Environmental DNA (eDNA) Approaches for Large-Scale Urban Biodiversity Assessment and Species Surveillance

Abstract

Abstract— The rapid urbanization is driving biodiversity decline, but traditional ecological surveys in the field are extremely labor intensive, time consuming, spatially constrained, and unable to reveal cryptic and rare species. This research tackles these challenges by offering a framework for the environmental DNA (eDNA) based assessment and monitoring of biodiversity at a large scale in urban settings. The main goal is to develop an integrated monitoring approach that will allow for the identification of species with high accuracy using eDNA metabarcoding, the mapping of biodiversity and ecological hotspots using high throughput sequencing, geographic information systems (GIS) and machine learning in different urban habitats. Samples from various urban ecosystems are taken, DNA is extracted from water, soil and sediment, the sequences are filtered for quality, taxonomically assigned, features are engineered and species classified using Random Forest. Experimental evaluation shows that an overall species identification accuracy of 97.14%, precision of 96.72%, recall of 96.38%, F1-score of 96.55% and ROC-AUC of 98.41% can be obtained, and that biodiversity hotspot mapping can map 95.83% of the area in agreement with the validated ecological observations. The proposed framework also provides a 18.6 % increase in the detection of rare species, and 42.8 % reduction in monitoring time compared to using conventional species survey methods. The novelty of this research is the combination of eDNA metabarcoding and artificial intelligence and spatial analysis for scalable and non-invasive urban biodiversity surveillance. The proposed framework brings a cost-effective, efficient and accurate decision support system for biodiversity conservation, ecological planning, monitoring invasive species and sustainable management of urban ecosystems.