The Computation Techniques For Flow-dependent Unbalanced Variances In Global 4DVAR And Their Applications

With the introduction of modern scientific and technological achievements and the rapid development of atmospheric science itself,weather forecasting has developed from traditional qualitative forecasting methods based on the principles of meteorology,mathematical statistics and the experience of forecasters,and developed theories based on atmospheric exploration and atmospheric science.Comprehensive use of new achievements in science and technology,modern numerical weather forecast system implemented on high-performance computers.The level of numerical weather forecast has become an important indicator of a country's meteorological business support capabilities.An accurate numerical weather forecast must be based on a high-quality initial field.Data assimilation to generate initial field data is one of the core key technologies of numerical prediction,and the background error covariance matrix is an indispensable and important part of the variational data assimilation system.With the development of numerical weather prediction models and data assimilation technology,correspondingly,it is required to continuously improve the background error covariance model in the assimilation system.Background error covariance plays an important role in data assimilation,and plays an important role in spreading information,smoothing information,embodying balanced attributes,and constructing a structure that changes with the flow field.By introducing the flow-dependent covariance estimated based on the aggregate information,the ability of the background error covariance to reflect the evolution information of the real air current field with the weather situation can be improved.In the process of assimilation of variational data,the increment of the model variable is generally regarded as the sum of the balance term represented by the balance variable and the remaining unbalanced term.The control variable transformation is carried out on the covariance of the huge background error of the scale.The balanced relationship between variables enables the model variables to be transformed into unrelated control variables.In the domestic independent YH4 DVAR system,the vorticity is used as the balanced variable,and the divergence,temperature and surface pressure are decomposed into two parts,the balanced and unbalanced terms that are balanced with the vorticity,and the specific humidity is regarded as an independent variable,but At present,only the flow-dependent variance estimation of the balance term is realized.Aiming at the problem of inadequate estimation of flow-dependent information in YH4 DVAR system,this paper proposes a method to simultaneously estimate the flow-dependent variance information of balanced and unbalanced terms.On the basis of the existing design,an estimation process of flow-dependent error variance of unbalanced terms is designed.The main idea is to use the background error samples generated by the assimilation of the aggregate data to estimate the background error variance of the unbalanced term in real time.In order to overcome the problem of inadequate background error samples in a short period of time,a short-term lag method was used to increase the sample size,that is,to select a combination of forecast field sets with different forecast aging of the current time set and different forecast aging of the most recent times.The advantage is that a sufficient number of background error samples are constructed in a short period of time,thereby effectively reducing sampling errors in error statistics.The main work and conclusions of this article are summarized as follows:(1)Review the research progress of data assimilation and background error covariance,introduce the research status of hybrid data assimilation method and flowdependent background error covariance,introduce multi-incremental four-dimensional variational method,YH4 DVAR assimilation system framework,spherical wavelet background Error covariance model,on this basis,a method for estimating the dependent variance information of the balanced and unbalanced term streams is proposed,and the estimation process of the dependent variance of the unbalanced term stream dependent errors is given.(2)In order to remove the multicollinearity,it is proposed to use flow-dependent samples to successfully estimate the variance of unbalanced flow-dependent errors of different variables,and to solve the multicollinearity problem.Based on the ridge regression estimation method,a new balance operator is used to eliminate and weaken larger increments in the increment field,showing a smoothing effect.By exploring the influence of balance coefficient on assimilation and forecast,the rationality of the new balance operator obtained by statistics is proved.Then,according to the new and old different balanced operators,two large-scale assimilation forecast experiments were carried out.The standard statistical test comprehensive scoring program recommended by the World Meteorological Organization was used.The statistical test results showed that the positive effect agianst climate state value was the forecast zone after using the new balance operator and the most obvious improvement in the forecast effect in the southern hemisphere.Analysing the character of the distribution of unbalanced variances,and concluding that unbalanced variances make up zonal distrtibution in the westerlies where the atmospheric motion is very active.Benefitting from the correction and filtering,unbalanced variances make up a similar distribution to total variances.While unbalanced variances reveal the character of spatial smoothing,the noises have been filtered effectively.(3)Taking typhoon ”Noru” as an example,three sets of global assimilation forecast experiments were designed according to different variance configurations,and the forecast effect was demonstrated.The test results show that: using updated vorticity and updated divergence,temperature,and surface pressure after the variance of the unbalanced term,the prediction effect of the typhoon center pressure has been significantly improved,which initially reflects the improvement brought by the introduction of the flow-dependent unbalanced variance.