The assimilation of extrapolated radar reflectivity using a new precipitation nudging scheme for improving short-term quantitative precipitation prediction in New Zealand

This paper demonstrates a method for improving short term quantitative precipitation forecasting (QPF) in New Zealand (NZ). Firstly, the Kessler warm rain processes were implemented within the Weather Research and Forecasting (WRF) Model Nudging System (K-Nudging) in order to improve QPF in NZ by making use of the observed high resolution radar reflectivity. The new scheme adjusts the model specific humidity vertical profiles, which are calculated from the equations derived from the Kessler warm rain processes based on observed precipitation, at each time step within an assimilation time window. Secondly, in order to provide more accurate model backgrounds for the first several hours' forecasts, the performance of assimilation with extrapolated radar reflectivity over different time scales (1, 3, 6 h) in NZ has been investigated by using the K-Nudging scheme.

A total of 17 cases with heavy precipitation occurred in NZ during the summer of 2011/2012 (Nov 2011 to Jan 2012) were selected for analysis in this paper. Synoptic analysis indicates that most of these events either resulted from an air mass developing in the Tasman Sea or are generated locally by the orographic impacts of South Alps, which means that these 17 cases have covered most common heavy precipitation generation mechanisms that occur during a NZ summer.

The results show that the accuracy of forecasts with the K-Nudging assimilation of observed reflectivity is improved on average for 6-9 hours ahead compared to the forecasts initialized using 3DVAR method and extrapolation (Cross-correlation) method. Furthermore, By assimilating extrapolated radar reflectivity with the K-Nudging, the results suggest that (1) the accuracy of forecasts largely depends on the accuracy of extrapolated reflectivity; (2) Generally, the accuracy of short term QPF could be further improved significantly by the assimilation of 1-3 h extrapolated radar reflectivity compared to the assimilation with observations only at the observation time.

However, all statistical scores show that, although radar observed/extrapolated reflectivity could be assimilated at the same resolution which the model is actually run, simple radar extrapolation scheme still should play an important role in 1-3 h precipitation nowcasting in operational applications.