In the present study, a newly designed radiative forcing scheme, i.e., impulsive radiative forcing scheme (IRFS), is applied in the one-dimensional Regional Ocean Modeling System-Carbon Silicate Nitrate Ecosystem (ROMS-CoSiNE) model to reproduce under-ice phytoplankton blooms (UIBs) in the Arctic Ocean. The model results obtained with the IRFS and the traditional continuous radiative forcing scheme (CRFS) are compared with observations of an UIB north of Svalbard. It is found that, the new IRFS performs much better than the traditional CRFS with lower biases, lower root-mean-square difference (RMSD), and a higher correlation coefficient (r), especially at the near-surface layer of the water column (0–15 m). It is also found that in the simulations with the CRFS, the UIB started earlier and maximum chlorophyll concentration was more than twice the observed data. However, the UIB with the IRFS is similar to observations in terms of UIB timing and magnitude. Since the IRFS allows more solar radiation to penetrate through the upper water column, phytoplankton growth is supported at greater water depth (10–40 m). Therefore, this model reproduces a deeper phytoplankton bloom and corresponding deepened subsurface chlorophyll maxima, which agrees well with the observations. Additionally, the introduction of melt ponds into the IRFS model plays a vital role in accelerating spring UIBs because ponded ice can transmit more light. However, the magnitude of integrated chlorophyll-a hardly changes between the model runs with and without melt ponds. This study proposes a more accurate averaging of the light field that simply changes the averaging procedure to take into account horizontal light variability at the sub-grid scale and provides an approach to improve radiative forcing scheme in the Pan-Arctic model case.