华南区域高分辨率数值模式前汛期预报初步评估

PRELIMINARY EVALUATION OF FORECAST SKILL OF GRAPES GUANGZHOU REGIONAL MODELING SYSTEM

关键词:  GRAPES_GZ  降水预报  检验评估  FSS评分

Key words:  GRAPES_GZ  precipitation forecast  verification and evaluation  Fraction Skill Score

摘要:

基于华南地区自动站逐小时观测资料，采用传统站点评分、邻域法等评估华南区域高分辨率数值模式(包括GRAPES_GZ_R 1 km模式和GRAPES_GZ 3 km模式)对降水、地面温度和风场等要素的预报能力。结果表明：GRAPES_GZ_R 1 km模式的降水预报技巧优于GRAPES_GZ 3 km模式，模式预报以正偏差为主。对于不同起报时间的预报，00时(世界时，下同)起报的预报效果优于12时。GRAPES_GZ_R 1 km模式的TS评分是GRAPES_GZ 3 km模式的两倍以上，对不同降水阈值的评分均较高。分数技巧评分(FSS)显示GRAPES_GZ_R 1 km模式6 h累计降水预报在0.1 mm、1 mm及5 mm以上的降水均可达到最低预报技巧尺度，对所检验降水对象的空间位置把握能力更好。2 m气温和10 m风速检验结果表明两个模式均能较好把握广东省温度的分布特征，GRAPES_GZ_R 1 km模式对2 m气温预报结果优于GRAPES_GZ 3 km模式，预报绝对误差更小；两个模式对风速的预报整体偏强，预报偏差在1~4 m/s之间，但相比之下GRAPES_GZ 3 km模式在风场预报上表现更好。GRAPES_GZ_R 1 km模式的2 m气温和10 m风速预报偏差随降水过程存在明显波动，强降水过后温度预报整体偏低，风速预报偏强，在模式产品订正、使用等需要考虑模式对主要天气系统的预报情况。总的来说，GRAPES_GZ_R 1 km模式的预报产品具有较好的参考价值。

Abstract:

Based on hourly observational data from automatic stations in south China, traditional site scoring and neighborhood methods are used to evaluate the forecast skill of GRAPES Guangzhou Regional Modeling System (including GRAPES_GZ_R 1 km model and GRAPES_GZ 3 km model) in forecasting precipitation, surface temperature, and wind fields. The analysis of the results shows that the precipitation forecasting skills of the GRAPES_GZ_R 1 km model are better than those of the GRAPES_GZ 3 km model, and the deviation of GRAPES_GZ_R 1 km forecasts from observations is mainly positive. The Threat Score (TS) of rainfall forecast by GRAPES_GZ_R 1 km is significantly improved at all thresholds and is more than twice that of GRAPES_GZ 3 km. But GRAPES_GZ_R 1 km only has the highest score in the first 3 hours, while its TS decreases gradually with the increase of integration time and precipitation threshold. The Fraction Skill Scores (FSS) of GRAPES_GZ_R 1 km show improvement in both 6 h and 24 h accumulated precipitation forecast. Meanwhile, GRAPES_GZ_R 1 km can achieve the lowest forecast skill scale for rainfall above 0.1 mm, 1 mm and 5 mm, while GRAPES_GZ 3 km usually fails to reach the lowest forecast skill scale. The forecast of location and intensity of rainfall has an overall improvement. Daily time evolution of root mean square errors for 2 m temperature forecast are similar, while the amount predicted by GRAPES_GZ_R 1 km is less than that by GRAPES_GZ 3 km. The rainfall and 2 m temperature forecast have made remarkable progresses. However, it is apparent that GRAPES_GZ 3 km performs better in forecasting 10 m wind fields, as mean bias and root mean square errors of 10 m wind field forecast are less than those by GRAPES_GZ_R 1 km. Mean errors are reduced by about 1～2 m/s. In addition, the bias of 2 m temperature and 10m wind speed forecasts by GRAPES_GZ_R 1 km fluctuates significantly with the precipitation process. The 2 m temperature forecast is generally lower and the 10 m wind speed is stronger than observations. In general, forecast products of GRAPES_GZ_R 1 km have good reference value.