**4. Discussion**

The method we chose, to distinguish between NL and FL, was based on the distance of observed lightning from the radar site. The results showed that the method can be applicable in general, however, it would be better if we could also identify in which part of the storm the vertical radar measurements are taken. We plan to focus on an analysis of the possibility to identify the position of radar within storms in future. Specifically, our aim is to study whether the radar is located on the frontal or back side of a storm or on its lateral sides. Such an analysis, however, needs a wider dataset of thunderstorms registered close to the radar, which we expect to obtain in future.

At present, the amount of data during two years of radar operation that meet the condition for NL is not sufficient to allow us to divide data into further subsets, which is necessary for the identification of where—in the thunderstorm—the radar is located. The division of the current dataset into further subsets would not lead to sufficiently robust results. Moreover, determining the position of the radar within a thunderstorm is not trivial and needs thorough investigations.

The distribution of data and the method of processing (averaging) inevitably led to a smoothing of AV. This smoothing (averaging) of vertical velocity resulted in low values of AV, since the dataset comprised both positive and negative velocities, which, after averaging, became naturally low. Nevertheless, regardless of AV smoothing, the profiles of mean AV for NL as compared to that for FL qualitatively correspond to our knowledge. During the mature state of a thunderstorm, (strong) positive as well as negative vertical velocities are supposed to be observed.

The applied technique for the recognition of hydrometeors (Hclass) provides meaningful results, although, of course, we cannot perform its exact verification. However, the structure of hydrometeors seems right during the maximum activity of the thunderstorm on 10 June 2019 (around 21:40 and later). At that time, there was a noticeable occurrence of all types of hydrometeors, including hail and supercooled cloud water throughout the vertical profile. The results obtained by processing all thunderstorms were also reasonable and explainable.

As far as the supercooled cloud water is concerned, there are no given rules on how to determine it exactly. In this study, we used a simple algorithm that allows the supercooled water to occur up to −40 ◦C under the condition that there are small values in the measured Doppler spectra and there is an updraft of the air motion. We are aware that such an identification of supercooled cloud water is burdened with uncertainties. However, we consider our results satisfying.

We studied how the LDR differs for stormy areas (NL) and non-stormy areas (FL) and in our opinion, we showed, in agreement with other works e.g., [25], that a strong electric field in a thundercloud can be identified by high LDR values. This opinion is based on the fact that the increased averages of LDR at altitudes of 4 to 6.5 km were reflected for NL only. Our analysis showed that it is unlikely that the increase in LDR would be solely related to the presence of hail or graupel. On the contrary, we believe that two processes occur almost simultaneously: (i) electric field formation due to collisions of graupel and ice particles and (ii) the alignment of ice particles in the electric field leading to high LDR. It should be mentioned that aligned targets may cause characteristic signatures in the differential phase between co- and cross-channel IQ signals, but the interpretation of these signatures is difficult and we would like to address them in the future.
