Forecasting the extreme rainfall, low temperatures, and strong winds associated with the northern Queensland floods of February 2019

Abstract

From late January to early February 2019, a quasi-stationary monsoon depression situated over northeast Australia caused devastating floods, killing an estimated 625,000 head of cattle in northwest Queensland, and inundating over 3 000 homes in the coastal city of Townsville. The monsoon depression lasted ~10 days, driving daily rainfall accumulations exceeding 200 mm/day, maximum temperatures 8–10 °C below normal, and wind gusts above 70 km/h. In this study, the atmospheric conditions during the event and its predictability on the weekly to subseasonal range are investigated. Results show that during the event, the tropical convective signal of the Madden-Julian Oscillation was over the western Pacific, and likely contributed to the heavy rainfall, however the El Niño-Southern Oscillation was not in the usual phase for increased rainfall over Queensland. Over the northern Tasman Sea, an anticyclone helped maintain a positive phase of the Southern Annular Mode and promote onshore easterly flow. Somewhat consistent with these climate drivers, the monthly rainfall outlook for February issued by the Australian Bureau of Meteorology on 31 January provided no indication of the event, yet forecasts, not available to the public, of weekly-averaged conditions by the Bureau's dynamical subseasonal-to-seasonal (S2S) prediction system were more successful. For the week of 31 January to 6 February the prediction system forecast a more than doubling of the probability of extreme (highest quintile) weekly rainfall a week prior to the event, along with increased probabilities of extremely low (lowest quintile) maximum temperatures and extreme (highest quintile) wind speeds. Ensemble-mean weekly rainfall amounts, however, were considerably underestimated by the prediction system, even in forecasts initialised at the start of the peak flooding week, consistent with other state-of-the-art dynamical S2S prediction systems. Despite this, one of the individual ensemble members of the Bureau's prediction system did manage to forecast close to 85% of the magnitude of the rainfall across the most heavily impacted region of northwest Queensland a week before the event. Predicting this exceptional event beyond two weeks appears beyond our current capability despite the dynamical system forecasts showing good skill in forecasting the broad-scale atmospheric conditions north of Australia a week prior. Keywords: Extreme, Rainfall, Temperature, Flood, Forecast, Skill, Madden-Julian Oscillation, Subseasonal