Comparison of the United States Precision Lightning Network (USPLN) and the Cloud-to-Ground Lightning Surveillance System (CGLSS)

Thumbnail Image
Jacques, Alexander Andrew
WSI Corporation requested a performance evaluation of their United States Precision Lightning Network™ (USPLN™), which is co-owned by TOA Systems, Inc. The USPLN is a national lightning detection network with over 160 sensors placed across the North American continent. Previous performance evaluations of the network had been limited to simulated lightning events and individual fixed tower case studies. Thus, a longer evaluation of the network had yet to be completed, which this study attempts to achieve. As a validation tool, the second generation of the Cloud-to-Ground Lightning Surveillance System (CGLSS-II) was selected. CGLSS-II is a local detection network used for critical lightning surveillance at Kennedy Space Center and Cape Canaveral Air Force Station (KSC/CCAFS). The network of six sensors has been certified by the U.S. Air Force since 1989, and is constantly monitored and evaluated. CGLSS-II and the USPLN share numerous similarities including: the processing of all lightning strokes, GPS timing, and the time-of-arrival technique for triangulating stroke locations. Stroke data for CGLSS-II and USPLN were acquired and quality controlled for the selected study period of 20 May 2008 to 31 August 2010. The study period was further divided into sub-periods based on changes to CGLSS-II performance, and data were restricted to a region surrounding KSC/CCAFS. A correlation procedure was selected which matched strokes between the two networks using time and distance thresholds, creating a comparative dataset. Data from the Four Dimensional Lightning Surveillance System (4DLSS) was also collected as a means to classify cloud-to-ground (CG) and intra-cloud (IC) strokes. Melbourne (KMLB) composite reflectivity radar imagery was also acquired to further evaluate USPLN performance. Several analyses of USPLN stroke detection efficiency (DE) and location accuracy were conducted to first determine average performance and then to examine specific case studies. Analyses of USPLN stroke DE revealed several strengths and weaknesses to the network. Simple weighted average analyses of each sub-period revealed that the USPLN failed to detect a significant portion of the CGLSS-II strokes. Logistic regression and plots of USPLN stroke DE versus CGLSS-II peak current (Ip) indicated that most of the missed detections were due to low current strokes, while the USPLN excelled at detecting high current strokes. A pseudo-flash DE analysis concluded that perhaps many of the undetected low current strokes were subsequent strokes in a lightning flash. Additionally, performance was found to be degraded when the USPLN sensor baseline was altered significantly by sensor outages. Temporally, the USPLN stroke DE improved with time until around 1 July 2010, after which performance decreased. Analyses of USPLN location accuracy showed a similar temporal performance improvement up to around 1 July 2010, with a significant decrease thereafter. The 95% confidence USPLN location accuracy metrics were on the order of 600 m during peak performance from February to June 2010. Similar decreases in performance were also discovered when USPLN sensor outages occurred, coinciding with the loss of DE. An analysis of directional variation between matching CGLSS-II and USPLN stroke locations revealed a prominent northeast and less prominent southwest bias for the USPLN over the study region. Four case study days were examined to determine potential causes for USPLN strokes that were not matched with CGLSS-II. Comparisons to 4DLSS data indicated that the majority of these unmatched strokes were truly IC strokes detected by the USPLN and falsely classified as CG strokes. Any phantom" strokes those which seemed to occur with little or no 4DLSS activity were examined with radar imagery. The radar analyses produced primarily inconclusive results. Additional case study analyses revealed that the USPLN rarely falsely categorizes a true CG stroke as IC or reports the incorrect polarity of a stroke. This study revealed several strengths and weaknesses for the USPLN. It is unclear exactly why low current strokes were frequently missed but hypotheses were introduced regarding sensor sensitivity and other factors. An investigation into the 1 July 2010 performance decrease uncovered a software change to the USPLN which likely was the cause for the performance decrease. Additionally an investigation into a poorly analyzed CGLSS-II event was also initiated during this study. Future work should include more performance evaluations of the USPLN with other local and national networks a review of the quality control methods the addition of new methodologies and perhaps additional CGLSS-II comparisons with other stroke-based networks.