以2022年9月影响东北地区的台风“梅花”残余系统降水为例,采用SAL(structure amplitude and location)空间检验方法对5种中国气象局数值业务模式(CMA模式)降水最强日(2022年9月16日08时至17日08时,热带低压)累计24 h降水预报进行空间...以2022年9月影响东北地区的台风“梅花”残余系统降水为例,采用SAL(structure amplitude and location)空间检验方法对5种中国气象局数值业务模式(CMA模式)降水最强日(2022年9月16日08时至17日08时,热带低压)累计24 h降水预报进行空间检验、偏差成因及预报调整分析。结果表明:此次过程中各模式在降水位置方面预报均较好,CMA-GFS在降水结构分布、强度方面预报表现最佳,除CMA-BJ外其他模式对于降水极值预测均偏小。CMA-GFS预报效果最佳的原因是其对于台风变性、850 hPa低空急流及暖式切变线位置、强度以及移速预报更准确,CMA-TYM预报较差主要是未能预报出台风变性,对切变线预报速度过快而导致暴雨落区偏大。CMA-GFS随着预报时效的临近,预报效果越来越好,优势主要体现在临近时效内,但在长时效预报中几乎无预报能力,CMA-TYM虽然在临近时段结构与强度预报效果较差,但在长预报时效是最早指示出强降水大致落区与量级的模式。展开更多
In this paper, the CMA-TRAMS tropical high-resolution system was used to forecast a typical hot weather process in Guangdong, China with different horizontal resolutions and surface coverage. The results of resolution...In this paper, the CMA-TRAMS tropical high-resolution system was used to forecast a typical hot weather process in Guangdong, China with different horizontal resolutions and surface coverage. The results of resolutions of 0.02° and 0.06° were presented with the same surface coverage of the GlobeLand30 V2020, companies with the results of resolution 0.02° with the USGS global surface coverage. The results showed that, on the overall assessment the 2 km model performed better in forecasting 2 m temperature, while the 6 km model was more accurate in predicting 10 m wind speed. In the evaluation of representative stations, the 2 km model performed better in forecasting 2 m temperature and 2 m relative humidity at the coastal stations, and the 2 km model was also better in forecasting 2 m pressure at the representative stations. However, the 6 km model performed better in forecasting 10 m wind speed at the representative stations. Furthermore, the 2 km model, owing to its higher horizontal resolution, presented a more detailed stratification of various meteorological field maps, allowing for a more pronounced simulation of local meteorological element variations. And the use of the surface coverage data of the GlobeLand30 V2020 improved the forecasting of 2 m temperature, and 10 m wind speed compared to the USGS surface coverage data.展开更多
为发展中国气象局地球系统数值预报中心CMA-MESO模式对流尺度集合预报,基于CMA-MESO模式设计了观测扰动构建技术,并利用该技术发展集合资料同化(ensemble data assimilation,EDA)初值扰动方法。开展观测扰动敏感性试验、EDA方法在CMA-M...为发展中国气象局地球系统数值预报中心CMA-MESO模式对流尺度集合预报,基于CMA-MESO模式设计了观测扰动构建技术,并利用该技术发展集合资料同化(ensemble data assimilation,EDA)初值扰动方法。开展观测扰动敏感性试验、EDA方法在CMA-MESO对流尺度集合预报中的应用试验,分析观测扰动构建合理性及影响特征,并对比传统的动力降尺度方法与EDA方法的效果,结果表明:观测扰动可有效表征同化中来源于观测资料的不确定性特征;观测扰动主要影响CMA-MESO模式短时效预报效果,随时效延长逐渐耗散;EDA方法可有效形成对流尺度集合预报初值扰动,相对于传统的动力降尺度,该方法可显著减少初值扰动中来自背景场的扰动分量,并增加观测扰动分量体现观测的不确定性;强对流降水个例试验也表明,EDA方法可有效提高降水概率预报效果。展开更多
As a prominent mode of variability in the tropical stratosphere on the interannual timescale,the Quasi-Biennial Oscillation(QBO)can significantly influence global atmospheric circulation and weather patterns.This stud...As a prominent mode of variability in the tropical stratosphere on the interannual timescale,the Quasi-Biennial Oscillation(QBO)can significantly influence global atmospheric circulation and weather patterns.This study explores the dynamic processes of QBO disruptions using the integrated climate model of the China Meteorological Administration(CMA)by nudging the tropical zonal winds toward observations.A comparative analysis with ERA5 reanalysis data shows that the nudged runs accurately replicate the general characteristics of the QBO,including the alternating QBO wind regimes and QBO disruption events.The evolution of the QBO winds is diagnosed using empirical orthogonal function and root-mean-square difference analyses,and the rarity of the disruption events is confirmed in the CMA model.Different aspects of the QBO disruptions and the relevant dynamics are present in the model.Firstly,the momentum budget analysis highlights the crucial roles of extratropical Rossby waves and non-orographic gravity waves in the transition from westerly to easterly winds during a disruption.Secondly,Kelvin waves and non-orographic gravity waves explain much of the transition from easterly to westerly winds near 40 hPa.Thirdly,the positive tendency from enhanced vertical advection further accelerates westerly momentum development via secondary meridional circulation.These findings underscore the importance of nudging techniques in understanding QBO dynamics,which provides valuable insights for future climate model improvements toward better forecasting QBO-related climate variability.Notably,due to model limitations,no QBO disruptions were simulated in the free-run experiments.展开更多
利用2020年夏季(6~8月)CMA-MESO模式逐日08:00(北京时,下同)起报的12~36 h逐时降水预报数据和地面—卫星—雷达三源融合逐时降水产品,着眼于小时尺度降水特征,细致评估了CMA-MESO对四川盆地及周边地区的降水预报性能。结果表明,CMA-MES...利用2020年夏季(6~8月)CMA-MESO模式逐日08:00(北京时,下同)起报的12~36 h逐时降水预报数据和地面—卫星—雷达三源融合逐时降水产品,着眼于小时尺度降水特征,细致评估了CMA-MESO对四川盆地及周边地区的降水预报性能。结果表明,CMA-MESO较好把握了夏季降水的空间分布特征,即小时平均降水量和降水频率的大值区位于四川盆地西部、北部和东部的高海拔山区,而降水强度大值区主要位于山脉迎风坡一侧,但CMA-MESO预报的降水量和频率大值区位置较观测偏南。CMA-MESO合理描述了研究区域内降水量和频率峰值时间位相自西向东逐步滞后的特征,能够把握区域平均的降水量和频率清晨主峰、傍晚次峰的双峰形态以及降水强度的单峰特征,但预报的降水日变化位相超前于观测。CMA-MESO预报的逐时降水量均大于观测,明显的降水量预报正偏差发生于夜间(21:00至次日03:00)和午后至傍晚(14:00~20:00),分别由一般性降水(0.1~10 mm h-1)预报偏差和强降水(≥10 mm h-1)预报偏差主导,其偏差大值区分别位于青藏高原东南缘至四川盆地西部和四川盆地以东、以南地区,模式对热力和动力场的预报偏差结合地形的影响是降水量预报偏差的成因。展开更多
研究模型选择对支持向量机(SVM)的泛化性能有着重要影响。针对传统梯度算法对初始值敏感及网格搜索法计算复杂的缺点,为了提高全面优化能力和分类精度,提出了一种基于协方差矩阵自适应进化策略(CMA-ES)的支持向量机(SVM)模型优化算法,...研究模型选择对支持向量机(SVM)的泛化性能有着重要影响。针对传统梯度算法对初始值敏感及网格搜索法计算复杂的缺点,为了提高全面优化能力和分类精度,提出了一种基于协方差矩阵自适应进化策略(CMA-ES)的支持向量机(SVM)模型优化算法,通过对SVM泛化性能界(Bounds on Generalization Performance)的优化求解,实现了基于CMA-ES算法的SVM模型选择。在标准数据集上的实验结果表明:相比遗传算法和梯度算法,上述方法能够在较小计算代价下得到更优的超参数,提高支持向量机的预测精度稳定性,尤其适合大样本数据条件下的模型选择。展开更多
文摘In this paper, the CMA-TRAMS tropical high-resolution system was used to forecast a typical hot weather process in Guangdong, China with different horizontal resolutions and surface coverage. The results of resolutions of 0.02° and 0.06° were presented with the same surface coverage of the GlobeLand30 V2020, companies with the results of resolution 0.02° with the USGS global surface coverage. The results showed that, on the overall assessment the 2 km model performed better in forecasting 2 m temperature, while the 6 km model was more accurate in predicting 10 m wind speed. In the evaluation of representative stations, the 2 km model performed better in forecasting 2 m temperature and 2 m relative humidity at the coastal stations, and the 2 km model was also better in forecasting 2 m pressure at the representative stations. However, the 6 km model performed better in forecasting 10 m wind speed at the representative stations. Furthermore, the 2 km model, owing to its higher horizontal resolution, presented a more detailed stratification of various meteorological field maps, allowing for a more pronounced simulation of local meteorological element variations. And the use of the surface coverage data of the GlobeLand30 V2020 improved the forecasting of 2 m temperature, and 10 m wind speed compared to the USGS surface coverage data.
文摘为发展中国气象局地球系统数值预报中心CMA-MESO模式对流尺度集合预报,基于CMA-MESO模式设计了观测扰动构建技术,并利用该技术发展集合资料同化(ensemble data assimilation,EDA)初值扰动方法。开展观测扰动敏感性试验、EDA方法在CMA-MESO对流尺度集合预报中的应用试验,分析观测扰动构建合理性及影响特征,并对比传统的动力降尺度方法与EDA方法的效果,结果表明:观测扰动可有效表征同化中来源于观测资料的不确定性特征;观测扰动主要影响CMA-MESO模式短时效预报效果,随时效延长逐渐耗散;EDA方法可有效形成对流尺度集合预报初值扰动,相对于传统的动力降尺度,该方法可显著减少初值扰动中来自背景场的扰动分量,并增加观测扰动分量体现观测的不确定性;强对流降水个例试验也表明,EDA方法可有效提高降水概率预报效果。
基金funded by the National Natural Science Foundation of China(Grant No.42275056).
文摘As a prominent mode of variability in the tropical stratosphere on the interannual timescale,the Quasi-Biennial Oscillation(QBO)can significantly influence global atmospheric circulation and weather patterns.This study explores the dynamic processes of QBO disruptions using the integrated climate model of the China Meteorological Administration(CMA)by nudging the tropical zonal winds toward observations.A comparative analysis with ERA5 reanalysis data shows that the nudged runs accurately replicate the general characteristics of the QBO,including the alternating QBO wind regimes and QBO disruption events.The evolution of the QBO winds is diagnosed using empirical orthogonal function and root-mean-square difference analyses,and the rarity of the disruption events is confirmed in the CMA model.Different aspects of the QBO disruptions and the relevant dynamics are present in the model.Firstly,the momentum budget analysis highlights the crucial roles of extratropical Rossby waves and non-orographic gravity waves in the transition from westerly to easterly winds during a disruption.Secondly,Kelvin waves and non-orographic gravity waves explain much of the transition from easterly to westerly winds near 40 hPa.Thirdly,the positive tendency from enhanced vertical advection further accelerates westerly momentum development via secondary meridional circulation.These findings underscore the importance of nudging techniques in understanding QBO dynamics,which provides valuable insights for future climate model improvements toward better forecasting QBO-related climate variability.Notably,due to model limitations,no QBO disruptions were simulated in the free-run experiments.
文摘利用2020年夏季(6~8月)CMA-MESO模式逐日08:00(北京时,下同)起报的12~36 h逐时降水预报数据和地面—卫星—雷达三源融合逐时降水产品,着眼于小时尺度降水特征,细致评估了CMA-MESO对四川盆地及周边地区的降水预报性能。结果表明,CMA-MESO较好把握了夏季降水的空间分布特征,即小时平均降水量和降水频率的大值区位于四川盆地西部、北部和东部的高海拔山区,而降水强度大值区主要位于山脉迎风坡一侧,但CMA-MESO预报的降水量和频率大值区位置较观测偏南。CMA-MESO合理描述了研究区域内降水量和频率峰值时间位相自西向东逐步滞后的特征,能够把握区域平均的降水量和频率清晨主峰、傍晚次峰的双峰形态以及降水强度的单峰特征,但预报的降水日变化位相超前于观测。CMA-MESO预报的逐时降水量均大于观测,明显的降水量预报正偏差发生于夜间(21:00至次日03:00)和午后至傍晚(14:00~20:00),分别由一般性降水(0.1~10 mm h-1)预报偏差和强降水(≥10 mm h-1)预报偏差主导,其偏差大值区分别位于青藏高原东南缘至四川盆地西部和四川盆地以东、以南地区,模式对热力和动力场的预报偏差结合地形的影响是降水量预报偏差的成因。
文摘研究模型选择对支持向量机(SVM)的泛化性能有着重要影响。针对传统梯度算法对初始值敏感及网格搜索法计算复杂的缺点,为了提高全面优化能力和分类精度,提出了一种基于协方差矩阵自适应进化策略(CMA-ES)的支持向量机(SVM)模型优化算法,通过对SVM泛化性能界(Bounds on Generalization Performance)的优化求解,实现了基于CMA-ES算法的SVM模型选择。在标准数据集上的实验结果表明:相比遗传算法和梯度算法,上述方法能够在较小计算代价下得到更优的超参数,提高支持向量机的预测精度稳定性,尤其适合大样本数据条件下的模型选择。
