Abstract
Rain nowcasting is an essential part of weather monitoring. It plays a vital role in human life, ranging from advanced warning systems to scheduling open air events and tourism. A nowcasting system can be divided into three fundamental steps, i.e., storm identification, tracking and nowcasting. The main contribution of this work is to propose procedures for each step of the rain nowcasting tool and to objectively evaluate the performances of every step, focusing on two-dimension data collected from short-range X-band radars installed in different parts of Italy. This work presents the solution of previously unsolved problems in storm identification: first, the selection of suitable thresholds for storm identification; second, the isolation of false merger (loosely-connected storms); and third, the identification of a high reflectivity sub-storm within a large storm. The storm tracking step of the existing tools, such as TITANand SCIT, use only up to two storm attributes, i.e., center of mass and area. It is possible to use more attributes for tracking. Furthermore, the contribution of each attribute in storm tracking is yet to be investigated. This paper presents a novel procedure called SALdEdA (structure, amplitude, location, eccentricity difference and areal difference) for storm tracking. This work also presents the contribution of each component of SALdEdA in storm tracking. The second order exponential smoothing strategy is used for storm nowcasting, where the growth and decay of each variable of interest is considered to be linear. We evaluated the major steps of our method. The adopted techniques for automatic threshold calculation are assessed with a 97% goodness. False merger and sub-storms within a cluster of storms are successfully handled. Furthermore, the storm tracking procedure produced good results with an accuracy of 99.34% for convective events and 100% for stratiform events.
Highlights
Long-range radars are used for weather monitoring, but these radars are power demanding, costly in terms of price and are not suitable for narrow valleys surrounded by high mountains, while X-band radars are quite efficient for monitoring localized rain fall events with small basins of interest
The multi-threshold procedure is adopted only for storm identification; for storm tracking and forecasting, the single threshold criterion is used for simplicity
The multi-level threshold for multi-level storm identification is separated by 5 dBZ
Summary
Long-range radars are used for weather monitoring, but these radars are power demanding, costly in terms of price and are not suitable for narrow valleys surrounded by high mountains, while X-band radars are quite efficient for monitoring localized rain fall events with small basins of interest. Like [4,5,6,7,8], are based on global thresholding. To overcome the problem of choosing a suitable initial threshold value, we have proposed fully-automated thresholding techniques (discussed in Sections 3.1.1 and 3.1.2) based on ThreshGW (G stands for Gonzalez and W for Woods) [9] and graythresh [10]. A multi-level thresholding technique, discussed, is used to overcome the problem of identifying sub-storms within a cluster of storm(s). Centroid-based algorithms can track individual storms more adequately and can provide more information about individual storms. Two storms in two consecutive time instances nearest to each other are candidates for matching Those algorithms that are based on the cross-correlation produce more accurate speed and direction information for large areas [8]. We have adopted first and second order exponential smoothing strategies in order to model our variables of interest
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
