The synoptic pattern over northeastern Australia is dominated in the warmer months by a ridge-trough system. Accurate prediction of the location of the system is a significant forecasting problem for regional and global operational models. The regional model which was operational at the time of this study exhibited two significant weaknesses characteristic of current operational models: a westward bias in the location of the coastal ridge; and errors in the location and strength of the inland trough. The present investigation had two aims: first, to compute model location errors of the ridge-trough system, from a large (six-month, twice-daily) data set of operational forecasts: the second, to explain these errors by evaluating a new regional model which has more accurate numerics and an enhanced representation of physical processes. The operational model had a marked mean westward bias of about 2 degrees of longitude in the location of both the trough and the ridge. Moreover, there was a noticeable latitudinal distribution in the trough errors, with the greatest errors in the north. Ridge location errors were much larger in the south. Overall,almost 60 per cent of errors were 2 degrees of longitude or greater. The new model was far more skillful in forecasting the ridge- trough system, with predicted locations of the both the ridges and the troughs being superior at greater than the 99 per cent confidence level. In the new model a mean westward error remained in the location of the ridges and troughs, but it was less than 1 degree, and the percentage of errors greater than 2 degrees of longitude dropped to about 20 per cent for the ridges and about 35 per cent for the troughs. Better representation of the steep coastal orography, and improved simulation of the heat low to the west of the coastal ranges, seem to be responsible for the decreased location errors in the new model. This was confirmed in three distinctly different case studies, at very high resolution (15km), using the new model but with the same operational data