AREM模式两种初值扰动方案的集合降水预报试验及检验

PRECIPITATION VERIFICATIONS TO AN ENSEMBLE PREDICTION SYSTEM USING TWO INITIAL PERTURBATION SCHEMES BASED ON AREM

基于我国科学家自主开发的AREM区域中尺度模式,考虑模式初值的不确定性,初步建立起11个成员37 km分辨率覆盖全国大部地区的短期集合预报系统。采用两套初值扰动试验方案(降尺度方案和BGM方案)分别进行了批量试验,试验结果表明:(1)通过对集合预报离散度和概率预报的检验,表明采用区域模式自身孵化循环产生的初始扰动明显优于直接使用全球集合预报提供的初始扰动。(2)无论采用哪种初值扰动方案,基于集合预报方法的集合平均预报、概率预报等均优于单一的确定性预报和业务预报。(3)离散度与集合平均预报误差的面平均有较好的对应关系;离散度与预报误差的空间相关系数平均为40%～45%。(4)目前集合预报系统离散度与预报误差相比偏小,其部分原因在于没有考虑模式物理过程的不确定性,这将在下一步的工作中逐步改进。

A Short-Range Ensemble Forecast（SREF） system is under development by perturbing initial conditions（ICs） based on a regional model,AREM（Advanced Regional Eta Model）,which has 11 members in 37 km horizontal resolution over China.Two strategies in perturbing IC are compared to find out an optimal approach in building a regional ensemble prediction system（EPS）：one is directly applying perturbations from a global EPS to the regional EPS（a downscaling of a global EPS,Sch-1）,and the other is generating perturbations by the regional model itself（BGM,Sch-2）.The following four results may be obtained：（1） By evaluating ensemble spread and probabilistic forecasts,it shows that the ＂Sch-2＂ seems to be superior to the ＂Sch-1＂ in this preliminary study（2） No matter what IC perturbation strategies are applied,ensemble-based forecasts,ensemble mean and probabilistic forecasts,are in general outperform single control and operational forecast.（3） The ensemble spread is positively correlated quite well with forecast skill on domain average.And the spatial correlation between spread and forecast error is about 40-45% on average.（4） Overall,the divergence ensemble spread in the current ensemble prediction system is small comparing with the forecast error,partially due to lack of consideration in model of physics uncertainty,which is a subject to be explored during the next phase of this EPS development.