Understanding past changes in Australian droughts and their drivers

Australia is frequently exposed to considerable impacts from severe and widespread droughts. Despite this, a comprehensive understanding of the past trends and drivers of Australian droughts remains elusive. Existing studies have often characterised past trends based on changes in mean values rather than the extremes. However, given Australia’s exceptionally variable climate, this may fail to capture the full nature of the country’s drought trends. Furthermore, studies often rely on a limited number of drought indicators and may not encompass the diverse meteorological, hydrological and ecological conditions contributing to drought.

This work explores past drought trends in Australia using multiple drought indicators. We analyse changes in traditional drought metrics, including precipitation, runoff and soil moisture, defining droughts as time periods below the 15^th percentile. We complement these metrics with an impacts-based drought indicator built from government drought reports using machine learning. We explore the drivers of drought trends using explainable machine learning methods, and consider multiple drivers including large-scale climate features, land surface properties and past conditions. Using a diverse range of metrics allows for a more comprehensive analysis of drought changes experienced over the past decades and will provide greater insight into the main drivers behind Australian droughts.