Assimilation Of Doppler Radar Observations With SVD-En4DVar Method

The hybrid assimilation scheme which couples the4DVar and EnKF is an important research branch of data assimilation and its benefits has been shown in a lot of existed tests. As a representative method to combine the ensemble assimilation with the variational assimilation, SVD-En4DVar has been tested successfully in many experiments. But so far most of the experiments are performed on the models with very simple physical processes and very small freedom, that is differ greatly with real operation models. In this dissertation SVD-En4DVar is used to assimilate Doppler radar observations on WRF model and the main goal is to solve the main problems which will be met when SVD-En4DVar is used in practical operation situation. There9chapters in this dissertation. Chapter1reviews the development of assimilation theory and introduces the advantage and some problems of the methods which are popular used. Then a review of SVD-En4DVar method is presented. The problems which will be met when SVD-En4DVar is used in practical operation especially for radar data assimilation are discussed and the ways to resolve the problems are also discussed.Chapter2introduces the theoretical basis of SVD-En4DVar. A detail derivation of the formula is given and the main feature of this method is discussed.In chapter3SVD-En3DVar is used to assimilate simulated Doppler radial velocity observations on WRF model. A new scheme producing initial perturbation samples is proposed. This scheme takes the pseudo-random perturbation fields of temperature and specific humidity as the observation innovation and3DVar system is utilized to yield the initial perturbation fields of all variables from the observation innovation. Experiments demonstrate that in the initial perturbation fields produced by using the new scheme the compatibility between different variables is better and the perturbation will not decay quickly in the forecast, so the spin-up time is cut down and the time interval of assimilation cycle can be shortened. The impact of different methods of producing initial perturbation and integration time lengths for forecast samples on the assimilation is emphatically investigated. A torrential rain case in south china is chosen to test the effectiveness of the samples produced by the new method.In chapter4, a localize scheme is introduced into SVD-En3DVar and is used to assimilate simulated radar assimilation with WRF model. Local and Global schemes are compared and the sensitivities to the localization parameters are investigated. The experiments show that localization can improve the skill of analysis and forecast by avoiding spurious analysis increment in the vacant area of observation, the loss of observation information caused by insufficient basis vectors is also alleviated after localization. The behaviors of the assimilation are dependent on the localization scale.Chapter5study the feasibility of using SVD-En3DVar for assimilating radar velocity observations with real observation data. Two torrential rain cases (June2008in south china and July2003in the Changj iang-Huaihe region) are chosen for the test and the18-hour forecast of rainfall is compared with that by WRF-3DVar assimilation. For the fist case (2008) the observational data from13radars are assimilated and the forecast of rainfall within18hours is improved after assimilation with SVD-En3DVar, but the improvement is not evident with WRF-3DVar assimilation. For the second case (2003) only single-radar observations are used and the forecast of rainfall is improved in the first6hours after assimilation with SVD-En3DVar, however the forecasts are not improved by using either SVD-En3DVar or WRF-3DVar in the subsequent12hours.In chapter6, an observation localization scheme is introduced into the ensemble-based3DVar assimilation method based on the singular value decomposition technique (SVD-En3DVar) to improve its assimilation skill. In this scheme, a point-by-point analysis technique is adopted and the observational weight decreases with distance between analysis point and observation point. Using the WRF model a set of numerical experiments on assimilating simulated Doppler radar data is designed to test the scheme. By comparing it with the two original schemes in SVD-En3DVar, which are global scheme and local patch scheme without observation localization, it is found that the observation localization scheme not only eliminates the spurious analysis increments in the area of observation vacancy, but also avoids discontinuous of analysis field existed in the local patch scheme. The new scheme can get more good analysis fields and more reasonable short-range rainfall forecast than the original schemes. Additional assimilation and forecast experiments on a heavy rain case, in which the real data from10radars are assimilated, indicate that the short-term forecast skill of precipitation can be improved by assimilating the radar data and the assimilation schemes with observation localization are better than the other two schemes.Simulated and real radar observations are used for assimilation cycle experiments in chapter7. Compared with the test which only assimilates observations at analysis time, it can be found the observation information is absorbed continuously and the analysis field is improved with the increasing assimilation cycles. The improvement of precipitation is also more obviously for assimilation cycle experiments.In chapter8SVD-En4DVar which can assimilate multi-time observations is used to assimilate radar data, the result is compared with SVD-En3DVar. Because SVD-En4DVar can assimilate observations within assimilation time windows simultaneously, and the forecast model is introduced as a kind of constraint at the same time, more observation information will be absorbed into background field in SVD-En4DVar. It's also found that the result of assimilation cycle experiments with SVD-En3DVar is batter than SVD-En4DVar from the comparison of precipitation forecasts.The summary and future work is discussed in chapter9, the main innovation points and problems still existed are also pointed out in this part.