Abstract: Enhancing the operational forecast ability for severe convection is of great significance to national disaster prevention and mitigation. Based on the 3 km China Meteorological Administration limited-area numerical prediction system and convection-allowing ensemble prediction system, this study constructed eight model climate datasets for four severe convective physical parameters—convective available potential energy (Convective Available Potential Energy, CAPE), 0–6 km vertical wind shear (0-6 km vertical wind SHeaR, SHR6), precipitable water (Precipitable WATer, PWAT), and a composite parameter (Composite Parameter, CP)—using multiple spatiotemporal extension schemes. On this basis, Extreme Forecast Index (EFI) products for the above four parameters were developed. Both statistical and case-based verification of the EFI forecasts were performed using severe convection observations, and the optimal model climate construction scheme, the most effective physical parameters for forecasting different types of severe convection, and the corresponding EFI thresholds for issuing severe convection warning signals were determined, with the threat score as the criterion and taking into account the performance of other scores at the same time. Results showed that the model climate distributions were reasonable, and the EFI products based on the monthly model climate scheme with a 1° × 1° spatial range and a ± 30-day temporal window performed the best. CAPE, PWAT, and CP demonstrated relatively good forecast performance for severe convection, whereas SHR6 performed poorly. PWAT (CAPE and CP) showed the best skill in forecasting short-duration heavy rainfall (thunderstorm high winds). The severe convection EFI products provided useful indications of the occurrence frequency of severe convection but very limited insight into its intensity. Case study indicated that CAPE, PWAT, and CP were able to provide early warning signals for severe convection, though false alarms and missing events existed in some regions. This study provided foundation for establishing convective-scale model climate datasets of severe convective parameters and EFI products for extreme severe convection in China, showing potential for operational application.