【Objective】This study aims to investigate the wind-induced response characteristics of trees across different terrains. It specifically examines to analyze how leaf quantity and turbulence profiles affect the acceleration probability distribution and extremes of the branches and trunks. By analyzing these factors, the study aims to provide valuable references for maintaining and managing protective forests. Understanding how environmental variables impact the aerodynamic behavior of trees can aid in designing better protective measures against wind damage.【Method】An aeroelastic model tree, designed to meet primary similarity requirements, served as the research foundation. Eight distinct crown shapes, varying in leaf quantity, were meticulously crafted to replicate realistic tree structures. These models were tested in wind tunnel terrains, constructed based on the Engineering Sciences Data Unit (ESDU) 85020 framework. Four terrain were replicated in the wind tunnel to measure wind-induced responses. Wind speed, along with branch and trunk accelerations, were recorded to explore the probability distribution and extremes of these accelerations. This comprehensive setup ensured a thorough investigation of how different wind conditions and tree structures affect wind-induced responses.【Result】Skewness and kurtosis values at the 80% cumulative probability were used to establish criteria for determining the Gaussian distribution of trunk acceleration. The number of leaves had a limited effect on the skewness and kurtosis values for trunk acceleration in the along wind and cross wind directions but had a strong effect on branch acceleration. The trunk accelerations in both wind directions generally followed a Gaussian distribution, while branch accelerations did not. Skewness and kurtosis values for branch acceleration significantly increased with turbulence intensity. Trunk acceleration samples in the along-wind direction approximately exhibited a bell shape, with a high peak and heavy tails. The Generalized Pareto distribution fited the central part of the sample data best, followed by the Gaussian distribution, while the Gumbel distribution fited the worst. Similarly, the Generalized Pareto distribution best fitted the tail data, followed by the Gaussian and Gumbel distributions. This indicated that the Generalized Pareto distribution was more effective for fitting the probability distribution of sample data than the Gaussian and Gumbel distributions. Among the three calculation methods, Generalized Extreme Value (Gumbel) method, Generalized Pareto method, and Peak Factor method, the Generalized Pareto method best fited extreme values of trunk accelerations in the along-wind direction. A quadratic polynomial effectively modeled the relationship between extreme trunk accelerations in the along wind direction and average wind speed. Trunk acceleration increaseed with both average wind speed and turbulence levels.【Conclusion】Based on wind tunnel tests with an aeroelastic model tree, this study provides insights into the probability distribution characteristics and extreme characteristics of branch and trunk accelerations at full scale.