Abstract: Wind loads acting on building surfaces often show a strong correlation. Studies have indicated that this correlation can exacerbate the extent of wind-induced damage to building envelopes. However, wind tunnel tests typically last 10~30 minutes, provide short-duration data, poses challenges in modelling multivariate extreme wind loads that account for the effects of correlation. This paper delves into an in-depth investigation of multivariate extreme wind load modeling under conditions of short-duration data. By introducing Kendall's tau to assess the nonlinear correlation of extreme wind loads, the correlation conversion between extreme wind loads at different epochs is derived. Furthermore, a method for reconstructing the joint distribution of multivariate extreme wind loads is proposed. Results show that the method proposed enables the modelling of multivariate extreme wind loads from short-duration data. When applied to wind-induced damage assessment of buildings, it reveals that the correlation of extreme wind loads can lead to a more dispersed and uncertain determination of the extent of building damage. The method proposed provides a rational consideration of extreme wind load correlation, contributing to the assessment of wind-induced damage of buildings.