The Direct Three-dimensional Variational Assimilation Of Wind Observations

The past studies and current operational data assimilation systems, e.g. GRAPES in CMA, EnKF in CMC,4DVAR in ECMWF, GDAS/GSI in NCEP and WRFDA in NCAR, do not consider wind direction errors when assimilating wind observations. Despite being collected in the form of speed(sp) and direction(dir), wind observations are transformed to their longitudinal and latitudinal components, u and v, prior to assimilation(asm_uv). Although the dir errors can impact the uncertainty of u and v, they cannot be considered during the assimilation process, and therefore have no independent influence on the assimilation results. Additionally, the conventional method assumes the u and v observation errors to be uncorrelated to simplify the observation error covariance matrix to a diagonal matrix. Since u and v components are not observed in the real world, but are both derived from observations of sp and dir, not only are u and v observation errors correlated, but they are both directly impacted by the dir observation errors. For observing systems directly observing sp and dir, it is safe to assume sp observation errors are not correlated with dir observation errors. Thus, a new method of directly assimilating sp and dir observations (asm_sd) is proposed in this paper, which is to assimilate the sp and dir directly and consider the dir observation error as an independent source from the sp observation error.TAMDAR are becoming a major data source for NWP because of the advantages of high spatiotemporal resolution and humidity measurements. In this study, the estimation of TAMDAR observational errors, and the impacts of TAMDAR observations with new error statistics on short-term forecasts are presented. The observational errors are estimated by a three-way collocated statistical comparison. This method employs collocated meteorological reports from three data sources:TAMDAR, radiosondes, and the6-h forecast from WRF model. The performance of TAMDAR observations with the new error statistics was then evaluated based on this model and WRFDA3DVAR system.The conclusions in this thesis can be summaried as following:(1) The calculation of observation errors:the standard deviation of the observational error of TAMDAR, which has relatively stable values regardless season, is comparable to radiosondes for temperature, and slightly smaller than that of RAOB for relative humidity. The observational errors in dir significantly depend on wind speeds. In general, at low wind speeds, the error in TAMDAR is greater than that of RAOB; however, the opposite is true for higher wind speeds. The impact of TAMDAR on both6-h and24-h WRF forecasts during the studied period is positive when using the default observational AIREP error statistics. The new TAMDAR error statistics presented here bring much further improvement compared with using the default error.(2) The advantages from the direct assimilation of wind observations:using a highly simplified configuration and WRFDA single observation pair experiments we have shown how asm_sd and asm_uv can produce very different analyses from the same background and the same observation. The new assimilation scheme regards the dir observation error as an independent error source, where the analyses will change with the dir observation errors corresponding to the quality of the different observation types. The formulation of asm_sd ensures that the analysis vector falls between the background and observation in terms of sp and dir. The asm_sd and asm_uv methods are further compared in a clean and more realistic framework, using simulated wind observations. Because truth was defined by the nature run, observation errors were computed and used for assimilation experiments. The results show that the asm_sd method produces a more accurate analysis, and as a result, can lead to better forecasts when compared to the asm_uv method.(3) The assimilation of operational dataset:the new methodology is further tested in real observation framework. As a positive effort to explore the potential benefit of the new system in operational forecasts, the full cycle scheme is employed, and GTS, TAMDAR and MDCRS dataset were assimilated. Given the real time environment where some perfect assumption in OSSE doesn’t make sense, the potential benefit is going to be less than that in OSSE. Whereas, the consistency of the qualltative conclusions between OSSE and real observations experiments does still exist. The results demonstrated that analyses from the new formulation were better than those of traditional ones in the term of RMS errors and bias against RAOB. The improved analyses by the new method made subsequently improved forecasts, especially for the forecasting bias. A storm case study was performed to further understand the forecast improvement, especially for the precipitation forecast. Although the same temperature and relative humidity datasets were used, the more accurate temperature forecasts and humidity forecasts were obtained from the direct assimilation methodology by the accumulated observations information in cycle assimilation, which helped get the more accurate precipitation forecasts. Actually, the impacts from direct assimilation on analysis and forecasts partly source from quality control process.