The sensitivity of object-based nowcasts to object threshold selection

Nowcasts that involve the identification and tracking of “objects” may be sensitive to the choice of threshold used to delineate an object. This paper explores the variability in forecast by looking at a number of cases using two nowcast schemes. These are the K-means cluster method used in the WDSS – II (Lakshmanan et al. 2003) and the Spectral Prognosis scheme (Seed 2003). A number of cases are examined representing a range of storm types.

It is found that the choice of reflectivity threshold used to define a cell can significantly impact the quality of a forecast, and that adapting the threshold value to the meteorological situation or storm-type can have benefits. Alternatively using a range of different threshold values to nowcast the same storm can efficiently generate an ensemble of nowcasts which can produce a robust product that captures some measure of the uncertainty in the storm track.