Quality Control Of Surface Pressure And Temperature For Numerical Weather Prediction

The spatial density of surface data is very high in China. There are more than2thousand national surface stations and more30thousand provincial automatic surfacestations. For the moment only a small percentage of surface data is assimilated in thecurrent operational system except for surface pressure observations. How to use theinformation of surface observation effectively is significant for improving numericalforecast skill especially for mesoscale weather system. If data with gross error areassimilated into numerical model, the forecast skill will decrease, unfortunately. Thus,proper quality control is the first step toward making better use of surfaceobservations. In this thesis, six hourly T639analysis on0.28125×0.28125degree andNCEP analysis on1.0×1.0degree of2-m temperature and surface pressure in winter(from December2009to February2010) and summer (from June to August2010) areused as background fields to study quality control method of2-m temperature andsurface pressure observations in China and its neighboring regions, respectively.To begin with, EOF (Empirical Orthogonal Function) quality control method andOMB (observation-minus-background) quality control method is employed to2-mtemperature in winter, respectively. Compared with OMB quality method, EOFquality control method could retain correct observations reporting unusual states ofatmosphere properly, although both methods could make background and observationincrements distribution more gaussian.Then, EOF quality control method is applied to2-m temperature in winter andsummer. Results show that outliers identified with NCEP background fields used aremore than that of T639background fields, especially in winter. Meanwhile, theseareas where more outliers are identified overlap with the areas where the biweight standard deviations of observation increment are higher. As value of biweightstandard deviations of observation increment in winter is larger than that in summer,the number of stations with higher outliers in winter is larger than that in summer.Additionally, it is difficult for EOF quality control method to pick out the data that aredotted with consecutive errors but showing small fluctuations.Furthermore, progressive EOF quality control method was proposed to beapplied to2-m temperature and surface pressure. Meanwhile, biweight mean methodwas proposed to correct background. Results show that2-m temperature observationsthat are dotted with consecutive errors but showing small fluctuations are removedproperly. Compared with Lapse Rate of Temperature (LRT) method, biweight meanmethod could remove the systematic errors generated by the model, allowingobservational increments (the difference between the observation and the background)closer to the normal distribution, and ensuring the data quality after the quality control.As to surface pressure, the number of outliers identified in winter is larger than that insummer, and biweight mean method is better than barometric height formula in theprocess of correcting backgrounds.Finally, WRF three-dimensional variational data assimilation system is applied.Taking T639data as backgrounds, observations before and after progressive qualitycontrol are assimilated. Results show that analysis fields have a more reasonable fit tothe observation fields.