Frequent monitoring of wild Nile crocodile (Crocodylus niloticus) populations to gather demographic and management data can be complicated by their cryptic behavior, occurrence in often inaccessible areas and their semi-aquatic nature (Behangana et al., 2017). In wild crocodile population surveys, it is often easier to observe larger animals, which are more reliably spotted and objectively classed into size categories (Behangana et al., 2017; Aubert et al., 2021; Utete, 2021). Until recently, crocodile population surveys relied exclusively on counts from fixed-wing aircraft and helicopters, or spotlight counts from boats at night; where waterways are accessible (Bayliss, 1987; Combrink, 2004; Ferreira and Pienaar, 2011; Wallace et al., 2013). These methods are potentially disruptive, time consuming, require costly fuel and equipment, and may result in observer-biased size categorization (Elsey and Trosclair, 2016; Ezat et al., 2018).

Uncrewed aerial vehicles (UAVs) have emerged as a valuable tool for conducting crocodile population and habitat surveys due to several advantages over traditional methods, including improved counting accuracy, cost-effectiveness, repeatability, reduced animal stress, and improved researcher safety (Aubert et al., 2021). The efficacy of UAV surveys is constrained by the visibility of animals from above (Aubert et al., 2021). Recent advancements in imaging sensor quality have resulted in higher-quality images, reducing the need for lower flight altitudes (Toffanin, 2019). Given sufficient drone batteries and suitable vehicle access to establish viable drone takeoff and landing points, large areas can now be surveyed whilst producing clear enough images for accurate morphological measurements. Basking behaviour in open systems (e.g., bedrock based, mature river reaches), especially during the winter season, offers the ideal UAV monitoring scenario as many crocodiles leave the water between 10h00 and 17h00 to thermoregulate (Downs et al., 2008). The effects of UAV presence, with a focus on auditory and visual disturbances, observer bias, ethical flight practices, and UAV-related animal behaviours, have been considered (Hodgson and Koh, 2016; Bevan et al., 2018; Appendix 1 ). The recommended flight altitudes for crocodilians range from 40 to 50 meters, depending on the imaging sensor used, to produce images with sufficient clarity and photogrammetric accuracy for crocodile population monitoring (Bevan et al., 2018; Ezat et al., 2018; Aubert et al., 2021; Myburgh et al., 2021).

Zoometric measures have ecological significance as they provide information about the size, growth, and health of individual crocodiles and populations (Salem, 2011; Warner et al., 2016a). Several zoometric measurements are associated with crocodilians (Edwards et al., 2017), a few of which can be recognized from the perspective of a UAV. Examples of standard morphometric measures for crocodiles are total length (TL: distance from the tip of the snout to the tip of the tail), snout vent length (SVL: distance from the tip of the snout to the third circumcircle scute layer that corresponds to the caudal margin of the cloaca), head length (HL: distance from the tip of the snout to the posterior ridge of the supraoccipital bone) and belly width (BW: transverse measurement across the widest part of a crocodile’s smooth ventral scutes) (Warner et al., 2016a; Edwards et al., 2017; Webb et al., 2021).

Despite advantages of UAV-based crocodile population census are clear, some factors limit the widespread application of the technique. In the case of UAV-based morphometrics, TL as a measure can be problematic due to the loss of tail ends, the difficulty to accurately distinguish the thin tail tip from its surroundings, or due to partial submersion of a crocodile in water (Mazzotti et al., 2012; Myburgh, 2021). Orthophotos with a ground sampling distance (GSD) >3cm/pixel can make sections of the tail difficult to distinguish. Previous studies have derived relationships between TL and SVL or HL of crocodiles as a means of accurately estimating TL from these more reliably captured measures. Morphometric measures hold importance for body condition and therefore welfare assessments, as well as population size estimates (Hutton, 1987; Salem, 2011; Warner et al., 2016a).

A body condition score is a measure of an animal’s relative “fatness”, from which health and well-being is inferred (Fujisaki et al., 2009; Mazzotti et al., 2012; Ojeda-Adame et al., 2020). Body conditions of crocodiles have been assessed in multiple studies using Fultons index (K) which relates length to mass, which is considered useful for comparisons across populations (Zweig, 2003; Fujisaki et al., 2009; Mazzotti et al., 2012; Brandt et al., 2016; Shirley et al., 2017; Webb et al., 2021). Ojeda-Adame et al. (2020) evaluated three condition score indices based on length and mass measurements of 26 American crocodiles (Crocodylus acutus); namely Fultons index, relative condition index and scaled mass index and suggested that the indices were difficult to interpret and potentially misleading. The proposed alternative was an in-depth condition score that assessed multiple morphological zones rather than assigning a relationship between length and volume measures (Ojeda-Adame et al., 2020). For a hands-on assessment this is the recommended condition scoring method as it assesses regions of the body known to be affected by nutritional status.

Hands -on condition scoring techniques may not always be desirable or possible when working with wild populations or particularly large crocodiles. UAV-based body condition assessments have the advantage of being non-invasive and have been applied to Fur seals, Antillean manatees and eastern North Pacific gray whales (Allan et al., 2019; Soledade Lemos et al., 2020; Ramos et al., 2022). In these studies, straight-line body length, belly girth and surface area, among other measures were considered. To our knowledge, no remote sensing body condition assessments have been attempted for crocodilians. A normalized UAV-based body condition score could be a useful health and well-being assessment tool in situations where large-scale assessments are required. These methods could be valuable in instances where rapid assessments of the well-being of a large crocodile population are required, a population is distributed over a large area, where crocodile populations occur in remote locations, or where capture/handling is not possible (i.e., capture difficulties, not wanting to cause the animals undue stress, illness is suspected in a population).

In this study, the relationships between Nile crocodile morphometric measures derived from high resolution rectified UAV imagery were assessed. A logical combination of these measures for the assessment of well-being/health is derived and discussed, in the form of a body condition score. The relationship between several crocodile morphometric parameters is determined for use in future UAV based crocodilian population monitoring.

Two hundred and eighty-eight crocodiles were assessed at two commercial crocodile farms in South Africa. One of the farms crocodiles appeared slimmer at time of the flights, when compared to the other farms crocodiles. The spectrum of thin to fat crocodiles was assessed by the following UAV-captured measures: total length (TL), snout hindlimb length (SHL), snout neck length (SNL), neck width (NW), belly width (BW), total surface area (SA) and perimeter. A drone disturbance assessment occurred prior to morphometrics capture and identified the drone model and flight altitudes that were utilized in the current study, the results of which can be found in Appendix 1 .

The current study assessed hundreds of farmed Nile crocodiles in a matter of days, at a relatively low cost. The crocodiles were counted, their morphometrics measured and their health inferred through a body condition score developed with data generated from this work. The variation in body condition scores between the two farms lends credibility to the measures. This method allowed a fast paced and repeatable examination of the sizes and conditions of two populations of captive crocodiles. This method has conservation value and could easily be applied to wild Nile crocodile populations.

When appropriate morphometric measures are used, UAVs enable accurate body condition analyses of crocodiles. The crocodiles assessed in the present study represented a range of body condition states, covering a wide range of body conditions that may be encountered during wild crocodile population surveys. NW was not considered a viable metric for use in wild population surveys due to the difficulty of identifying the neck of crocodiles in lower resolution imagery (Myburgh, unpub. data). BW is considered a more suitable UAV-derived metric for estimating body condition, as it is easier to identify/measure and could be better suited for future applications of emerging technologies. Artificial intelligence, for example, requires distinguishable features for automated detection. Although SHL could be used for estimating body condition relationships, the use of TL was intentional, because previous studies have used TL and its relation to other length-based morphometrics (SVL and HL traditionally, SHL and SNL in this study). Therefore, a UAV-based body condition score was calculated (Equation 2) using TL and BW, which ranked the assessed crocodiles between 1 (comparatively thin) and 5 (comparatively fat). Applying this simple formula to UAV-derived imagery could offer an accurate, fast-paced, and non-invasive health and welfare monitoring tool for large, wild or captive crocodile populations.

Relationships between various zoometric parameters for crocodilians have been investigated through traditional capture and measurement methods (Hutton, 1987; Salem, 2011; Warner et al., 2016a), but none have assessed UAV derived morphometrics. Multiple equations for predicting UAV-based length morphometrics from one another were presented in the current study (equations 3-5), all with R² values ≥0.92. TL was more reliably estimated from SNL and SHL for Farm A than from farm B, and it is postulated that variations in body condition status may have influenced the observed differences in R-squared values.

These equations can be useful in generating population-wide morphometric statistics, even when not all body parts of a Nile crocodile can be seen in UAV imagery. The strong positive and significant Pearson correlation coefficients between the length-based morphometric measures are likely indicative of allometric growth. Relating these morphometrics to body weight and manual morphometric measurements would add further precision to UAV-based morphometric data collections and body condition scoring in future studies.

Morphometrics measured in this study were only possible for crocodiles >1m TL due to image distortion for smaller crocodiles (measurement landmarks not clearly visible). The tendency for smaller crocodiles to pile (especially in pens on commercial farms) and thereby prevent accurate measurements, and the skittish nature of small (<1m total length) crocodiles impedes image quality. For farmed Nile crocodile populations specifically, the measurement of crocodiles <1m in TL is still a potentially useful endeavor in terms of continuous morphometric and condition assessments if a viable method for surveying them can be developed. Further assessments are ongoing regarding UAV-based morphometrics for farmed grower (<1m total length) Nile crocodiles (D Viljoen, unpub. data).

Possible confounding factors for health predictions conducted through monitoring morphometrics (i.e., a body condition index/score) would be chronic diseases, toxicological problems, pathological changes, being gravid, and food availability. Pansteatitis for example, where crocodiles appear obese due to the swelling of all fat depots (Lane et al., 2013), would likely confound condition scoring within a population. To our knowledge, the effects of egg-bearing on crocodilian morphometric relationships have yet to be explored. Emaciation is suggestive of animals experiencing health or welfare issues; however, sightings of emaciated crocodiles are not uncommon in the wild (Swanepoel et al., 2000; Warner et al., 2016a, b). The low BCS of 1 or 2 for farmed crocodiles may be a more natural state for wild crocodiles where feeding is infrequent and health issues are not as controlled.

Future studies could consider assigning a cut-off value for “healthy” morphometrics per size class, which may differ for wild and farmed crocodiles, allowing the identification of discrepancies/differences between obese, emaciated, and gravid crocodiles. Considering pansteatitis specifically, this might allow particularly obese crocodiles to be identified as early warning signs of the disease. Future studies utilizing UAV-based morphometrics should also consider comparing UAV measures with true measures from captured crocodiles and including weight recordings to expand the UAV body condition methods capabilities and accuracy, as well as its comparison to traditional body metrics.

The datasets presented in this article are not readily available because of the sensitive nature of the data and confidentiality agreements with the farms involved. Requests to access the datasets should be directed to the corresponding author.

The animal studies were approved by the Animal Ethics Committee, University of Pretoria. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent was obtained from the owners for the participation of their animals in this study.

DV and AM designed the study and conducted the experiments/data collection. DV, EW and AM contributed to the statistical analyses. DV composed the initial and subsequent drafts of the manuscript. All authors reviewed and edited the manuscript, providing critical feedback and approving the final version.