Spin-up characteristics with three types of initial fields and the restart effects on forecast accuracy in the GRAPES global forecast system

Abstract

Abstract. The spin-up refers to the dynamic and thermal adjustments made at the initial stage of numerical integration in order to reach a statistical equilibrium state. The analyses on the characteristics and effects of spin-ups are of great significance for optimizing the initial field of the model and improving its forecast skills. In this paper, three different initial fields are used in the experiments: the analysis field of four-dimensional variational (4D-VAR) assimilation, the 3 h prediction field in the operational forecasting system, and the Final (FNL) Operational Global Analysis data provided by National Centers for Environmental Prediction (NCEP). Following this, the characteristics of spin-ups in the version 2.3.1 of GRAPES (Global/Regional Assimilation and Prediction System) global forecast system (GRAPES_GFS2.3.1) under different initial fields are compared and analyzed. In addition, the influence of the lost cloud-field information on the spin-up and forecast results of the GRAPES model in the current operation is discussed as well. The results are as follows. With any initial field, the spin-up of GRAPES_GFS2.3.1 has to go through two stages – the dramatic adjustment in the first half-hour of integration and the slow dynamic and thermal adjustments afterwards. The spin-up in GRAPES_GFS2.3.1 lasts for at least 6 h, and the adjustment is gradually completed from lower to upper layers in the model. Therefore, in the evaluation of the GRAPES_GFS2.3.1, the forecast results in the first 6 h should be avoided, and the GRAPES_GFS2.3.1 with its own analysis field performs better than the one using FNL reanalysis data for the cold start in the spin-up because the variations in amplitude of the temperature and humidity tendency are smaller and the spin-up time is slightly shorter. Based on the 4D-VAR assimilation analysis field, the forecast in the operational model is artificially interrupted and restarted after 3 h of integration. In this process, as the cloud-field information is not retained, the spin-up should repeat in the model. The characteristics of spin-up are mostly consistent with those using the 4D-VAR assimilation analysis field as the initial field. However, as the cloud-field information is not retained in the current operation, the hydrometeor content in the atmosphere at the early stage of the forecast is underestimated, affecting the calculation accuracy of the radiation and causing a systematic positive bias of temperature and geopotential height fields at 500 hPa. In addition, the precipitation is also underestimated at the early stage of the simulation, affecting the forecast of typhoon tracks.