CAJ | 학술논문

提利用安徽砀山气象站的2001—2013年冬半年（10月至次年4月）的观测资料，探讨霜生与气温、地温、水汽压和风速等气象要素的相关性，并基于 Bayes 判别方法，采用逐步判别分析，建立多套霜生自动判别模型。结果表明：（1）霜是否出现与日最低及夜间不同观测时次的气温、地表温度显著相关，当夜间气温或地表温度越低，低于霜点的可能性越大，结霜的可能性也越大。（2）通过回算性检验和独立样本的预报性检验，基于 Bayes 判别法的霜生模型，对霜未发生的平均判别准确率达到86．5％，对霜发生的平均判别准确率达到92．7％，其中用日最低地温、当日07时水汽压和当日07时风速所建立的三要素模型最优，对霜发生的判别准确率可达到90％以上。因此，可以将 Bayes 霜生判别模型与图像识别技术相结合应用于霜的自动化观测。
제이용안휘탕산기상참적2001—2013년동반년（10월지차년4월）적관측자료，탐토상생여기온、지온、수기압화풍속등기상요소적상관성，병기우 Bayes 판별방법，채용축보판별분석，건립다투상생자동판별모형。결과표명：（1）상시부출현여일최저급야간불동관측시차적기온、지표온도현저상관，당야간기온혹지표온도월저，저우상점적가능성월대，결상적가능성야월대。（2）통과회산성검험화독립양본적예보성검험，기우 Bayes 판별법적상생모형，대상미발생적평균판별준학솔체도86．5％，대상발생적평균판별준학솔체도92．7％，기중용일최저지온、당일07시수기압화당일07시풍속소건립적삼요소모형최우，대상발생적판별준학솔가체도90％이상。인차，가이장 Bayes 상생판별모형여도상식별기술상결합응용우상적자동화관측。
The correlations between frost and temperature,surface temperature,vapor pressure,wind speed and other meteorological factors are discussed in this paper by using the observation data from Anhui Dangshan Weather Station in the winter half-year (from October to April of the next year)from 2001 to 2013.Using stepwise discriminant analysis method,multiple sets of frost automatic discriminant models for the occurrence of frost are built based on Bayes discriminant method.The results show that:(1)The occurrence of frost is significantly correlated with daily minimum temperature,night temperature of differ-ent observation time and surface temperature.The lower the night temperature or the surface temperature is,the larger the possibility of the temperature is lower than the frost point and the greater the possibility of the frost occurrence.(2)Through the back calculation test and prediction test of independent samples, the average accuracy rate of un-occurred frost discriminated by the frost model is 86.5% based on Bayes discriminant method and the average accuracy rate of the seen frost is 92.7%.The three factor models based on the daily minimum temperature,the daily vapor pressure at 07:00 BT and the daily wind speed at 07:00 BT are optimal.The accuracy rate of discriminating the frost occurrence by the three factor models is more than 90%.Therefore,we can combine the Bayes frost discriminant model with image recognition technology,and apply the new technology to frost automatic observation.