Effects of open boundary bias correction and data assimilation in a regional ocean circulation model for the East Sea using observed temperature and salinity data

Abstract

Boundary conditions are important constraints on the interior solution when solving partial differential equations. Global ocean prediction models have been used to provide boundary conditions for regional ocean models. However, in the Korea Strait, the simulated salinity by the global models was lower than the in-situ observation data in winter and higher in summer. The simulated temperature from the global model also has biases with respective to the observation. These boundary biases have caused high uncertainties in the prediction of temperature and salinity distribution in the East Sea by nested regional ocean circulation models. To reduce prediction errors in the interior of the nested regional model domain, the open boundary bias correction (BC) and the data assimilation (DA) were performed using observation data. The effects of BC and DA were quantified from four numerical experiments: free run without BC and DA (EXP 1) and simulations with BC (EXP 2), with DA (EXP 3), and with both BC and DA (EXP 4). The root mean square errors (RMSEs) of temperature from the EXP 1, 2, 3, and 4 were 1.59℃, 1.55℃, 0.88℃, and 0.92℃ at 10-m depth in 2019, respectively. The RMSEs of salinity from the four experiments were 0.36, 0.30, 0.33, and 0.27 at 10-m depth in 2019, respectively. EXP 3 and 4 reduced temperature errors by 44% and 42%, respectively, indicating that DA improves the prediction of temperature. EXP 2 and 4 reduced salinity errors by 17% and 26%, respectively, implying that BC reduced the boundary errors and improved the interior solution of salinity. DA and BC were effective in the northwestern Ulleung Basin in February and at the Korea Strait in August, respectively. Therefore, the BC and DA are expected to improve the interior solution of salinity and temperature in regional ocean models.