The European Centre for Medium Range Weather Forecasting (ECMWF) and the UK Met. Office (UKMO) (also known locally as the Bracknell Fine Mesh (BFM)) models are widely used by forecasters and research meteorologists in New Zealand. It is therefore important that the performance of these two models in this region be evaluated. A summary of the performance of the ECMWF and BFM weather forecast models in forecasting systems that often produce heavy rains over New Zealand is presented here. For New Zealand these rain producing weather systems can be divided into five types, A) Low pressure systems that develop and/or traverse the North Tasman Sea., (B) frontal systems that bring strong Northwesterly flow across New Zealand, (C) “mesoscale or convective systems” in “less-organised” synoptic situations, (D) subtropical depressions, and (E) ex-tropical cyclones Unfortunately, for two of the storm types (mesoscale and ex tropical cyclones) there were too few cases within the study period (July 1998 – December 2000) to produce statistically useful results.

The results of our analysis can be briefly summarized as follows. Using bias and rms error scores the ECMWF model was judged to be more skillful in forecasting North Tasman lows, northwesterly fronts, and sub-tropical depressions. Errors in forecasting the strength of the systems contributed more to the rmse erros than it errors in storm location. Both models were on average positively biased (by about 0-2 hPa) for forecast hours 24, 48, and 72. However, when looking at errors in forecasting the central pressures of the lows (North Tasman and sub-tropical depressions only), it was found that the ECMWF had more of a tendency to over-deepen the lows, and on two occasions forecast bombs that did not occur. Examing, the distribution of location errors, it was found both models tended to shift the low pressure centres to the SE quadrant. Finally, there was a tendency (though not pronounced) that during specific storms for one model (after average biases were removed) to consistently out-perform the other.