Analytical Constrained Ensemble Forecast Initial Disturbance Construction Scheme Research

Whether ensemble perturbation can accurately describe the uncertainty of atmospheric motion is the core issue of ensemble forecasting research.A reasonable perturbation structure and amplitude should be able to reflect the characteristics of forecast error about the state of atmospheric motion.With the in-depth research and understanding of ensemble perturbation and data assimilation,the research about initial perturbation scheme and ensemble data assimilation(EDA)are closely integrated and developed together.Based on GRAPES-REPS system,which independently developed by the Numerical Weather Prediction Center of CMA,to solving the inconsistence between the initial perturbation and the forecast error,this paper considers the characteristics of forecast error from different spatial scale weather systems,effectively combines the ensemble perturbation which describing forecast uncertainty and analysis increments from EDA which representing the uncertainty of observation and prediction.Then,this study puts forward an analysis constrains scheme to improve the quality of the ensemble initial perturbation,provides a basis for the optimization of the ensemble forecast quality.After full consideration of the analysis increments'spatial structure and magnitude characteristics,we construct the constant local analysis-constrained function w₁,the self-adaptive local analysis constrains function w₂and the self-adaptive overall analysis constrains function w₃,to realize the identification and adjustment of unreasonable perturbation,then analyzes the spatial physical structure and spatiotemporal evolution characteristics of ensemble perturbation,spread and perturbation energy,thereafter carry out batch trials to explore the universality,and draw the following conclusions.(1)The analysis increments include observation uncertainty and prediction uncertainty,which can reasonably reflect the forecast error structure and amplitude of meso-and micro-scale systems,and is in good agreement with the forecast uncertainty represented by ensemble perturbation.(2)The analysis constrains scheme established by the analysis increments and the ensemble initial perturbation structure and amplitude characteristics can effectively absorbs the analysis uncertainty information,has a good adjustment effect on the physical structure and amplitude of the initial perturbation,and can effectively eliminate the unreasonable false perturbation information.The constraint effect of the self-adaptive analysis constrains scheme is obviously better than that of the constant constrain scheme.However,the self-adaptive local adjustment scheme has significant discontinuity.The overall self-adaptive adjustment scheme not only has better continuity,but also has a significant improvement effect on the initial perturbation,which is more scientific.(3)After the initial perturbation is adjusted by different analysis constrains schemes,the structure and amplitude of the wind field and mass field forecast perturbation have been reasonably improved,which is more consistent with the development and evolution of the forecast error.The results of the ensemble average precipitation forecast is also closer to the reality.At the same time,the growth and evolution characteristics of spread and perturbed energy are more reasonable.Among them,the improvement of the overall self-adaptive analysis constrains scheme is the most significant.(4)Batch test shows that the overall self-adaptive adjustment scheme can improve the performance of regional ensemble prediction.From the perspective of energy structure,spread and RMSE structure and their consistency,the self-adaptive overall analysis constrains scheme has significantly improved the prediction effect of basic physical quantities;from the perspective of the consistency of prediction and observation,the improvement of temperature is more obvious,and it has not improved the altitude;the scheme has a certain improving effect on the reliability and identifiability of the forecast,but the results still need to be improved.