This paper presents a novel automatic mode identification and tracking algorithm designed for long-term Structural Health Monitoring (SHM) applications. The proposed framework directly utilizes covariance-based Stochastic Subspace Identification (cov-SSI) outputs and employs a self-adapting, streaming clustering procedure to identify and track structural modes over time. By minimizing reliance on traditional Automatic Operational Modal Analysis (AOMA) steps, the routine significantly reduces computational overhead and complexity while enhancing robustness. The algorithm is validated on a six-month continuous monitoring dataset of a timber bridge in Norway, characterized by complex dynamics, which encompasses substantial variability of environmental and operational variables (EOVs). Results demonstrate that the method can reliably track all automatically identified modes, promptly re-identify modes that temporarily disappear, and retain high detection rates over the entire monitoring period. The tracked modal properties of the bridge reveal pronounced dependencies on EOVs, emphasizing the necessity of accounting for these influences in long-term SHM applications. Overall, the proposed approach offers a robust and computationally efficient solution for automatic modal-based SHM, enabling more accurate characterization of EOV effects and supporting the development of damage detection strategies with reduced risk of false alarms.