Landfalling typhoons can cause disasters over large regions.The government and emergency responders need to take measures to mitigate disasters according to the forecast of landfall position,while slight timing error ...Landfalling typhoons can cause disasters over large regions.The government and emergency responders need to take measures to mitigate disasters according to the forecast of landfall position,while slight timing error can be ignored.The reliability of operational model forecasts of typhoon landfall position needs to be evaluated beforehand,according to the forecasts and observation of historical cases.In the evaluation of landfalling typhoon track,the traditional method based on point-to-point matching methods could be influenced by the predicted typhoon translation speed.Consequently,the traditional track evaluation method may result in a large track error even if the predicted landfall position is close to observation.The purpose of this paper is to address the above issue using a simple evaluation method of landfalling typhoon track forecast based on the time neighborhood approach.In this new method,the timing error was lessened to highlight the importance of the position error during the landfall of typhoon.The properties of the time neighborhood method are compared with the traditional method based on numerical forecast results of 12 landfalling typhoon cases.Results demonstrated that the new method is not sensitive to the sampling frequency,and that the difference between the time neighborhood and traditional method will be more obvious when the moving speed of typhoon is moderate(between 15−30 km h^(−1)).The time neighborhood concept can be easily extended to a broader context when one attempts to examine the position error more than the timing error.展开更多
随着全球气候变暖的加剧,极端强降水事件发生频率明显增加,对经济社会发展及人民生命财产安全构成重大威胁。开展短时强降水的预报研究对于防灾减灾具有重要意义。基于湖北省区域自动站降水资料、短时强降水概率预报产品和中尺度高分辨...随着全球气候变暖的加剧,极端强降水事件发生频率明显增加,对经济社会发展及人民生命财产安全构成重大威胁。开展短时强降水的预报研究对于防灾减灾具有重要意义。基于湖北省区域自动站降水资料、短时强降水概率预报产品和中尺度高分辨率数值模式资料,采用邻域最优概率法和多模式融合技术对湖北省1~12 h短时强降水的落区进行预报与检验评估。结果表明,邻域法明显提高了中尺度数值模式对短时强降水的预报能力,其中面积邻域法的效果优于单点邻域法,CMAMESO、CMA-SH9和WH-RUC模式的最优面积概率均为5%,最优邻域半径分别为50、60、60 km;多模式融合预报方法较单模式单点邻域法表现出明显优势,2023年、2024年4—9月短时强降水的1~12 h TS评分均表现为正技巧,分别提高0.014、0.020;改进后的多模式融合方法对短时强降水的命中率有大幅提升,尤其是在湖北省2023年8月7日和2024年6月28日的多次强对流过程预报中均有提前精准预报。展开更多
近年来,基于变分水平集方法的心脏医学图像分割在图像处理中得到广泛的应用,然而,由于图像灰度不均匀性和梯度下降法中的符号距离函数导致图像在分割中有计算复杂、运算成本较高的问题.为了解决这些问题,本文在自适应局部拟合(Adaptive ...近年来,基于变分水平集方法的心脏医学图像分割在图像处理中得到广泛的应用,然而,由于图像灰度不均匀性和梯度下降法中的符号距离函数导致图像在分割中有计算复杂、运算成本较高的问题.为了解决这些问题,本文在自适应局部拟合(Adaptive local fitting,ALF)模型的基础上修改并加入边缘检测函数,提出了一种改进的活动轮廓模型,并与图像分割的快速计算算法——乘子交替方向法(Alternating direction method of multipliers,ADMM)相结合来求解水平集方程.本文提出的新水平集图像分割模型,包含了图像的邻域信息,可以更好的解决图像不均匀的问题;利用传统的梯度下降法来分割图像会有耗时长、计算成本高等问题,而用ADMM算法代替传统算法,原本复杂的问题可以被拆分成若干个简单的子问题,逐一解决这些子问题能够更快速并准确地解决整个问题,进而解决了传统模型存在耗时长、计算复杂、计算成本高的问题.实验结果表明新模型不仅对灰度不均匀的图像具有较强的鲁棒性,还具有更高的分割效率和精度.展开更多
基金supported by the National Natural Science Foundation of China(Grant Nos.U1811464,U2142213)Guangdong Basic and Applied Basic Research Foundation(Grants Nos.2020A1515110275,2020A1515110040,2022A1515011870)the Special program for innovation and development of China Meteorological Administration(CXFZ2021Z006,CXFZ2022P026).
文摘Landfalling typhoons can cause disasters over large regions.The government and emergency responders need to take measures to mitigate disasters according to the forecast of landfall position,while slight timing error can be ignored.The reliability of operational model forecasts of typhoon landfall position needs to be evaluated beforehand,according to the forecasts and observation of historical cases.In the evaluation of landfalling typhoon track,the traditional method based on point-to-point matching methods could be influenced by the predicted typhoon translation speed.Consequently,the traditional track evaluation method may result in a large track error even if the predicted landfall position is close to observation.The purpose of this paper is to address the above issue using a simple evaluation method of landfalling typhoon track forecast based on the time neighborhood approach.In this new method,the timing error was lessened to highlight the importance of the position error during the landfall of typhoon.The properties of the time neighborhood method are compared with the traditional method based on numerical forecast results of 12 landfalling typhoon cases.Results demonstrated that the new method is not sensitive to the sampling frequency,and that the difference between the time neighborhood and traditional method will be more obvious when the moving speed of typhoon is moderate(between 15−30 km h^(−1)).The time neighborhood concept can be easily extended to a broader context when one attempts to examine the position error more than the timing error.
文摘对一种降水预报跨量级通用综合评价方法(precipitation accuracy score,PAS)进行了邻域法改进。改进方案通过预报与观测资料匹配技术,采用距离权重评分统计方法,旨在减轻双重惩罚问题,同时确保评分系统能合理表征位置预报的准确性。研究应用邻域PAS方法,基于2021年汛期中国气象局智能网格实况产品,对江苏本地精细化天气分析预报系统进行整体和典型个例检验,同时引入均方根误差(root mean square error,RMSE)、结构相似性(structural similarity,SSIM)、峰值信噪比(peak signal-to-noise ratio,PSNR)、概率空间中的稳定公平误差(stable equitable error in probability space,SEEPS)等跨量级检验指标进行了对比。结果表明,邻域PAS方法显著回避了原方法在位置预报上的双重惩罚问题,更符合预报人员的主观预期和预报应用服务的要求,具有明显优势。邻域PAS评分、PAS评分与现有多个跨量级指标均表现出良好的相关性,多方验证了方法的有效性。同时,相较于RMSE,该方法更有效地平衡了对不同量级降水的敏感性;而与SSIM和PSNR相比,则展现了更强的可解释性,多个个例显示评分结果更符合预报员的认知。邻域PAS方法相比于SEEPS技巧评分保留了对大量级降水的检验分辨能力,同时减轻了PAS方法检验大量级降水产生的严重双重惩罚,一定程度平衡了大量级降水检验的两难问题。
文摘随着全球气候变暖的加剧,极端强降水事件发生频率明显增加,对经济社会发展及人民生命财产安全构成重大威胁。开展短时强降水的预报研究对于防灾减灾具有重要意义。基于湖北省区域自动站降水资料、短时强降水概率预报产品和中尺度高分辨率数值模式资料,采用邻域最优概率法和多模式融合技术对湖北省1~12 h短时强降水的落区进行预报与检验评估。结果表明,邻域法明显提高了中尺度数值模式对短时强降水的预报能力,其中面积邻域法的效果优于单点邻域法,CMAMESO、CMA-SH9和WH-RUC模式的最优面积概率均为5%,最优邻域半径分别为50、60、60 km;多模式融合预报方法较单模式单点邻域法表现出明显优势,2023年、2024年4—9月短时强降水的1~12 h TS评分均表现为正技巧,分别提高0.014、0.020;改进后的多模式融合方法对短时强降水的命中率有大幅提升,尤其是在湖北省2023年8月7日和2024年6月28日的多次强对流过程预报中均有提前精准预报。
文摘近年来,基于变分水平集方法的心脏医学图像分割在图像处理中得到广泛的应用,然而,由于图像灰度不均匀性和梯度下降法中的符号距离函数导致图像在分割中有计算复杂、运算成本较高的问题.为了解决这些问题,本文在自适应局部拟合(Adaptive local fitting,ALF)模型的基础上修改并加入边缘检测函数,提出了一种改进的活动轮廓模型,并与图像分割的快速计算算法——乘子交替方向法(Alternating direction method of multipliers,ADMM)相结合来求解水平集方程.本文提出的新水平集图像分割模型,包含了图像的邻域信息,可以更好的解决图像不均匀的问题;利用传统的梯度下降法来分割图像会有耗时长、计算成本高等问题,而用ADMM算法代替传统算法,原本复杂的问题可以被拆分成若干个简单的子问题,逐一解决这些子问题能够更快速并准确地解决整个问题,进而解决了传统模型存在耗时长、计算复杂、计算成本高的问题.实验结果表明新模型不仅对灰度不均匀的图像具有较强的鲁棒性,还具有更高的分割效率和精度.
