Terrestrial-aerial bimodal vehicles (TABVs) have air-ground advantages but face localization bottlenecks due to FoV differences, abrupt mode-transition motion, unstable ground attitude and visual SLAM limits.An environment-aware robust localization system for passive-wheel-based TABVs is proposed, fusing data from stereo cameras, IMUs, encoders and single-point laser sensors.It includes a terrain-aware odometer model and an anomaly-aware method, with extensive experiments proving its effectiveness in accurate ground localization.