The province of British Columbia, located on the western seaboard of Canada, is renowned for its diverse and dramatic physical geography along its extensive coastline. Spanning over 27,000 km, the coastline features a complex network of fjords, inlets, and islands, including the prominent Vancouver Island and Haida Gwaii (Taylor et al., 1997). The coastal waters are characterized by a mixture of cold, nutrient-rich currents and deep ocean upwellings that support a diverse and vibrant marine ecosystem including many marine mammals. There are 31 species of marine mammals recorded in British Columbian waters, including 25 cetaceans (whales, dolphins, and porpoises) and 6 species of carnivores (seals, sea lions, and sea otter) (Ford, 2014).

Marine mammal strandings in British Columbia are monitored province-wide by the British Columbia Marine Mammal Response Network (BCMMRN) coordinated by the Department of Fisheries and Oceans Canada (DFO). Necropsies for this study were conducted in cooperation with Indigenous Community Members, Fisheries Officers, Canadian Coast Guard (CCG), research scientists, non-government organizations, and the Provincial Animal Health Centre in Abbotsford, British Columbia.

In May 2022, data collection for strandings that were conducive to a postmortem examination and 3D reconstruction within the territorial waters of British Columbia was initiated. In this case series, 11 species and 13 specimens were examined, including 8 cetacean and 3 carnivore species.

While in-depth descriptions of the background theory for the methods used can be found elsewhere (see Gomes et al., 2014; Kang et al., 2020; Ma and Liu, 2018; Zhou et al., 2024), given the novelty of the application, a short overview is presented here.

Light Detection and Ranging (LiDAR) is a remote sensing technology that employs laser pulses to measure the distance between the instrument and a target surface or object. Distance is determined through a measure of the time it takes for the pulses generated by the instrument to travel to the surface and be reflected to the sensor. This process generates precise, high-resolution data about the physical environment. The resulting output, typically a 3D point cloud, provides a detailed representation of surfaces and objects, making LiDAR a versatile tool for applications ranging from topographic mapping to structural analysis (Wandinger, 2005).

LiDAR systems are available in various types, including terrestrial, RPAS, airborne, and handheld devices, each suited to specific applications. In recent years, LiDAR technology has been miniaturized and integrated into consumer-grade mobile devices, such as the latest iPhones and iPads (Vacca, 2023). Third-party applications make use of these iOS integrated devices to generate 3D point clouds and solid meshes with minimal user training needed. They are also relatively low-cost and more accessible than survey-graded systems. Although these built-in systems lack the range and resolution of professional-grade devices, they are sufficient for many applications, including conservation (Abbas et al., 2024), small-scale mapping (Abdel-Majeed et al., 2024), and preliminary structural analysis (Tondo et al., 2023). Handheld and mobile LiDAR devices are particularly efficient for data collection in diverse environments, from controlled indoor spaces to rugged field settings (Bauwens et al., 2016; Chio and Hou, 2021; Desai et al., 2021). Advanced techniques, such as simultaneous localization and mapping (SLAM) algorithms, also enhance the functionality of these systems, enabling real-time processing of spatial data, even in dynamic or GPS-deprived environments (Al-Tawil et al., 2024).

While professional LiDAR systems can be costly, with tripod-based survey devices typically ranging in excess of US$20,000, iOS mobile devices with integrated LiDAR provide a lower-cost alternative (Yen et al., 2011). For instance, an iPhone equipped with LiDAR can be purchased for under US$1,500 (Apple, Cupertino, California), offering researchers and professionals an affordable way to perform 3D scanning and mapping tasks.

In contrast to LiDAR, structure-from-Motion Multi-View Stereo (SfM-MVS) photogrammetry is a technique used to create 3D models from 2D photographs (Eltner and Sofia, 2020). By capturing a series of overlapping photographs from different angles, SfM software can identify common points within the images and reconstruct their 3D positions. This process involves camera calibration, feature detection, and matching, followed by the generation of a sparse point cloud that represents the object’s or scene’s geometry (Fathi and Brilakis, 2011). Further refinement through dense point cloud reconstruction is achieved through algorithms such as MVS (Strecha et al., 2012).

A 3D point cloud is a collection of discrete points in 3D space, each representing a specific location on the surface of an object or scene. These points collectively describe the object’s geometry but lack information about the surface’s continuity (Liu et al., 2019). In contrast, a 3D mesh provides a more structured representation, consisting of vertices, edges, and faces that form a continuous surface (Rassineux, 1997). Figure 1 shows what the two different formats of point cloud (a) and 3D mesh model (b) represent visually for reference on an inflatable test whale.

Over the study period, 13 marine mammal specimens composed of 11 species were collected or examined for this study. A summary of the stranding information and 3D data collection for each case is provided in Supplementary Table 1 , and the geographic location of each specimen is shown in Supplementary Figure 1 . In addition to the specimens listed in Supplementary Table 1 , LiDAR and SfM data from an inflatable test whale were also collected to validate the process of data collection for beach cast animals.

LiDAR data were collected using an iPhone 12 Pro (Apple, Cupertino, CA) using the free Scaniverse application (Niantic, San Francisco, CA). The LiDAR sensor is available on the Pro and Pro Max models of the iPhone 12 and later, in addition to iPad Pro models from 2020.

The acquisition was achieved by walking around the animal with the iPhone in hand, carefully covering the entire extent of the animal and ensuring to not exceed the adjustable LiDAR range (i.e., 0.3–5 m) in distance from the surface of the animal while scanning. This range was adjusted as needed per animal, with the larger whales generally requiring a farther range in order to enable the data collector to step around the animal and not lose contact with the animal surface. The details of the range used for each scan is contained in Supplementary Figure 1 . For large whales, data collection took up to 10 min, where with the smaller marine mammals, this process was completed in a maximum of 2–3 min.

The Scaniverse app collects both LiDAR and photographs from the phone’s main camera. Processing through the application generates a realistic triangular 3D mesh as the primary output. The scanning systems on iOS device applications use a form of SLAM that continuously tracks the sensor’s position and orientation in three dimensions over time. SLAM relies on optical data overlaps, utilizing previously observed features to establish relative coordinates and maintain accurate image registration (Lehtola et al., 2021). The Scaniverse interface and an example workflow for marine mammal data collection and subsequent export capabilities using the Scaniverse application are depicted in Figure 2 .

For large whale necropsy conditions that allowed for collection of UAV photogrammetry, a Mavic Professional [Da-Jiang Innovations (DJI), Shenzhen, China] UAV was used. Data collection consisted of capturing photographs of the exposed surfaces of stranded whales and the surrounding area. The UAV was flown manually as opposed to the widely used preprogrammed orthogonal flight path transects as no previous understanding of the stranding sites was available to create a detailed flight plan. During the manual flights, the best effort was made to create orthogonal flight lines with high overlap (~80%) in a cross-hatched pattern. See Supplementary Table 2 for flight details. Pix4DMapper (Pix4D, Prilly, Switzerland) was used to implement the SfM-MVS workflow. The resultant 3D mesh output for these large whale cases was used primarily to provide additional context for the large whale strandings.

One important consideration in the creation of realistic models is the determination of scale. The LiDAR system measures distances accurately and therefore evaluates the appropriate scale of the models within Scaniverse without the need for input of a known reference measurement in the scene. In the case of UAV photogrammetry here, the geotagging of each image used to create the model provides this distance measurement within a scene that is required to produce accurate scale. For non-georeferenced images, a known reference measurement is required to add these scale constraints to the model.

The acquired marine mammal datasets were generally categorized into two main groups, each corresponding to distinct use cases. The first is the application of 3D data collection in the instance of remote or inaccessible necropsies that benefit from the ease of use of these sensors and software. This is the case where trained regional partners are most likely to use this technology to collect 3D reconstruction information in the field where the animal is too large (for large whales) or too remote to be transported for a full postmortem examination or for pathologists to travel to the area. In this scenario, the mesh output of the Scaniverse workflow from a handheld device ( Figure 2 ) provided an acceptable reconstruction without postprocessing. When possible, the use of UAV photogrammetry for large whales was implemented, to provide more environmental context and auxiliary data to the handheld LiDAR collection alone. This procedure could only be done if appropriate personnel and equipment were available. The second case is more exploratory in using this technology in the context of animals that were stranded or were transported to resourced areas where a full necropsy can be conducted, with all organ systems sampled and excised with a 3D dataset collected. Both data collection protocols followed a similar workflow, which is outlined in depth in Supplementary Figure 2 .

Following processing and inspection of the resulting model, the meshes were exported to both.obj (geometry definition file format) and.las (RGB 3-dimensional point cloud data) files from the Scaniverse application to further refine in post-processing.

Post-processing the Scaniverse output was done by the free open-source 3D point cloud and mesh processing software CloudCompare Version 2.13. The two main tools used were to crop and segment data as well as merge multiple scans of the same specimen. For the cases where UAV photogrammetry data were collected, Pix4DMapper Version 4.8 was used to process the UAV data into meshes that could then be combined with the Scaniverse mesh outputs in CloudCompare.

The key morphometric quantities of interest for cetaceans during postmortem exams and photogrammetry (and in general) are straightline length (from rostral limit of upper jaw), appendage lengths, girth, and blubber thickness. There are 12 straightline length measurements, 4 appendage measurements, 4 girth measurements, 2 throat pleat measurements (if applicable), and 9 blubber measurements typically recorded for complete morphometric characterization during postmortem exams ( Supplementary Figure 3 ). For carnivores, morphometrics typically include straightline and curvilinear length (tip of snout to tip of tail), two appendage measurements, three girth measurements, and three blubber measurements (if applicable) (Rowles et al., 2001; Burns and Gillespie, 2003; IJsseldijk et al., 2019).

In this case series, digital morphometric measurements (Euclidean distance) were made in Scaniverse using a straightline length or using a polyline measurement for curvilinear measurements of length or girth. For all carcasses, the multiple digital measurements are compared against the manually collected morphometric data as a proof of concept. One important consideration in this work is that there is no way to verify the accuracy of the manual measurements; experienced personnel were primarily tasked with these measurements, but there is an unquantifiable source of error intrinsic to manual measurements that must be mentioned. To minimize the impact of this, only qualified personnel with appropriate necropsy experience were taking measurements in these cases, with the veterinary pathologist supervising if unable to take the measurement themselves.

The generated 3D reconstructions of the specimens and auxiliary information were stored in an accessible web platform for visualization and further investigation. The exported. obj meshes were uploaded to a server using the open-source 3D Heritage Online Presenter (3DHOP) framework (Visual Computing Laboratory - ISTI-CNR, Pisa, Italy) for interactive web presentations of high-resolution 3D models. These models can be evaluated through this medium for their quality, reproducibility, and reusability.

Uploading and incorporating the models into VR was implemented using the ENGAGE VR (Immersive VR Education, Ireland) platform. This platform allows for the uploading of.glb files, which, unlike.obj, provide a packaged single file containing all textures and shaders of a model. Within the platform, a presentation was created to view the output of these scans in lifelike sizes, providing an opportunity to view what happens during a necropsy through VR for multiple species types.

The 5.00-m measuring tape next to the inflatable whale was used as a reference and measured over 15 scan iterations within the Scaniverse app to be 4.99 ± 0.03 m. This demonstrates that barring any obvious distortions, errors in scanning, or errors in using the measurement tool within Scaniverse, measurements achieved through reconstructions have approximately a 1% uncertainty. These three error sources must be closely monitored to obtain accurate measurements from a 3D reconstruction. An example of the reference measurement as well as three example morphometric measurements of the inflatable whale including dorsal fin height, flipper anterior length, and half tail fluke width are shown in Figure 3 . The proportions and overall shape of the inflatable whale are not entirely comparable to a real animal; however, the reconstruction can clearly distinguish the structure of the fins, as well as the body shape.

During testing, it was found that for most cases, using “detail mode” (photogrammetry) for processing within Scaniverse occasionally resulted in large artifacts (e.g., two tails) especially for the large whales, while in “area mode”, the structural data provided by LiDAR allow for the best representation of true form in all cases. For smaller animals and excised organ systems, detail mode was able to better represent the surface texture, which is important for diagnostics.

The fin whale that presented dead in Prince Rupert had ample subcutaneous and visceral adipose stores and the animal is well muscled as seen in Figure 4 . Pre-necropsy morphometrics via 3D data showed a measured straightline length of 19.45 m, which differed by 3.14% from the straightline length determined from the manual measurement (20.07 m) seen in Figure 4A . The tail fluke width differed by 2.25% ( Figure 4C ), while the anterior flipper length differed by 0.43% ( Figure 4D ). The dorsal fin height differed by 2.60% ( Figure 4E ). Externally, the right dorsolateral aspect of the head and neck shows localized swelling and asymmetry, consistent with acute subcutaneous hemorrhage and edema caused by the suspected vessel strike. Multifocal skin erosions and occasional ulcerations appear as irregular surface textures and depressions on the body. The reduced blubber thickness in the ventral third, indicative of moderate atrophy, is evident, particularly in cross-sectional views. Post-necropsy in Figure 4B , the internal trauma to the head and neck is clearly visible, with areas of hemorrhage and edema in the underlying musculature appearing as disrupted and discolored tissue. Additionally, muscle degeneration and necrosis in the skeletal musculature are apparent as irregular and separated muscle fibers with fluid accumulation. Together, the 3D datasets capture the extent of trauma and pathological findings consistent with a suspected vessel strike. The gross necropsy findings of edema (and hemorrhage) are borne microscopically and also consistent with cause of death due to a traumatic incident.

The gray whale that was reported dead after a 4-month entrapment in a lagoon was severely emaciated but presented with some postmortem bloating. Pre-necropsy morphometrics via 3D data in addition to a post-necropsy representation of the animal that displays several internal organs are shown in Figure 5 . With full exposure of the animal at necropsy, an accurate straightline length can be determined directly from Scaniverse. The measured straightline length (12.35 m) was <1% different from the value measured manually in the field (12.30 m). The axillary girth, doubled at half, was different by a factor of 1.4%. This can partially be explained by the discrepancy when using a polyline measurement in Scaniverse as opposed to a tape measure that can adequately lie flush to the animal. In addition, postmortem bloat, liquefaction of the blubber, posture of the carcass on the beach, and other factors may have contributed to this discrepancy. In this case, polyline measurements at less than 10 cm between vertices were used. The tail fluke measurement had a discrepancy of 0.4 m or approximately 10% if using a straightline fluke length, reduced to 2% using a six-part polyline measurement. This is an important consideration in this case and other strandings where the animal’s fluke was in a curved shape, meaning a straightline length as shown in Figure 5B will underestimate this value if the morphometric measurement was done with a tape measure flush to the fluke length as was done here. A 13% discrepancy can be seen with the anterior flipper measurement as a small portion of the pectoral was submerged, another consideration when compiling morphometrics from 3D data. The animal had marked reductions in subcutaneous and visceral adipose stores and was poorly muscled. There was salmon discoloration of the blubber. This emaciation manifests in a “peanut head” appearance, or a marked bilaterally symmetric narrowing of the caudal aspect of the skull, which is readily apparent in the 3D representation of the animal ( Figures 5D, E ). The salmon discoloration of the blubber has been observed in other gray whales, and it is difficult to infer whether this may be related to dietary changes (abundant carotenoids), endogenous pigment production, or some other process. The animal also had generalized sloughing of the skin, attributed to postmortem change.

The gray whale stranded on Nootka island was bloated, and upon incision of the abdominal musculature, there was marked deflation of the abdominal cavity. Figure 6 displays the UAV 3D model output obtained during this necropsy and 3D dataset collection conducted before and after necropsy. The animal’s total length was estimated at 9.9 m from nadir UAV photographs; the fluke of the animal was still submerged at low tide. The analysis included a correcting factor given the estimated 20° decline of the tail in the water. An anterior flipper measurement from the pre-necropsy LiDAR in Scaniverse was collected to be 1.44 m, within 2% of the manual measurement of 1.47 m. In the post-necropsy scan, it is apparent that, upon incision, a moderate amount of brown–red fluid was drained from the animal as a result of advanced autolysis. The small intestines and colon were distended with gas, which was well characterized by the 3D reconstruction. The post-necropsy dataset of this animal did not characterize the head adequately. This is due to the LiDAR collection losing contact (out of range) halfway through the scan. The processing software cannot register where the data are collected from and must re-register from a new reference point. This resulted in this scan essentially having two lower halves of the animal separated by approximately 15°, which was rectified in postprocessing via segmentation of the scan at the point of deviation and subsequent re-registration through the use of equivalent reference points that are present in both segments (Xu et al., 2023). Despite the visual success shown in Figure 6D , this scan is not suitable for use in morphometric analysis, only a visual assessment.

The stranded humpback whale in Haida Gwaii was in dorsal recumbency and the UAV photogrammetry and 3D dataset are presented in Figure 7 . The UAV photogrammetry shown before and after necropsy especially outlines the steep tidal gorge the animal was in that precluded much of the normal necropsy procedure for safety concerns. LiDAR data were collected after necropsy and showed that on incision of the musculature, multiple tense loops of gas that inflated the small intestine were extruded. The flukes and distal third of the peduncle were submerged. The oral cavity was also only partially exposed from the water. Morphometric measurements were collected solely through assessing the 3D reconstruction and UAV images due to the difficult environment, although some measurements such as that of the anterior flippers could be verified via a measuring tape. With some areas of the animal inaccessible to the scanner being underwater, the straightline length could not be calculated directly in Scaniverse. The animal’s length was estimated at 12.2 m from these scans and associated photographs in postprocessing. Other morphometrics were collected from Scaniverse, including an example anterior fluke measurement at 2.47 m, consistent with the manual measurement (2.50 m) to within 2%. The 3D reconstruction adequately displays the animal’s moderate body condition, prominent mammae, and open abdominal cavity, including a moderate amount of ingesta primarily consisting of herring bones.

Stranded killer whales are rare in British Columbia relative to the large whales (Barbieri et al., 2013) and each event provides a unique opportunity to showcase 3D reconstructions of this species. For the adult pregnant female transient that was stranded near Zeballos, this was especially true. This animal initially was live stranded but subsequently died despite efforts to refloat her, and a postmortem examination was undertaken within 36 h. Figure 8 displays the pre-necropsy scan of the near-term pregnant killer whale. Extremely wet and humid conditions caused specular anomalies of the scanner operator and other surrounding areas to be present on the surface of the whale in the 3D reconstruction. This did not impact the LiDAR information obtained from the scan (and thus ability to collect morphometrics). The straightline length is measured at 6.02 m with a 6.00-m tape measurement in the field. The straightline length of the fetus was measured to within 1% of the field morphometric measurement, with the 1.50-m measurement in Figure 8E comparable to the 1.53 m measured in the field.

The Risso’s dolphin specimen was found with extensive scavenging and poor postmortem condition, with extensive skin sloughing. The animal was 2.21 m in Scaniverse compared to 2.23 m during the necropsy with a difference of less than 1%. The pre-necropsy scan of the animal is shown in Figure 9 . The suboptimal nutrition of this animal coupled with the lack of ingesta and blubber atrophy found during the necropsy is consistent with a negative energy balance that would have been severe enough to cause the loss of this animal. There were no other apparent internal or external lesions.

The live stranded striped dolphin is shown through its 3D reconstruction in Figure 10 . The animal had superficial abrasions on the leading edge of the dorsal fin and edges of the tail flukes, possibly as a result of the live stranding event. The straightline length of the animal was measured by Scaniverse at 1.91 m, with a field measurement of 1.95 m, an approximately 2% discrepancy. This animal and the remaining carcasses included in this case series had complete necropsies and tissues were scanned in situ and excised as was logistically feasible. The following strandings with full necropsies were less time- and resource-constrained and allowed for more focused systematic visualization of the stranded animal and associated internal anatomy. This procedure facilitated review of the stranded animal as a 3D dataset to compare against different specimens or measure/visually assess areas, including that of organ systems. The sequential removal of all tissues and organs during the full necropsy is displayed for select organs. The topographic features and unique structure of the organs and stomach, pluck including goosebeak and thoracic viscera, as well as the kidney were well depicted by the scan and are displayed in Figure 10 . There were no significant lesions in the adipose tissue, peripheral vasculature, penis, urinary bladder, trachea, kidney, thyroid gland, adrenal gland, diaphragm, or liver.

The stranded Dall’s porpoise is shown via its reconstruction in Figure 11 . The animal was measured to be 1.97 m straightline length by Scaniverse, a difference of 3.5% from the manual measurement of 2.04 m. The animal had reduced subcutaneous and visceral adipose stores and was fairly muscled. There were no apparent internal or external lesions that would have contributed to the loss of this animal. All organs were sampled with select scans of organ systems shown in Figure 11 .

Two harbor porpoise reconstructions are displayed in Figure 12 . The adult was moderately fleshed with possible otitis interna. Incised cutaneous wounds are visible on this animal and suggestive of a possible vessel strike interaction. These wounds are visible on the scan and can be subsequently analyzed by various characteristics of the injury (Byard et al., 2012). The animal was measured with a straightline length of 1.46 m, within 3% of the manual measurement of 1.50 m. All organs were sampled with select organ systems shown in Figures 12a–g . The harbor porpoise calf was in good postmortem and body condition with no apparent internal or external lesions. The animal was measured at a straightline length of 0.81 m, within 2% of the manual measurement of 0.83 m. Selected scans of the calf’s overall body condition and organ systems are shown in Figures 12A–E .

All carnivore full-body reconstructions are displayed in Figure 13 . The use of “detail” mode in Scaniverse improved the resolution of the fur texture in some scans. The sea otter was autolyzed, with a pendulous fluid filled abdomen. The abdomen was fluctuant with a prominent fluid line on ballottement. The reconstruction confirmed moderate abdominal distention with fluid accumulation in Figure 13a . The straightline length of the animal was 1.09 m using the reconstruction, a difference of 4% from the manual measurement during the examination of 1.14 m. The subcutis is a dull pale green brown and tacky, and the liver is moderately enlarged. Oysters in the stomach contents can be viewed in a scan of the digestive system, along with scans of various organs presented in the web application ( Supplementary Table 3 ). Unfortunately, postmortem change hampered a gross examination of the animal although the suspected cause of death is trauma to the head. The northern fur seal specimen was also in a state of advanced autolysis, although was fairly fleshed ( Figure 13A ). There is mild swelling of the head and vulva, and the lungs are mottled pale to dark red. The animal was measured at 0.77 m, approximately a 3% difference from that of the manual measurement at 0.80 m. The reconstructed harbor seal pup is presented in Figure 13α ; the reconstruction measurement was 0.78 m with a difference of approximately 1% from the manual morphometric measurement of 0.79 m.

A summary of each specimen and associated measurement types, manual and model morphometrics, and the percentage difference between the two is compiled in Table 1 .

All the scans presented here in addition to those of other organ systems or stages of the necropsy are in Supplementary Table 3 , where the scans can be visualized, explored, and zoomed into. These manipulations are a major advantage of 3D data versus a series of two-dimensional photographs. Future strandings where LiDAR or photogrammetry data are collected will continue to be added to this dataset, which is intended for public access.

The VR implementation of these scans was achieved through a short presentation in ENGAGE VR, which highlights the importance of marine mammal necropsies and the possibility of VR in improving the science and education of necropsies. The users within the VR environment can walk around and experience several perspectives of necropsies at close to life size while a video of the stranding response and pathologists dissecting the whale is displayed above. The application provides a unique sense of the scale of necropsies to those that have never witnessed them in this context. An example screenshot of navigating the VR environment is shown in Figure 14 .