1 Introduction

Accurate collection of biometric data is important for meticulously monitoring marine species within aquarium environments, providing vital insights into the physiological and behavioral patterns of aquatic organisms, and facilitating a comprehensive understanding of their health, growth, and overall well-being. Through the systematic acquisition of data such as length, weight, and morphological features, researchers can discern subtle variations indicative of individual development, nutritional status, and potential health issues. Additionally, these data may help in understanding the biology and ecology of organisms in their natural ecosystems (Simpfendorfer and Dulvy 2017; Ripple et al. 2021; Lewis et al. 2023). Traditional methods to obtain body length data often involve capturing, handling, or stressing specimens, which is unsuitable for studying populations and can disrupt the ecological balance (e.g., Raoult et al. 2019). Alternative less intrusive methods such as visual census have also been used for decades (e.g., Brock 1982) though they have reported biases and limitations (Harvey et al. 2004; May et al. 2019). As a result, researchers have been striving to develop noninvasive techniques to accurately measure the length of marine megafauna, including large sharks and rays for which nonintrusive length estimates are crucial. Most of these species are currently facing severe global declines due to overexploitation and habitat loss, making accurate data imperative for conservation efforts (Dulvy et al. 2014; Edwards et al. 2019; Boggio-Pasqua et al. 2022). Furthermore, these nondisruptive methodologies contribute to optimizing resources by allowing the measurement of numerous animals with minimal staff involvement. This reduces costs and minimizes risks. Additionally, the images collected can be analyzed at any time, providing flexibility, and enhancing data accuracy.

Stereo-video photogrammetry has been used for underwater measurements since the 1980s (see the review by Shortis, Harvey, and Seager 2007) for many purposes such as estimating biodiversity, abundance, fish behavior, reactions to anthropogenic pressures (e.g., Griffin et al. 2016) and along with other techniques like paired lasers, has been utilized to obtain in situ length estimates for marine megafauna (Klimley and Brown 1983; Rohner et al. 2015; Perry et al. 2018). Although it has a long history of use in aquaculture environments (Harvey et al. 2003; Silva, Aires, and Rodrigueset 2023; Naiberg et al. 1993), this technique is scarcely reported for use in aquaria, being mentioned by Harvey et al. (2003) in a study where 17 reef fish from six species were measured in an aquarium and posteriorly in situ, by stereo-video and divers allowing the comparison of these two methods. Later, Letessier et al. (2013) reported this technique in a study where Antarctic krill was measured, and more recently, Ikeya et al. (2022) published a study reporting the measurement of six individuals of catfish using stereo-video and the direct linear transformation method.

Stereo-video has proven to perform better than other noninvasive methodologies, particularly for larger swimming animals, by overcoming the limitations and biases of other methods (Harvey et al. 2004; Sequeira et al. 2016). This method can provide measurements of sharks with minimal error of ≈1% (Harvey et al. 2010) when both ends of a target can be clearly seen in high-definition stereo-video imagery. This error might increase but still be considered acceptable (< 3%) when using a low-cost approach to compare stereo-video estimates with tape-measured sharks (e.g., Delacy et al. 2017). Stereo-video utilizes two cameras positioned in the same direction with a slight convergence inward, providing precise measurements of subjects up to 8 m away. Custom software enables the cameras to be calibrated to allow the decentering and radial distortions within a lens to be calculated as well as the focal lengths of both cameras, the distance between the cameras, and the orientation of the cameras in their housings relative to one another (Harvey and Shortis 1995; Shortis 2015). To warrant accurate and repeatable length estimates, measurements that exceed a certain error limit are recommended to be rejected. In the wild, the root mean square (RMS) error is indicated to be up to 10 mm (Harvey et al. 2010) and should be constrained to a maximum of 20 mm. The recommendable distance for length measurements extends up to 8 m, or the minimum visibility necessary for standardization. In instances where elevated RMS values are observed (> 10 mm), it is imperative to verify the accurate synchronization of the left and right images, as well as to confirm the utilization of the correct camera calibration files (Harvey et al. 2010; Goetze et al. 2019).

In this study, we used stereo-video photogrammetry as a low-cost and noninvasive procedure to collect biometric data as a husbandry routine in a public aquaria multi-species exhibit. The significance of our study lies in the use of an in-house designed system, and the scarcity of research utilizing this technique in aquariums. Despite the recurrent use of biometric data in these institutions as part of husbandry or as a contribution to conservation initiatives, no literature has explored to our knowledge the application of this methodology for measuring multiple species in a large aquarium setting. Our innovative approach seeks to highlight the diver-operated stereo-video system as a valuable tool for aquariums to invest in. We demonstrate its applicability to both individual and group assessments of elasmobranch and teleost fish, while reducing costs and risks, and improving efficiency. We further comment on the amount and accuracy of the recorded data when using a custom in-house device or sophisticated equipment.

2 Methods

To analyse the effectiveness and accuracy of diver-operated stereo-video equipment for measuring rays, sharks, and teleost fish, a pilot study was conducted at Oceanário de Lisboa (Oceanário). The base of the apparatus was constructed from stainless steel, with a length of 700 mm and a width of 280 mm. Two cameras GoPro Hero 4 Silver were placed 560 mm apart and converged inwards at 4°, using adhesive mounts and GoPro protective housings. A stainless-steel stopper was incorporated to ensure the cameras' housings' immobility (Figure 1). Before and after each recording session, the diver confirmed that the protective housings were pressed against the mentioned stopper, ensuring that they remained in the same position. The set of floaters underneath each camera provided neutral buoyancy to the equipment. The central U-shaped part allowed the diver to grip the apparatus steady during the dive (Figure 1). GoPro Hero 4 cameras were used as they were available and mentioned with good results for similar studies in the wild, using the software from SeaGis (e.g., May et al. 2019) and other programs currently available and at a reduced cost (e.g., Lewis, Dawson, and Rayment 2023) and their settings were defined according to software Event Measure manufacturer recommendations. After constructing the apparatus, camera calibration was performed in a freshwater pool and involved specific hardware (a cube and a distance bar) and software (CAL program, https://www.seagis.com.au/bundle.html), and it was performed following the original procedures described by Harvey and Shortis (1995) and later by Boutros, Shortis, and Harvey (2015). Recordings from the calibration session were validated in the software, to proceed to the recording session in the aquarium.

Recordings were conducted during day hours in the Oceanário's main aquarium, targeting all present elasmobranchs and emblematic teleost species in this 5 billion litres and 7 m deep exhibit, where optimal light and visibility conditions are kept stable. A code was established to correctly identify the individuals, mostly from elasmobranch species: the number of the individual was hand-signaled to the cameras when the animal approached the observer. The fingers in the upright position signaled for males and in the horizontal position, for females. A clapping hand movement was used at the beginning of the session to allow the cameras' synchronization when later processing the images. The recordings were posteriorly analyzed with the SeaGis software Event Measure (https://www.seagis.com.au/event.html). Whenever possible, measurements were attempted for the fork and total lengths of sharks and teleost fish, while for rays, the total length, disc length, and disc width were measured. One or two measurements were performed, except for teleost fish monitored as a group, due to time and access constraints in image processing. RMS and precision values were retrieved from the software to evaluate data accuracy. Measurements were performed by the same observer, who was also the one collecting the images in the aquarium.

The total list of analyzed species and their measurements is available upon request. Results presented here were filtered for targeted groups for which real measurements were available, including rays, sharks (monitored as individuals), and teleost fish species which were recorded both at the individual level or as a group. Real measurements are typically collected during routine veterinary exams to minimize handling and subsequent stress for the individuals. The animal is captured using rubber nets and placed inside a stretcher by two to three divers, then transported to a submerged platform, where veterinarians observe them. During this procedure, elasmobranch specimens are usually kept ventilated using an Eheim Universal submersible pump (https://eheim.com/en_GB/aquatics/technology/pumps/universal/universal-2400), which has a regulation valve to adjust the water flow according to each specimen's respiratory rate. Biometric data collection is performed using a tape measure, with the animal restrained in the straightest position possible. A single measurement was recorded as the true length for that date to keep the animals' restrained time to a minimum. Afterward, the animal was released and observed until it resumed normal behavior. Real measurements used for assessment were collected within a year of the stereo-video session date to allow comparison (Table 1).

Table 1. Selected measurement examples with the stereo-DOV and comparison with real measurements (in bold), for rays (Mobula hypostoma and Glaucostegus cemiculus), shark (Stegostoma tigrinum), teleost fish (Mola mola) monitored as individuals and for teleost fish (Scomber colias) monitored as a group. When multiple measurements were possible the average ± SD of the measurement and errors is presented.

Measurement screenshot example	Species/IUCN Cat.	Type	Measurements (mm) (n)	Date (dd/mm/yy)	RMS (mm)	Precision (mm)	Error (%)	Error (mm)
	Mobula hypostoma/EN	DW	1109.2 (n = 1)	21/10/21	40.895	37.663	2.701	−30.8
			1140 (n = 1)	07/09/21
		DL	581.8 (n = 1)	21/10/21	54.657	6.385	1.389	−8.2
			590 (n = 1)	07/09/21
	Glaucostegus cemiculus/CR	TL	1834.5 ± 84.8 (n = 2)	21/10/21	31.207 ± 1.4	7.906 ± 1.4	1.370	−25.488
1860 (n = 1)			18/03/21
	Stegostoma tigrinum/EN	TL	1780.137 ± 3.4 (n = 2)	21/10/21	23.501 ± 3.8	6.778 ± 2	4.805	−89.863
			1870 ± 14.1 (n = 2)	07/01/21 07/09/21
	Mola Mola/DD	BH	661.2 (n = 1)	21/10/21	17.740	2.679	5.550	−38.8.848
			700 (n = 1)	30/11/20
		TH	1246.182 ± 57.6 (n = 2)	21/10/21	32.399 ± 13.6	11.437 ± 10.5	1.875	−23.818
			1270 (n = 1)	30/11/20
	Scomber colias/LC	FL	353.310 ± 56.8 (n = 9)	21/10/21	39.359 ± 5.5a	15.506 ± 4.1a	1.858	−6.690
		TL	360 ± 84.8a (n = 2)	21/10/21 01/11/21

• Note: Errors are presented as root mean square (RMS) and precision obtained from Event Measure, percentages, and millimeters by comparison with real measurements. The number of measurements (n) and dates are given for each species and its current IUCN Red List Category (IUCN Cat.).
• Abbreviations: BH, body height; DL, disc length; DW, disc width; FL, fork length; SD, standard deviation; stereo-DOV, diver-operated stereo-video systems; TH, total height; TL, total length.
• ^a Total lengths from necropsy data, the available real measurements for comparison.

3 Results

The stereo-DOV equipment allowed the successful recording of 101 individuals from 47 species (16 elasmobranch and 31 teleost species) in a single dive of 2 h and 15 min total time (data available upon request). Approximately 8 h were needed to measure all individuals on Event Measure.

The animals were recorded at a minimum range of 965 mm and a maximum of 6082 mm of distance from the cameras. The RMS values were below 20 mm in 26% of the measurements while for the targeted species (Table 1) these values were all higher, except for Mola mola (Linnaeus, 1758) body height measure (17.74 mm). From the comparison with real measurements, results show a maximum difference of 1.3% (8.2 mm) in disc length (DL) and of 2.7% (30.8 mm) in disc width (DW) for the individual ray Mobula hypostoma (Bancroft, 1831) and 1.8% (34.5 mm) in total length (TL) for Glaucostegus cemiculus (Geoffroy Saint-Hilaire, 1817). Regarding the Stegostoma tigrinum (Forster, 1791) shark, exceptionally two true measurements were considered (taken 8 months apart for an adult individual) once this species has a particularly long tail and is complex to measure ensuring no body flexion. The measurement error was therefore the highest for elasmobranchs, reaching 4.8% (89 mm) for an ≈1800 mm TL specimen. The selected species to represent individual teleost (M. mola) revealed an error of 1.8% (≈23 mm, n = 2, for a ≈ 1250 mm total height fish) up to 5.5% body height (BH). Representing teleosts monitored as a group, the species Scomber colias (Gmelin, 1789) presented an error of 1.8% (6.6 mm) even when comparing the available real TL (n = 2) with the fork length (FL) estimates (n = 9).

4 Discussion

This study aimed to investigate whether a diver-operated stereo-video system can be used in an aquarium setting, to accurately collect biometric data from multiple marine species. This method allowed the measurement of over 100 animals in a nearly 10-h time window, saving considerable time and costs while being safer for animals and staff.

The stereo-DOV was designed and built at Oceanário during the coronavirus pandemic following a low-cost approach. Although the software CAL validated the calibration session, before the recording session, some issues were later identified with the design of the stereo-DOV. For the distance range at which the animals were recorded (from 965 to 6082 mm), a larger camera separation would be recommendable (at least 800 mm and as far as 1400 mm), as well as a higher conversion angle for the cameras (as of 8° degrees instead of the 4° used) (Harvey and Shortis 1995; Shortis 2015; Goetze et al. 2019). The third and probably more impacting issue was the camera's immobility. The GoPro Hero 4 original housings do not guarantee the camera's immobility inside the housings. It is possible that even without relocating the cameras from the housings, a few millimeters move (during the single recording session) could contribute to poorer calibration stability when analyzing the frames on Event Measure, thus resulting in high RMS and Precision values (Harvey et al. 2010). More recent models of similar cameras, such as the GoPro Hero 8 and newer, are available and do not require additional housing to operate at depths of up to 20 m.

One single dive session provided data for 101 individuals from 47 species. The main indicator of measurement accuracy (RMS) was higher than recommended in most measurements (Harvey and Shortis 1995; Harvey et al. 2010). Despite this evidence, real differences between stereo-video estimates and real measurements of the selected species ranged from 1.37% to 5.5% (8 to a maximum of 89 mm). This data is of great importance, not only to the institution caring for the species but to a broader scale as it increases the demographics knowledge for many marine species (Goetze et al. 2018, Lewis, Dawson, and Rayment 2023). For S. colias, monitored as a group, individual lengths and weights are crucial for extrapolating data for the average school. Standardized measurements are essential to compare estimated lengths with true values. For similar teleost species, true measurements are typically obtained from necropsies, as capturing the fish for measurement is stressful, potentially requiring sedation, and impractical for large collections. Repeated length estimates can build robust length-weight relationships in controlled environments, subsequently used to estimate biomass and adjust feeding, supplements, or medication dosages (e.g., Boggio-Pasqua et al. 2022). These relationships are crucial for emblematic and under-studied species like Mola mola, the fastest-growing bony fish in the world. Frequent length and weight estimates are pivotal for their husbandry requirements and welfare assessment (Howard et al. 2020, Chapter 13). Note that accurate length measurements are particularly important when estimating fish biomass, fecundity, or evaluating population trends (Harvey et al. 2010; Boutros, Shortis, and Harvey 2015) and even reduced errors might result in a large influence on fecundity analysis or length-to-weight estimates. Future studies with enhanced equipment should aim to improve measurement quality, targeting errors close to 1% and RMS and precision values < 10 mm for more precise biomass estimates.

Even when working with true lengths, performed with a tape measure, and the animals restrained to their straightest position, there are resulting intrinsic errors. In practical terms and for the use of increasing knowledge and improving husbandry practices on many endangered species, the length estimates obtained in this study are acceptable and usable being close to the actual dimensions. Differences of a few millimeters can be insignificant, mainly in large animals if they serve their purpose of allowing to access growth, and filling gaps in the species' biology knowledge (Delacy et al. 2017). For this particular study and considering the entire data set, first estimates (besides visual) were obtained for several critically endangered shark species such as Carcharias taurus, Charcarhinus plumbeus) which due to their size and probably exacerbated negative reaction to handling stress, were never measured. Estimates of the blacktip reef shark, Carcarhinus melanopterus were performed years after they were introduced as juveniles, confirming they reached adult size.

In conclusion, this work demonstrates that stereo-DOVS are an effective technique to be used across many aquariums, provided that good visibility of the individuals, adequate space for diver operation, animal desensitization to the environment (thus less flight risk), and sufficient technical and financial resources are ensured. This method is nondisruptive, cost-effective, and safer for both animals and staff. It enables the rapid collection of large amounts of data, which can be revisited as needed, and requires fewer staff compared to traditional, intrusive methods such as capture. Large, accredited aquariums house hundreds of specimens, many of which are endangered or protected species. These institutions prioritize conservation and education, replicating natural habitats, and supporting species survival through various programs that require extensive multi-species data collection (Ripple et al. 2021). The volume and quality of data that can be gathered using this method have the potential to significantly enhance the conservation efforts of aquariums and similar institutions, thereby contributing more effectively to scientific research.

Conflicts of Interest

The authors declare no conflicts of interest.