Post provided by Nathan Hirtle (he/him)

A whale’s size not only provides information on its own health, but also the status of the marine ecosystem in which it lives. So how do scientists weigh the biggest creatures on planet? In this blog post, Nathan Hirtle shares insight from his recent paper on measuring the volume of whales using drone images.

Science of animal shape

Studying the morphology of an animal – that is, the shape and size of its body – is a simple yet effective way to understand its health, age, and role in an ecosystem. For instance, a heavier individual is likely to have been more successful at finding food than a lighter individual. Likewise, a larger individual (measured from head to tail) might be older than a smaller individual. By collecting size and shape data from different individuals we can gain insight on wider population trends through a process called morphometric (‘morpho’ meaning ‘shape/form’ and ‘metric’ meaning ‘measurement’) analysis.

While it can be relatively easy to collect and measure many terrestrial species, sampling species in aquatic ecosystems can be more difficult, especially without the use of nets. This method won’t work for whales, in part because of the potential for harming them but also just because of how large they are!

Aerial image of a whale taken under permit number (NMFS GA No. 21889). Credit: Julia Stepanuk

Weight of whaling

Historically, industrial whaling ships had infrastructure that allowed harvested whales to be weighed and measured. These measurements are still used today to make comparisons with modern whale populations. Given that the widespread harvesting of whales caused rapid global population declines, current whaling practice has almost entirely ceased. This was certainly necessary to save these species from extinction, though it also made measuring whales incredibly difficult.

Whales are important consumers in marine ecosystems- they eat a lot of food, store a lot of energy in their bodies, live a long time, and travel great distances. These life history traits mean that stored in the body of every whale is a lot of information about the marine ecosystem, which has led to some to refer to large whales as ‘ecosystem sentinels’, because by studying whales, we can learn a lot about their environment.

So how do we measure an animal that we can’t catch? We get creative.

Modelling whales

Recently, there has been a lot of work in the field focused on measuring animals remotely, many of which rely on the use of unoccupied aerial vehicles, or drones. This process is very useful for whale biologists because it allows us to measure whales without even touching them.

We wanted to know if we could use these drone-based measurements in combination with three dimensional (3D) models to create a virtual representation of a humpback whale (think about it like a video game character). This virtual, 3D model would then allow us to estimate the volume of the real whale. Volume is a measure of how much space something occupies, which can tell us a lot about how animals move and how much they have been eating.

(a) top and side images aligned along the body axis of a humpback whale, scaled so that the total length of the whale in both images is the same. The three-dimensional mesh with ellipses at length intervals is overlaid onto the side (b) and top (c) images. Ellipses extend from 1 – 85% total length. Credit: Hirtle et al., 2022

Volume is usually measured in animals by putting individuals in a container with water in it and calculating how much water the animal displaces, which is of course impossible to do for free-living whales! Estimating volume from 3D models has previously been achieved for stranded pilot whales, but not for living animals.

The first step in this process was creating a 3D model. We used a free software called Blender to build our model. First, we identified some really nice images of humpback whales from the top and from the side. We then traced the outline of the top profile and the side profile of the humpback whale using a 3D mesh that starts off as nothing more than a cylinder. By dragging the different points of the 3D mesh around to match the images, we were able to “sculpt” the cylinder into a shape that looks like a whale. Thankfully, the body shape of a humpback whale is pretty simple – because I am no artist! We didn’t include the fins or tails of the whales in the model, because of how complex they are.

At this point, the size of the model does not matter. All that matters is that it LOOKS like a whale. Next, the drone measurements were used to make this model the correct size. This is done using a Python script in Blender. Basically, the total length and width measurements taken from the drone are used to squish and pull the model into the correct size.

(a) Width measurements (white arrows) and total length measurements taken from drone images. Numbers between width measurements indicate percent total length from head to tail notch. (b) The base 3D model. (c) The base model (blue) is scaled to length (d) The model is scaled to match corresponding width measurements and height-width ratios, resulting in the final model (purple). Black arrows indicate regions of greatest change relative to the base model. Credit: Hirtle et al., 2022

On the scale of a whale

But you might be questioning the feasibility of this; after all, that is a lot of measurements you need, and it can’t be easy to get all of those from one whale, especially when you think about things like waves, the whale exhaling water, the whale bending and moving…. the list goes on. And you would be right!

One of the main objectives of this work was to determine how many and which of these measurements were necessary to accurately represent the whale. The way we did this was to create models that used all of the different combinations of width measurements available to us. With 17 different width measurements, that’s 131,072 different combinations…. for each whale.

Although code can do this scaling process much faster than a human, it’s still a lot of individual calculations being done for each whale. So, this was a really challenging exercise in improving computational efficiency while still getting the results we wanted. In this case, that meant that I had to rework the code from scratch- but it paid off. We were pretty impressed with our results as we were able to create accurate 3D models of humpback whales using as few as 5 width measurements, as compared to the full 17 used in the ‘full’ model- although some of the models were certainly not the prettiest (see the video below).

Research application

This will be really helpful moving forward, because it means that we can use images of whales that are less than perfect to create accurate 3D models and estimates of volume from those models. So, if a few parts of the whale are obscured by, say, the whale’s blow, it’s not a problem. It’s also important that these calculations can be done pretty quickly, which could be important in helping entanglement rescuers for calculating sedation dosage. Also, we can use these models to measure the same whale repeatedly to see how the whale may be changing from year to year, which can then tell us how the ecosystem is doing more broadly- all without touching the whale!

We think this method is really cool, and part of this process for me was learning the importance of code documentation, because we want other people to be able to use these methods for their own study species. All of the code and guidelines for making 3D models was part of the product for this work, so if you are interested, please check out the paper and see the code archived on GitHub.

About the lead author

Nathan Hirtle is a PhD candidate at the School of Marine and Atmospheric Sciences at Stony Brook University in New York working with Lesley Thorne. Nathan is interested in understanding baleen whale energetics and distributional shifts in odontocetes in the Western Atlantic associated with climate change.

Read the full article “Integrating 3D models with morphometric measurements to improve volumetric estimates in marine mammals