Statistical Model Limitations of Ground Flash Density for Lightning Risk Assessment

Ground flash density, a critical parameter for assessing lightning risks in power grids, is typically modeled with Poisson-distributed assumptions that oversimplify actual lightning dynamics. This study challenges these assumptions by analyzing 6-year lightning locating system (LLS) data from Guangxi Power Grid, China. Through stochastic distribution simulation and non-parametric statistical analyses, we systematically compare observed lightning activity patterns with theoretical Poisson expectations. Our analysis demonstrates: (1) Actual lightning flash activity exhibits significant spatial clustering and temporal persistence, contradicting the spatiotemporal independence assumption of Poisson models. (2) Non-parametric tests further reveal that annual ground flash distribution patterns exhibit strong interyear correlations and do not originate from the same distribution population, violating the Poisson model's independence and identical distribution criteria. These results highlight that Poisson-based frameworks neglect critical spatial and temporal heterogeneity, leading to underestimated risks in localized ground flash hotspots (spatial regions with significantly higher frequencies) during periods of heightened lightning activity. Consequently, it is necessary to revise lightning protection strategies by developing localized spatiotemporal statistical models, rather than relying solely on averaged ground flash parameters from large-scale models.