Abstract: Reanalysis data and model prediction products are frequently used in studying the convective potential index and its thresholds. CMA global reanalysis (CRA) developed by China Meteorological Administration and CMA-Global Ensemble Prediction System (CMA-GEPS) products are assessed to evaluate the capability in depicting convective potential index, and to establish a potential prediction threshold for thunderstorm weather across China. A comparison of the convective potential index derived from ERA5 reanalysis data, CRA reanalysis data, and control forecast products of CMA-GEPS with radiosonde observations is conducted during the warm season from July to September in 2023. The threshold for thunderstorm potential is determined using lightning location data from the National Lightning Detection Network. It's found that ERA5 reanalysis data induces the smallest forecasting errors and the highest correlation coefficients with radiosonde observations for convective available potential energy (CAPE), and the control prediction products of CMA-GEPS are next in line. However, CRA reanalysis data exhibits obvious positive errors for CAPE values, and the distribution of CAPE index appears to be relatively discrete. The composite CAPE-shear parameters are found to correlate with CAPE value. Consequently, it suggests that ERA5 reanalysis data and CMA-GEPS control prediction products can effectively substitute for actual radiosonde observations in the study of convective potential index. Correlation coefficients of vertical wind shear among ERA5 reanalysis data, CRA reanalysis data, CMA-GEPS control prediction products, and radiosonde observations are found to be high but with small negative mean errors. In addition, compared to 0-6 km vertical wind shear, correlation coefficients of 0-1 km vertical wind shear between reanalysis data, ensemble prediction products and radiosonde observations are found to be lower, and the mean errors are found to be larger. Consequently, using boundary layer wind from reanalysis data or prediction products is considered inappropriate for studying the convective potential index. The CAPE value, along with composite parameters that integrate CAPE with 0-6 km vertical wind shear, can be utilized to some extent for discriminating between thunderstorm and non-thunderstorm events. It is important to note that the potential thresholds vary with different datasets and should be regarded as probabilistic indicators. However, it shows that 0-6 km vertical wind shear index is ineffective in distinguishing thunderstorm events from non-thunderstorm events.