Back to search

Boston University

Investigating positional behavior and locomotion of wild Bornean orangutans using behavioral sampling and computer vision

Abstract

dc:description.abstract

Non-invasive locomotion research of the critically endangered Bornean orangutan (Pongo pygmaeus wurmbii) has traditionally been difficult, with work primarily capturing small samples through cross-sectional studies. In situ studies are crucial to better understanding orangutan locomotor patterns as they can aid in promoting appropriate conservation practices, such as the maintenance of rehabilitation programs or the development of effective habitat corridors. Moreover, orangutans serve as useful extant animal models to study the evolution of human bipedalism and reconstruct the locomotion of fossil hominoids. This study applies traditional behavioral sampling and computer vision to investigate positional behavior and locomotion of wild orangutans in Gunung Palung National Park in Borneo, Indonesia. In Chapter 1, I investigate the ontogeny of positional behavior and locomotion in Bornean orangutans. I use an extensive behavioral dataset, collected over 30 years, to analyze differences in positional category diversity (a combination of behavioral contexts, positional behaviors, and substrate use) and broad behavioral categories between both age and sex classes. I show that orangutan positional category diversity and average positional category distribution differ by age class, with a peak during juvenility noted in both positional category diversity and unique position categories observed, after which positional category diversity decreases into adulthood. Further, I present a case study looking at positional category diversity aging in one female. I find that positional category diversity continues to decrease even after adulthood is reached. In Chapter 2, I introduce a novel, robust pose estimation model for fine-grained, non-invasive kinematic analyses of wild orangutans (DeepHutan). I present an extensively annotated image dataset, used to train DeepHutan, and systematically explore the use of transfer learning during model training. I do not find improved results when using a pre-trained primate model as a starting point for DeepHutan. I also explore the effect of substrate type on orangutan terrestrial movement as an example application of DeepHutan. I use joint predictions from DeepHutan on videos of orangutan terrestrial bouts on both irregular and flat substrates and collect corresponding joint angles. I manually collect duty factor and cycle duration gait parameters. I find, interestingly, that orangutans exhibit more extended joints while traveling on irregular surfaces, with no other significant differences between other gait parameters. In Chapter 3, I continue to use DeepHutan to study quadrupedal terrestrial movement in orangutans and compare it to chimpanzees. I train a small pose estimation model for chimpanzee joint estimation and use the two models to capture, automatically, gait and kinematic differences between the species. From the sagittal plane, I find no significant differences in gait parameters. However, from the transverse plane, I demonstrate differences in the joint trajectories of the non-weight-bearing limbs between the two species; orangutans exhibit more movement in the hindlimbs, though am limited by sample size. Finally, in Chapter 4, I hypothesize that arboreal adaptations prime safe downclimbing and facultative terrestrial bipedal postures, specifically arguing that an arboreal hominid ancestor’s environmental niche could have facilitated immediate bipedal stances. I use DeepHutan to extract joint angles during downclimbing bouts in orangutans compared to chimpanzees and mangabeys. Orangutans exhibit more extended forelimbs, indicating a controlled descent. Additionally, I compare the behavioral contexts behind facultative bipedalism in orangutans and chimpanzees, finding that foraging is the most prevalent context. Results from this thesis extend knowledge of wild Bornean orangutan positional behavior and locomotion, both arboreally and terrestrially. Moreover, the tool introduced, DeepHutan, provides a framework for continued non-invasive investigation of locomotion in endangered species.

Degree

thesis:*
Name thesis:degree_name
Master of Science
Level thesis:degree_level
masters
Discipline thesis:degree_discipline
Biology
Grantor
Boston University
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Sibal, Ritika
Advisor dc:contributor.advisor
  • Knott, Cheryl D.
ORCID dc:identifier.orcid

Subjects

dc:subject × 9

Rights

dc:rights
Statement dc:rights
  • Attribution-ShareAlike 4.0 International
Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/53422

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Sibal, Ritika. Investigating positional behavior and locomotion of wild Bornean orangutans using behavioral sampling and computer vision. masters thesis, Boston University, 2025. https://hdl.handle.net/2144/53422