Stochastic reconstruction of non-stationary downburst winds using spectral proper orthogonal decomposition

Designing safe and resilient structures begins with understanding their response to wind forces under realistic conditions. Performance-based wind design requires reliable estimation of probabilistic performance metrics, such as structural collapse reliability, to support continuous occupancy objectives. These estimates typically rely on Monte Carlo or stratified sampling methods, which demand numerous realizations of stochastic wind loads with consistent statistical properties. Because repeating thousands of experimental realizations is often prohibitively costly, efficient stochastic reconstruction of experimental data provides a practical alternative. Building on the authors' spectral proper orthogonal decomposition method (Lin et al., 2025; Mech. Syst. Signal Process. , vol. 235, p. 112876), which improved reconstruction accuracy for stationary wind events, this study extends the method to model non-stationary downburst winds. This framework incorporates a stationarity check to verify the applicability of the method and offers practical guidance for noise truncation, moving-window size selection, and interpolation. Two illustrative examples of structures with profoundly different aerodynamic properties are presented to demonstrate the implementation and versatility of the proposed method. The framework's accuracy and effectiveness are assessed by comparing the variance and power spectral density of the reconstructed data to measured values, highlighting its potential for modeling stochastic non-stationary wind loads in performance-based wind design. • Proposed an SPOD-based framework for stochastic simulation of non-stationary downbursts. • Achieved higher accuracy and efficiency compared to baseline models. • Integrated a stationarity check for broad applicability in wind engineering. • Provided two illustrative examples for clear implementation guidance to future users. • Offers a reliable basis for performance-based wind design under non-stationary loads.