Generate seasonal climate outlooks and assess historical forecast skill.
Rainfall & temperature seasonal forecast
How Likely? is used to generate probabilistic seasonal rainfall and temperature outlooks for 3 to 6 month periods at user-defined locations, accompanied by an assessment of past forecast skill. A statistical methodology is used to calculate the forecast probabilities based on the understanding that there is a correlation between Australian climate and selected sea surface temperature data that replicates patterns of the El Niño Southern Oscillation and the Indian Ocean Dipole, and using data back to 1949.
Provides rainfall and temperature forecast probabilities for the next 3-6 months based on a “discriminant-analysis” (statistical methodology) of local climate and ocean temperature patterns. An estimate of forecast reliability (skill) is also provided.
How Likely? is designed for anyone who wants to explore seasonal (3 to 6 month) rainfall or temperature forecasts, and the historical skill of these forecasts for different periods of the year.
- What is the probability of temperatures being greater than the median in the next 6 months?
- What is the rainfall outlook for the next four months?
- What periods of the year are these forecasts most reliable?
The user selects (1) a variable (rainfall or monthly average temperature); (2) a conditional operator (greater than or less than); (3) a target statistic (median or tercile threshold); and (4) a duration of 3 to 6 months. How Likely? uses long term climate/weather data from SILO, and ocean temperature anomaly data from the Australian Bureau of Meteorology.
The results are presented in a range of graphical and textural formats focusing on the probability of an event occurring, and how reliable these predictions have been in the past. Forecasts are presented graphically using a “fire-gauge” chart to show the likelihood of a specific event occurring, while more traditional “chocolate-wheels” (tercile or medianal pie-charts) show a range of forecast probabilities for different rainfall categories.
In addition to generating a forecast, How Likely? provides an assessment of historical forecast “skill” for the current forecast period, as well as at other times of the year. Users can then assess how reliable these forecasts have been in the past, as well as identifying periods of the year where there is more skill. Given the statistical nature of the forecasts, it is possible that How Likely? can show an optimistic forecast associated with low or poor skill. Such a forecast should be viewed with caution as it has not been reliable in the past.
How Likely? is based on methodology developed by the Australian Bureau of Meteorology in their statistical operational forecast system. It has been adapted from stand-alone software SCOPIC (Seasonal Climate Outlooks in Pacific Island Countries, BOM) and FlowCast (Seasonal forecast analysis tool, Queensland Governemnt) for the iOS Smart Phone software, by RPS Australia East and DHM Environmental Software Engineering Pty Ltd for Managing CliMate Variability Research and Development Program.
Pie chart representation of risk builds on experience from RainMan and CliMate workshops and presented by SARDI. Dr Peter Hayman has been influential in this form of presentation.
SCOPIC – (Seasonal Climate Outlooks for Pacific Island Countries) from Australian Bureau of Meteorology. http://www.bom.gov.au/climate/pi-cpp/scopic.shtml
FlowCast (2009) from Queensland Government, Queensland Climate Change Center or Excellence.