be a challenging, but rewarding and fun experience (which it is). However, I underestimated both ends of the spectrum. I simultaneously did not expect just how hard it would be and could not imagine the thrill of working so close to whales in a beautiful place. One part that I really did not consider was the pre-season phase. Before we actually get out on the boats, we spend months preparing for the work. This prep work involves buying gear, revising and developing protocols, hiring new people, equipment maintenance and testing, and training new skills. Regardless of how many successful seasons came before a project, there are always new tasks and challenges in the preparation phase.

For example, as the GEMM Lab GRANITE project team geared up for its seventh field season, we had a few new components to prepare for. Just to remind you, the GRANITE (Gray whale Response to Ambient Noise Informed by Technology and Ecology) project’s field season typically takes place from June to mid-October of each year. Throughout this time period the field team goes out on a small RHIB (rigid hull inflatable boat), whenever the weather is good enough, to collect photo-ID data, fecal samples, and drone imagery of the Pacific Coast Feeding Group (PCFG) gray whales foraging near Newport, OR, USA. We use the data to assess the health, ecology and population dynamics of these whales, with our ultimate goal being to understand the effect of ambient noise on the population. As previous blogs have described, a typical field day involves long hours on the water looking for whales and collecting data. This year, one of our exciting new updates is that we are going out on two boats for the first part of the field season and starting our season 10 days early (our first day was May 20th). These updates are happening because a National Science Foundation funded seismic survey is being conducted within our study area starting in June. The aim of this survey is to assess geophysical structures but provides us with an opportunity to assess the effect of seismic noise on our study group by collecting data before, during, and after the survey. So, we started our season early in order to capture the “before seismic survey” data and we are using a two-boat approach to maximize our data collection ability.

While this is a cool opportunistic project, implementing the two-boat approach came with a new set of challenges. We had to find a second boat to use, buy a new set of gear for the second boat, figure out the best way to set up our gear on a boat we had not used before, and update our data processing protocols to include data collected from two boats on the same day. Using two boats also means that everyone on the core field team works every day. This core team includes Leigh (lab director/fearless leader), Todd (research assistant), Lisa (PhD student), Ale (new post-doc), and me (Clara, PhD student). Leigh and Todd are our experts in boat driving and working with whales, Todd is our experienced drone pilot, I am our newly certified drone pilot, and Lisa, Ale, and myself are boat drivers. Something I am particularly excited about this season is that Lisa, Ale, and I all have at least one field season under our belts, which means that we get to become more involved in the process. We are learning how to trailer and drive the boats, fly the drones, and handling more of the post-field work data processing. We are becoming more involved in every step of a field day from start to finish, and while it means taking on more responsibility, it feels really exciting. Throughout most of graduate school, we grow as researchers as we develop our analytical and writing skills. But it’s just as valuable to build our skillset for field work. The ocean conditions were not ideal on the first day of the field season, so we spent our first day practicing our field skills.

For our “dry run” of a field day, we went through the process of a typical day, which mostly involved a lot of learning from Leigh and Todd. Lisa practiced her trailering and launching of the boat (figure 1), Ale and Lisa practiced driving the boat, and I practiced flying the drone (figure 2). Even though we never left the bay or saw any whales, I thoroughly enjoyed our dry run. It was useful to run through our routine, without rushing, to get all the kinks out, and it also felt wonderful to be learning in a supportive environment. Practicing new skills is stressful to say the least, especially when there is expensive equipment involved, and no one wants to mess up when they’re being watched. But our group was full of support and appreciation for the challenges of learning. We cheered for successful boat launchings and dockings, and drone landings. I left that day feeling good about practicing and improving my drone piloting skills, full of gratitude for our team and excited for the season ahead.

Figure 1. Lisa (driving the truck) launching the boat.
Figure 2. Clara (seated, wearing a black jacket) landing the drone in Ale’s hands.

All the diligent prep work paid off on Saturday with a great first day (figure 3). We conducted five GoPro drops (figure 4), collected seven fecal samples from four different whales (figure 5), and flew four drone flights over three individuals including our star from last season, Sole. Combined, we collected two trifectas (photo-ID images, fecal samples, and drone footage)! Our goal is to get as many trifectas as possible because we use them to study the relationship between the drone data (body condition and behavior) and the fecal sample data (hormones). We were all exhausted after 10 hours on the water, but we were all very excited to kick-start our field season with a great day.

Figure 3. Lisa on the bow pulpit during our first sighting of the day.
Figure 4. Lisa doing a GoPro drop, she’s lowering the GoPro into the water using the line in her hands.
Figure 5. Clara and Ale collecting a fecal sample.

On Sunday, just one boat went out to collect more data from Sole after a rainy morning and I successfully flew over her from launching to landing! We have a long season ahead, but I am excited to learn and see what data we collect. Stay tuned for more updates from team GRANITE as our season progresses!

Defining Behaviors

Clara Bird, PhD Student, OSU Department of Fisheries and Wildlife, Geospatial Ecology of Marine Megafauna Lab

When I started working on my thesis, I anticipated many challenges related to studying the behavioral ecology of gray whales. From processing five-plus years of drone footage to data analysis, there has been no shortage of anticipated and unexpected issues. I recently hit an unexpected challenge when I started video processing that piqued my interest. As I’ve discussed in a previous blog, ethograms are lists of defined behaviors that help us properly and consistently collect data in a standardized approach. Ethograms form a crucial foundation of any behavior study as the behaviors defined ultimately affect what questions can be asked and what patterns are detected. Since I am working off of the thorough ethogram of Oregon gray whales from Torres et al. (2018), I had not given much thought to the process of adding behaviors to the ethogram. But, while processing the first chunk of drone videos, I noticed some behaviors that were not in the original ethogram and struggled to decide whether or not to add them. I learned that ethogram development can lead down several rabbit holes. The instinct to try and identify every movement is strong but dangerous. Every minute movement does not necessarily need to be included and it’s important to remember the ultimate goal of the analysis to avoid getting bogged down.

Fundamental behavior questions cannot be answered without ethograms. For example, Baker et al. (2017) developed an ethogram for bottlenose dolphins in Ireland in order to conduct an initial quantitative behavior analysis. They did so by reviewing published ethograms for bottlenose dolphins, consulting with multiple experts, and revising the ethogram throughout the study. They then used their data to test inter-observer variability, calculate activity budgets, and analyze how the activity budgets varied across space and time.

Howe et al. (2015) also developed an ethogram in order to conduct quantitative behavior analyses. Their goals were to use the ethogram and subsequent analyses to better understand the behavior of beluga whales in Cook Inlet, AK, USA and to inform conservation. They started by writing down all behaviors they observed in the field, then they consolidated their notes into a formal ethogram that they used and refined during subsequent field seasons. They used their data to analyze how the frequencies of different behaviors varied throughout the study area at different times. This study served as an initial analysis investigating the effect of anthropogenic disturbance and was refined in future studies.

My research is similarly geared towards understanding behavior patterns to ultimately inform conservation. The primary questions of my thesis involve individual specialization, patterns of behavior across space, the relationship between behavior and body condition, and social behavior (check out this blog to learn more). While deciding what behaviors to add to my ethogram I’ve had to remind myself of these main questions and the bigger picture. The drone footage lets us see so much detail that it’s tempting to try to define every movement we can observe. One rabbit hole I’ve had to avoid a few times is locomotion. From the footage, it is possible to document fluke beats and pectoral fin strokes. While it could be interesting to investigate how different whales move in different ways, it could easily become a complicated mess of classifying different movements and take me deep into the world of whale locomotion. Talking through what that work would look like reminded me that we cannot answer every question and trying to assess all exciting side projects can cause us to lose focus on the main questions.

While I avoided going down the locomotion rabbit hole, there were some new behaviors that I did add to my ethogram. I’ll illustrate the process with the examples of two new behaviors I recently added: fluke swish and pass under (Clips 1 and 2). Clip 1 shows a whale rapidly moving its fluke to the side. I chose to add fluke swish because it’s such a distinct movement and I’m curious to see if there’s a pattern across space, time, individual, or nearby human activity that might explain its function. Clip 2 shows a calf passing under its mom.  It’s not nursing because the calf doesn’t spend time under its mom, it just crosses underneath her. The calf pass under behavior could be a type of mom-calf tactile interaction. Analyzing how the frequency of this behavior changes over time could show how a calf’s dependency on its mom changes over as it ages.

In defining these behaviors, I had to consider how many different variations of this behavior would be included in the definition. This process involves considering at what point a variation of that behavior could serve a different function, even without knowing the function of the original behavior. For fluke swish this process involved deciding to only count a behavior as a fluke swish if it was a big, fast movement. A small and slow movement of the fluke a little to the side could serve a different function, such as turning, or be a random movement.

Clip 1: Fluke swish behavior (Video filmed under NOAA/NMFS research permit #16111​​ by certified drone pilot Todd Chandler).
Clip 2: Pass under behavior (Video filmed under NOAA/NMFS research permit #16111​​ by certified drone pilot Todd Chandler).

The next step involved deciding if the behavior would be a ‘state’ or ‘point’ event. A state event is a behavior with a start and stop moment; a point event is instantaneous and assigned to just a point in time. I would categorize a behavior as a state event if I was interested in questions about its duration. For example, I could ask “what percentage of the total observation time was spent in a certain behavior state?” A point event would be a behavior where duration is not applicable, but I could ask a question like “Did whale 1 perform more point event A than whale 2?”. Both fluke swish and pass under are point events because they only happen for an instant. In a pass under the calf is passing under its mom for just a brief point in time, making it a point event. The final step was to name the behavior. As I discussed in this blog, the name of the behavior does not matter as much as the definition but it is important that the name is clear and descriptive. We chose the name fluke swish because the fluke rapidly moves from side to side and pass under because the calf crosses under its mom.

Frankly, in the beginning, I was a bit overwhelmed by the realization that the content of my ethogram would ultimately control the questions I could answer. I could not help but worry that after processing all the videos, I would end up regretting not defining more behaviors. However, after reading some of the literature, chatting with Leigh, and reviewing the initial chunk of videos several times, I am more confidence in my judgment and my ethogram. I have accepted the fact that I can’t anticipate everything, and I am confident that the behaviors I need to answer my research questions are included. The process of reviewing and updating my ethogram has been a rewarding challenge that resulted in a valuable lesson that I will take with me for the rest of my career.


Baker, I., O’Brien, J., McHugh, K., & Berrow, S. (2017). An ethogram for bottlenose dolphins (Tursiops truncatus) in the Shannon Estuary, Ireland. Aquatic Mammals, 43(6), 594–613.

Howe, M., Castellote, M., Garner, C., McKee, P., Small, R. J., & Hobbs, R. (2015). Beluga, Delphinapterus leucas, ethogram: A tool for cook inlet beluga conservation? Marine Fisheries Review, 77(1), 32–40.

Torres, L. G., Nieukirk, S. L., Lemos, L., & Chandler, T. E. (2018). Drone up! Quantifying whale behavior from a new perspective improves observational capacity. Frontiers in Marine Science, 5(SEP).