-
Successful Olympic services cannot be separated from continuous improvement to operational technology. Besides the scientific achievements mentioned above, several real-time operational systems were also improved and used for the Beijing 2022 Winter Olympics according to the Science and Technology Winter Olympics Special Program. Based on both operational and experimental modeling systems, a whole set of seamless, refined numerical forecast model technologies and forecasting systems were established (Table 1). These NWP systems are highlighted below.
Model Domain Horizontal resolution (km) Forecast length (hour) Run frequency (hourly) Main technological improvements CMA-GEPS Global 50 360 6 Vertical coordinate upgrades; enhancements of initial SV perturbations CMA-GFS Global 25 240 6 Correction algorithm to mass conservation; 4DVar; satellite data assimilation; microphysics and cumulus parameterization CMA-TYM Asia 9 120 6 Static data (MODIS); SAS convection scheme; cloud analysis CMA-REPS China 10 84 6 3D rescaling masks of initial perturbation; revisions to background fields; SPP model perturbation scheme CMA-MESO_3km China 3 72 3 Terrain filtering; 3D reference profile; boundary layer scheme based on C-P grid CMA-HREPS Huabei region 3 36 6 Error analysis over complex terrain; data assimilation–related ensemble techniques; multi-scale blending CMA-MESO_1km Huabei region 1 24 1 Dynamics (more vertical levels; hybrid coordinates); 3DVar (B Matrixes; multi-analysis; IAU initialization); microphysics (new Noah, new static data; double-moment microphysical scheme Downscaling Chongli and Yanqing competition area 0.1 (100 m) 24 1 CALMET wind downscaling; small-scale atmospheric dispersion forecasts Table 1. Atmospheric model configurations for the Beijing 2022 Games.
CMA-GFS is the CMA’s global forecasting system, wherein a new hybrid vertical coordinate (87 vertical layers with 0.1 hPa as the model top) was implemented. The new coordinate improved the forecasting capability at upper levels. The horizontal resolution is 25 km globally, and a 4DVar data assimilation system uses the maximum amount of satellite and conventional observations from global sources and generates ICs for the global forecasts. It makes a 10-day forecast at 0000/1200 UTC and a 5-day forecast at 0600/1800 UTC.
CMA-GEPS stands for CMA Global Ensemble Prediction System, in which a total energy norm–based approach with singular vectors (SVs) is implemented to create initial perturbations. These SV perturbations are closely related to the development of the CMA global 4DVar system. The Stochastically Perturbed Parameterization Tendencies (SPPT) scheme was applied to represent the model uncertainty. There are 31 ensemble members with a forecast lead time of 15 days.
CMA-TYM is the CMA’s operational Typhoon Forecast Model, based on the CMA-MESO software infrastructure. The key techniques include a vortex initialization scheme, physical process tuning, and highlights on data assimilation. It replaced the original 10-km resolution CMA-MESO and extended its forecast domain from China to Asia and the western Pacific.
CMA-REPS is the CMA’s Regional Ensemble Prediction System. It uses an Ensemble Transform Kalman Filter for IC perturbation, and the SPPT scheme for model perturbation. A cloud analysis scheme was added to each ensemble member, which assimilates radar and satellite data. The model’s horizontal resolution is 0.1°, and the forecast length is 84 h.
CMA-MESO_3km is a 3-km fast-cycle assimilation version of CMA-MESO. The main improvements include better calculation accuracy and stability of the model dynamic framework, improved microphysics for the high-resolution model, and the establishment of a convection-resolving assimilation system and land surface data assimilation system to make use of conventional and unconventional local dense data such as those from radar, wind profile radar, FY-4A, surface precipitation, and near-surface measurements.
CMA-HREPS is an experimental 3-km cloud-resolving High-Resolution Ensemble Prediction System. Using the improved CMA-MESO model, data assimilation system, IC perturbation scheme, and model physics perturbation schemes developed for the Beijing 2022 Winter Olympics weather forecasting and research program, CMA-HREPS was established for the Beijing 2022 Games. The first-order Markov process-based model stochastic perturbation technique was used to perturb the model physics (Li et al., 2008; Fan et al., 2022). It has 15 members and runs two cycles per day (0000 and 1200 UTC) to 36 h (extended to 60 h at a later stage). Bias correction of ensemble forecasts, probabilistic products, and verification are also part of this ensemble prediction system.
CMA-MESO_1km is a 1-km hourly cycle assimilation version of CMA-MESO. In order to overcome the difficulties resulting from complex terrain and high-frequency observations, a special design involving ground temperature/humidity observation algorithms, high-resolution static data, optimization of the radiation parameterization scheme, and a mixed-scale scheme was introduced during the development of this hourly cycle assimilation system. With these improvements, the quality of 2-m temperature, 10-m wind field, and precipitation forecasts was obviously improved.
A 100-meter dynamical downscaling model and a fine-scale fireworks dispersion forecasting model were two other capabilities developed for the Beijing 2022 Games. The forecast wind fields from CMA-MESO_1km were downscaled to different height levels based on the complex mountain conditions of sporting venues. It ran 24 times per day to provide forecasters with very high spatial (100 m) and temporal resolution (every 10 min) gridded forecasts, 24 h per cycle. During the Winter Olympics, it provided “100 m-level, minute-level” wind field forecasts for Chongli station and the Yanqing competition sites. The fireworks dispersion forecasting used small-scale Lagrangian smoke diffusion forecasting technology. The 100-m resolution fireworks diffusion products were applied to some special events like the opening and closing ceremonies.
Model | Domain | Horizontal resolution (km) | Forecast length (hour) | Run frequency (hourly) | Main technological improvements |
CMA-GEPS | Global | 50 | 360 | 6 | Vertical coordinate upgrades; enhancements of initial SV perturbations |
CMA-GFS | Global | 25 | 240 | 6 | Correction algorithm to mass conservation; 4DVar; satellite data assimilation; microphysics and cumulus parameterization |
CMA-TYM | Asia | 9 | 120 | 6 | Static data (MODIS); SAS convection scheme; cloud analysis |
CMA-REPS | China | 10 | 84 | 6 | 3D rescaling masks of initial perturbation; revisions to background fields; SPP model perturbation scheme |
CMA-MESO_3km | China | 3 | 72 | 3 | Terrain filtering; 3D reference profile; boundary layer scheme based on C-P grid |
CMA-HREPS | Huabei region | 3 | 36 | 6 | Error analysis over complex terrain; data assimilation–related ensemble techniques; multi-scale blending |
CMA-MESO_1km | Huabei region | 1 | 24 | 1 | Dynamics (more vertical levels; hybrid coordinates); 3DVar (B Matrixes; multi-analysis; IAU initialization); microphysics (new Noah, new static data; double-moment microphysical scheme |
Downscaling | Chongli and Yanqing competition area | 0.1 (100 m) | 24 | 1 | CALMET wind downscaling; small-scale atmospheric dispersion forecasts |