Abstract:
The imbalance of the analysis variables created by introducing irregular high frequency and high density observations into the high-resolution analysis gives rise to unphysical gravity-inertia waves in the model integration, and thus affects the forecast quality and stability of operational models. In this paper, the diabatic digital-filtering initialization and incremental digital filter initialization are applied to 3 km resolution GRAPES regional model to suppress noises in the early forecast hours, and the impact on analysis and forecast with different cutoff periods of initializations are evaluated. The analysis and forecast experiments show that the filtering effect relies on the cutoff period. If the cutoff period is longer, more high frequency signals are removed. In the 3 km resolution modelling, both 30 min and 15 min cutoff periods initialization can remove high frequency noises and retain meaningful analysis information. The impact produced by different cutoff period initialization schemes in the GRAPES-MESO system on the model variables can be neglected, whereas the accumulated effects of initialization schemes are gradually observed in the GRAPES-RAFS system. The results show that wind and precipitation forecast qualities are obviously improved with a 15 min cutoff period initialization compared to that with the 2 h cutoff period. The impact of digital filter initialization scheme on high-resolution model prediction depends on the cutoff period. The cutoff period of 30 min or 15 min is suitable for 3 km resolution model, and the initialization scheme of incremental digital filter demonstrates obvious advantages.