Abstract

Earthworm behavioral biomarkers provide more sensitive and comprehensive assessments for environmental contaminants. However, natural soil opaqueness has limited understanding of the continuous three-dimensional (3D) characteristics of earthworm burrowing activity. In this study, we developed an integrative method combining a transparent medium with a deep learning-based trajectory reconstruction algorithm to quantitatively investigate species-specific burrowing strategies. Earthworms thrived in the transparent medium, as shown by biomass increase (0.62 to 0.72 g) and >80% survival over the 28-day trial, while algorithm operational stability (83.65% coordinate extraction rate) allowed continuous tracking of the earthworm spatial coordinates in 3D locations. Comparative analysis across five earthworm species revealed species-specific patterns: Eisenia fetida, Amynthas pectiniferus, and Eudrilus eugeniae preferred surface layers and moved laterally, whereas Perionyx excavatus and Metaphire guillelmi created vertical burrows. These findings are valuable for quantifying soil health assessments using different earthworm species. Moreover, this method successfully captured avoidance behavior in arsenate-contaminated soils, as arsenic exposure inhibited cumulative burrowing length. These results establish a critical tool for advancing ecological risk assessment and management of soil pollutants by offering visual and quantitative insights into 3D earthworm burrowing dynamics, and provide potential applications in ecology, environmental pollution monitoring, and remediation.