Abstract: In the southeastern hilly regions of China, extreme rainfall frequently triggers landslide disasters, characterized by their rapid occurrence and clustering. Establishing effective rainfall thresholds for early warning systems is crucial for regional disaster prevention and mitigation. The empirical rainfall index R′, which is based on antecedent effective rainfall and triggering rainfall, has been successfully utilized for early warnings and forecasting of landslide disasters in Hiroshima, Japan. Considering the strong similarities in disaster-prone environments between the coastal-hilly areas of Fujian Province and Hiroshima Prefecture, applying this methodology in Yongtai County, Fuzhou, is of significant practical value. This study involved statistical analysis of historical rainfall and disaster data from typical typhoon-induced rainstorm events, such as the Typhoons Haikui and Nepartak. Key parameters of the R′ model were established, and typical rainfall processes and landslide disaster points were analyzed for inverse verification, to propose a rainfall warning threshold suitable for Yongtai County. The results show that: 1) key parameters for the R′ model are determined, including the baseline values for antecedent rainfall (R₁)=120 mm, triggering rainfall (r₁)=135 mm, a rainfall weight factor (a)=2.5, and an effective rainfall reduction coefficient (α)=0.85. 2) a rainfall warning threshold of R′=156mm was proposed for typhoon rainstorm-induced landslides in Yongtai County. This threshold has proven effective in fully predicting landslides triggered triggered by Typhoon Nepartak, and it can also provide early warning for isolated landslide events, with a recommended lead time of 30 minutes. The rainfall index R′ model and its established threshold demonstrate excellent applicability for early warnings of typhoon rainstorm-induced landslides in Yongtai County, serving as a valuable reference for meteorological early warnings of geological disasters in similar coastal areas across southeastern China.