A Preliminary Study on the Application of High-frequency Cycling Updating Approaches to the GRAPES-TYM Model System

The GRAPES-TYM model system, an operational tropical cyclone numerical forecast system run at the China Meteorological Administration (CMA), is a cold start forecast with an initial field which is based on the GFS analysis data from the National Centers for Environmental Prediction (NCEP) and employs an extra vortex relocation and bogus scheme. Recently, more observations with higher spatial and temporal resolutions have become available and the request for more detailed precipitation and winds structure forecasts of tropical cyclone (TC) near the shore and inland are proposed, so that using these observation data effectively to improve TC numerical forecast is very important. Consequently, the current process of GRAPES-TYM operational system may not satisfy the demands on higher resolution prediction of typhoon and be unsuitable in the future. To improve the typhoon numerical forecast skills, a more wide and frequent application of the observation data to the GRAPES-TYM system should be considered. In this study, a series of sensitive experiments has been designed to estimate the impacts of high-frequency cycling updating and warm start forecast approaches in the GRAPES-TYM system on typhoon numerical forecast. There are mainly three strategies of high-frequency cycling updating and warm start forecast are employed in the GRAPES-TYM model system to design these sensitive experiments that the full cycling, the partial cycling and the dynamic initialization (DI) approaches. And this study also discusses some considerations primarily for these high-frequency cycling updating approaches being applied in the GRAPES-TYM system.

When applying the partial cycling approach to the GRAPES-TYM system, the scheme of both using vortex relocation and data assimilation techniques at the time of warm start analysis is more effective, especially to the weaker typhoon cases. While the typhoon forecast seems to not benefit more from the more outer loops procedure during the data assimilation. When applying the full cycling approach in the GRAPES-TYM system, the scheme of using vortex bogusing technique during each warm start analysis often produces a much stronger typhoon. The data assimilation cycles at 6-h intervals facilitate more improvement of typhoon intensity forecast skills in the 24-h early warning area than that at 3-h intervals. Both of the full cycling and partial cycling strategies improve the 24-h precipitation forecast of typhoon in the location and pattern aspects obviously, and also make a significant improvement in the intensity forecast of a typhoon in the 24-h early warning area, but do not enhance the forecast skills of a typhoon outside of the 48-h early warning area. Applying the DI approach often decreases the model integrate stability. Applying the DI approach alone or with the partial cycling approach together both enhance the spiral feature of typhoon 24-h precipitation forecast.