Horizontal dynamic displacement is important to the structural health monitoring and safety assessment. This study proposes a dynamic displacement estimation method which fuses strain-derived displacement and acceleration data to accurately estimate displacement. The proposed method derives displacement from strain directly and thus avoids analysis error of supertall structures. The dynamic displacement is first derived directly from longitudinal strains via geometric deformation. The strain-derived displacement is then combined with acceleration data via multi-rate Kalman filtering. The accuracy and applicability of the proposed data fusion method is verified through a field test of the 335 m Wuhan Yangtze Navigation Centre (WYNC) under ambient excitation. The proposed data fusion displacement is more accurate than the GPS-measured displacement both in the time and frequency domains.