CAJ | 학술논문

介绍了国际上一种新的降水检验方法——概率空间中的稳定公平误差（stable equitable error in probability space, SEEPS）的原理、计算方法和误差特征，并应用于我国定量降水预报检验进行评估试验。SEEPS方法在评分意义、降水分类、评分计算及评分应用等方面，比传统检验评分更加灵活，具有更清晰的实际意义。利用两个降水概率阈值，SEEPS方法将降水气候概率分布划分为“干”、“小雨”、“大雨”三类；该方法基于降水概率计算误差矩阵，根据站点分布密度计算区域平均评分权重系数。SEEPS在不同降水概率下具有不同的误差评分特征，使其能够自动适应不同的降水气候。SEEPS不仅可以定量化给出降水预报能力的高低，还可以通过分析不同观测和预报分类组合的误差评分，给出造成评分高低的成因。利用2007年3月—2013年12月24h累积降水逐日观测资料，对中央气象台预报员定量降水预报进行了SEEPS检验试验，并与传统的检验评分进行比较。结果表明：预报员定量降水预报的误差主要来源为两类——预报“小雨”对观测“大雨”的漏报和预报“小雨”对观测“干”类型的空报，合计占到了总误差的近七成；前者说明预报员降水预报量级较实际降水偏小，后者说明预报员对“小雨”的大范围、高频率空报也可以导致总体预报效果的明显下降。SEEPS方法对降水预报能力的评估结论与传统检验评分总体相当，但SEEPS检验指标更简单直接，便于管理层和决策层面使用，具有较好的推广应用价值。
개소료국제상일충신적강수검험방법——개솔공간중적은정공평오차（stable equitable error in probability space, SEEPS）적원리、계산방법화오차특정，병응용우아국정량강수예보검험진행평고시험。SEEPS방법재평분의의、강수분류、평분계산급평분응용등방면，비전통검험평분경가령활，구유경청석적실제의의。이용량개강수개솔역치，SEEPS방법장강수기후개솔분포화분위“간”、“소우”、“대우”삼류；해방법기우강수개솔계산오차구진，근거참점분포밀도계산구역평균평분권중계수。SEEPS재불동강수개솔하구유불동적오차평분특정，사기능구자동괄응불동적강수기후。SEEPS불부가이정양화급출강수예보능력적고저，환가이통과분석불동관측화예보분류조합적오차평분，급출조성평분고저적성인。이용2007년3월—2013년12월24h루적강수축일관측자료，대중앙기상태예보원정량강수예보진행료SEEPS검험시험，병여전통적검험평분진행비교。결과표명：예보원정량강수예보적오차주요래원위량류——예보“소우”대관측“대우”적루보화예보“소우”대관측“간”류형적공보，합계점도료총오차적근칠성；전자설명예보원강수예보량급교실제강수편소，후자설명예보원대“소우”적대범위、고빈솔공보야가이도치총체예보효과적명현하강。SEEPS방법대강수예보능력적평고결론여전통검험평분총체상당，단SEEPS검험지표경간단직접，편우관리층화결책층면사용，구유교호적추엄응용개치。
The principles, computation and error characteristics of a new method newly developed internationally for precipitation verification, i.e., SEEPS (Stable Equitable Error in Probability Space), are briefly introduced in this paper. With regard to score meaning, precipitation classiifcation, score calculation and application, SEEPS is more lfexible and has clearer practical meaning than traditional verification scores. The SEEPS method is applied in the verification and assessment experiment for quantitative precipitation forecasts in China region, and some issues encountered in the application are talked about. Climatic precipitation probability distribution is divided into three classiifcations, ‘dry’, ‘light’ and ‘heavy’ by two thresholds in SEEPS. Error matrices are determined by the precipitation probability. Area mean is weighed based on the station density. Possessing different error score characteristics under varying precipitation probabilities makes SEEPS automatically adapt to various kinds of precipitation climate. The ability of precipitation forecasts is quantiifed by the SEEPS values, and different elements of SEEPS can also be analyzed to ifnd the reason why the SEEPS score is high or low. Daily observations of 24-hour accumulated precipitation from Mar. 2007 to Dec. 2013 are used in the SEEPS veriifcation test for quantitative precipitation forecasts (QPF) of forecasters in the National Meteorological Center of CMA. Results show that the two main categories of forecasters’ QPF errors are misses of heavy precipitation with light precipitation forecasts, and false alarms of “dry” with light precipitation forecasts. These two kinds of errors account for about 70 percent of all errors. The former explains the order of forecast precipitation is lower than observations and the latter states that false alarms with large areas and high frequencies can also obviously deteriorate the whole forecast performance. Conclusions of SEEPS are approximately equivalent to those of traditional verification scores. But the SEEPS score is simpler and more suitable for use in management and decision making, and has more value for popularization and application.