Original paper
Localization and flow-dependency on blending techniques
Atencia, Aitor; Wang, Yong; Kann, Alexander; Meier, Florian
Meteorologische Zeitschrift Vol. 29 No. 3 (2020), p. 231 - 246
36 references
published: Oct 16, 2020
published online: Dec 17, 2019
manuscript accepted: Oct 17, 2019
manuscript revision received: Sep 27, 2019
manuscript revision requested: Sep 2, 2019
manuscript received: Jul 3, 2019
Open Access (paper may be downloaded free of charge)
Abstract
Lagrangian extrapolation nowcasting is still the most accurate technique for predicting precipitation for the first hours despite the improvements achieved by data assimilation in high-resolution numerical weather prediction models. Yet, the combination of both sources by blending techniques is necessary for creating a seamless forecast for the nowcasting range. The blending is usually carried by adding one forecast to the other one multiplied previously by a weight. This weight can be computed climatologically or taking into account parameters such as the synoptic situation or the intensity of precipitation from the different sources of forecast. In this paper, a general formulation for computing the blending weight is introduced. This new formulation allows not only to introduce new factors such as the localization (pixel scale) but also to isolate the effect of some parameters such as the flow-dependency (synoptic situation). A sensitivity study of these parameters, separated and combined, is carried out for a whole month. The results show the relative importance separately of each effect (localization and flow-dependency) but, most importantly, the improvements in terms of combining the two sources of information when these two factors are used together in the computation of the blending weights.