Monitoring rhinoceroses in Namibia’s private custodianship properties

Permissions statement

The authors obtained verbal and written permission to conduct this study from the landowners and the Ministry of Environment and Tourism of Namibia.

Terminology

Census: A complete count, or estimate, of individuals in a population

Monitor: To follow the demographics and distribution of a population over time, by repeated censuses or other assessments conducted at regular intervals

Footprint: An impression in the substrate made by the foot.

Trail: An unbroken series of footprints made by one individual animal.

Census techniques

This pilot study was conducted at three custodianship properties in Namibia in May-June 2018; sites A, B and C. Custodian sites were chosen by the MET as those that performed regular monitoring and were presumed to have reliable estimates of rhino numbers. Sites A and B were predominantly dry bushveld and Site C was mixed grassland and dry bushveld. Site locations are not disclosed to maintain rhino security. Each site hosted a small black rhino population and C also hosted white rhino. Custodian estimates of rhino populations were obtained prior to the survey and are reported in Table 1. While each custodian was confident that they knew how many rhinos they had on their property from anti-poaching and routine surveillance patrols, we regarded their estimates as a baseline to be tested by our Method One. Our further aim was to test the consistency of the three methods relative to each other and the Custodian’s estimate and compare the utilities of the three different census techniques, all based on footprint identification.

Sites A and C were commercial hunting farms, A on mostly flat sandy terrain while C had flat sandy, largely grassed terrain transitioning to rocky undulating bushveld terrain. Site B was a recreational tourism site on flat sandy terrain. At each site, rhinos were monitored regularly through anti-poaching and infrastructure repair patrols, and through observations with clients. At all three sites, roads, game tracks, and bare areas around waterpoints provided generally good substrate for tracking. There was no rain during the study period which took place in the middle of the dry season.

The three methods

Footprints were collected by photography from a thorough search of roads, tracks, and waterholes as the common component of all three methods.

Method 2. Identification by heel pattern only (HP)

This method involved a census to provide a population estimate just using the heel lines visible in all footprint images collected that were sufficiently clear. Without using any reference to known individuals, the heel lines were matched visually from the footprint images alone and processed off-site by co-author SA, who was not involved in the field work. This method is potentially quicker than Method 1, but since rhinos were not tracked to a visual, it would not match footprints to individuals. Figure 1 shows the obvious differences in heel pattern between two individuals but their similarity in different footprints of the same individual.

Field protocol

Field notes

At site C, we focused first on the white rhino, beginning on 29 May 2018. By 9 June we had achieved visual identities on each of the 17 white rhinos known to the staff, sexed the two most recent calves (previously unsexed), and unambiguously photographed and matched footprints to each of these individuals. Every ‘trail identity’ established in the field was matched to a known rhino.

The black rhino at site C had been introduced recently as a group of six. We collected footprint images for black rhino at site C from June 1 through 10, focusing most of our attention on the black rhino from June 5 through 10. We obtained visuals on four of the six, identified the footprints of one other seen by a professional hunter (identified by ear notches), and had unambiguously photographed trails from these five black rhinos plus another trail we did not match to these five. These black rhinos were walking extensively along roads, still apparently exploring their new landscape. This fact allowed us to capture many images from single trails for some of these black rhinos.

Although the black rhinos were long established at site A, they did not confine their movements to small ranges, though there were parts of site A that rhinos apparently did not utilize. Between 21 and 29 June, we obtained visuals on all four known adult black rhino at site A and obtained the first visual of a recently born calf for which only spoor had previously been reported, thereby identifying its mother. We also obtained unambiguous spoor photographs for each adult rhino and all ‘trail identities’ were matched to known rhino.

Between 13 and 19 June we obtained visuals on each of the eight black rhinos known to staff at site B, obtained unambiguous footprint images for each of these individuals, and matched each ‘trail identity’ to a known rhino.

Analysis

More than 1,500 footprints were collected from the three sites, over a five-week period. 608 photographs of white rhino footprints and 331 black rhino footprints were collected at site C over 17 days. 402 photographs of footprints of black rhino were collected at site B over 7 days and 300 at site A over 7 days. Over the entire period of five weeks, 45 separate visuals (sometimes of multiple groups of white rhinos, and for both species occasionally of a bull accompanying, or near, a female rhino) were achieved. 23 visuals resulted from tracking footprints; the remainder began as a visual from a vehicle. Footprints were analysed as described above, and grouped by individual identity, according to the three methods. The individual groupings for the animals at each site are presented in Table S1 (Site A), S2 (Site B) and S3,4,5 (Site C)

Table 1 presents, for each site, the number of rhinos estimated by each of the three different methods along with the custodian’s prior estimate of the number of their rhino. Each method takes the pool of trails employed in its analysis and clusters them into groups, with trails within a group classified as like and trails in different groups classified as unlike. The resulting number of groups provides the estimate of the number of individuals.

At all sites, Method One confirmed the Custodian estimates. At site A, methods Two (HP) and Three (FIT) each estimated an additional individual. At site B, FIT estimated the same number of rhinos as FIDP, but HP estimated one additional rhino. At site C, all three methods agreed for black rhino. For white rhino, HP gave the same estimate as FIDP while FIT estimated 18 individuals; however, with the benefit of FIDP, it is known that no trail for one of the known white rhinos proved suitable for FIT analysis so in fact FIT recognized 16 of the known white rhinos but also created two additional identities.

Comparing and implementing the three different footprint methods

Figure 2 (see also Tables S1–S5) compares the performance of the three different methods across the three sites and indicates that the individual rhino groupings of trails generally agreed across all three methods at each site. In a few instances the trail identities disagreed, resulting in the groupings diverging from each other. These divergences were mostly due to low numbers of footprints in a trail. Method 3 (FIT) was the most demanding in terms of footprint clarity, and not all the trails that could be assessed by methods 1 and 2 were accessible to method 3 resulting in some groupings not extending to the far right of Fig. 2 that shows the FIT classification.

The most intensive method, FIDP, provided results that agreed with the custodian estimates and is also most informative about a population of rhinos. As the only method of the three that matches footprint images to visual confirmation of an individual, and thus provides the custodian with a reference set of images matching each individual to its footprints, it also allows custodians to continue to monitor individuals without prior knowledge of the population. On the other hand, the data required for methods 2 and 3 can be collected as part of the custodianship’s routine fieldwork, requiring only that this be extended to include photographing of footprints, an activity which in itself improves knowledge of the whereabouts and movements of rhino, and thereby the effectiveness of the routine monitoring. It would also draw attention to individuals whose footprints were not being observed once a library of footprints of all individuals is established.

That every trail photographed for the FIDP method was ultimately matched to a rhino known by a visual identity not only confirmed the Custodian estimates but provides evidence for its utility as a complete census. Accepting the resulting censuses provided by FIDP, method 2 (HP) erred at sites A and B by inferring one extra identity at each. Each of these identities consisted of a single trail not matched by HP to any other trail in the site’s database. For site A, that trail was photographed under a midday sun, and despite shading the footprint and using a reflector to attempt to draw out contrast, the photographs show little detail of the heel pattern, and less than was observed in the field. This trail was together with the trail of another rhino. We had seen and photographed two trails together two days previously and were able, with the detail observable in the field, to match these two trails to the previous ones; in particular, the trail constituting the additional identity in method 2 was recognized by the field workers as belonging to the bull at site A. At site B, the additional identity under method 2 also consisted of a single trail. Through method 1, it was known to the field workers that this trail was collected as the trail of the calf in a cow-calf pair. Each trail manifests variation in the heel pattern impression and highlights the difficulty that can be encountered with relying on heel patterns alone.

Method 3, FIT, agreed with method 1 at site B and for black rhino at site C. At site A, FIT, assigned an extra identity, which consisted of a single trail known to belong to Cow8 through tracking. For white rhino at site C, no trail for an adult female with calf designated cow4 proved acceptable for FIT analysis so this identity does not occur for FIT, but FIT did recognize the other 16 white rhino. In addition, two trails whose identities were known by FIDP were not matched to any other trail in the site C database by FIT and thus each constituted a separate identity.

Thus, each fictitious identity produced by methods 2 and 3 as judged by method 1 consisted of a single trail only, for which method 1 provided additional information through tracking and visuals to identify. In this pilot study we sometimes photographed trails that were less than ideal for either methods 2 or 3 because with supplementary field information they were useful in method 1. The fact that methods 2 and 3 never created a fictitious identity consisting of more than one trail suggests that the trails yielding the fictitious identities can be considered outliers for either methods 2 or 3 and that the precision of these methods can be improved by stricter adherence to the protocols for collecting photographs for these methods. At the same time, identities consisting of a single trail of less than ideal quality can be highlighted as suspect and deserving follow-up investigation. We also note that the incidence of this error did not increase with population size; specifically, the per capita rate decreased with population size.

Table 2 compares the three different methods in terms of speed of delivery, accuracy, and relative cost. It should be noted that method 2 does involve some subjective judgement and expertise. In this study method 2 was performed by co-author SA, who has years of experience examining rhino footprints. While demanding in its protocol for collecting footprints, FIT offers the attraction of being completely objective. As already noted, methods 2 and 3 exploit different information in the footprint and one may prove more effective on some substrates than the other.

Recommendations on the application of methods 1–3

Custodians are required to obtain regular sightings of their rhinos and report to the Ministry of Environment and Tourism. If an individual is not seen within the allocated time, an intensive search is required. To fulfill these demands, they must make a strategic choice of monitoring strategy depending on their resources including staff availability and expertise, and the budget available for monitoring. Terrain suitability and climate will also dictate the method(s) chosen. We recommend that a full ID protocol (method 1) should be undertaken at each custodianship annually if possible. In between times, and at least monthly, custodians can either use their own team or host a team trained in the protocol to do either a quick onsite survey (method 2) or take images using the FIT protocol to submit to an off-site survey (method 3). Any or all of these approaches can and should be combined and/or augmented using camera-traps and regular traditional patrols on foot or waterhole counts to ensure that all rhinos are seen regularly (see below). Clearly it is in the interests of the custodians to be self-sufficient in rhino monitoring, and the WildTrack non-profit program aims to provide training and a JMP software license without cost to all custodians who are interested in using FIT to monitor their rhinos. The training can be for data collection, or for collection and analysis, depending on custodian staff resources.

During the fieldwork, other demographic observations were made. All custodians shared some natural uncertainty about the sex of new calves and the identity of their dam. The ground survey teams were able to locate all the new calves, sex some of them from direct observation and identify their mothers. At site A, trackers occasionally attributed a rhino identity to a trail based on its location, rather than careful inspection of the pad impression. We have encountered this behavior in previous field work. New light has been shed on black-rhino recursive movement patterns (Seidel et al., 2019) that could help explain these tracker assumptions. By including tracking into the monitoring regime, movements of rhino would be better known to those charged with monitoring, leading to a greater familiarity with their recursive movements within their home range. This information not only aids in keeping tabs on rhinos but also provides useful demographic information, such as which males are spatially co-extensive with which females and whether a single bull is dominating over others.

Footprints, in the form of the three methods tested in this study, not only provide a monitoring tool but also current information such as ranging and social interactions. When the team arrived at site B, staff reported seeing footprints on their walks and were unsure of the identity of the animal that made them. Tracking and FIT identified these as from the younger bull at this location. Method 1 will provide the most information of this sort and consequently, the most accurate representation of the population.

Comparison with other monitoring options

Although a detailed analysis of costs and other benefits is outside the remit of this paper, it is worth considering, from the perspective of a custodian, how footprint identification compares with other available techniques. Table 3 compares the performance of five different methods of monitoring. Although the costs of some monitoring techniques have been published for different species (Christie et al., 2016; Claus, 2017), absolute comparisons are very difficult to make over different sized areas, different ecosystems, times of year, providers and users. The comparison in Table 3, based on our experience of many years working in this field, gives a best estimate of relative costs and benefits of these different methods as a baseline for custodians to explore further with local variables.

There are several advantages to the use of footprints compared with other current alternatives. In terms of a cost-benefit ratio, footprint identification ranks relatively highly because it requires only minimal equipment (a camera, GPS and pencil/paper/ruler) and can engage scouts who are already on site for anti-poaching/tourist work. However, it is only effective where the substrate will hold clear footprints, where expert trackers are available, and where the weather does not compromise the integrity of the footprints. Also, our experience indicates that as an animal ages, the heel patterns and footprint morphometrics may change. In our experience of work at the same site over several years (Alibhai, Jewell & Law, 2008) this does not appear to affect the accuracy of the FIT algorithm if monitoring is conducted at least once a year. To date, the largest population we have surveyed is 40 white rhinos over three years at this same site. We anticipate that the technique would scale up reasonably well if monitoring is carried out routinely.

Where trackers are available, the use of a footprint-based method can also act to leverage their traditional ecological knowledge (TEK) and, if trainees are engaged, revive these skills for the benefit of wildlife conservation. Trackers are often employed by hunting or photographic tourism sites specifically to locate an animal, or for anti-poaching duties. However, where they are employed as part of an ongoing rhino monitoring strategy their TEK about rhino ecology and behavior can be integrated into the process, contributing to local capacity in wildlife custodianship and helping develop sustainable monitoring in the absence of external experts. Footprint monitoring techniques can thus be particularly cost-effective in leveraging scout TEK for monitoring at the same time as guiding and anti-poaching.

Footprint surveys, consisting of a tracker and photographer, can provide a reliable census figure with minimal equipment (cameras, GPS and rulers) in about a week for a typical custodianship. They also provide a means of matching each individual rhino on a property to its unique footprints, so that if patrols cover the ground regularly it can provide a cost-effective ongoing monitoring approach. If deployed effectively, and ideally in combination with other ground sensors (camera-traps, regular patrols), it is also a robust anti-poaching tool. With effective monitoring rhino populations will increase and animals disperse. Footprint monitoring need not be compromised by large population sizes, however, if geographic location, social groups and known movement patterns are included as prior knowledge to assist in the identification of more animals.

Aerial counts are more expensive but have the advantage of rapid deployment. However, rhinos are not always easy to spot, particularly with vegetation cover. The advent of UAV monitoring (Christie et al., 2016) will certainly reduce the costs of aerial monitoring once the technology is fully developed, but the challenge of aerial identification remains. The more traditional approach of water hole counts are typically undertaken by local staff during a full-moon to provide at least some lighting to aid visual identification during the hours when rhino come to drink, but which is still challenging. We compared footprint identification with waterhole visual counts in Etosha National Park (Alibhai & Jewell, 1997). Footprints clearly indicated that some animals were visiting more than one waterhole each night but these different waterhole encounters with the same rhino were usually counted as distinct rhino in the waterhole counts, resulting in overcounting. Fitting instrumentation, such as horn-implants, ankle bracelets and collars, is the most expensive method as it requires the considerable costs (and risks to both animals and humans) of immobilization, instrumentation, routine (often aerial) follow-up monitoring and often repetition of these tasks if the instrumentation fails (Alibhai & Jewell, 2001a; Alibhai, Jewell & Towindo, 2001b). It will also only ‘count’ those animals already instrumented and is therefore not a full monitoring method unless every animal is instrumented and maintained in that state.