Research On GPS Radio Occultation Data Assimilation

GPS Radio Occultation (RO) technology offers us a new type of observation datawith the advantage of near real time, no cloud and rain impact, high vertical resolution,no need to check and global averaged distribution. High time-space resolution GPS ROdata can improve the temperature fields, humidity fields and other initial fields. GPSRO data have positive effects to the model forecast skill scores.This thesis focuses on GPS RO data assimilation methods. We provided the basicalgorithms, concrete implementations and the experiments in detail and we discussedthe results, and then draw conclusions in a scientific way.1D Local refractivity operator is the most nonaccurate but the easiest implementedoperator. There also exists the refractivity operator in the bending angle operators. So itis necessary to discuss the value of refractivity parameters. The three term SW53expression is more accurate than two term expression. RU02parameters has someproblems in theory, we need more experiments to research on it.The emphasis and main work of this thesis is the variational assimilationimplementation of1D bending angle operator. In a4D-VAR system, we can not onlycare about forward model operator. The implementation of tagant linear operator andadjoint operator is the most important part. After introducing GPS RO data in globalvariational data assimilation experiments, it has obviously positive impact on thegeopotential hight of south hemisphere and north hemisphere. The RMS of globalmedium-range forecast is reduced. The improment of anormaly correlation coefficientsis statistically significant at the10%level. In the typhoon forecast impact experiment,the path and strength of the typhoon with GPS RO data si actually better than thatwithout GPS RO data in24hours forecast and120hours forecast.2D bending angle operator is more accurate than1D operator in low and middletroposphere, but its computation cost is actually high. And1D operator is perfectlyadequate enough in upper troposphere and stratosphere. As the bending angleobservations are mainly in upper troposphere and stratosphere, there is no significantimprovement in the system.Finally, the most important part of GPS RO data assimilation is the observationerror model. It is time and latitude dependence. We discussed and built the observationerror model of refractivity and bending angle. This will be the emphasis of my futureresearch.