To investigate impacts of initialization through DA with GSI in HWRF, three sets of cycle forecast experiments were conducted. Each set contains 73 five-day forecast experiments. The HWRF configuration is almost the same as the recent operational HWRF (H217). One set includes a HWRF cycle run which includes vortex initialization without GSI, second run is HWRF with GSI , and third is HWRF with GSI and high-resolution AMV derived from Himawari-8 (H8AMV).
H8AMV produced the smallest track errors for forecast hours of 84-126 among these experiments. One of the factor for the improvement can be explained by reduction of northward track bias. However, the intensity forecast error was the largest for almost all forecast lead times in H8AMV. This could be caused by larger negative (weak) bias of intensity in H8AMV as compared to the other two experiments. The track forecast error of GSI is a little larger than HWRF and the intensity error is comparable with HWRF.
Our preliminary results indicate that the assimilation of high resolution AMV leads to improved track forecasts but a remedy for weak bias is needed. Detailed differences in TC development and structure among the three experiments will be presented.