Figure 5. (a) Raw, un-geolocated ISS LIS background image from approximately 2044 UTC on 1 June 2017. (b) The same ISS LIS background image but geolocated, showing ISS LIS events detected in the domain of the image during the overpass. The suspect LIS flash not seen near CATS-inferred cloud also is indicated, and LIS and CATS boresight ground tracks are shown.
Of the 8246 LIS/CATS matchups, 105 (1.3%) have no CATS-identified cloud within 50 km along the CATS ground track (further constrained by the flash centroid needing to be within 25 km distance perpendicular to the CATS ground track). Of these 105 candidate false alarms, 65 occurred during daytime (777.4 nm backgrounds during night have limited utility for this paper’s analysis) and have nearly coincident ISS LIS backgrounds (within approximately 30-60 seconds, the frequency with which LIS provides background imagery, as discussed in Section 2.3).
Manual review of these geolocated backgrounds, with LIS lightning overlaid, was performed. This analysis found that the vast majority of candidate false alarms were similar to Fig. 5, where obvious convection (in this case, over Hispaniola) is in the LIS field of view and the presence of lightning is reasonable to infer. This case demonstrates the fundamental limitations of co,paring 2D horizontal lightning observations with lidar nadir curtains: Sometimes, obvious connections are just missed. Tightening distance thresholds obviously would improve on this, at the cost of reducing the total size of the already small LIS/CATS dataset. At the same time, the fact that only ~1% of LIS/CATS comparisons with the chosen distance thresholds couldn’t find at least some cloud in the vicinity of lightning is highly encouraging for the approach adopted in this study.