CALIBRATING UAS PHOTOGRAMMETRY TO DERIVE DELPHINID POPULATION DEMOGRAPHIC PARAMETERS:
PROJECT PHASES 1&2
This project aims to develop a method to rapidly assess delphinid population age structure. This will ultimately provide an assessment of population growth and survival rates, and early insight into whether there may be cause for concern for specific populations.
The power to detect population increases or declines using traditional photo-identification methods typically requires decades of data collection before changes in vital rates (e.g. survival, fertility etc.) manifest themselves as changes in population size. The age structure of a population represents an informative parameter to evaluate the health and sustainability a population. New technologies are available that enable assessment of group composition and population age structure, as well as measures of individual animal health that can ultimately provide assessment of population growth and survival rates, and early insight into whether there may be cause for concern for these populations. This project consists of a four-phase study to assess the feasibility of using small Unoccupied Aerial Systems (UASs, or drones) to determine the population age structure of mid-sized delphinids. For this project, we will focus on the Hawaiian population of spinner dolphins, off the Kona coastline (Hawaii).
Project phase 1 and 2
The long-term objective of this project is to define the age structure of the spinner dolphin population off the Kona coastline (Hawaii). Understanding the age structure of this population will provide a means by which to assess its trajectory and health status One of the first step of this project (link to previous page) is to be able to estimate the length of free-ranging dolphins. However, dolphins rarely spend much time at the surface, and almost never display a full straight body. Therefore, the first objectives of this project were to find a mean by which to estimate the full length of the dolphins. Here we present the two first phases of the four-phase project (see figure 1 below).
Figure 1. Schematic overview of the first two components of the study to assess the feasibility of using multi- rotor Unoccupied Aerial System (UAS) -photogrammetry to determine free-ranging delphinid group age composition and population age-structure.
The aims of the first phase of this project are to (Figure 1, phase 1):
Collect hands-on physical measurements of five adult male bottlenose dolphins (Tursiops truncatus) at a public facility (Dolphin Quest Oahu, Hawaii).
Hands-on measurements consisted of collecting total length (from the tip of the rostrum of the dolphin to the notch in its flukes) and the length of the distance between the center of the blowhole to the insertion of their dorsal fin. Recent studies have shown that the distance between the blowhole and the insertion of the dorsal fin represents a good proxy for total length in bottlenose dolphins. Hence, obtaining the length of this proxy would allow for the estimation of total length. These measurements were collected using a regular tape measure.
Collect aerial images of the same five dolphins, performing two behaviors, at five altitudes (16-50m)
Aerial images were collected using Unoccupied Aerial Systems (UASs, or drone) during two behavior exercises. First the animals were asked to remain stationary and fully extended at the surface of the water (this allows for the comparison with the hands-on measurements collected previously, see figure 1). Then the animals were asked to swim freely to allow for the sampling of natural surfacing events (corresponding to the time when the animals fully expose both their blowhole and dorsal fin at the same time, see figure 1). Surfacing events allow for the estimation of the total length of these animals using the distance between the blowhole and dorsal fin insertion as a proxy for total length.
Collect additional measurements for individual dolphin body health assessment (more information at the end of this page)*
Compare hands-on measurements with UAS-derived measurements to test the accuracy and the precision of UAS-photogrammetry in measuring dolphins.
Once both the hands-on and UAS-derived measurements were collected, a direct comparison between the two was realized. This comparison tests the precision and accuracy performances of UAS-photogrammetry in estimating real lengths.
First, the UAS bias in estimating stationary total length was calculated. Then, the UAS bias in estimating total length using the blowhole to dorsal fin insertion as proxy for total length was calculated. The latest shows that UAS-photogrammetry is precise and accurate enough to estimate the total length, based on a proxy, of free-ranging dolphins.
The aims of the second phase of this project are to (Figure 1, phase 2):
Test the sensitivity of UAS-photogrammetry in converting length estimates (meters) to the age of an individual.
To establish the age-composition of a group of dolphins, and ultimately assess the age-structure of its population, there is a need to test whether UAS-derived total length estimates based on a proxy allow for the accurate age-classification of an individual. In other words, what is the probability of correctly aging an individual based on its length?
To answer this question, we are collaborating with the Sarasota Dolphin Research Program who has been collecting hands-on measurement of free-ranging bottlenose dolphins in Sarasota for the past four decades. This datasets contains three important information for 90+% of the Sarasota population: the total length, the blowhole-dorsal length and the age of each individual.
First, biased total lengths will be estimated for the Sarasota dolphins by applying the UAS-bias estimate that was calculated previously on the physical total length measurements. As individuals of a same age can vary in length, the population will be divided into different age-class bins (e.g., individuals of age 0 to 2 belong to the same age-class). Different age-class bin scenarios will be defined, and each dolphin will be classified into a specific age-class bin based on the length estimates of those age-class bins. The same exercise will be performed for the true lengths physically measured. Comparing the two results will allow for the testing of the sensitivity of the UAS-bias in correctly aging all individuals sampled within a population (figure 2). The precision of UAS-photogrammetry in aging free-ranging dolphins will give us confidence when it comes to assess the age structure of the spinner dolphin population of Kona, Hawaii (See phase 3 and 4 of the project).Link
Figure 2: Schematic representation of the age structuration of bottlenose dolphins (Tursiops truncatus) using Unnocupied Aerial System (UAS) -photogrammetry and fourty years of hands-on morphometric datasets.
*Additional measurements collected at Dolphin Quest Oahu, Hawaii, during phase 1 of the project:
In addition to the total length and blowhole-dorsal lengths, the following measurements were collected for all animals: the girth (at specific body positions to allow for body volume estimates), the distance between these girth axes, the weight, and a 3-D model (one dolphin only).
From the UAS images collected previously, a cylindrical body volume for each animal is estimated using the methodoogy developed by Christiansen et al. 2018. This method relies on the measurement of the animal's body width (figure 3, yellow lines) at every 5% of the total body length (figure 3, pink line).
Figure 3: Aerial image of a bottlenose dolphin (Tursiops truncatus, Dolphin Quest Oahu, Hawaii). Unnocupied Aerial System (UAS) -photogrammetry was used to estimate a cylindrical body volume of the animal based on width measurements (yellow lines) and total length (pink line).
To further understand the link between UAS-derived and hands-on volume estimates, we realized a 3-D model of a live dolphin (Figure 4) using a 3-D scanner; a courtesy of Dr. Joshua Madin. From the 3-D model of the dolphin a 3-D volume will be estimated. Additionally, the model can be partitioned to estimate the volume a specific section (figure 4 bottom row) based on the position of the girth axes used above. Finally, the 3-D volume is expected to be the closest estimate to the true volume of the animal and will therefore allow for the direct comparison with the cylindrical volume estimated with the UAS-photogrammetry and hands-on measurements. With this, we will be able to refine the UAS-photogrammetry used to estimate the volume of the animals.
Figure 4: 3-D model of a live bottlenose dolphin (Tursiops truncatus) housed at Dolphin Quest Oahu, Hawaii.
Lars Bejder - MMRP
Fabien Vivier - MMRP
Cormac Booth - SMRU Consulting
Erin Oleson - NOAA/PIFSC
Amanda Bradford - NOAA/PIFSC
Marie Hill - NOAA/JIMAR
Kristi West - Stranding network/MMRP
Jason Baker - NOAA
Aude Pacini - MMRP
Julie Rocho-Levine - DQ
Randall Wells - Sarasota Dolphin Research Program