气象
氣象
기상
METEOROLOGICAL MONTHLY
2010年
1期
54-58
,共5页
胡胜%罗兵%黄晓梅%梁巧倩%沃伟峰
鬍勝%囉兵%黃曉梅%樑巧倩%沃偉峰
호성%라병%황효매%량교천%옥위봉
风暴识别%核抽取%风暴追踪%风暴预报%平均绝对误差
風暴識彆%覈抽取%風暴追蹤%風暴預報%平均絕對誤差
풍폭식별%핵추취%풍폭추종%풍폭예보%평균절대오차
storm identification%cell nucleus extraction%storm tracking%storm forecast%mean absolute error
介绍了临近预报系统"SWIFT"(Severe Weather Integrated Forecasting Tools)中的风暴产品的设计,包括风暴识别、风暴追踪和风暴预报.在识别风暴时,采用了多反射率因子阈值、特征核抽取和相近单体处理技术,并保留远距离上的强的2D风暴.该方法在面对成串或成簇多单体时.能够分离多个单体核,并准确定位.在风暴追踪和预报算法中,对当前时刻识别出来的风暴,利用匹配方案,将其与前1时刻的风暴建立对应关系,追寻历史轨迹,匹配方案是在空间位置相关的前提下,按照相似原则进行;风暴预报采用TREC(Tracking Radar Echoes by Correlation)技术获取的移动矢量场进行外推,提供未来1小时内的风暴移动位置.在北京奥运会天气预报示范项目(Forecast Demonstration Project,简称FDP)第二次测试期间,该风暴产品得到应用.分析表明:在预报时效为30分钟时,风暴产品在X轴和Y轴上的平均绝对误差为7.1和6.2 km,样本数为3891个;随着预报时效的增加,风暴产品的平均绝对误差增大,且在经向上的误差略大于纬向上;在径向上,风暴产品的预报出现了系统性的偏慢,而在纬向上,预报出现了系统性的偏快.
介紹瞭臨近預報繫統"SWIFT"(Severe Weather Integrated Forecasting Tools)中的風暴產品的設計,包括風暴識彆、風暴追蹤和風暴預報.在識彆風暴時,採用瞭多反射率因子閾值、特徵覈抽取和相近單體處理技術,併保留遠距離上的彊的2D風暴.該方法在麵對成串或成簇多單體時.能夠分離多箇單體覈,併準確定位.在風暴追蹤和預報算法中,對噹前時刻識彆齣來的風暴,利用匹配方案,將其與前1時刻的風暴建立對應關繫,追尋歷史軌跡,匹配方案是在空間位置相關的前提下,按照相似原則進行;風暴預報採用TREC(Tracking Radar Echoes by Correlation)技術穫取的移動矢量場進行外推,提供未來1小時內的風暴移動位置.在北京奧運會天氣預報示範項目(Forecast Demonstration Project,簡稱FDP)第二次測試期間,該風暴產品得到應用.分析錶明:在預報時效為30分鐘時,風暴產品在X軸和Y軸上的平均絕對誤差為7.1和6.2 km,樣本數為3891箇;隨著預報時效的增加,風暴產品的平均絕對誤差增大,且在經嚮上的誤差略大于緯嚮上;在徑嚮上,風暴產品的預報齣現瞭繫統性的偏慢,而在緯嚮上,預報齣現瞭繫統性的偏快.
개소료림근예보계통"SWIFT"(Severe Weather Integrated Forecasting Tools)중적풍폭산품적설계,포괄풍폭식별、풍폭추종화풍폭예보.재식별풍폭시,채용료다반사솔인자역치、특정핵추취화상근단체처리기술,병보류원거리상적강적2D풍폭.해방법재면대성천혹성족다단체시.능구분리다개단체핵,병준학정위.재풍폭추종화예보산법중,대당전시각식별출래적풍폭,이용필배방안,장기여전1시각적풍폭건립대응관계,추심역사궤적,필배방안시재공간위치상관적전제하,안조상사원칙진행;풍폭예보채용TREC(Tracking Radar Echoes by Correlation)기술획취적이동시량장진행외추,제공미래1소시내적풍폭이동위치.재북경오운회천기예보시범항목(Forecast Demonstration Project,간칭FDP)제이차측시기간,해풍폭산품득도응용.분석표명:재예보시효위30분종시,풍폭산품재X축화Y축상적평균절대오차위7.1화6.2 km,양본수위3891개;수착예보시효적증가,풍폭산품적평균절대오차증대,차재경향상적오차략대우위향상;재경향상,풍폭산품적예보출현료계통성적편만,이재위향상,예보출현료계통성적편쾌.
The storm series algorithms in the SWIFT (Severe Weather Integrated Forecasting Tools), including storm cell identification, storm tracking and storm forecast, are discussed. Storm cell identification algorithm tests the intensity and continuity of the objective echoes by multiple-prescribed thresholds to build 3-D storms. It uses multiple reflectivity thresholds, newly designs the techniques of cell nucleus extraction and close-spaced storms processing, and therefore is capable of identifying embedded cells in multi-cellular storms. The strong area components at a long distance are saved as 2-D storms. Storm cells identi-fied in two consecutive volume scans are associated temporally to determine the cell tracking. The distance between the centroid of each cell detected in the current volume scan and each of the first-guess location is calculated to check distance correlation. Those similar storms with distance correlation are matched, The motion vector for each storm is computed by using the technique of TREC (Tracking Radar Echoes by Correlation), and storm locations in the next hour are provided. During the second trial of the FDP (Forecast Demonstration Project) in 2007, these algorithms have been applied. It is found that 3891 storms are identified and the mean absolute errors in the X-axis and Y-axis for 30-min storm forecast are 7. 1 and 6.2 km respectively. With the increase of forecast time length, the mean absolute errors of the storm product become larger, and the X-axis error is greater than that in the Y-axis. The statistical analysis also shows that the mean forecast velocity in the X-axis is less than the mean actual velocity of storms, but the conclusion is contrary in the Y-axis.