McGill Algorithm for Precipitation nowcasting by Lagrangian Extrapolation (MAPLE) applied to the South Korean radar network. Part 1: Sensitivity studies of the Variational Echo Tracking (VET) technique

A Variational Echo Tracking (VET) technique has been applied to four months of archived data from the South Korean radar network in order to examine the influence of the various user-selectable parameters on the skill of the resulting 20-min to 4-h nowcasts. The latter are computed over a (512x512) array at 2-km resolution. After correcting the original algorithm to take into account the motion of precipitation across the boundaries of such a smaller radar network, we concluded that the set of default input parameters initially assumed is very close to the optimum combination. Decreasing to (5x5) or increasing to (50x50) the default vector density of (25x25), using two or three maps for velocity determination, varying the relative weights for the constraints of conservation of reflectivity and of the smoothing of the velocity vectors, and finally the application of temporal smoothing all had only marginal effects on the skill of the forecasts. Changing the time interval between maps, from 20 to 10 minutes, and significantly increasing the reflectivity threshold from 15 to 30 dBZ had a more noticeable reduction on the forecast skill. Comparisons with the Eulerian “zero velocity” forecast and with a "single" vector forecast have also been performed in order to determine the accrued skill of the VET algorithm. Because of the extensive stratiform nature of the precipitation areas affecting the Korean peninsula, the increased skill is not as large as may have been anticipated. This can be explained by the relative reduced size of the radar coverage domain as compared to the coverage of the precipitation systems.