Improving the nowcasting of precipitation in an Alpine region with an enhanced radar echo tracking algorithm

Nowcasting for hydrological applications is discussed. The tracking algorit hm extrapolates radar images in space and time. It originates from the patt ern recognition techniques TREC (Tracking Radar Echoes by Correlation, Rine hart and Garvey, J. Appl. Meteor.. 34 ( 1995) 1286) and COTREC (Continuity of TREC vectors, Li et al., Nature, 273 (1978) 287). To evaluate the qualit y of the extrapolation. a parameter scheme is introduced, able to distingui sh between errors in the position and the intensity of the predicted precip itation. The parameters for the position are the absolute error, the relati ve error and the error of the forecasted direction. The parameters for the intensity are the ratio of the medians and the variations of the rain rate (ratio of two quantiles) between the actual and the forecasted image. To ju dge the overall quality of the forecast, the correlation coefficient betwee n the forecasted and the actual radar image has been used. To improve the forecast, three aspects have been investigated: (a) Common m eteorological attributes of convective cells, derived from a hail statistic s, have been determined to optimize the parameters of the tracking algorith m. Using (a), the forecast procedure modifications (b) and (c) have been ap plied. (b) Small-scale features have been removed by using larger tracking areas and by applying a spatial and temporal smoothing, since problems with the tracking algorithm are mainly caused by small-scale/short-term variati ons of the echo pattern or because of limitations caused by the radar techn ique itself (erroneous vectors caused by clutter or shielding). (c) The sea rching area and the number of searched boxes have been restricted. This lim its false detections, which is especially useful in stratiform precipitatio n and for stationary echoes. Whereas a larger scale and the removal of smal l-scale features improve the forecasted position for the convective precipi tation, the forecast of the stratiform event is not influenced, but limitin g the search area leads to a slightly better forecast. The forecast of the intensity is successful for both precipitation events. Forecasting the vari ation of the rain rate calls for further investigation. Applying COTREC imp roves the forecast of the convective precipitation, especially for extrapol ation times exceeding 30 min. (C) 2000 Elsevier Science B.V. All rights res erved.