Figure 1. (a) Time-height plot of CATS feature mask with ISS LIS flashes overlaid for half an ISS orbit on 15 March 2017. ISS LIS flash times are plotted vs. distance from CATS ground track (not flash height, which LIS does not measure). (b) Time-height plot of CATS cloud phase with ISS LIS flashes overlaid for half an ISS orbit on 15 March 2017. ISS LIS flash times are plotted vs. distance from CATS ground track (not flash height, which LIS does not measure).
In addition to the above statistical analysis, a manual review of combined LIS/CATS quicklooks (e.g., Fig. 1) was performed. A major focus of this review was to identify LIS lightning that did not appear to occur near CATS-identified clouds. These potential false alarms were further studied using manual inspection of daytime LIS/CATS matchups using geolocated ISS LIS backgrounds, which are available every 30-60 seconds from LIS. These backgrounds (Blakeslee, 2020b) were geolocated using the ISS_Camera_Geolocate open-source software (Lang, 2019), originally developed in support of the Schultz et al. (2020) study.
2.4 Other datasets used
To assess the distribution of radar-observed echo-top heights, for comparison to CATS-measured thunderstorm cloud-top heights, we examine TRMM and Global Precipitation Measurement (GPM) Precipitation Features radar precipitation features (RPFs; Liu et al., 2008) for March through October 1998-2014 (TRMM) and 2014-2020 (GPM). The RPFs are defined as contiguous areas of at least 4 pixels of precipitation as detected by the TRMM Precipitation Radar (PR) or the GPM Dual-Frequency Precipitation Radar (DPR) Ku-band radars. These radars have minimum threshold reflectivities of 17 dBZ and 12 dBZ, respectively. Pixel size is roughly 20 km2. For each RPF, radar and passive microwave characteristics (e.g., echo-top heights, reflectivity profiles, and brightness temperatures) are provided, but in this study the focus was primarily on echo-top heights. In order define features with lightning, the coincident LIS data were used for TRMM RPFs. The GPM Core satellite does not have a lightning imager onboard, and therefore GPM RPFs were co-located with World Wide Lightning Location Network (WWLLN; Virts et al., 2013) data within a +/- 10-minute window in the boundary of the feature (Liu, 2020).
3 Results
3.1 Histograms of cloud- and echo-top height
Figure 2 shows a 2D histogram of CATS-measured maximum cloud-top-height vs. latitude within 50 km along the CATS ground track of at least one LIS-detected flash (Fig. 2a). The histogram is normalized but no adjustments for sampling frequency have been made due to the short time period of analysis (< 8 months). March-October 2017 primarily covers the spring, summer, and early fall seasons (i.e., warm season) in the northern hemisphere. Within 10° S to 20° N latitude, thunderstorm maximum cloud-top height was most frequently 16-17 km MSL. A downward sloping in maximum cloud-top height occurred toward the northern mid-latitudes, reflecting the general downward sloping of the tropopause toward the poles (Santer et al., 2003). Of course, as is well known thunderstorm heights are not fully constrained by the tropopause (Liu & Zipser, 2005). Regardless, within 35-50° N the thunderstorm cloud-top height most commonly ranged between 10 and 14 km MSL. Note the relatively few samples south of 10° S, reflecting the LIS/CATS dataset’s focus on the northern hemisphere’s warm season.
To help validate the conclusions implied by Fig. 2a, histograms for TRMM (Fig. 2b) and GPM RPF (Fig. 2c) maximum 20-dBZ echo-top height vs. latitude are shown for all RPFs with at least one LIS- or WWLLN-detected lightning flash during March through October (1998-2014 for TRMM, 2014-2020 for GPM). Due to the longer time periods for analysis, the TRMM and GPM histograms were normalized by sampling frequency. Note that TRMM was inclined at a much lower orbital angle (35°) than either ISS (57°) or GPM (65°), so latitudinal coverage was reduced relative to the other two platforms. Regardless, both TRMM and GPM suggest that thunderstorm 20-dBZ echo-top heights are lower by approximately 2 km than lidar-inferred cloud-top heights. The downward sloping of the echo-top heights toward the poles (with approximately the same slope as Fig. 2a), as well as the preference for thunderstorms in the northern hemisphere during May-October, are also apparent. This suggests the added value of using a lidar over a radar to obtain a more accurate measurement of cloud-top height, but also implies a way to use lidar observations to help scale radar echo-top heights to cloud-top heights if only a radar is available (at least from an average global perspective). Moreover, this analysis also helps validate the matching criteria for the LIS/CATS comparison.