Urban Air Mobility (UAM) networks pose challenges to confidentiality, availability, and integrity within security and safety concerns. Successfully planning and executing today's complex operations in defense of national interests requires timely, accurate, trusted, and unambiguous communications up, down, and across an extended chain of command spanning multi-national air, ground, sea, and cyber forces. Understanding the underlying reasons for predictions in machine learning (ML)-based spoofing detection is crucial for gaining insights into attack tactics while improving the detection accuracy, and informing mitigation strategies. ML prediction enhances the interpretability, trust, and accountability in unmanned systems security and enables proactive defense measures and informed decision-making. In this paper, we introduce a Trust AI-based Decentralized Anomaly Detection (TADAD) framework to enhance the UAM capability to allow secured data accessing and sharing with low-latency and high explainable reliability among aircraft and Air Traffic Service (ATS) providers or ground stations within dynamic network environments. TADAD extends the existing airborne networks to heterogeneous open networks with advanced and autonomous detection at tactical edges. TADAD utilizes a software-in-the-loop (SITL) simulator to emulate the data communication messages (e.g., MAVLink) among UAVs and generates both benign and attack samples. An ML-based approach is then developed to incorporate explainable artificial intelligence (AI) techniques, specifically Shapley Additive Explanations (SHAP), to analyze why a signal is classified as an anomaly. Our experimental study validates the performance of TADAD in detecting accuracy and explainability against typical cyber-attacks (e.g., GPS spoofing).