LAUSR

Abstract As the global wind engineering community works towards the determination of a robust and accurate thunderstorm wind model to be implemented in national wind loading codes, there is a need to identify not only the regions most impacted by extreme non-synoptic winds, but the magnitude and frequency of such events. The Brazilian wind code, NBR 6123 (ABNT 1988), is currently undergoing a process of revision, allowing for the opportunity to study historical meteorological data with the aim to propose basic wind speed maps for both synoptic and non-synoptic winds. The separation of extreme wind events into classes is not a simple task and is made difficult by the presence of false observations which often take the form of sudden and sharp rises in wind speed, not dissimilar to real thunderstorm events. This study details the datasets and methodologies used in the creation of algorithms which were used to separate extreme wind events into one of three classes: non-synoptic, synoptic and suspect. Challenges and limitations faced by the implementation of such algorithms are identified and recommendations are made to potential future improvements.