Abstract: The performance of atmospheric dispersion models varies across different scenarios, making it crucial to validate these models in multiple scenarios to quantify their range of performance. This paper integrates the California Meteorological Model (CALMET) with the Lagrangian splitting puff model (SPUFF) to perform the local-scale dispersion simulation based on the multiple wind tunnel experiments under neutral atmospheric condition at the Sanmen nuclear power plant (NPP) scenario conducted in the multi-complex terrains. The simulations are compared in detail with the measurements from networked sites and individual axial sites at ground level, aiming to further evaluate the performance of SPUFF against multiple dispersion scenarios, both qualitatively and quantitatively. The evaluation results reveal that SPUFF demonstrates sufficient dispersion across multiple scenarios, which effectively covers most of the measurement sites. Meanwhile, the numerical evolution of the plume closely matches the trends observed in the measurements, accurately reproducing the values at the axial sites. The statistical results indicate that the performance of SPUFF is excellent. All four statistical metrics, including the fraction of simulations within a factor of two of the measurements (FAC2), the fractional mean bias (FB), the normalized mean-square error (NMSE), and the normalized absolute difference (NAD) meet model acceptance criteria at both networked sites and individual axial sites. Among them, FAC2 in all scenarios reaches a good level of 0.51 and 0.83 for networked sites and individual axial sites, respectively; while FB achieves excellent performance at -0.03 and -0.06, respectively. The above results demonstrate that SPUFF is suitable for various dispersion scenarios under neutral atmospheric condition in complex NPP sites with multiple terrains, making it an important tool for future early response to nuclear emergency.

Key words: nuclear power plant sites with complex terrain, atmospheric dispersion, CALMET, SPUFF, puff splitting, wind tunnel experiment evaluations with multiple scenarios