P72     Upgrading from WRF 3.1.1 to WRF 3.4.1 operationally at Metservice:  Comparison of the two versions for eleven significant weather events

 

Davis, Cory, Dominikus Heinzeller, Graham Rye, Paul Shucksmith, Peter Kreft, John Crouch, Heath Gullery, Paul Mallinson, Chris Noble, Liz Walsh, Brooke Lockhart, and Chelsea Russ, Metservice

 

At Metservice we run 24 WRF models over New Zealand, Australia, UK, Europe and

Asia, at various horizontal resolutions, using GFS, ECMWF, and UKMO NWP

data for initial and boundary conditions. Since 2009, we have been using a

slightly modified version of WRF 3.1.1. To benefit from recent improvements in

the WRF system, and to enable more effective research on improving our

implementation of WRF, we upgraded to version 3.4.1 operationally in May 2013.

A critical application of our WRF output is its use as guidance by all forecasting desks (severe weather, public, marine, and aviation) in our National Weather Service. It was necessary to get forecaster feedback and acceptance before making the upgrade operational. This was done by running the new and old versions over 11 historical cases that were selected by forecasters. To ease the transition to

the new version, no changes were made to physics options. The historical cases

included severe air mass convection and embedded convection events, broad scale

wind, rain, and snow events, marginal wind and rain events, and a major fog

event. A customized imagery viewer was created, allowing easy comparison, and

forecasters from all desks were invited to comment on an internal wiki page. Overall the new version was found to better represent the cases considered, although in an airmass convection case, the low level wind field was significantly different in the new version, misplacing an area of convergence, and removing the signal of a severe convection event. Differences in the low level wind field were noticeable in several other cases.

 

This poster describes the historical cases, the configuration of WRF, and the

relative performance of versions 3.1.1 and 3.4.1.