A Study On The Correction Of East Asian Surface Air Temperature Forecast Based On Modal Projection

Numerical weather prediction models are an important measure of meteorological forecasting.Considering that there are deficiencies in the assimilation of initial field data and parameterization of physical processes,accompanied by the inherent chaotic characteristics of earth system,it is diffifult for numerical models to accurately reflect the development of weather and its accuracy needs to be further improved.Weather forecasts with the lead times of1-7 days are an important part of the operational weather forecasting,which play a significant role in issuing early warnings and assisting decision-making for the governments.In this study,the 1-7 day temperature(daily maximum/minimum temperature)forecasts in East Asia are choosed as the research object,and the mean absolute error(MAE),the two-degree hit rate(HR2),the pattern correlation coefficient(PCC)and other evaluation methods are used to evaluate forecast skills of the European Centre for Medium-Range Weather Forecasts(ECMWF)? the National Centers for Environmental Prediction(NCEP)? the Japan Meteorological Agency(JMA)and the United Kingdom Meteorological Office(UKMO).The temporal and spatial distribution characteristics of errors are illustrated.Then,the Stepwise Modal Projection Method(SPPM)and the Neighborhood Modal Projection Method(NPPM)are investigated to improve the forecasting skill of the best single model,and the Decaying Average Method(DAM)is selected as the benchmark.In addition,the Mean square Error(MSE)decomposition method is used to diagnose the error of each forecasting scheme,analyze the error source of each forecasting scheme,and reveal the correcting effects of different schemes on the errors of different sources.The main conclusions of the paper are as follows:(1)In general,ECMWF and JMA have higher temperature forecasting skills in East Asia than NCEP and UKMO,and ECMWF has the best overall performance.The large error of each model are mainly concentrated in the plateau area,and gradually show the characteristics of high in the north and low in the south with the increase of forecast time.In the Tmax forecast,the forecast skill of each numerical model does not change much with the seasons,while in the Tmin forecast,the forecast skill varies significantly with the seasons,and corresponds to the highest forecast skill in summer.In addition,ECMWF showed more cold bias overall,JMA showed more warm bias,and NCEP and UKMO were more evenly distributed.(2)The correction study on the optimal single-model ECMWF temperature forecast found that DAM,SPPM,and NPPM can all improve the temperature forecasting skills of ECMWF to a certain extent,and NPPM has the best performance overall.All statistical post-processing models showed the best correction effect in the plateau area.After the NPPM correction,the cold bias phenomenon of ECMWF was significantly improved,and the proportion of absolute errors less than 1°C was significantly increased.In two cases(heat wave and cold wave),both DAM and SPPM can effectively improve the forecasting skills of Tmax in heat wave events,but they do not perform well in cold wave events.On the other hand,NPPM showed excellent forecasting skills in both heat wave and cold wave events,and significantly improved the warm and cold biases of ECMWF.(3)The MSE decomposition study of ECMWF temperature forecast shows that: the sequence error(Sequence)is the main contributor to the total error,which increases significantly with the lead time,while the bias(Bias2)is the second contributor to the error,which does not change significantly with the lead time.The distribution error(Distribution)accounts for a relatively small contribution to the total error.DAM,SPPM and NPPM can significantly reduce Bias2 and Distribution,and the correction effect of each scheme is basically the same.In terms of Sequence,only NPPM shows obvious advantages.In the two case events,Bias2 became the main source of error.In the heat wave event,all three post-processing models could effectively reduce Bias2,but only the NPPM performed well in the cold wave event.This study aims to use various post-processing methods such as pattern projection methods to effectively improve the temperature forecasting skills of 1-7 lead days,and to diagnose and analyze the forecast errors of different schemes.In the current background that numerical grid forecasting has become the mainstream trend of meteorological business forecasting,it provides an important reference for the use of numerical model prediction results in business departments and scientific research,and has important scientific significance and application value.