Effect of Length Scale Tuning of Background Error in WRF-3DVAR System on Assimilation of High-Resolution Surface Data for Heavy Rainfall Simulation

We investigated the impact of tuning the length scale of the background error covariance in the Weather Research and Forecasting (WRF) three-dimensional variational assimilation (3DVAR) system. In particular, we studied the effect of this parameter on the assimilation of high-resolution surface data for heavy rainfall forecasts associated with mesoscale convective systems over the Korean Peninsula. In the assimilation of high-resolution surface data, the National Meteorological Center method tended to exaggerate the length scale that determined the shape and extent to which observed information spreads out. In this study, we used the difference between observation and background data to tune the length scale in the assimilation of high-resolution surface data. The resulting assimilation clearly showed that the analysis with the tuned length scale was able to reproduce the small-scale features of the ideal field effectively. We also investigated the effect of a double-iteration method with two different length scales, representing large and small-length scales in the WRF-3DVAR. This method reflected the large and small-scale features of observed information in the model fields. The quantitative accuracy of the precipitation forecast using this double iteration with two different length scales for heavy rainfall was high; results were in good agreement with observations in terms of the maximum rainfall amount and equitable threat scores. The improved forecast in the experiment resulted from the development of well-identified mesoscale convective systems by intensified low-level winds and their consequent convergence near the rainfall area.