Based on the dual-polarization radar particle classification products and cloud-to-ground lightning location data from Fuzhou, Xiamen and Longyan from April 2022 to June 2024, in this study we select 54 convective storm samples with or without cloud-to-ground lightning and investigate the temporal variation characteristics of graupel particle layer thickness during the developing processes of these samples. The results show that the graupel particle layer thickness of convective storms with cloud-to-ground lightning reaches at least 2.26 km at the initiation moment of the first cloud-to-ground lightning, while 95% of the samples without cloud-to-ground lightning, have graupel particle layer thickness less than 2.2 km throughout their lifecycles. Considering both the lead time and accuracy of cloud-to-ground lightning initiation forecasts, we propose to use a graupel particle layer thickness greater than 2 km as a forecast indicator for cloud-to-ground lightning initiation in Fujian Province. The forecast potential is evaluated, and the calculated TS score of the sample is 0.864 and the average forecast lead time is 28.13 min. Taking four convective storms on 14 June 2022 in southern Fujian Province as examples, the above-mentioned forecast indicators are applied and analyzed in these cases. The results show that the proposed forecast indicator can accurately forecast whether cloud-to-ground lightning will occur in the four convective storms, and the forecast lead time for cloud-to-ground lightning initiation is 6 min.