Hydrometeor classification from polarimetric radar measurements:a clustering approach

[摘要] A data-driven approach to the classification of hydrometeors frommeasurements collected with polarimetric weather radars isproposed. In a first step, the optimalnumber ofhydrometeor classes (n_opt) that can be reliably identified from a large setof polarimetric data is determined. This is done by means of anunsupervised clustering technique guided by criteria related both todata similarity and to spatial smoothness of the classifiedimages. In a second step, the n_opt clusters are assignedto the appropriate hydrometeor class by means of humaninterpretation and comparisons with the output of otherclassification techniques.The main innovation in the proposedmethod is the unsupervised part: the hydrometeor classes are notdefined a priori, but they are learned from data.Theapproach is applied to data collected by an X-band polarimetricweather radar during two field campaigns (from which about 50 precipitation eventsare used in the present study).

Seven hydrometeor classes (n_{opt = 7}) have been found in thedata set, and they have been identified as light rain (LR), rain(RN), heavy rain (HR), melting snow (MS), ice crystals/smallaggregates (CR), aggregates (AG), and rimed-ice particles (RI).

[学科分类] 几何与拓扑