Lightning, as a natural discharge phenomenon, is often accompanied by strong convective weather, which can lead to human and animal casualties, facility damage and economic losses. With the socio-economic development, there is an increasing demand for accurate prediction of lightning distribution. In this study, an empirical orthogonal function (EOF) analysis combined with a geographically weighted regression model is used to analyze the distribution pattern of lightning activity in Fujian Province and to make spatial predictions. The study utilizes the actual monitoring data of Fujian Power Grid LLS for the past 40 years (1982-2021), decomposes the ground flash records into mutually orthogonal spatial eigenvectors and time coefficients based on the EOF method, and analyzes the spatial heterogeneity of the lightning activity using a geographically weighted regression model. The results show that the cumulative variance contribution of the first three modes of the EOF decomposition of thunderstorm days in Fujian Province reaches 75.45%, of which the first mode contributes 57.43%, which is mainly characterized by the negative phase distribution in the province; the overshooting lag correlation coefficient between thunderstorm days and ENSO index is 75.07%, of which the correlation coefficient with the El Niño event is as high as 83.53%, which is significantly higher than the correlation coefficient with the La Niña event (35.79%); the correlation coefficient with the ENSO index is 75.07%. The correlation coefficient with El Niño event is 83.53%, which is significantly higher than that with La Niña event (35.79%); 94.67% of the lightning intensities in Fujian are less than 50 kA, 63.46% of the lightning intensities are less than 15 kA, and the interval with the highest frequency of the lightning current amplitude is 7-17 kA. The study concludes that the lightning activity in Fujian Province is mainly affected by the topographic and climatic factors, and the mountainous and hilly areas are the main hotspot of the lightning activity and there is a significant spatial heterogeneity. The geographically weighted regression model can effectively predict the spatial distribution of lightning in complex terrain.
