A Comparative Analysis Of The Revised Methods Of Minimum Temperature Forecast In Fuxin Area

In order to improve the accuracy rate of minimum temperature forecast in Fuxin area and make this area meet the assessment criteria of the operation.At the same time,daily minimum temperature is an important meteorological factor affecting crop low temperature disasters,and its accurate prediction is of great significance for preventing low temperature disasters.Therefore,the accuracy rate and error of EC-thin 2m temperature prediction products in 2017 and 2018 are analyzed in this paper.Then,direct correction,fixed correction period correction,sliding correction period correction,interpolation correction,Kalman filter correction and deep neural network correction are applied to correct the products.The results are as follows:1.In Fuxin area(Fuxin County and Zhangwu County),the European model minimum temperature forecast accuracy rate is the best in summer(June,July and August),and the worst in December.The system error of the model can be corrected by direct correction method.The largest range is in December(Zhangwu County),which is revised from 25.8% to 74.2%.The range of errors in summer is relatively small,which is basically concentrated in the range of-3? to 4?.It is consistent with the distribution of annual error frequency.Positive errors are obviously more than negative errors.The negative errors in June and August are within the range of business requirements(which can deviate from 2? positively and negatively).The largest range of errors is April,from-6? to-15?,followed by December,from-3? to 10?,with the most negative errors.In the remaining months(January to March,May,September to November),the errors are mainly concentrated between-2? to 8?.Except in summer,the errors are scattered.Compared with the two counties,the error range of Zhangwu County is smaller than that of Fuxin County.2.According to the order of the revised annual accuracy rate from high to low,the correction methods of Fuxin County and Zhangwu County are deep neural network correction method,fixed correction period correction method,sliding correction period correction method,Kalman filter regression correction method and interpolation method.3.According to the accuracy rate of each month,In Fuxin County the months used the deep neural network correction method to revise are February to May,July to December,and used fixed correction period correction method to revise are January and June.In Zhangwu County the months revised by the deep neural network correction method are March,October to December.January,February,April to June and September with the fixed correction period correction method,April,May and August with the interpolation correction method,and July with other correction methods except the deep neural network.4.Different correction methods were used to revise the 2m minimum temperature forecast of the European model,which met the operational requirements(accuracy 81%)from June to August in Fuxin County and from March to October in Zhangwu County.The annual forecast accuracy rate reaches the operational standard of Zhangwu County.The pressure field reached the standard in Zhangwu County is low pressure,inverted trough and before high pressure,in Fuxin County is low pressure and inverted trough.Through this study,we hope to provide a basis for forecasters to revise the temperature forecast,and provide technical support for fine and objective lattice prediction of temperature product inspection.