A Novel Approach To Improve The Numerical Weather Prediction Skills Within15Days

With the development of computer hardware, the improvement of the numerical modeland the application of satellite and radar data, the accuracy of numerical weather prediction(NWP) is improved considerably. However, the uncertainty of the NWP result still exists,which leads to numerical prediction being far from the operational forecasting demand. In thispaper, the model dependence on initial conditions and the simulation ability within15-day arefirstly tested with the Community Atmosphere Model version3.0(CAM3.0). Then annumerical approach named ANO (Anomaly Numerical-correction with Observations) isdeveloped to correct the model errors by using the concept of anomaly integration andhistorical observation data. The cases of winter and summer are simulated respectively for theverification of the ANO method. Finally, the errors of the ECMWF daily numerical predictionare also corrected with ANO method. The results and conclusions are as following,(1) The ERA-interim reanalysis data are used as the initial conditions for CAM3.0(T42)model in the30numerical tests during January1982-2011. Each case is simulated for15daysand its ACC (Anomaly Correlation Coefficient) and RMSE (Root Mean Square Error) arecalculated. It is found that the1-6d ACC of500-hPa geopotential height field on globeexceeds0.6and the RMSE shows to be less than90gpm.The initial condition is the mainfactor to NWP in this period. The ACC and RMSE of6-11d prediction shows to be0.2-0.6and90-120gpm respectively, while it displays to be less than0.2and higher than120gpm,respectively during11-15d. In these two periods, the role of external forcing becomes moreand more important. The general physics and parameterized schemes are included in theCAM3.0model, which shows its ability to simulate the weather processes within15-day.Similar results are found on two sensitivity experiments of15-day NWP to the initialconditions with ERA-interim and NCEP reanalysis data.(2) The winter case of severe weather in southern China at9-24January2008is taken asan example of testing the improvement of ANO method. The forecast skills of200hPa globalgeopotential height, temperature and wind are all improved significantly. The improvement insouthern hemisphere is better than that in northern hemisphere. The same results are alsofound on geopotential height at500hPa. The correction of500hPa temperature shows1Kdecrease of15-day average RMSE. Besides, much more reasonable horizontal distribution and vertical structure are achieved in ANO model geopotential height, temperature, relativehumidity, and horizontal wind components in comparison to reanalysis data. In comparisonwith another model error correcting method, the ANO method is found more effective withsmaller error in geopotential height.(3) The CAM3.0(T85) model is used for the15-day numerical weather prediction in30summer cases during June1982to2011, and the performance of the ANO method isdisplayed in comparison with the original results. The corrected global forecasts of200-hPa,500-hPa and850-hPa geopotential height exhibit decrease of root mean square errors and alsosignificant increase of ACC. The general improvement is also confirmed with the ACC onfifth and tenth day of integration in30cases. In addition to the circulation pattern, watervapor transport as well as its convergence is well described in the persistent heavy rainfallcases with the ANO method. In comparison with another model error correcting method, theANO method is found more effective with smaller error in geopotential height, temperature,wind vector, and moisture.(4) The ECMWF model products from20to30July2012are corrected with the ANOmethod. It is found that the500-hPa and850-hPa geopotential height ACC is higher than theoriginal forecasts, while the RMSE are smaller. The corrected temperature and relativehumidity at500-hPa and850-hPa are more similar to the observation data compared with theoriginal forecasts, especially for850-hPa. In a word, there is a possibility of really applyingthe ANO method to the operational weather forecasting.