Vetnews | September 2025 28 « BACK TO CONTENTS Find Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats Lucia Gordon1 , Nikhil Behari2 , Samuel Collier1, Elizabeth Bondi-Kelly2 , Jackson A. Killian1 , Catherine Ressijac1 , Peter Boucher1 , Andrew Davies1 and Milind Tambe1 1 Harvard University 2 Masssachusetts Institute of Technology luciagordon@g.harvard.edu, nbehari@media.mit.edu, scollier1@g.harvard.edu, ecbk@umich.edu, {jkillian, catherine ressijac, pboucher, andrew davies}@g.harvard.edu, milind tambe@harvard.edu ABSTRACT Much of Earth’s charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos’ movement is crucial to their protection, but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos’ spatial behaviour valuable to antipoaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labelling time when used on a similarly sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7. Introduction Vertebrates are going extinct 100x faster than before the Anthropocene, indicating that the Earth is undergoing the sixth mass extinction [Ceballos et al., 2015]. Among the human threats to biodiversity, poaching has endangered myriad species, particularly the rhino. The poaching epidemic is a barrier to achieving the United Nations’ Fifteenth Sustainable Development Goal: Life on Land [UN, 2023b]. In particular, Target 15.7 calls for“urgent action to end poaching and trafficking of protected species of flora and fauna” [UN, 2023b]. Poaching, as well as habitat loss, has exterminated African rhinos across much of their historical range [WWF, 2023]. In South Africa, one of the rhino’s last strongholds, poaching has driven a 59% decline in Kruger National Park’s rhino population since 2013 [Save the Rhino International, 2022]. As rhinos have disproportionately large impacts on ecosystem structure and function due to their role as megaherbivores [Ferreira et al., 2015], it is imperative to strengthen efforts to protect them. Protecting endangered species like the rhino involves studying their patterns of habitat use [Johnson and Gillingham, 2005], but this is challenging due to rhinos being reclusive and dangerous to observe in the wild [Linklater et al., 2013]. Moreover, efforts to study rhinos are often constrained by the limited human and financial resources available in the African savanna landscapes rhinos inhabit, which can be large and inaccessible [Anderson and Gaston, 2013]. Rather than study rhinos directly, we propose for the first time to use rhino middens, depicted in Figure 1, to improve our understanding of rhinos’ spatial behaviour. Middens are communal defecation sites used for territorial marking and social communication [Owen-Smith and Smith, 1973]. Thus, understanding midden spatial patterning is a noninvasive means for gaining insight into rhino locations and movement, providing conservation practitioners with spatial information key to effective poaching prevention, management efforts, and reintroduction plans for rhinos. Unfortunately, rhino midden locations have not yet been mapped because they are distributed throughout huge areas. Manually locating all the middens from the ground is practically impossible, and whether drones can be used for this purpose is untested. Against this background, this paper provides the first results from harnessing remotely sensed data to detect rhino middens along with the first spatial map of middens. Middens are considerably easier to detect in remotely sensed imagery than the elusive rhinos themselves, as the former are stationary and larger. Figure 1: Left: A white rhino midden next to a road in iMfolozi [Marneweck, 2013]. Right: A midden in our dataset photographed by a drone. Middens boxed in green
RkJQdWJsaXNoZXIy OTc5MDU=