*3.3. LDR during Analyzed Thunderstorms*

Averages of LDR are depicted in Figure 10 for NL, as compared to FL. For both NL and FL, the melting layer is not pronounced in LDR averages; there is no obvious increase in LDR averages in lowest gates. This is very likely related to the fact that the height of the melting layer depends on current atmospheric conditions, which change from one thunderstorm to another and might also change during one particular thunderstorm. As a consequence, the height of the melting layer becomes smooth in averaging, making it imperceptible in the figure.

**Figure 10.** Vertical profile of mean LDR during thunderstorms observed at the Milešovka observatory for NL (red curve) as compared to FL (blue curve). Y-axis represents the height [m] above the cloud radar situated at an elevation of 837 m a.s.l.

The character of the curves in Figure 10 (their oscillation) is influenced by the number of averaged cases (Figure 9). This is especially true for the red curve representing NL discharges. The isolated maxima of LDR averages are probably random. However, Figure 10 clearly depicts that at an elevation of 4 to 6.5 km approximately, there are large LDR averages, which show little oscillations for NL, thus they do not correspond to random processes. These averages are much larger than the LDR averages for FL. As in Section 3.1, we attribute it to electrification by collisions and alignment of ice crystals.

To better assign the cause of increased LDR averages in the middle troposphere, Figure 11 shows 1 km layers of frequency of LDR in profiles with similar distribution of graupel and hail (i.e., rounded hydrometeors). Because concentrations of graupel and hail are similar in both NL and FL (ice or snow being present almost everywhere in these 1 km layers, Figure 8), it is obvious that there had to be another process that made the higher LDR more frequent in the case of NL as compared to FL. We suggest that the additional process could be the alignment of ice crystals observed by other

researchers, e.g., by Melnikov et al. [25]. However, we are aware that this hypothesis cannot be exactly verified.

**Figure 11.** Frequency of LDR (f[−]) at an elevation of: (**a**) 3–4 km, (**b**) 4–5 km, (**c**) 5–6 km and (**d**) 6–7 km during thunderstorms observed at the Milešovka observatory for NL (red curve) as compared to FL (blue curve), when taking into account similar vertical profiles of graupel and hail concentrations. Y-axis represents the height [m] above the cloud radar situated at an elevation of 837 m a.s.l.

At an elevation between 8 and 9 km, significantly higher averages of LDR for NL, as compared to FL, could be rather random because of high oscillations of that for NL. On the other hand, the high oscillations of LDR averages for NL can also be related to the orientation of aligned ice crystals in an electrified field. LDR can increase if the particles align at an angle close to 45◦ from both the co- and cross-channels, while it can decrease if the particles align along with the co-channel (LDR reaches large negative values). Thus, the LDR of non-spherical targets, such as ice crystals, can have strongly different values (large and small) depending on the azimuth direction to the channels. Therefore, the variability of LDR may be increased in the case of NL.

The results also suggest that the clouds producing lightning in the vicinity (NL) are vertically massive and higher than clouds producing FL, at least in our dataset.
