The rollout of 5G has increased spatial and temporal variability in RF-EMF exposure due to higher frequency bands, beamforming, and denser deployments, challenging conventional maps that rely only on simulations or sparse measurements. We present a hybrid framework that fuses deterministic ray-tracing with heterogeneous data (static sensor node data gathered over a two-year period, a mobile spectrum analyzer (SA) campaign, and personal exposimeter measurements) to generate high-resolution spatio-temporal exposure maps for a dense urban area of 2.5 km2. At its core is the Mean-Matched Gradient-Preserving Optimization (MMGPO), which estimates a smooth multiplicative correction field over the simulated grid to remove bias while preserving logarithmic gradients. When validated against 28 independent SA measurements, MMGPO achieves an RMSE=5.47 dB and correlation r=0.70 for the total electric field exposure over all frequency bands. By incorporating temporally varying node data, this approach provides policymakers, researchers, and the public with reliable, high-resolution exposure assessments across multiple time windows, capturing spatial detail while reflecting temporal variations in modern wireless networks.
