Date of Award
Master of Science (MS)
Christopher J. Schultz
GOES (Meteorological satellite), Lightning., Severe storms., Thunderstorms.
The objective of this study is to implement the two-sigma lightning jump algorithm (LJA), initially developed using Lightning Mapping Arrays (LMAs), with GOES-16 Geostationary Lightning Mapper (GLM) flashes, evaluate its performance, and identify any needed adjustments to the algorithm to optimize operational skill. The GLM is projected to have lower detection efficiency (DE) (70-90 percent) than operational LMAs (95-99 percent). The reduced GLM DE coupled with the coarser spatial resolution of the GLM could have impacts on flash rates and trends that could affect the LJA in various ways. The LJA between the two systems are looked at in both a context of severe storm reports and the radar intensity metrics of MESH and VIL. Initial comparisons between the GLM and LMA on a subset of 5 severe and 2 non-severe storms show a number of differences in flash rates and trends between the two. These differences are maximized in extraordinarily intense storms with high flash rates in the LMA. It was found that LMA was better correlated with MESH (VIL) at 0.498 (0.484) than the GLM was at 0.225 (0.238). Flash rate differences affected the jumps using the two lightning measurement systems in various ways with the most important being the increase in jump frequency in the GLM relative to the LMA during weak non-severe storms and decaying storms. This increase in jumps in the GLM for non-severe and decaying storms are an indication that the conceptual model of the LJA is failing on the GLM and is likely due to one or both of the following classes of error: GLM flash measurement error or GLM jump algorithm error. These differences and multiple sources of potential error suggest that a larger sample size study must be conducted in order to properly evaluate and optimize the LJA with the GLM. This larger study analyzed 930 storms and when verified against storm reports at its original configuration yielded a POD of 67.9%, FAR of 82.2%, and CSI of 0.16. While the POD was close, but slightly lower, to other studies using LMA and GLM proxy data, the FAR was 20% higher than any prior study. The high FAR is an important finding and is looked into more in-depth. It was found that false alarms in the GLM are found to be mainly associated with weaker MESH and VIL values than hits indicating that the high FAR is likely influenced by physical model, GLM measurement or jump algorithm errors.
Curtis, Nathan L., "An analysis of the lightning jump algorithm using the GOES-16 geostationary lightning mapper" (2018). Theses. 233.