CAJ | 학술논문

预测数值模式对于中期天气预报能力是集合预报的重要应用之一。为研究集合预报在中国区域中期时效（96360h）预报技巧的预测能力，本文基于欧洲中期天气预报中心（ECMWF）全球集合预报系统2007-2011年500hPa高度场和850hPa温度场预报数据，采取两种不同的离散度-预报技巧关系表征方法进行对比分析，结果表明：（1）以均方根误差为表征的预报技巧（RMSENS）和离散度（SPRMSE）表现出季节循环特征，即冬（夏）季值高（低），这是大气内在性质的表现。而以距平相关为表征的预报技巧（ACENS）和离散度（SPAC）没有显著的内在季节变化特征。（2）对比分析两种不同表征的离散度-预报技巧关系可知，以距平相关为表征的离散度-预报技巧关系更能反映中期时效的预报技巧，且850hPa温度场较500hPa高度场二者的相关性更好。（3）定量分析离散度-预报技巧关系表明，小离散度情况下更能体现高的预报技巧，但这种关系从96360h样本百分比下降了20％左右，而在大离散度情况下离散度-预报技巧关系相对弱一些，且随预报时效的延长样本百分比没有显著的降低。（4）样本统计显示中期各时效SPRMSE和SPAC二者一致的样本占59％66％，并没有显示较高的一致性特征。上述分析结果为集合预报在中期时效预报技巧预测方面提供定性和定量的参考。
예측수치모식대우중기천기예보능력시집합예보적중요응용지일。위연구집합예보재중국구역중기시효（96360h）예보기교적예측능력，본문기우구주중기천기예보중심（ECMWF）전구집합예보계통2007-2011년500hPa고도장화850hPa온도장예보수거，채취량충불동적리산도-예보기교관계표정방법진행대비분석，결과표명：（1）이균방근오차위표정적예보기교（RMSENS）화리산도（SPRMSE）표현출계절순배특정，즉동（하）계치고（저），저시대기내재성질적표현。이이거평상관위표정적예보기교（ACENS）화리산도（SPAC）몰유현저적내재계절변화특정。（2）대비분석량충불동표정적리산도-예보기교관계가지，이거평상관위표정적리산도-예보기교관계경능반영중기시효적예보기교，차850hPa온도장교500hPa고도장이자적상관성경호。（3）정량분석리산도-예보기교관계표명，소리산도정황하경능체현고적예보기교，단저충관계종96360h양본백분비하강료20％좌우，이재대리산도정황하리산도-예보기교관계상대약일사，차수예보시효적연장양본백분비몰유현저적강저。（4）양본통계현시중기각시효SPRMSE화SPAC이자일치적양본점59％66％，병몰유현시교고적일치성특정。상술분석결과위집합예보재중기시효예보기교예측방면제공정성화정량적삼고。
Predictive ability of numerical model for medium-term weather forecast is one of the important applications of ensemble forecast.To study the predictive ability of forecast skill of ensemble prediction in medium-term period (96-360 h)in China,our analysis applies forecast data covering the period 2007-2011 from ECMWF global ensemble prediction system and chooses 500 hPa geopotential height and 850 hPa temperature as the variables,and then a comparative analysis is carried out by two different measures of spread-skill relations.The results show that:(1)Forecast skill (RMSENS)and spread (SPRMSE)repre-sented by root mean square errors show a seasonal cycle,i.e.,winter (summer)is high (low),which is an inherent atmospheric property.However,the forecast skill (ACENS )and spread (SPAC )represented by anomaly correlations have no clear inherent seasonal cycle.(2)By comparing the two different measures of spread-skill relations,it is known that the spread-skill relations represented by anomaly correlations can reflect forecast skill better in medium-term period forecast,and this spread-skill relations based on T850 is stronger than based on Z500.(3)Quantitative analysis of spread-skill relations indicates that good forecast skill can be reflected when spread is small,and this relationship declines about 20%from 96 to 360 h sam-ples,while spread-skill relations is weaker in the case of large spread,and this relationship does not de-crease significantly with the increase of the valid time of forecast.(4)Statistic data shows that samples of the consistency of SPRMSE and SPAC account for 59% to 66% in medium-term period,which does not show high consistency.The above analysis results can provide qualitative and quantitative reference for the fore-cast skill of ensemble prediction in medium-term forecast.