No abstract available

<div>An automatic algorithm for classifying wind gust events has been developed at Geoscience Australia, utilizing 1-minute weather observations from Automatic Weather Stations (AWS). This algorithm employs a comprehensive dataset of wind, temperature, dew point, and pressure measurements within a two-hour timeframe centred on the peak wind gust.&nbsp;&nbsp;</div><div> The classification methodology effectively segregates wind gust events into convective and non-convective categories. Initial development entails a subset of stations, employing visual classification verified by contemporaneous observer reports and weather radar data, to create a robust training dataset. The algorithm, based on the analysis of almost 1000 visually-classified events, demonstrates the capability to classify over 150,000 events in a matter of minutes.&nbsp;</div><div> Utilizing wind gust events from past 20 years via our algorithm, the spatial distribution, diurnal cycle and seasonal variation are investigated across Australia. Moreover, a comparative analysis of spatial and temporal disparities, along with radar characteristics, has been conducted for convective and non-convective gust events. Finally, the extreme values of wind gust events, including the 1% annual exceedance probability wind speed (using the Generalized Pareto Distribution) across Australia is shown in this presentation. &nbsp;</div> Presented at the 30th Conference of the Australian Meteorological and Oceanographic Society (AMOS) 2024

<div>An automatic algorithm for classifying wind gust events has been developed at Geoscience Australia, utilizing 1-minute weather observations from Automatic Weather Stations (AWS). This algorithm employs a comprehensive dataset of wind, temperature, dew point, and pressure measurements within a two-hour timeframe centred on the peak wind gust.&nbsp;&nbsp;</div><div> The classification methodology effectively segregates wind gust events into convective and non-convective categories. Initial development entails a subset of stations, employing visual classification verified by contemporaneous observer reports and weather radar data, to create a robust training dataset. The algorithm, based on the analysis of almost 1000 visually-classified events, demonstrates the capability to classify over 150,000 events in a matter of minutes.&nbsp;</div><div> Utilizing wind gust events from past 20 years via our algorithm, the spatial distribution, diurnal cycle and seasonal variation are investigated across Australia. Moreover, a comparative analysis of spatial and temporal disparities, along with radar characteristics, has been conducted for convective and non-convective gust events. Finally, the extreme values of wind gust events, including the 1% annual exceedance probability wind speed (using the Generalized Pareto Distribution) across Australia is shown in this presentation. &nbsp;</div> Presented at the 30th Conference of the Australian Meteorological and Oceanographic Society (AMOS) 2024

<div>This database presents classified wind gust events for all Australian Automatic Weather Stations, based on semi-automatic classification of 1-minute observations of wind gust speed, temperature, dew point and station pressure. Wind events are classified based on the temporal evolution of the weather variables, using convolutional kernel transforms. Additional attributes include a number of derived variables (e.g. rainfall preceding and following the gust event), contemporaneous weather phenomena and binary classifications from a range of authors. </div><div><br></div><div>The main classification is described by Arthur, Hu and Allen (submitted to <em>Natural Hazards</em>, 2024). </div><div><br></div><div>Weather observation data are provided by the Bureau of Meteorology. Lightning data (2004-2024) was provided by TOA Systems Global Lightning Network. </div>