Smart cities increasingly rely on interconnected data infrastructures, yet emergency management systems—particularly for wildfire evacuation—remain disconnected from these capabilities. This paper proposes a four-layer GPS-based smart city mobility data framework that integrates crowd-sourced location data as a permanent component of urban evacuation planning infrastructure. The framework encompasses passive GPS sensing and ingestion, behavioral inference through origin detection and departure timing analysis, dynamic multi-criteria routing with fire-risk awareness, and governance-compliant decision support. Proof-of-concept validation using 181 million GPS records from the 2023 McDougall Creek wildfire in Kelowna, British Columbia, demonstrates that GPS-calibrated behavioral inputs combined with hybrid dynamic routing reduce evacuation clearance time by 39% and fire-risk exposure by 61% compared to conventional approaches. We discuss data integration challenges, accuracy considerations across GPS and cellular sources, transferability conditions, and governance requirements for operational deployment in wildland–urban interface communities.