强弱强迫条件下对流尺度集合扰动增长特征及对降水预报性能的影响

Convective-scale ensemble perturbation growth and precipitation predictability under strong and weak synoptic forcing conditions

摘要:

如何准确刻画弱天气强迫下的中小尺度扰动增长特征,并构建能够合理表征强、弱天气强迫下预报不确定性的初始扰动,是发展集合预报初始扰动技术亟待解决的关键问题。基于中国气象局对流尺度集合预报系统(China Meteorological Administration Regional Ensemble Prediction System,CMA-REPS V4.0),本文选取发生在同一时期分别受强、弱天气强迫影响的降水事件作为研究对象,除进行动力降尺度初始扰动的控制试验外,还设计了以余弦分析约束方法改善初始扰动质量的对照试验,以进一步探究其对后续集合扰动增长及两类天气强迫下降水预报性能的影响。研究结果表明,两组对流尺度集合试验均以中β尺度(20～200 km)扰动增长最为明显,与弱强迫相比,在强强迫条件下两组试验的较小尺度(≤200 km)集合扰动非线性增长明显更快。在两类天气强迫条件下,在采用余弦分析约束方法调整扰动后的试验中,较小尺度扰动升尺度过程的完成均较控制试验有所提前。两类天气强迫对应的区域平均降水预报结果显示,包含更丰富中小尺度扰动信息的初始扰动有利于CMA-REPS V4.0降水预报性能的提升,且初始扰动的质量对可预报性较高的强强迫降水前期预报具有更积极的影响。

Abstract:

Characterizing the nonlinear growth of small-scale ensemble perturbations remains a major challenge in developing effective initial perturbation techniques for convective-scale ensemble prediction systems.Understanding the general characteristics of ensemble perturbations,particularly under different synoptic forcing conditions,is essential for constructing more representative initial perturbations and improving forecast uncertainty quantification.Although previous studies have investigated ensemble perturbation growth,the differences between strong and weak synoptic forcing conditions remain insufficiently understood.
In this study,two concurrent precipitation events over different regions of China are examined:one over northern China under strong synoptic forcing,and the other over southern China under weak synoptic forcing,characterized as a warm-sector heavy rainfall event.To minimize the influence of lateral boundary conditions and emphasize smaller-scale ensemble perturbations,two ensemble experiments were conducted using the China Meteorological Administration Regional Ensemble Prediction System (CMA-REPS V4.0) over a domain spanning (70.0°—145.0°E and 10.0°—60.1°N).In the control experiment (CTRL),14 ensemble members were generated by adding perturbations to both the initial conditions (ICs) and lateral boundary conditions (LBCs),which were downscaled from the CMA Global Ensemble Prediction System (CMA-GEPS;0.5°×0.5° resolution).The ICs and LBCs for the control member were downscaled from the National Centers for Environmental Prediction Global Forecast System (NCEP-GFS;0.5°×0.5° resolution).In a second experiment (CONS),the cosine analysis constraint method was applied to optimize the initial perturbations in CTRL,allowing assessment of their sensitivity and influence on precipitation predictability under different synoptic forcing regimes.
The results show that meso-β scale ensemble perturbations exhibit more pronounced evolution,with faster nonlinear growth under strong forcing than weak forcing.In the CONS experiment,the growth of smaller-scale perturbations is enhanced under both forcing conditions.This indicates that incorporating more small-scale perturbations improves the ability of CMA-REPS V4.0 to represent forecast uncertainty,leading to better agreement with observed precipitation tendencies and improved forecast performance for heavy rainfall events.Region-averaged results confirm that precipitation under weak forcing has lower predictability than that under strong forcing.Correspondingly,forecast performance improvements are more evident in strong-forcing conditions.Compared with CTRL,the CONS experiment exhibits greater ensemble spread and a stronger ability to capture precipitation uncertainty under both forcing regimes,with some members successfully reproducing observed variability.
However,the lower predictability of weak-forcing precipitation suggests that limitations in simulating heavy rainfall cannot be attributed solely to initial perturbation design.The overestimation of precipitation in both events,particularly during the early stage of the weak-forcing case,highlights the important role of physical parameterization schemes.Therefore,improving forecast skill requires not only optimized initial perturbations but also the selection and development of appropriate microphysics and cumulus parameterization schemes to better represent moist convection processes.Although not the primary focus of this study,such improvements are essential for enhancing CMA-REPS performance in forecasting extreme precipitation events.