Abstract | Spatial uncertain quantities such as the geometric imperfections are found of paramount importance when investigating the behaviour of thin-walled structures. The present work addresses the probabilistic estimation of the buckling capacity of single-bolted members from plain angle sections with stochastic geometric imperfections. The simulation is based on detailed experimental data of imperfections measured by an inhouse built set-up. The quatro-variate non-stationary stochastic processes herein combine the spectral representation method along with the method of separation, incorporating correlations between the imperfection components. The spectral separability check is also performed through a minor investigation. A large-scale nonlinear finite-element problem is defined to estimate the stochastic buckling capacities (in the frame of Monte Carlo simulation) for a bolted angle-section column. More uncertain parameters are then included, namely material properties and lateral load (wind pressure), towards an extended range of applicability. To this end, in order to improve the computational performance, the Latin hypercube sampling method is adopted. The resulting density distributions, presented in both tabular and graphical forms, reveal the influence of each parameter (and their combinations) on the buckling load variability and can be used also for probabilistic assessment.