The accuracy of operational tropical cyclone (TC)prediction has improved considerably in recent years. For example, in the Australian region, mean operational forecast position errors were reduced below 200 km at 48 hours for the first time in the 1999/2000 tropical cyclone season. However, many tropical cyclones remain for which large forecast errors still occur. An entire tropical cyclone season can have a large mean error as a result of one or a few large errors in track prediction. Many of these larger errors occur as a tropical cyclone approaches or makes landfall, which is also the least desirable time for such errors to occur.

In this study we have examined the generation, characteristics and use of high spatial, temporal, and spectral resolution wind data in a 4D-VAR assimilation scheme to provide a better initial state for a high resolution NWP model. Initial results are also presented of the impact on precipitation forecasts of the 4D-VAR assimilation of radar data, and a new method for determining the error characteristics of atmospheric motion vectors is introduced. It was found that high resolution AMV data with appropriate error characterization, and used in a 4D-VAR scheme, considerably reduced tropical cyclone track forecast errors in an NWP model, for cases where large model errors occurred in operations. A total of 15 TCs was examined, and the high-resolution satellite wind data, assimilated using 4D-VAR, led to 48 hour track forecast errors considerably below the average for the operational forecasts of the tropical cyclones considered here. In particular, track errors were significantly below the average operational forecasts from the Australian Bureau of Meteorology.

We have also used these data in the assimilation and prediction system to assess the potential for improving TC intensity prediction. Forecasts for one case of a rapidly developing Category 5 tropical cyclone, TC Gwenda, are presented. The results are very encouraging with the intensity prediction matching closely the estimated central pressure of TC Gwenda.

Overall, the results have shown using high-resolution wind data and modern continuous assimilation techniques, in this case 4D-VAR, and modelling at an appropriately high resolution, provides significantly improved TC track and intensity forecasts. In this study the improved forecasts were for cases chosen because they had large operational forecast errors. In addition, it has also been shown that the assimilation of radar data can improve the initial and predicted rainfall rates and patterns for TCs and tropical storms. The improvement reported here for just two cases of short-term rainfall predictions has demonstrated such potential benefits for operations that a research initiative is currently underway to extend the work.