Entanglement-Enhanced Quantum-Informed PINNs for Predictive Troposcatter Link Adaptation in Anomalous Atmospheric Ducts

While Low Earth Orbit (LEO) satellite constellations face increasing vulnerabilities in contested radio frequency environments, tropospheric scatter (troposcatter) communications provide a resilient, space-independent Beyond-Line-of-Sight (BLOS) alternative for tactical networks. However, escalating climate volatility increasingly induces non-stationary atmospheric ducting layers—such as marine evaporation ducts and thermal inversions—that severely alter the vertical modified refractivity profile, causing catastrophic multipath trapping, deep fading, and link outages. Classical numerical wave solvers suffer from prohibitive computational complexity that prevents real-time deployment on tactical edge modems, whereas pure data-driven deep learning models lack physical inductive biases and fail during out-of-distribution weather anomalies. To address these fundamental limitations, we propose a hybrid Quantum Physics-Informed Neural Network (QPINN) integrated within a hierarchical Space-Ground Environmental Digital Twin (SG-EDT) architecture for real-time troposcatter link adaptation. The QPINN encodes classical atmospheric telemetry into a variational quantum circuit with entangling gates, leveraging Hilbert-space representation learning to evaluate concurrent multi-variable cross-correlations. Crucially, the model embeds exact 2D narrow-angle Parabolic Wave Equation constraints alongside a Huberized thermodynamic gradient penalty bounded by physical atmospheric trapping boundaries directly into the autograd optimization loop. Evaluated on tactical edge hardware, our framework eliminates loss penalty explosions, achieves high-fidelity refractivity profile reconstruction, and executes low-latency forward inference, enabling preemptive Adaptive Transmit Power Control (ATPC) and spatial nulling before multipath link degradation occurs.