Multi-week prediction of northern Australian summer rainfall bursts in ACCESS-S1

Multi-week prediction of northern Australian summer rainfall bursts in ACCESS-S1

Rainfall bursts are relatively short-lived events that occur over a timeframe of a few days up to a week, and over tropical northern Australia, they are strongly associated with broad-scale convective activity (Wheeler & McBride, 2005). Bursts are crucial for sustaining the pasture growth throughout the Oct-Apr wet season, allowing the livestock and agricultural sectors to prosper across northern Australia's semi-arid regions. Coinciding with a prototype burst forecast product release for the 2020/21 summer season, we evaluate rainfall bursts across northern Australia in gridded observations and calibrated multi-week hindcasts from the Bureau of Meteorology's seasonal prediction system, ACCESS-S1 (Australian Community Climate and Earth-System Simulator, Seasonal version 1). Here a burst is defined as when there is an accumulation of 30 mm over 3 consecutive days, similar to the "green break" rule used by northeastern Queensland graziers to mark the first decent rainfall of the wet season (Balston & English, 2009).

Based on observations, there has been a significant increase in the number of December burst days over northwest Australia, with now approximately three extra burst days today than in the 1960s. Across the tropical northeast, La Niña is associated with more early season burst days, whereas the conditions in the equatorial Pacific make very little difference in the number of burst days in January and February. There is a tendency for ACCESS-S1 to overestimate the total number of burst days for regions north of 20S, including positive biases in the proportion of summer rainfall from bursts as well as longer bursts. For assessing hindcast skill, we simplify our approach, focusing on the prediction of any burst event in the forecast period to match with what the information provided in the burst prototype product. We find, based on a Brier Skill Score, that ACCESS-S1 shows good skill in burst event prediction out to week 4 over the Top End and Cape York. Initial results also suggest there is improved skill in burst event prediction out to four weeks over the Top End alongside a prediction of strong Madden-Julian Oscillation (MJO) amplitudes. However, for the far northeast, there is better burst forecast skill for when weak MJO amplitudes are predicted. The ability of ACCESS-S1 to skillfully forecast bursts beyond three weeks would be of significant interest to northern Australia livestock producers, where important weekly management decisions on large cattle stations very much depends on accurate multi-week rainfall forecasts, particularly for wide-spread wetting events.

References:

Balston, J., & English, B. (2009). Defining and predicting the “break of the season” for north-east Queensland grazing areas. Rangeland Journal, 31(1), 151–159. https://doi.org/10.1071/RJ08054

Wheeler, M. C., & McBride, J. L. (2005). Australian-Indonesian monsoon. In Intraseasonal Variability in the Atmosphere-Ocean Climate System (pp. 125–173). Berlin, Heidelberg: Springer Berlin Heidelberg. https://doi.org/10.1007/3-540-27250-X_5

There are no files associated with this item.