Using Landmarks to Verify Complex Meteorological Phenomena

Precipitation forecasts from convection-resolving simulations are difficult to verify by any measure due to their complex, variable and ever shifting features.  Most pattern verification strategies take an Eulerian approach by identifying and matching specific events in the predicted and observed fields.  In this work, a more Lagrangian approach was taken by subjectively measuring forecast performance with respect to specific landmarks.  Locations were displayed on a map along with concentric range rings and radials, much like a radar display.  Observed and predicted events that directly affected each site were judged for the quality of the match in both time and space.  General information about the average storm cell size, precipitation intensity, mean direction of motion, and time lag was collected over a number of cases over Florida.  The results provided valuable insight about the nature and timing of convection, as well as the ability of the model to reproduce complex events such as secondary interactions between outflow boundaries.