Hydrometeor Identification Method and Its Application Using a Hierarchical Agglomerative Clustering Algorithm based on Cloud Radar Observations

To overcome the subjectivity and uncertainty inherent in hydrometeor identification based on fuzzy logic, this study proposes a novel identification method that integrates hierarchical agglomerative clustering (HAC) with fuzzy logic, tailored for Ka-band cloud radar observations. This method utilizes multiple radar-derived parameters—reflectivity (Z_H), radial velocity (V_R), spectral width (S_w), and linear depolarization ratio (LDR)—along with temperature from reanalyzed data to construct a multidimensional dataset. The hierarchical agglomerative clustering algorithm is first applied to classify hydrometeor phases and extract their physical characteristics. Based on these clustering results, the fuzzy logic method is used to identify the vertical distribution of hydrometeor phases within cloud systems. The classification reliability is then verified using two-dimensional cloud particle images obtained from airborne microphysical probes. The proposed method was applied to spring mixed-phase clouds in Central China and autumn stratiform clouds in Northeast China. The identified vertical distributions of ice crystals, snow, and mixed-phase hydrometeors in the negative temperature region were consistent with the observed evolution of macro- and microphysical cloud characteristics. The estimated heights of each hydrometeor phase agreed well with in situ airborne measurements, outperforming both the traditional fuzzy logic and the K-means clustering–fuzzy logic methods, and aligned with the physical laws of cloud development and precipitation formation.