Abstract: Using WRF-Electric model coupled with electrification and discharge schemes, experimental predictions are carried out on the regional lightning activity from 2015 to 2017. By establishing a lightning activity prediction verification method, experiment results are classified and evaluated. Taking operational prediction as a reference, the ability to predict regional lightning activity with the numerical model is evaluated objectively. Main problems of the model are identified through verification, which provides a basis for its further improvement.The major region of lightning activity could be predicted well by the meso-scale electrification and discharge model. In the strict point-to-point verification, the CSI of the model prediction almost reach the operational prediction level during the main flood season (June-August). Quantitative verification results over North China also show that the prediction performs best with the forecast time of 6-12 hours. For small-scale thunderstorms, CSI of the model prediction is higher than that of the operational prediction. With expansion of the thunderstorm scale, the model prediction gradually loses its advantage. Therefore, the model is more valuable for predicting localized and small-scale thunderstorms.The range of the lightning activity predicted by the model is small and relatively concentrated, and some scattered lightning activity is often missed. Thus, in the parameterization of discharge, the threshold should be decreased at the initial time of lightning to improve the performance in relatively weaker electrification region. The lightning flash density predicted by the model is obviously greater than observed. To reduce the predicted flash density in the strong electrification area, the amount of neutralization charge of a single lightning should be consistent with the observation in the discharge scheme.CSI is relatively low with both the operational and this model prediction. In some cases, the prediction can achieve a relatively high CSI, but in long-term prediction experiments it's difficult to maintain high score using the strict point-to-point verification method. On the other hand, for weather phenomena with strong randomness in their occurrence position, the predictability is usually poor.Although the model can forecast the lightning activity area well, reaching the level of operational prediction, many problems remain in terms of the flash density forecast. How to parameterize lightning reasonably in a meso-scale model is still unresolved and extremely challenging. Currently, numerical models can predict precipitation successfully, while the ability to quantitatively predict the flash density is far behind. Improvement in lightning parameterization schemes and the selection of relevant thresholds in models relies on new methods and a large number of experiments being conducted.