Comparing Manual and Automated Spatial Tracking of Captive Spider Monkeys Using Heatmaps.

Animal welfare assessments increasingly aim to quantify enclosure use and activity to support naturalistic behavior and improve Quality of Life (QoL). Traditionally, this is achieved through manual observations, which are time-consuming, subject to observer bias, and limited in temporal resolution due to short observation periods. Here, we compared manual tracking using ZooMonitor with automated pose estimation (SLEAP) in a mother-son pair of black-headed spider monkeys (Ateles fusciceps) at Aalborg Zoo. We collected manual observations on six non-consecutive days (median daily duration: 62 min, mean: 66 min, range: 52-90 min) and visualized this as spatial heatmaps. We applied pose estimation to the same video footage, tracking four body parts to generate corresponding heatmaps. Across most days, the methods showed strong agreement (overlap 83-99%, Pearson's r = 0.93-1.00), with both highlighting core activity areas on the floor near the central climbing structures and by the door with feeding gutters. Both methods also produced comparable estimates of time spent being active, with no significant difference across days (p = 0.952). Our results demonstrate that computer vision technology can provide a reliable and scalable tool for monitoring enclosure use and activity, enhancing the efficiency and consistency of zoo-based welfare assessments while reducing reliance on labor-intensive manual observations.