Abstract: As an emerging sea state monitoring device, HF radar can provide observations of large-scale sea surface currents along the coastal region. However, a single HF radar can only provide radial flow observation. This paper develops a two-dimensional variational method (2DVAR), which blends radial current observation with surface ocean current vector from the ROMS (regional ocean model system) model of Taiwan Strait to obtain a blended pro duct. The results show that the average relative error of the blended product and the radar radial flow is reduced from 9.70% to 1.54% compared to the average relative error of the model output and the radar radial flow. To further examine the blending method, two independent sampling experiments were designed using the original data. In the first experiment, the radar radial flows were evenly divided into blending samples and independent samples in space. The blending samples of radial velocity observation were then blended with the model output. The mean root mean square error (RMSE) of the blended result is 0.19 m/s compared with the RMSE of model output of 0.41 m/s. In the second experiment, both the radar radial flow and the model output were spatially evenly divided into blending samples and independent samples, and the blending samples of the two kinds of data were blended. The average error variance of the blended result, estimated using the TC (triple collocation) method, is the smallest. And the error variance of the blended result is significantly reduced at the far shore region compared with the error variance of the independent radar radial flow as well as the error variance of the model output. All the above results show that the blended product is closer to the true current, which proves the effectiveness of the 2DVAR method for blending sea surface currents.