Statistical Evidence for Factors Associated With Positive Leader Branching in Long Air‐Gap Discharges

Abstract

Abstract Branching is a common feature in positive leader development, which changes the channel topology and significantly influences discharge behavior. However, the factors associated with its occurrence remain poorly understood. In this study, an event‐based analytical framework was established using synchronized optical observations and electrical measurements. Five physically motivated variables were constructed, including the injected charge Q , the rise rate of applied voltage dU / dt , the background field E , the background‐field direction angle θ 1 , and the prior‐channel deflection angle θ 2 . Their independent, nonlinear, and interaction effects were analyzed using multivariable logistic regression, restricted cubic splines, and random‐forest‐assisted interaction mining with regression validation. Primary and sensitivity analyses show that Q , dU / dt , and E are consistently and positively associated with branching occurrence, whereas θ 1 and θ 2 provide weaker and less stable information. Q shows the most reproducible nonlinear association, although its curve shape is partly affected by original‐scale skewness. The interaction rule of E > 275 kV/m and dU / dt > 8.1 kV/μs is identified as the most reproducible candidate interaction signal, suggesting that branching is more likely when stronger background‐field conditions coincide with faster dU / dt ; however, its statistical support is sensitive to variable definition and sample composition. These results suggest that positive leader branching is associated with multiple interacting field‐related factors rather than a single fixed threshold. The proposed event‐based proxy‐variable framework provides a feasible approach for quantitatively analyzing branching‐related factors under the resolution limits of the present experiments.