In recent years, the investigation of leader tortuosity in long air discharge is brought into focus, for it is the main factor that influents the 50% breakdown voltage and can be used to validate and test the discharge modes in the design of long air gap dimension. In order to find out whether the test data of the whole leader channels are accuracy to represent the leader propagation paths, a discharge test of a 4 m rodplane air gap subjected to switching impulse voltage is carried out. And an observation platform is set up to synchronously record the voltage waveforms, the discharge currents, the leader propagation processes and the whole leader channels. A new method is proposed to subtract the final-jump part from the whole leader channel by identifying the leader tip height at the initiation of the final jump (LTH for short) from the CCD image. Then the test data of LTH and the leader propagation path are statistically obtained. And probability density functions (PDFs for short) of three different angles between consecutive segments of the leader propagation path are used to quantitatively describe the leader tortuosity. According to the analysis of the results, the mean value of LTH is nearly half the gap length in the gap axial direction. And an obvious difference can be discovered when comparing PDFs of the leader propagation paths with those of the whole leader channels. A conclusion to account for the difference is deduced that the leader paths during the final jump are much straighter than those during leader propagation. In the end, the present work recommends that the final-jump part should be subtracted from the whole leader channel when the test data of leader tortuosity are used in the numerical model for positive long air discharge. INDEX TERMS Air discharge, design of long air gap dimension, leader tortuosity, long rod-plane air gaps, 50% breakdown voltage, leader propagation, final jump.