Home | About Us | Contacts | Help | Feedback | SEARCH Global | Australia | NSW | Vic. | Qld | WA | SA | Tas. | ACT | NT | Ant. | Weather & Warnings | Hydrology | Climate | Numerical Prediction | About Services | Learn About Meteorology | Registered User Services |

WARNING

The following are experimental results from research systems developed in BMRC. These results highlight some of our current work and should be of interest to other researchers. They are not part of the Bureau of Meteorology's products and services. Only moderate skill may be expected from the forecast displayed.

---

• CURRENT BMRC FORECAST
• HISTORICAL SKILL OF FORECAST
• VERIFICATION OF OPERATIONAL FORECASTS
• SCIENTIFIC DOCUMENTATION PLUS FUTURE DEVELOPMENT
• SUMMARY OF EIGHT INTERNATIONAL FORECASTS

BACKGROUND

A coupled ocean-atmosphere model for the prediction of eastern tropical Pacific ocean temperatures has been developed by Dr. Richard Kleeman and co-workers within the Bureau of Meteorology Research Centre in Melbourne Australia. Warmer than normal temperatures are usually referred to as El Nino while cooler than normal conditions are referred to as La Nina. The index NINO3 is used to measure these temperature deviations. If the index is greater than 1.5 then strong El Nino conditions are said to exist. Conversely values less than -1.5 indicate that strong La Nina conditions exist. Within Australia, El Nino conditions are normally associated with drought conditions in eastern Australia particularly in inland Queensland and NSW. The reverse applies during La Nina.

CURRENT FORECAST

The vertical yellow line indicates when the forecast begins. Ocean thermal data is used up to and including the previous month. Wind data is used up to this month which is the month before the forecast is issued. The assimilation period shows the ingestion of past data while the forecast period shows the coupled model prediction for the next two years. A postscript version of this forecast can be found here here

 The current forecast shows cool conditions over the next year.

SKILL OF FORECASTS

The skill of the coupled model is comparable with the best models currently available internationally. It has been tested by making forecasts every 3 months for the period 1982 to 1991 inclusively. The resulting forecasts were compared with the observations and the correlation and root-mean-square (RMS) error calculated for forecast lead times out to 23 months. For comparison the skill of persistence is also calculated. A persistence forecast is the assumption that the current temperature deviation will persist into the future. The correlation skill estimate can be found here, while the RMS error of the model is here. A postscript version of these skills can be found here here.

VERIFICATION OF FORECASTS

The current configuration of the coupled model has been run routinely every month since the middle of 1994. In order to check the recent performance of the model the six and nine month predictions made since 1994 have been verified against the observed values. The six month prediction verification may be found here, while the corresponding nine month verification is here.

SCIENTIFIC DOCUMENTATION AND FUTURE DEVELOPMENTS

The coupled model is described in the following two places:

Kleeman R., 1993: ``On the dependence of hindcast skill on Ocean Thermodynamics in a Coupled Ocean Atmosphere Model'', Journal of Climate , Vol 6, pp2012-2033.

Kleeman R., A.M.Moore and N.R.Smith, 1995: ``Assimilation of sub-surface thermal data into a simple ocean model for the initialisation of an intermediate coupled ocean-atmosphere forecast model'', Monthly Weather Review , Vol 123, pp3103-3113.

A further list of internationally refereed publications by Richard Kleeman, some of which are relevant to the forecast model, may be found here.

In the future we intend to introduce a measure of the reliability of a given forecast. This will be based on the stability of the forecast initial conditions to optimal perturbations. More details can be found in:

Moore A. M. and R. Kleeman, 1996, ``Skill Assessment for ENSO using Ensemble Prediction'', submitted to Q. J. R. Met. Soc..

In this manuscript we show that there is a relationship between the spread of forecasts which have their initial conditions optimally perturbed and the skill of the forecast without perturbation. A plot of this relation can be found here. Note that MAD is the mean average deviation of the ensemble of predictions subject to initial condition perturbation while the skill shown refers to the correlation skill averaged over a two year forecast. All quantities are calculated using the NINO3 index and the contouring and shading is for probability density.