消除偏差集合平均在黄海渤海大风预报中的应用_英文_

消除偏差集合平均在黄海渤海大风预报中的应用_英文_ Applications oBf ias-removed Ensemble Mean in the Gale Forecasts over the Yellow Sea and the Bohai Sea 1,2 1* 3ZHU Hua,ZHI Xie-fei,YU Yong -qing 1, Key Laboratory of MeteorologicaDil saster of Ministry of Education,Nanjing University of InformationS cience , Technology,Nanjing 210044, China; 2, Public MeteorologicalSer vice Center of Anhui Meteorological Bureau,Hefei 230061,China; 3, Shengli Oil Field Meteorological Ob- servatory,Dongying 257000,China Abstract Based on the daily sea surface wind field prediction data of Japan Meteorological Agency ( JMA) forecast model,National Centers for Environmental Prediction ( NCEP GFS) model and U, S, Navy Operational Global Atmospheric PredictionSys tem ( NOGAPS) model at 12: 00 UTC from June 28 to August 10 in 2009,the bias-removed ensemble mean ( BRE) was used to do the forecast test on the sea surface wind fields,and the root-mean-squaree rror ( RMSE) was used to test and evaluate the forecast results, The results showed that the BRE considerably reduced the RMSEs of 24 and 48 h sea surface wind field forecasts,and the forecast skill was superior to that of the single model forecast, The RMSE decrea- ses in the south of central Bohai Sea and the middle of the Yellow Sea were the most obvious, In addition,the BRE forecast improved evidently the forecast skill of the gale process which occurred during July 13 ) 14 and August 7 in 2009, The forecast accuracy tohfe wind speed and the gale lo- cation was also improved, Key words Bias-removed ensemble mean; Gale over the Yellow Sea and the Bohai Sea; Forecast skill; China The scientific research of the offshore gale has alwaysplicity, In this paper,we have carried out the BRE forecaste x- been rather important, How to use the numerical forecastp rod- periments of the sea surface wind fields over the Yellow Sea ucts,especially the multi-model ensemble numerical forecast and Bohai Sea during the summer of 2009,which improved products,to improve the accuracy of gale forecast is a new considerably the accuracy of offshorega le forecast, ,1, problem at present, Lorenz firstly put forward the thought of ensemble numeri- 1Data and method,2, cal forecast,Le ith combined the nonlinear theory of Lorenz 1, 1 Data The data used in this study include the ensemblewith the dynamic random forecast t heory proposed by Ep- forecastda ta of the Japan Meteorological Agency ( JMA) fore- ,3, steinand put forward Monte Carlo method which could be cast model,the National Centers for Environmental Prediction ,4, used into the operational forecast,Si nc e then,the ensemble ( NCEP GFS) model,U, S, Navy Operational Global Atmos- forecastha s found its application, In the 1990s,the ensemble pheric Prediction System ( NOGAPS) model and NCEP / NCAR forecast technique was used in many national meteorological reanalysis data, The ensemble forecast data of the daily sea ,5 ) 6, services,K rishnamurti et al, firstly put forward the super- surface wind field forecastsa t 12: 00 UTC were taken from the ensemble forecast method based on the multi-model forecast above mentioned three models during the period from June 28 ,7,results, The training period proved that the multi-model en- to August 10 in 2009 over the Yellow Sea and Bohai Sea, The semble technique was the efficient method which sufficiently forecast region of the JMA model is located in the area of used the forecast results of multi-center models to reduce the ,7 ) 9, 115, 0?) 130, 0? E,25, 0? ) 42, 5? N with a spatial resolution of systematic bias of models,Z hi Xi efei et al, used TIGGE 2, 5? ×2 , 5?, The forecast region of the GFS and NOGAPSdata to conduct the multi-model ensemble forecast test of the model is located in the area of 116, 0? )13 0, 0? E,26, 0? )surface temperature in the Northern Hemisphere and found that 42, 0? N with a spatial resolution of 1, 0? × 1, 0?, The forecast the forecaste rror of the bias-removed ensemble mean ( BRE) ,10 ) 11, time was 24 ) 48 h with an interval of 24 h, NCEP / NCAR rean- method was smaller than that of each single model,I n alysis data were taken as the "observed value" ,T he daily glob-addition,the BRE forecast not only successfully reduced the al sea surface wind fields with a spatial resolution of 192 × 94forecaste rror,but also was favorable for its application in the Gauss grid at 12: 00 during the period from June 29 to Augustoperational forecast ofw eather observatories owing to its sim- 11 in 2009 were selected to verify the forecasts, 1, 2 MethodReceived: November 5,2010 Accepted: December 13,2010 N 1 Supported by Chinese Meteorological Administration's Special Funds1, 2, 1 Root-mean-square error ( RMSE) ,R MS E =,( F )? ii =1 N ( Meteorology ) for Scientific Re search on Public Causes 1 22thO) , ( i = 1,2,…,N) ,Fis the forecastv alue of the i sam- ( GYHY200906007) ; Gale Forecast Item of the Shengli Oil Field Ob- i i thservatory ( 2008001) , ple,and Ois the observed value of the i sample, i * Corresponding author, E-mail: zhi@nuist, edu, cn 1, 2, 2 Bias-removed ensemble mean ( BRE) forecast, The N 1 Sea ( 116? ) 128? E,26? ) 42? N) ,T he RMSE of the sea sur- formula oft he BRE forecasti s written as: E = ?( F ) +i = 1 N face wind forecast was calculated to evaluate the forecast skill F)i O, E is the forecast v alue of the bias-removed ensembleof the bias-removed ensemble mean forecasta s well as that of mean,O is the mean observed value in the training period,Fis the single model forecasts, As shown in Fig, 1,for 24 and 48 h i th the forecastv alue of the imodel,F is the mean forecast value forecast,the BRE forecast significantly reducedth e forecaste r- ith ror,and the forecast skill was superior to that of each single of the imodel in the training period,and N is the number of model, models which were taken for the multi-model ensemble forecasts, 2, 1, 2 Geographical distribution otfh e RMSE of the sea sur- face wind forecast, Seen from Fig, 2,the bias-removeden sem- 2Result and analysisble mean forecastr educed the area with RMSE , 2, 9 m / s sig- 2, 1 The BRE forecast experimentsnificantly,thus leading to considerable decrease of the forecast 2, 1, 1 The RMSE of the sea surface wind forecast, The nu-errors, Compared with NCEP GFS forecast,the decrease of merical forecast data of the sea surface fields taken from thethe BRE forecast e rror was very noticeable, and the area JMA,NCEP GFS and NOGAPS models were selected forc on- where the RMSE , 3, 5 m / s in the NCEP GFS forecastd isap- ductingt he multi-model ensemble forecasts duringth e period from peared on the whole, Compared with JMA model whose fore- June 28 to August 10 in 2009, The period from June 28 to July 27 cast skill was the best,the BRE also reduced the forecast er- was chosen to be the training period,and the period from July 28 ror,and the area with RMSE , 1, 7 m / s was reduced to August 10 was the forecast epriod, NCEP / NCAR reanalysis evidently,Th e RMSE decreased to less than 1, 5 m / s in the data were taken as the "observed value" during the same peri- south of cen- tral Bohai Sea, od,The test area was selected in the Yellow Sea and Bohai Fig, 1 The RMSEs of the 24 h ( a) and 48 h ( b ) sea surface wind speed forecast produced by JMA,NOGAPS,NCEP GFS and the bias-removed ensemble mean ( BRE) averaged over the area of 116? ) 128? E,26? ) 42? N from July 28 to Au- gust 10 in 2009 ( unit: m / s) Note: a, BRE; b, JMA; c, NOGAPS; d, NCEP GFS, Fig, 2 Geographical distribution othf e 24 h forecast RMSE of the sea surface wind over the Yellow Sea and Bohai Sea during the period from July 28 to August 10,2009 ( unit: m / s) reduction was to a large extent similar to that of the 24 h fore-As shown in Fig, 3,compared with JMA model,the BRE 48 h forecast reducedt he forecaste rror of the sea surface wind cast, The shaded areas of the NCEP GFS and NOGAPS model over the south ofc entral Bohai Sea,and the area with RMSE , with relatively large forecaste rrors were significantly reducedi n the BRE forecast, 1, 7 m / s was considerablyr educed, The distribution otfh e error Note: a, BRE; b, JMA; c, NOGAPS; d, NCEP GFS, Fig, 3Geographical distribution of the 48 h forecast RMSE of the sea surface wind over the Bohai Sea during the period from July 28 to August 10,2009 ( unit: m / s) Note: a, The observed wind field; b, BRE; c, JMA; d, NOGAPS; e, NCEP GFS, Fig, 4 Comparison of the 24 h wind forecasts issued at 12: 00 UTC on July 13 in 2009 with the observed values ( unit: m / s) 2, 2 A case of the deterministic gale forecast AccordingGFS,NOGAPS models as well as the BRE forecastdu ring the to NCEP gale / NCAR reanalysis data,a gale of force 6 ) 7 on the process,it was found that the wind speed forecast oft he JMA model ( 9 m / s) was smaller than the observed value on Beaufort scale occurred in most areas of the Yellow Sea during July 13,and it failed to forecastt he maximal wind speed center the period from July 13 to July 14 in 2009 ( Fig, 4) ,A t 1 2: 00 UTC on July 13,the maximal wind speed center was located in the central Yellow Sea ( Fig, 4c) ,T he wind speed forecast of the NCEP GFS model ( 13 m / s) was larger than the observed over the central Yellow Sea,and the wind speed maximum value ( Fig, 4e) ,T he BRE forecasti s close to the observed val- reached 12 m / s, The largest wind speed over the southern ue in terms of the wind speed and the location of the maximal Yellow Sea reached 11 m / s, At 12: 00 UTC on July 14,the wind speed center ( Fig, 4b) ,T he BRE forecast oft he wind maximal wind speed center moved slightly toward the south, speed on July 14 was relatively accurate,but the position oft he and the largest wind speed became 12 m / s ( Fig, 4a) ,C om - pared the forecastr esults of the National Meteorological Center maximal wind speed center slightly leaned to the north, Seen from Table 1,in the 24 h forecast ofa gale process issued by with the 24 h forecasts oft he sea surface wind given by JMA, the National Meteorological Center at 12: 00 UTC on July 13, lue, However,the GFS and NOGAPS than the observed va the wind speeds in most areas of the Bohai Sea and the Yellow forecasts were only 16 m / s and smaller than the observed val- Sea on July 14 reached force 7 ) 8 on the Beaufort scale,but ue, The BRE forecast oft he wind speed was relatively accu- the observed wind speed over the Bohai Sea was less than rate,but the gale region of more than 18 m / s was slightly force 6 on the Beaufort scale, It was thus apparent that the smaller than the observation, BRE forecasts were relatively accurate, However,the maximal forecast over the Bohai Sea and the Yellow Sea is-Gale Table 1wind speed center of the forecastsli ghtly leaned to the north, sued by National Meteorological Center Due to the influence of typhooinn summer,the determinis- Wind force Report timeSea areatic forecast oft he gale over the Yellow Sea and the Bohai ea S level is more difficult, At 20: 30 UTC on August 5 in 2009,the Na- Bohai Sea) 54 12: 00 UTC on July 12 tional Meteorological Center issued an early warning of typhoon Northern Yellow Sea5 ) 6Morakot, According to the forecast,a gale of force 7 ) 8 on the Central Yellow Sea 6 ) 7Beaufort scale appeared in the southern Yellow Sea and most 6 ) 7Southern Yellow Sea sea areas of East China Sea on August 7,2009, As shown in 7 ) 8Bohai Sea 12: 00 UTC on July 13 Northern Yellow Sea 7 ) 8Fig, 5,the wind speed in the northern Yellow Sea at 12: 00 UTC Central Yellow Sea 7 ) 8on August 7 reached 18 m / s, Compared the 24 h sea surface 6 ) 7Southern Yellow Sea wind forecast ofJ MA,GFS and NOGAPS with the BRE fore- cast,the JMA forecast oft he wind speed was 4 m / s larger Note: a, The observed wind field; b, BRE; c, JMA; d, NOGAPS; e, NCEP GFS, Fig, 5 Comparison of the 24 h wind field forecast issued at 12: 00 UTC on August 7,2009 with the observation ( unit: m / s) conductedo ver the Yellow Sea and the Bohai Sea during theOverall,the deterministic prediction oft he gale using the period from July 28 to August 10 in 2009, The forecastskil l of BRE forecast t echnique considerably improves the forecast the BRE technique was compared with that of each single mod- skills of the wind speed and the location of the maximal wind el,and the deterministic predictions of the sea surface gale speed regions over the single model forecasts, process during July 13 ) 14 and August 7 in 2009 were verified,Th e results showed that the root-mean-square error 3Conclusion and discussions( RMSE) had been significantly reduced for the2 4 and 48 h BRE Based on the forecasts taken from the JMA,NCEP GFSforecast of the sea surface wind and the forecast skill was and NOGAPS models, the bias-removed ensemble mean superior to that of the single models, The BRE method ( BRE) forecaste xperiment of the sea surface wind has been evidently reduced ,10,ZHI XF，LIN CZ，BAI YQ，et al, Multimodel superensemble fore- the area with RMSE , 2, 9 m / s, The error decrease of the wind speed forecast in the south of central Bohai Sea and central casts of surfacet emperature using TIGGE Datasets / /Collection of ndndYellow Sea was most significant, For the deterministic predic- abstracts of the 2 THORPEX-Asia science workshop,R,, The 2 THORPEX-Asia Science Workshop，2009,tion of a sea surface gale process,the BRE forecastw as supe- ,11, LIN CZ，ZHI XF，HAN Y，et al, Superensemble forecasts ofh terior to the single model forecasts in terms of wind speed and surface temperature based on TIGGE data,J,, Quarterly Journal ofthe gale region,and the gale forecastw as also relatively accu- Applied Meteorology，2009，20( 6) : 710 )712, ( in Chinese) ,rate under the influence of typhooinn summer, ,12,YANG XY，YAO HY，TIAN GY，et al, The climatic change charac- teristics ofs trong wind and the response to global climate warming References in Northeast China,J,, Journal of Anhui Agricultural Sciences，,1, DUAN MQ，WANG PX, Advances in researches and applications of 2010，38( 33) 1889: 4 )18896，18903, ( in Chinese) ,ensemble prediction,J,, Journal of Nanjing Institute ofM eteorology， ,13,FEI JF，LU HC，HAN R, The ensemble forecasting otfr opical cy- 2004，20( 2) 20: 8 )288, ( in Chinese) ,clone intensity based on En-KF data assimilation and bias-correc- ,2,LORENZ EN, A study of the predictability o2f8 -variable atmosphere tion,J,, Acta Meteorologica Sinica，2010，68( 1:) 79 )87, ( in Chi- model,J,, Tellus，1969，21: 739 )759, nese) , ,3,LEITH CE, Theoretical skill of Monte Carlo forecasts,J,, Mon Wea,14, XIONG YM，WU S，JIANG YZ，et al, Guilin thunderstorm and Rev，1974，102: 409 )418,strong wind characteristics and weather situation analysis,J,, Jour- ,4, EPSTEIN S, Stochastic dynamic prediction,J,, Tellus，1969，21:nal of Anhui Agricultural Sciences，2010，38( 5): 2520 ) 2521, ( in 739 )759,Chinese) , ,5,TOTH Z，KALNEY E, Ensemble forecasting at NMC: The generation ,15,SUN FM，HONG RC，WU XQ, Model-set design method with rep- of perturbations,J,, Bull Amer Meteor Soc，1993，74( 12) :231 7 ) resentative points of probability distribution,J,, Control Theory , 2330,Applications，2009，26( 5:) 475 )480, ( in Chinese) , ,6, ZHU YJ, Ensemble forecast: A new approach to uncertainty and ,16,LI J，GU SS, Analysis of a disaster-causing wind process in Anhui th predictability,J,, Adv Atmos Sci，2005，22( 6): 781 )788, on June 5，2009,J,, Journal of Anhui Agricultural Sciences， ,7,KRISHNAMURTI TN，KISHTAWAL CM，LAROW T，et al, Improved 2010，38( 14) 744: 3 )7445，7457, ( in Chinese) ,weather and seasonal climate forecasts from multimodel superen- ,17,LUO QJ，HU RB，WANG LJ，et al, Mass flux distribution in system semble,J,, Science，1999，285: 1548 )1550, pipe of continuous reformingf urnace,J,, Journal of China Universi- ,8, CARTWRIGHT TJ，KRISHNAMURTI TN, Warm season mesoscale ty of Petroleum: Edition of Natural Science，2005，29( 1): 98 )100, superensemble precipitation forecasts in the southeastern United ( in Chinese) ,States,J,, Weather and Forecasting，2007，22: 873 )886, ,18,YANG GJ，GUO ZY, Strong wind climate characteristics and circu- ,9,ZHI XF，LIN CZ，BAI YQ，et al, Superensemble forecasts oft he lation analysis in Jinzhou,J,, Journal of Anhui Agricultural Sci- surface temperature in Northern Hemisphere middle latitudesences，2010，38( 23) :1234 7 )12349，12369, ( in Chinese) , ,J,,Sc ientia Meteorologica Sinica，2009，29 ( 5) : 56 9 )57 4, ( in Chi- nese) , 消除偏差集合平均在黄海渤海大风预报中的应用 1,2 1* 3,,朱 桦智协飞俞永庆 ( 1, ,210044; 2, ,230061; 3, 南京信息工程大学气象灾害省部共建教育部重点实验室江苏南京 安徽省气象局公共气象服务中心安徽合肥 胜利油田气 象台,山东东营 257000) 2009 GYHY200906007) ; ( 2008001) 。( ) ( 年度公益性行业气象科研专项胜利油田气象台大风预报项目 基金项目 ( 1984 ) ) , ,,,,。* ,,,,作者简介 朱桦女内蒙古呼和浩特人助理工程师硕士从事数值天气预报研究通讯作者教授博士博士生导师从事数值天气 、,E-mail: zhi@ nuist, edu, cn。预报短期气候预测研究 2010-11-052010-12-13修回日期 收稿日期 欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁欁 nese) ,( From page 3) ,7,QU Y，JIN W，ZHAO LW，et al, Study on temperature precipitation ,5, IPCC, IPCC third assessment reportijing: MeteorologyBeclimatic zone in summer in Liaoning,J,, Liaoning Meteorology， ,M,,P ress，2001, ( in Chinese) , 2004( 4) :4 )5, ( in Chinese) , ,6,WEI FY, Modern climatic statistics diagnose forecastet chnique,M,, Beijing: China Meteorological Press，2007: 6，37 )3 9, ( in Chi- REOF 基 于 研究辽宁省极端最高气温时空分布 1 2,王 震王颖 ( 1, ,110001; 2, ,110016)辽宁省气象局辽宁沈阳 辽宁省沈阳区域气候中心辽宁沈阳 ( 1975 ) ) ,,,,,,E-mail: wwwzz@to m, com。作者简介 王震男吉林辽源人工程师在读硕士从事天气气候业务和研究 2010-09-032010-10-28收稿日期 修回日期