We present UAV-CacheTrace v1.0, a 140-million-row synthetic mobility–content trace that fuses a 70-day, 100 k-device urban trajectory corpus with time-varying video popularity derived from multi-regional trending statistics. Each record encodes a user identifier, 30-min time slot, 500 m  \(\times \)  500 m geo-cell, requested content ID, category, and file size, yielding a reproducible benchmark for cooperative caching, trajectory planning, and multi-agent reinforcement learning (MARL) in next-generation aerial edge networks. To mimic real workloads, content popularity follows a Zipf-like decay calibrated on viral bursts, while a density-aware spatial partition preserves hotspot heterogeneity. Validation confirms that the trace reproduces circadian and weekly cycles ( \(r_{24\,\textrm{h}}=0.978\) ) and attains a Zipf exponent \(\alpha \approx 0.8\) –values consistent with field measurements. The full CSV ( \(\approx \) 5 GB), generation scripts, random seeds, and baseline MARL policies are released under CC-BY 4.0, enabling transparent comparison across algorithms and laboratories. By bridging realistic human mobility with non-stationary multimedia demand, UAV-CacheTrace lowers the simulation-to-deployment gap for studies on edge caching, mobility prediction, federated learning, and energy-aware UAV swarms.