3种全球环流预报场近地层风速突变预报能力评估

Forecast Performance Verification of Surface Wind Ramp for Three Global Circulation Forecast Fields

本文采用Bernaola—Galvan启发式分剖算法对3种全球环流预报场预报风速序列以及气象站和测风塔的实测风速序列分别进行风速突变事件的挑选，并综合挑选结果对3种全球环流预报场对我国风速突变事件的预报能力进行检验和评估。结果显示：ECMWF全球环流预报场风速突变事件的预报效果最优，其次是GFS，然后为T639，但是在青藏高原和甘肃北部地区，GFS要优于ECMWF，在我国沿海尤其是南部沿海地区，T639为最佳；ECMWF在新疆北部、甘肃北部、内蒙古中西部、河北和山西风速突变预报准确率较高，约为30％以上，其次为东北、山东、河南和江苏，约为10％～30％，其余地区预报准确率基本上在20％以下，尤其在西藏和四川大部分地区预报准确率为0．T639预报场在沿海地区的预报准确率基本在30％左右；ECMWF预报场在我国大部分地区预报相位偏差在3h以内，幅度偏差基本上在3m／s之内，仅在新疆北部、甘肃北部、青海北部、内蒙古及辽宁地区相位偏差为3-6h之间，幅度偏差大于3m／s，在新疆北部和内蒙古东部幅度偏差达到6m／s。

The heuristic segmentation algorithm proposed by Bernaola-Galvan was used to detect wind ramp of wind speed time series from three global circulation forecast fields （ECMWF, GFS and T639） and observational data. Forecast performance of wind ramp was verified in China from June 2010 to May 2011. We found that forecast performance of the ECMWF field was the best overall, followed by GFS and T639. The forecast performance of GFS was better than ECMWF in the Tibetan Plateau and northern Gansu, and the forecast performance of T639 was best in coastal areas, especially in southern coastal areas. The forecast accuracy of ECMWF was higher （about 30%） in northern Xinjiang, northern Gansu, and western and central Inner Mongolia, Hebei and Shanxi. The next was about 10%-20% in the Northeast, Shandong, Henan and Jiangsu areas. There was less than 20% in other areas, especially in Xizang and Sichuan, the forecast accuracy of most stations was zero. Forecast accuracy of T639 was about 30% in coastal areas. The forecast phase bias of ECMWF was less than 3h in most parts of China, and the amplitude bias was less than 3m/ s. In some areas of northern Xinjiang, northern Gansu, northern Qinghai, and most areas of Inner Mongolia and Liaoning, phase bias and amplitude bias were 3-6h and greater than 3rrds, respectively, especially in northern Xinjiang and east of Inner Mongolia where the amplitude bias was greater than 6 m/s.