Impacts of Multigrid NLS-4DVar-based Doppler Radar Observation Assimilation on Numerical Simulations of Landfalling Typhoon Haikui (2012)

Abstract

We applied the multigrid nonlinear least-squares four-dimensional variational assimilation (MG-NLS4DVar) method in data assimilation and prediction experiments for Typhoon Haikui (2012) using the Weather Research and Forecasting (WRF) model. Observation data included radial velocity (Vr) and reflectivity (Z) data from a single Doppler radar, quality controlled prior to assimilation. Typhoon prediction results were evaluated and compared between the NLS-4DVar and MG-NLS4DVar methods. Compared with a forecast that began with NCEP analysis data, our radar data assimilation results were clearly improved in terms of structure, intensity, track, and precipitation prediction for Typhoon Haikui (2012). The results showed that the assimilation accuracy of the NLS-4DVar method was similar to that of the MG-NLS4DVar method, but that the latter was more efficient. The assimilation of Vr alone and Z alone each improved predictions of typhoon intensity, track, and precipitation; however, the impacts of Vr data were significantly greater that those of Z data. Assimilation window-length sensitivity experiments showed that a 6-h assimilation window with 30-min assimilation intervals produced slightly better results than either a 3-h assimilation window with 15-min assimilation intervals or a 1-h assimilation window with 6-min assimilation intervals.