Radar Data Assimilation for a Heavy Rainfall Case over the Korean Peninsula Using an Adjoint Sensitivity-based Data Assimilation Method

An Adjoint Sensitivity-based Data Assimilation (ASDA) method is proposed, and it is applied to a heavy rainfall case over the Korean Peninsula. The heavy rainfall case, which occurred on 26 July 2006, caused torrential rainfall over the central part of the Korean Peninsula. Synoptic environments related to the case were favorable for the development of Mesoscale Convective System (MCS). The MCS affecting the Korean Peninsula can be classified as Training Line/Adjoining Stratiform (TL/AS)-type for the earlier period, and Back Building (BB)-type for the later period, based on the morphological analyses of radar reflectivity data. In the ASDA method, an adjoint model is run backwards with forecast-error gradient as an input, and the output of the adjoint-model run, i.e., adjoint sensitivity of forecast error to initial conditions, is scaled by an optimal scaling factor. The optimal scaling factor is determined by minimizing observational cost function of Four Dimensional Variational (4D-Var) data assimilation method, and the scaled sensitivity is added to the original initial condition. The optimal settings for the ASDA-method implementation are determined from sensitivity experiments.

The simulated rainfall distribution is shifted northeastward compared to the observations when no radar radial velocity data are assimilated, or radar data are assimilated using Three Dimensional Variational (3D-Var) method. However, the rainfall distribution and time series of hourly rainfall are improved when radar data are assimilated using the 4D-Var or ASDA method. In addition, simulated atmospheric fields such as zonal wind, meridional wind, temperature, and water vapor mixing ratio are also improved via the 4D-Var or ASDA method, when verified against reanalysis and observational data. Analysis increments of 4D-Var and ASDA experiments are investigated to find out the reason for the improved forecasts. Due to the improvement of the analysis, subsequent forecasts of 4D-Var and ASDA experiments appropriately simulate the observed features of TL/AS- and BB-type MCSs and heavy rainfall. Although the quality of the analysis and subsequent forecast for the ASDA method is similar to those for the 4D-Var method, computational cost is significantly reduced in the ASDA method. The proposed ASDA method will be applied to a variety of cases over the Korean Peninsula and statistical analysis will be conducted to obtain robustness of the method.