Mesoscale Verification

LONG-TERM GOALS

The APL atmospheric sciences group is working to improve forecaster performance at Navy operational weather forecast detachments afloat and ashore. This work encompasses broad research and technology development in areas of visualization, human factors, human-machine interaction, and model and forecast verification with an emphasis on mesoscale ensembles and visualization of uncertainty. The verification effortÕs long-term goal is to develop an automated, objective verification technique for assessment of very high-resolution mesoscale predictions which accurately accounts for spatially or temporally misplaced features, false alarms and misses (Brown 2002).

OBJECTIVES

The overall objective of this effort is to develop a highly automated, rapid, object-oriented, mesoscale numerical weather prediction (NWP) verification tool for use by forecasters and model developers. The verification technique should consider distortion errors (phase/timing, rotation, and stretching) as well as the normal amplitude errors. It is intended to test the verification tool on the University of Washington Short Range Ensemble Forecast System (SREF) and with a version of the Navy COAMPS model implemented at APL.

APPROACH

Recently, a number of techniques from spatial statistics have shown promise in verification problems (Marzban and Sandgathe 2005; Harris and Foufoula-Georgiou 2005). Among them there exists one technique which not only allows for a novel verification method, but also addresses a central issue in our mesoscale verification technique (MVT) - the "box size" (or characteristic length scale as discussed in Hoffman et al 1995 and Nehrkorn et al 2003). The tool is the variogram and our proposal is to use the variogram to address the length-scale issue, both for verification and to address the 'feature' background error covariance.