NMOC Quarterly Summary April-June 2002

Summary of System Performance
Wave Forecasting Verifications

Analyses & Numerical Prediction | About Products | Map/Image/Chart Archives

The following diagrams show the verification of the significant wave height forecasts from the operational sea state models at wave recording sites around the Australian coast for the period April-June 2002.

Site details

Chart details

The first series of graphs shows the mean bias (average error) and RMS error of the global (1° resolution), Australian region (1° resolution) and the SE mesoscale region (0.25° resolution) wave models at Sydney and Cape Naturaliste averaged over the three months. Also included for comparison are the verifications of the UK wave model (blue diamonds). At Cape Naturaliste the major contribution to the error is the negative bias in forecasting of the extreme events whereas at Sydney there is very little bias evident over this period. At sites exposed to the mid-lattitude westerlies (such as Cape Naturaliste) the local models perform slightly worse than the UK model whereas the performance of all models is similar on the east coast.

The following charts show the performance of the 12 hour and 24 hour forecasts of the operational models (glbass = Global Model, regass = Australian region model, mesass = SE mesoscale model) and the UK wave model (UK_glo) together with wave height observations on a daily basis at all of the above sites.

At Sydney there is very good agreement for the whole of the period dislpays. Two major events are evident. The storm of June 18, caused by a strong south-easterly stream onto the New South Wales coast, was slightly underforecast by the Australian region model and the event of June 29, caused by an intense low pressure system moving across the Tasman sea, was extremely well forecast by all models.

At Cape Sorell the phase of the predictions was generally quite good although the local models tended to underpredict the wave heights. The major event of the period on June 10 which was caused by a very intense low pressure system south of Tasmania and associated strong westerly winds was very well forecast by all models.

At Cape Naturaliste the phase of predictions is again well forecast but the intensity is underpredicted by all models. The major events of 7.5 metres on May 7, caused by an intense low pressure system moving to the south of Western Australia, and 7.3 metres on 11 May, caused by the passage of a frontal system through the south west were not forecast well by the local models.

The final graph shows the number of observations per month and the verification of the 24 hour forecast from the Australian Region Wave Model since the beginning of 1999.

Running monthly statistics and time series plots of model performance at all the verifying sites are available on the Bureau's internal web site.

Wind Verification

The surface wind fields from the various models used to force the wave models have been compared with observations reported from moored and drifting buoys. The results at the North Rankin Oil Platform Station (Woodside Petroleum Ltd - WMO number 56002) are presented in the following tables and graphs.

GASP vs Station 56002

Start Date: 1-4-2002
End Date: 30-6-2002

Fcast Int	Bias (m/sec)	RMS Error (m/sec)	Correlation Coeff
0	-1.64	2.66	0.76
12	-1.38	2.49	0.77
24	-1.26	2.50	0.74
36	-1.27	2.51	0.74
48	-1.10	2.66	0.67

LAPS PT375 vs Station 56002

Start Date: 1-4-2002
End Date: 30-6-2002

Fcast Int	Bias (m/sec)	RMS Error (m/sec)	Correlation Coeff
0	-1.08	2.55	0.70
12	-1.20	2.34	0.77
24	-1.29	2.54	0.73
36	-1.48	2.67	0.72
48	-1.57	2.84	0.68

MESO LAPS PT125 vs Station 56002

Start Date: 1-4-2002
End Date: 30-6-2002

Fcast Int	Bias (m/sec)	RMS Error (m/sec)	Correlation Coeff
0	-0.45	2.35	0.73
12	-1.01	2.25	0.78
24	-1.10	2.42	0.74
36	-1.21	2.57	0.71

For this period all models are exhibiting a negative bias at this site, with the errors from the global model comparable to those from the limited area models. The low wind bias in the models has been under investigation by NMOC and the modeling group in BMRC and results are discussed in a brief report in this bulletin.