By introducing the statistics concept of "degree of instability" and dividing Southwest China into three kinds of areas, namely, the plateau, the side slope and the basin area in consideration of the terrain complexity there, numerical model GRAPES_GFS and EC's prediction stabilities, their temporal and spatial variations, the influence of the terrain on the models' performance as well as the difference between the two models were analyzed subjectively and objectively based on their predictions over the period from July, 2016 to September, 2017. The results showed that GRAPES_GFS's prediction instability of geopotential height and temperature fields appeared higher in the north of the area, while higher instability of relative humidity and wind speed was around the Tibet Plateau to the north and east. The instability had obvious seasonal fluctuation with different phases and amplitude as different elements. Terrain impacted mainly on the prediction error value of temperature and wind direction. For both relative humidity and wind speed, its impact is more on the rate of error growth. The "lower prediction" led to a high failure rate of GRAPES_GFS in 2 m temperature prediction, and "missing" was its major error source of weather systems prediction. GRAPES_GFS predicted the location of rainfall in the Southwest China by about effective rate of 50%, although its prediction of rainfall intensity was usually lower than the observation. The two models had little essential difference in the characteristics of prediction error. Compared to GRAPES_GFS, EC had less errors and higher stability.
JIANG Li-Juan, et al
.2019. Prediction stability of GRAPES_GFS in Southwest China and the relationship between its error and the terrain[J].
Torrential Rain and Disasters, 38(1): 59-65.