Using a 3D lightning radiation source locating system （LLS）, three pairs of associated lightning discharges （two or more adjacent lightning discharges following an arbitrary rule that their space-gap was less than 10 km and their time-gap was less than 800 ms） were observed, and the interaction between associated lightning discharges was analyzed. All these three pairs of associated lightning discharges were found to involve three or more charge regions （the ground was considered as a special charge region）. Moreover, at least one charge region involved two lightning discharges per pair of associated lightning discharges. Identified fl＇om electric field changes, the subsequent lightning discharges were suppressed by the prior lightning discharges. However, it is possible that the prior lightning discharge provided a remaining discharge channel to facilitate the subsequent lightning discharge. The third case provided evidence of this possibility. Together, the results suggested that, if the charges in the main negative charge region can be consumed using artificial lightning above the main negative charge regions, lightning accidents on the ground could be greatly reduced, on the condition that the height of the main negative charge region and the charge intensity of the lower positive charge region are suitable.
This paper reports the investigation of the location accuracy of a three-dimensional （3D） lightning-radiation-source locating system using sounding balloon measurements. By comparing the information from the balloon-borne VHF transmitter flight path and locations using simple geometric models, the location uncertainties of sources both over and outside the network were estimated. For radiation sources inside the network and below an altitude of 7 km, the horizontal uncertainty was 12-48 m and the total mean value was 21 m （rms）, and the vertical uncertainty was 20-78 m and the total mean value was 49 m （rms）. Outside the network, the location uncertainties increased with distance. The geometric model showed that range and altitude errors increased as a function of the range squared whereas the range errors increased parabolically with distance, and that was confirmed by the covariance calculation results. The standard deviation was used inside the network and covariance was used outside the network. The results indicated that location errors from a simple geometric model exhibited good agreement with standard experimental data. The geometry of the network, set of measurements, and calculation method were verified as suitable. The chi-square values of the least squares goodness of fit algorithm were verified and the timing error （A/rms） Of the fitting formula was estimated. The distribution of the chi-square values was less than 5, corresponding to a timing error of 50-66 ns （rms）.
