**2. Materials and Methods**

#### *2.1. GPM Satellite and Precipitation Data*

GPM DPR was designed by the Japanese Aerospace Exploration Agency (JAXA) and the National Institute of Communication Technology (NICT). It consists of a Ku-band precipitation radar (KuPR, 13.6 GHz) and a Ka-band precipitation radar (KaPR, 35.5 GHz). The detection area can cover 65◦S–65◦N on the Earth [24–26]. The KuPR has only one scanning mode (normal scan, NS), and its minimum detectable reflectivity is 14.5 dBZ. The KaPR has two scanning modes, matched scan (MS) and high-sensitivity scan (HS). The minimum detectable reflectivity is 16.7 dBZ and 10.2 dBZ, respectively [27]. The dropletsize parameters of hydrometeors can be inverted by different responses of the KuPR and the KaPR to hydrometeors [28]. Using ground-based observations, scholars have verified the reliability of the droplet-size distribution retrieved by DPR [29]. We used the GPM dual-frequency precipitation product 2ADPR from 2014 to 2019, including the rain type, near-surface rain rate, storm-top height, and droplet-size distributions (DSDs). Above the melting layer, for mixed-phase and solid-phase hydrometeors, the liquid-equivalent DSDs were retrieved from the official 2ADPR dataset [30].

### *2.2. GPM Products for Rain Types*

The variable *typePrecip* in the Classification Module (CSF Module) of GPM data provides rain-type classification made by various methods: the single-frequency horizontal pattern method (H-method), vertical profiling method (V-method) and measured dualfrequency ratio method (DFR*m*-method) [19]. In the V-method, the melting-layer bright band (BB) is first detected. The detection of the BB is to determine whether the vertical profile of radar reflectivity satisfies certain conditions which are typical for the profile of the radar reflectivity factor when a BB exits by examining the vertical profile of the radar reflectivity factor. When a BB is detected, if the reflectivity factor in the rain area does not exceed the convection threshold (46 dBZ), then the rain type is stratiform. When no BB is detected and the reflectivity factor exceeds the conventional convective threshold (40 dBZ), the rain type is convective. If rain type is neither stratiform nor convective, the rain type is other. In the H-method, the horizontal distribution of representative radar reflectivity factors, i.e., the maximum value of the reflectivity factor along the considered radar beam in the rain region, is detected. Rain-type classification adopts a modified University of Washington convective/stratiform separation method, which is divided into three categories: stratiform, convective, and other. Detection of convective precipitation is made first. The rain type is stratiform if it is not convective, unless the reflectivity factor is very small and has almost identical noise. If rain type is neither stratiform nor convective, it is other.

For the same precipitation profile, due to the different rain-type classification methods, the results may be different. Therefore, the variable *typePrecip* unifies rain types by the above classification methods, that is, the single-frequency method and the dual-frequency method are used to unify the rain types. In the dual-frequency method, the DFR*m*-method is used to detect the BB, which classifies rain into three types: stratiform, convective, and transition. The dual-frequency method merges the rain type of the DFR*m*-method with the single-frequency Ku-band rain type, and it outputs a unified rain type: stratiform, convective, and other. The features for the unification of the dual-frequency method are as follows: When the DFR*<sup>m</sup>* rain type is convective or stratiform, if no BB is detected, the unified rain type follows the DFR*<sup>m</sup>* rain type. If a BB is detected, however, the unified rain type is stratiform. When the DFR*<sup>m</sup>* rain type is transition or the DFR*<sup>m</sup>* processing is skipped, the single-frequency Ku-band rain type is the unified rain type. If heavy ice precipitation (HIP) or winter convection is detected, some stratiform type is changed into a convective type.

Since the rain-type identification by the DFR*m*-method considers observations from two bands, the rain types are probably more reliable. Therefore, the rain types in the present study are retrieved from the dual-frequency method. Figure 1 shows the spatial distribution of 10.4 μm brightness temperature from the Japanese Himawari-7 meteorological satellite, as well as the rain type, storm-top height, and radar echo at 6 km height from GPM DPR for the NCCV at 1200 UTC on 12 June 2014. The brightness temperature of the cloud system in the region of 24–28◦N and 136–140◦E is low (Figure 1a), and the radar echo intensity at 6 km is higher than 20 dBZ (Figure 1d). However, this region is classified as stratiform by GPM in Figure 1b. From the vertical profile of the radar echo of the NCCV (Figure 2), there is a clear BB inside the cloud system in this region, and the echo intensity in the BB is larger than that in the upper and lower layers. The BB is the main feature of the stratiform precipitation echo, indicating that the airflow in the stratiform precipitation is stable and there is no obvious convective activity. For the convective precipitation on the southeastern edge of the stratiform region, the storm-top height is less than 5 km. It can be seen in Figure 2d that there is no BB in the precipitation system, and there is a strong echo higher than 40 dBZ in the radar reflectivity factor. Therefore, the rain types from the dual-frequency method in GPM have relatively high reliability from the vertical and horizontal distribution of radar echoes.

**Figure 1.** (**a**) The 10.4 μm infrared brightness temperature of the Himawari-7 satellite (shading, k), (**b**) the rain type identified by GPM DPR (shading, unitless), with yellow (blue) representing convection (stratiform), (**c**) the storm-top height (shading, km), and (**d**) the reflectivity at 6 km height (shading, dBZ) within 2000 km distance of the NCCV center at 1200 UTC on 12 June 2014 with GPM orbit No.001631. (The purple lines represent the swath of GPM DPR; the '+' represents the NCCV center; and lines AB, CD, EF, and GH represent the section position in Figure 2).

**Figure 2.** Vertical sections of the radar echo along lines (**a**) AB, (**b**) CD, (**c**) EF, and (**d**) GH in Figure 1d. The *X* axis represents the latitude along the cross-section direction. The yellow and blue dots at the 11 km altitude represent the convective and stratiform precipitation retrieved from the dual-frequency method in GPM, respectively.

#### *2.3. Reanalysis Data*

The 500 hPa geopotential height fields are extracted from the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5) data with a temporal resolution of 1 h and a spatial resolution of 0.25◦ × 0.25◦ [31]. They were used to correct the initial 6 h NCCV centers provided by Chen et al. [1] which were identified from a reanalysis dataset with a temporal resolution of 6 h and a spatial resolution of 1◦ × 1◦. Firstly, NCCV centers at hourly resolution are obtained from the initial 6 h ones by a linear interpolation method. Secondly, the 1◦ resolution NCCV centers are refined to 0.25◦ resolution using the ERA5 500 hPa geopotential height field. The NCCV center is corrected by the lowest geopotential height within 1◦ from the original NCCV center. In this way, we can obtain the NCCV centers at hourly resolution, and match the NCCV centers with GPM data at the same time.
