Study On Surface Automatic Weather Station Data Quality Control And Its Three Dimension Variational Assimilation

Surface automatic weather station(AWS) data has the characteristics of having high spatial and temporal resolution, being real time data and the observational elements are the same with the model variable. However, its poor quality results in the low utilization rate. Systematic quality control procedures should be applied at the automatic weather station data to make it more representative and accurate so that based on the technique of data assimilation, we can extract some effective information from the observations to form a better initial background field for the numerical weather prediction.In this thesis, a feasible AWS data quality control(QC) module has been designed based on the advantage of other AWS QC module of advanced countries and the requirement of the high resolution Numerical Weather Prediction operational model. Many quality control techniques have been integrated into the QC module, such as the internal consistency check, the second iterative space consistency check, the consistence check with first guess, the temporal consistency check and the decision-making arithmetic. The AWS data quality control module has been applied into hourly AWS data in July2007of Jiangsu Province. Then by use of the mesoscale numerical model WRFV3.2and its3D-VAR system, some experiments on the assimilation of AWS data and sounding observations have been researched. Based on a heavy Meiyu front rainstorm occurred at Changjiang-Huaihe region during7th and8th July2007, Eight comparison experiments have been designed to study the assimilation of AWS data, sounding observations and the sensitivity to the assimilation interval of observations. Subjective verification and the numerical simulation results show that:(1) Wrong data in the observations can be distinguished effectively from the AWS data and the monthly average difference between NCPE data and AWS observations after quality control of every element is smaller than that with the AWS data before quality control. The error rate of wind field is the highest followed by the relative humidity field, while the error rate of temperature field and pressure field are relatively low. Further more, the stations with sensors of systematic bias can be detected. Overall, the observation data after quality control reflects the atmospheric status more precisely.(2) During the numerical simulation, the tests with data assimilation of AWS data have made great adjustment to the rain band structure. Nevertheless, assimilation of AWS data before quality control doesn’t significantly improve the rainfall comparing with the CONTROL experiment due to many false information in the initial data, which also increases the root mean square error of some fields in the forecast. In contrast, because of the promising quality assurance in the temperature, pressure, wind and relative humidity fields, the initial analysis field can be distinctly improved in the test with assimilation of AWS data after quality control, which undoubtedly results in a much better simulation of the drop zone and intensity of precipitation.(3) The sounding data can make up the defect of the adjustment of the initial field limited in the lower atmosphere with the AWS data assimilation. Further improvement in the entire precipitation intensity and region forecast has been achieved with the assimilation of both AWS data and sounding data.(4)Based on the physical fields of the numerical tests, we found that data assimilation could make better use of the available observation data to simulate the hard rainfall. Basically speaking, the data assimilation tests with AWS data after quality control could improve the simulation of the transfer of water vapor, the instability energy and the dynamic fields, which furthermore result in better precipitation forecast.