Figure 2. (a) Two-dimensional histogram of the frequency of CATS maximum cloud-top height vs. latitude, when at least one ISS LIS flash is detected, for March-October 2017. (b) Two-dimensional histogram of maximum 20-dBZ echo-top height vs. latitude for TRMM RPFs with at least one LIS-detected flash during March-October 1998-2914. (c) Two-dimensional histogram maximum 20-dBZ echo-top height vs. latitude for GPM RPFs with at least one WWLLN-detected flash during March-October 2014-2020. Frequency values below 0.0005 are not shown in any of these subplots.
3.2 Comparison of thunderstorm characteristics
It is also of interest to compare LIS-detected lightning to CATS-inferred thunderstorm properties, such as IWP, dominant hydrometeor type, optical depth, and other characteristics. For this analysis, it is understood that a lidar cannot penetrate deeply into optically thick thunderstorms, and that statistical correlations thus will be significantly reduced compared to standard radar-lightning comparisons (Wiens et al., 2005; Carey et al., 2019). However, Rutledge et al. (2020) demonstrated that passive infrared measurements of cloud properties (again, largely limited to near-cloud-top characteristics) can still provide useful physical insights into thunderstorm properties. Thus, the success criteria for this lidar-based analysis were modest, and the primary goals were to demonstrate weak yet statistically significant (and physically meaningful) correlations between lightning and cloud properties.
First, the CATS hydrometeor feature mask was analyzed when lightning was detected by LIS. For this, the analysis took advantage of the fact that the CATS hydrometeor mask uses an increasing index scale ranging from 0 (no cloud), to 1 (liquid cloud), to 2 (undetermined cloud phase), to 3 (ice cloud), and determined the frequency of the highest value index within 50 km of lightning along the CATS ground track. The results are shown in Fig. 3. Over 90% of profiles with lightning were associated with ice-phase or undetermined (likely mixed-phase) cloud. Only ~7% of profiles were associated with liquid cloud. Assuming that CATS is not providing evidence of solely warm-phase clouds producing lightning, this suggests that CATS’ hydrometeor mask can identify ice-containing cloud with better than 90% accuracy (notable given validation of radar-based hydrometeor identification is difficult; e.g., Ryzhkov et al., 2005), and thus demonstrates the value of using lightning observations to validate remote sensing of hydrometeor type. Moreover, given that lidar cannot penetrate deeply into optically thick clouds, one cannot rule out the presence of ice below the cloud-top-biased lidar observations, nor can one rule out the fundamental differences in LIS and CATS sampling geometry contributing to any discrepancies. The very small number of “no cloud” observations are also notable and will be analyzed in detail later in the next subsection.