Search

Reliable medium-range weather forecasts are crucial for both science and society. Although weather predictions primarily rely on numerical weather models, the artificial intelligence (AI) weather big models have shown potential for accurate weather forecasts with reduced computational costs. However, prediction skills of big models remain uncertain, particularly in East Asia, which limits further application of weather AI models. To systematically evaluate predictive capabilities of Pangu, FuXi, and GraphCast models over East Asia, their prediction results are focusing on 500 hPa geopotential height, 2 m air temperature, 10 m wind speed, precipitation, and track of tropical cyclones.ECMWF reanalysis V5 (ERA5) datasets are utilized to provide the initial conditions for big models, and to assess their predictive skill. Additionally, precipitation observations and China Meteorological Administration tropical cyclone datasets are utilized to access big models as well. FuXi shows the highest forecasting skills among 3 big models for 500 hPa geopotential height. The forecast from FuXi is reliable for up to 9.75 days, while the forecasts from Pangu and GraphCast are reliable for 8.75 days and 8.5 days, respectively. For 2 m air temperature forecasting, FuXi presents higher skills with an averaged temporal anomaly correlation coefficient (TCC) ranging from 0.48 to 0.91, while TCCs of Pangu and GraphCast are 0.43-0.91 and 0.38-0.83, respectively. Among 3 models, only FuXi and GraphCast provide precipitation forecasts. FuXi shows higher prediction skill compared to GraphCast in forecasting precipitation, light rain, and moderate rain; however, GraphCast has advantage in heavy rain forecast. As the lead time increases, the threat scores (TSs) of FuXi for rainfall, light rainfall and moderate rainfall are 0.22-0.41, 0.15-0.24 and 0.06-0.22, respectively. The model demonstrates higher skill in the northern and southeastern regions of China. For predicting the track of cyclones, Pangu model demonstrates superior predictive skill. As the lead time increases from 6 hours to 240 hours, biases of Pangu's prediction track increase from 17.5 km to 1850 km.The study focuses on the prediction skill of various AI big models through TCC, spatial anomaly correlation coefficient, and TS. Generally, the performance of FuXi is superior for most elements. And reasonable evaluation of AI model is helpful for the development of AI models.

Using daily precipitation model data produced by the second version of Climate Forecasting System of NCEP (CFSv2),the ensemble prediction method is used to carry out regional summer precipitation forecast. Three methods,i.e.,the root mean square error(RMSE) and accuracy(ACCU) of the onset and the ending time of Meiyu,the RMSE of the length of Meiyu,and anomaly sign consistency rate (Pc) of the average rainfall amount of Meiyu,are used to evaluate the forecast ability of the model to the Meiyu characteristics in Hubei province. The results show that the onset of Meiyu forecast is 13 days in advance,the ACCU is above 0.5,and the RMSE is less than 3 days. The ending of Meiyu forecast is 14 days in advance,the ACCU is above 0.5,and the RMSE is less than 3 days. The length of Meiyu period forecast is 14 days in advance,and the RMSE is less than 5 days. The average rainfall amount of Meiyu forecast is 14 days in advance,and the Pc is above 0.5. In general,the aging time of Meiyu characteristics forecast is about 14 days. The CFSv2 model shows certain forecast skill for regional summer precipitation during the Meiyu extended-range period.

Based on the convection-allowing scale high resolution (3 km) ensemble prediction system which has been operationally implemented in meteorological services of Chongqing and includes initial perturbation, model physical process perturbation and lateral boundary perturbation in it, the perturbation to model static topographic height field is realized by adopting different topographic interpolation schemes and smoothing schemes to generate static topographic height field of each ensemble members to express the uncertainty of the topographic conversions in the numerical model to some extent. Using these perturbations, we have performed a month consecutive ensemble prediction tests and the comparative analysis with respect to the ensemble forecast results without model topography perturbation. The results show that (1) after adding the model topography perturbation scheme into the ensemble prediction system, the topographic height spread among ensemble members has a good correspondence with the real topographic height gradient, and both spatial distribution characteristics are very similar. The topographic height spread in the plain area is small, while that in the plateau area or mountainous area with complicated terrain is large. (2) The addition of model topography perturbation scheme increases the ensemble different total energy (DTE) whose increase ratio in the low-level is larger than that in the mid- and upper-level. The enhancement of DTE in the short forecast lead time (12 h) is the most significant. It gradually decreased with the forecast lead time becoming longer. Large enhancement centers of DTE mainly appear in the areas with complicated topography and the great topographic height spread among the ensemble members. (3) In general, the addition of model topography perturbation scheme can improve the probability forecast skill and ensemble mean forecast skill of 24 h accumulated rain to some extent. It can optimize the ensemble distribution of all the elements in upper air, and 2 m air temperature, 10 m zonal wind, 10 m meridional wind and other elements near the ground level to a certain extent without the negative effect in the ensemble mean forecast error and ensemble spread.