In this project sponsored by the Managing Climate Variability Research and Development Program, we aim to provide skilful forecasts of the onset of the north Australian wet season. Wet season onset is defined to occur when a threshold rainfall accumulation of 50 mm is reached from 1st September. This amount has been shown to be useful for many agricultural applications as this is about the amount of rainfall that is required to generate new plant growth after the usually dry winter.
A statistical forecast system using logistic regression with the Southern Oscillation Index as the predictor was developed by Lo et al (2007), but not implemented operationally. Here we use the seasonal version of POAMA (P2.4) to predict wet season onset.
To develop and test the skill of the predictions we use hindcasts from POAMA for the 50 years 1960-2009. An ensemble of 10 hindcasts beginning on 1st July (Lead 2 months), 1st August (Lead 1) and 1st September (Lead 0) are generated using three different versions of the model (labelled as P24a, P24b, and P24c) for each of the 50 years. A separate climatology is calculated for each model version and for each lead time. This climatology reveals that all versions at all lead times have a dry bias over northern Australia. We correct for this dry bias by computing the probability of early onset relative to the corresponding model climatology. This is done by counting the number of early onsets in each model version, relative to the climatology of that version, and converting that number (out of 10) into a probability. These probabilities from each model version are also combined for a 30 member multi-model ensemble (MME).
These multi-member ensemble hindcasts, together with the verifying observed onset dates are available here. This includes the 2010 season, but this was not used in the verification since all the observed rainfall data was not available at time of the calculation.