Determination of Detection Efficiency of Lightning Detection Systems using Bayesian Analysis

Given the multitude of lightning detection systems, researchers are often interested in how different systems compare to each other. Usually, one system is assumed to be “truth” and is used to determine the relative detection efficiency of another system. While useful, this method does not fully consider the information available. Since no system detects all lightning in practice, it is more instructive to determine the probability that one system detects a discharge, given the other system detected a discharge. Then, the reverse probability is also determined: given the second system detects a discharge, the probability the first system detects the discharge is found.

In the context of a Bayesian analysis, this leads to a relative probability of detection for the two systems. Hence the relative information content of one system relative to another can be determined. Further application of probability theory can yield an upper limit on the absolute detection efficiency of each system.

We present results comparing the Earth Networks Total Lightning Network (ENTLN) and the Lightning Imaging Sensor (LIS). We do not compare flashes; since lightning processes that result in detectable light from space also produce detectable electric field changes, the base unit of each data set are directly comparable. By using ENLTN pulses and LIS groups, we minimize the fundamental bias that exists in flash sorting algorithms.