臭氧浓度的预测对于大气环境治理、空气质量改善等起到了重要的作用。本文提出了一种交互差分时空LSTM网络预测模型(ST-IDN)来挖掘臭氧浓度历史数据的时间相关性和空间相关性,并成功将其应用到网格化臭氧浓度数据预测上。在该模型中,首...臭氧浓度的预测对于大气环境治理、空气质量改善等起到了重要的作用。本文提出了一种交互差分时空LSTM网络预测模型(ST-IDN)来挖掘臭氧浓度历史数据的时间相关性和空间相关性,并成功将其应用到网格化臭氧浓度数据预测上。在该模型中,首先交互模块(IC)可以通过一系列的卷积操作来捕捉短期上下文信息,其次层融合模块(LF)可以融合不同层的空间信息来获得上一时刻丰富的空间信息,最后差分时空LSTM模块(DSTM)将捕捉到的时间信息和空间信息进行统一建模实现臭氧浓度预测。所构建模型分别与卷积LSTM网络(ConvLSTM)、预测循环神经网络(PredRNN)以及Memory in Memory网络(MIM)模型在河北省气象局提供的臭氧浓度数据上进行了对比分析,ST-IDN模型的平均绝对误差分别降低了19.836%、12.924%、7.506%。实验结果表明,所提出的模型能够提高臭氧浓度的预测精度。展开更多
Long-term trends of yearly and seasonal averages of tropospheric ozone over the whole country and some important regions of China during 1979-2005 are analyzed,based on tropospheric ozone residue(TOR) data retrieved f...Long-term trends of yearly and seasonal averages of tropospheric ozone over the whole country and some important regions of China during 1979-2005 are analyzed,based on tropospheric ozone residue(TOR) data retrieved from satellite measurements.The relationship between the TOR and Southern Oscillation Index(SOI) is studied.The results show that,over the whole country,there is a slight increase of TOR in summer and a decrease in other seasons,while the overall trend for the whole period is insignificant.There are decreasing trends of TOR over the Pearl River Delta and the Sichuan Basin.Significant increasing trends of TOR are found over the North China Plain(NCP) for all seasons except for winter,with a maximum rate of 1.10 DU per decade for summer.There are significant correlations between TOR and SOI for some regions in China but not for the NCP,suggesting that the observed increasing trend of TOR over the NCP may not be linked with changes in atmospheric circulations.展开更多
This paper analyzes multiple structural changes by GMDH (Group Meth- ods of Data Handling), which have obvious advantages. Our method extends the model of Lumsdaine & Papell[1] (1997), and it could be applied to ...This paper analyzes multiple structural changes by GMDH (Group Meth- ods of Data Handling), which have obvious advantages. Our method extends the model of Lumsdaine & Papell[1] (1997), and it could be applied to the case of more than two structural changes. Because of simultaneously considering every structural change of the hypothesis, it is likely to be of particular relevance in practice. And it can decrease large investigation costs by MATLAB programming. What is more, we can select the criterion value of F incremental statistic to control the significance of the breaks, based on kinds of investigation intentions. And the empirical evidences on Shenzhen Composite Index are presented to illustrate the usefulness of our method.展开更多
文摘臭氧浓度的预测对于大气环境治理、空气质量改善等起到了重要的作用。本文提出了一种交互差分时空LSTM网络预测模型(ST-IDN)来挖掘臭氧浓度历史数据的时间相关性和空间相关性,并成功将其应用到网格化臭氧浓度数据预测上。在该模型中,首先交互模块(IC)可以通过一系列的卷积操作来捕捉短期上下文信息,其次层融合模块(LF)可以融合不同层的空间信息来获得上一时刻丰富的空间信息,最后差分时空LSTM模块(DSTM)将捕捉到的时间信息和空间信息进行统一建模实现臭氧浓度预测。所构建模型分别与卷积LSTM网络(ConvLSTM)、预测循环神经网络(PredRNN)以及Memory in Memory网络(MIM)模型在河北省气象局提供的臭氧浓度数据上进行了对比分析,ST-IDN模型的平均绝对误差分别降低了19.836%、12.924%、7.506%。实验结果表明,所提出的模型能够提高臭氧浓度的预测精度。
文摘利用2005年1月至2017年12月搭载在美国环境监测Aura卫星上的臭氧监测仪(Ozone Monitoring Instrument,OMI)数据和NCEP气象资料,在夏季风环流指数定义方法的基础上,重新定义了南亚区域冬季风环流指数,并分别计算了南亚夏季风和冬季风环流指数.结合冬夏两季环流的强弱变化采用相关分析、合成分析和奇异值分解(Singular Value Decomposition,SVD)等方法,探讨了环流异常形势下臭氧的时空变化特征.结果表明:(1)南亚夏季纬向环流与经向环流的强度变化存在一致性,冬季经向环流与纬向环流的强度变化差异较大.(2)南亚臭氧柱总量的季节变化明显,且近13年来臭氧柱总量整体呈上升趋势.(3)夏季(冬季)风环流指数与对流层中低(中高)层和平流层中低层臭氧的相关性显著,但夏季平流层和对流层的相关趋势相反.(4)夏季风环流增强对应青藏高原-伊朗高原上空及南侧区域的上升运动增强,对臭氧的输送作用是造成对流层臭氧分布呈现差异的原因.(5)冬季风环流强弱期的垂直上升和下沉运动中心的移动,以及南北向、东西向气流交汇区的差异是造成臭氧分布不同的原因.
基金supported by the Chinese National Science Foundation(40775074)China Meteorological Administration(GYHY(QX)200706005).
文摘Long-term trends of yearly and seasonal averages of tropospheric ozone over the whole country and some important regions of China during 1979-2005 are analyzed,based on tropospheric ozone residue(TOR) data retrieved from satellite measurements.The relationship between the TOR and Southern Oscillation Index(SOI) is studied.The results show that,over the whole country,there is a slight increase of TOR in summer and a decrease in other seasons,while the overall trend for the whole period is insignificant.There are decreasing trends of TOR over the Pearl River Delta and the Sichuan Basin.Significant increasing trends of TOR are found over the North China Plain(NCP) for all seasons except for winter,with a maximum rate of 1.10 DU per decade for summer.There are significant correlations between TOR and SOI for some regions in China but not for the NCP,suggesting that the observed increasing trend of TOR over the NCP may not be linked with changes in atmospheric circulations.
文摘This paper analyzes multiple structural changes by GMDH (Group Meth- ods of Data Handling), which have obvious advantages. Our method extends the model of Lumsdaine & Papell[1] (1997), and it could be applied to the case of more than two structural changes. Because of simultaneously considering every structural change of the hypothesis, it is likely to be of particular relevance in practice. And it can decrease large investigation costs by MATLAB programming. What is more, we can select the criterion value of F incremental statistic to control the significance of the breaks, based on kinds of investigation intentions. And the empirical evidences on Shenzhen Composite Index are presented to illustrate the usefulness of our method.