Can the climate background of western North Pacific typhoon activity be predicted by climate model?

Based on the observation and reanalysis data through 1948-2004, the vertical shear of zonal wind, outgoing Iongwave radiation, and divergence fields in the lower and upper troposphere during summer are revealed to correlate significantly with the concurrent western North Pacific （WNP） typhoon frequency, and they therefore can be regarded as predictors for the WNP typhoon activity anomaly. After that, the 34-year （1970-2003） ensemble hindcast experiments are performed by the nine-level atmospheric general circulation model developed at the Institute of Atmospheric Physics Under the Chinese Academy of Sciences （IAP9L-AGCM）, aiming to investigate the numerical predictability of the summer vertical shear of zonal wind and divergence field in the lower troposphere. It is found that the temporal correlation coefficients between the hindcast and observation are 0.70 and 0.62 for the vertical shear of zonal wind and divergence field, respectively. This suggests that the model possesses a large potential skill for predicting the large-scale climate background closely related to the WNP typhoon activity, and the model is therefore capable of performing the real-time numerical prediction of the WNP typhoon activity anomaly to some extent.     

A physically-based statistical forecast model for the middle-lower reaches of the Yangtze River Valley summer rainfall

A new approach to forecast the middle-lower reaches of the Yangtze River Valley summer rainfall in June-August （JJA） is proposed in this paper. The year-to-year increment of the middle-lower reaches of the Yangtze River Valley is forecasted and hence the summer precipitation could be predicted. In this paper, DY is defined as the difference of a variable between the current year and the preceding year （year-to-year increment）. YR denotes the seasonal mean precipitation rate of the middle-lower reaches of the Yangtze River Valley summer rainfall. After analyzing the atmospheric circulation anomalies in winter and spring that were associated with the DY of YR, six key predictors for the DY of YR have been identified. Then the forecast model for the DY of YR is established by using the multi-linear regression method. The predictors for the DY of YR are Antarctic Oscillation, the meridional wind shear between 850hPa and 200hPa over the Indo-Australian region, and so on. The prediction model shows a high skill for the hindcast during 1997-2006, with the average relative root mean square error is at 18%. The model can even reproduce the upward and downward trends of YR during 1984--1998 and 1998--2006. Considering that the current operational forecast models of the summer precipitation over the China region have the average forecast scores at 60%--70% and that the prediction skill for the middle-lower reaches of Yangtze River Valley summer precipitation remains quite limited up to now, thus this new approach to predict the year-to-year increment of the summer precipitation over the Yangtze River Valley （and hence the summer precipitation itself） has the potential to significantly increase the operational forecast skill of the summer precipitation.     

基于粒子群-投影寻踪和遗传-神经网络集成的预测模型

  针对预测对象和预测因子存在复杂的线性和非线性关系的特点,利用自然正交展开方法进行线性降维,以及用粒子群 投影寻踪方法进行非线性降维,将高维的非线性数据投影到低维子空间上,构造了一种遗传 神经网络预测模型。在此基础上,应用该预测模型对影响华南的台风频数进行了预测试验,并将预测结果与统计回归模型的预测结果进行对比分析。结果表明,文中构建的非线性集预测模型,对台风频数有较好的预测效果,5 年预测的平均绝对误差为0.81个, 平均相对误差为13%,预测结果比统计回归模型有明显的改进。该文的结果可为进一步探索研究其他领域的预测建模提供了一种新的参考思路和方法。     

A new advance in_~El Nio prediction──The next El Nio will occur in 1997-1998

A statistic model of predicting El Nio based on the multiple regression analysis is developed by using sea surface temperature anomalies in the NINO3 region (5°S-5°N), 90-150°W) as a predictant and the Asian meridional circulation index as the predictors. This model cannot only simulate most of El Nio during 1950-1980 (except for El Nio in 1965), but also hindcast 1982/1983, 1986/1987, 1991/1992 and 1994 El Nio and predict 1997/1998 El Nio six months in advance. At the same time, three different sample series (periods of 1950-1969, 1970-1989, 1950-1989) are used to examine the stability and skill of the statistic model. The models have successfully hindcasted and predicted El Nio before and after the selected samples except for the 1965 El Nio. In addition, the forecast in the El Nio year is clearly superior to that in the following year. The model prediction reaches the lowest skill during March to April.     

A new advance in El Nino prediction

A statistic model of predicting El Niño based on the multiple regression analysis is developed by using sea surface temperature anomalies in the NINO3 region (5°S-5°N), 90–150°W) as a predictant and the Asian meridional circulation index as the predictors. This model cannot only simulate most of El Niño during 1950–1980 (except for El Niño in 1965), but also hindcast 1982/1983, 1986/1987, 1991/1992 and 1994 El Niño and predict 1997/1998 El Niño six months in advance. At the same time, three different sample series (periods of 1950–1969, 1970–1989, 1950–1989) are used to examine the stability and skill of the statistic model. The models have successfully hindcasted and predicted El Niño before and after the selected samples except for the 1965 El Niño. In addition, the forecast in the El Niño year is clearly superior to that in the following year. The model prediction reaches the lowest skill during March to April.     

一种面向工程应用的台风强度统计预测方法

为实现台风强度的年变化预测,提出了一种基于灰色均生函数模型的预测方法。选取最大风速作为台风强度的指标,利用中国气象局发布的西北太平洋台风资料,建立台风登陆和未登陆情况下的预测模型,做5 a预测,并与观测结果对比。结果表明:灰色均生函数模型对台风强度的预测具有较高精度,年平均值最大绝对误差为4.4 m/s,低于2012年各站25次台风强度的数值模式预测结果平均值;年极值最大绝对误差为11.1 m/s,低于上海台风模式对台风个例的预报结果。     

基于近邻成分分析的短期风电功率集成预测

针对短期风电功率预测关键气象因素影响程度的差异和单一模型预测精度不足的问题,提出一种基于近邻成分分析(neighborhood components analysis, NCA)特征加权和Stacking集成预测的短期风电功率预测模型。考虑气象特征对风电功率影响程度不同,利用NCA对气象特征进行加权,将加权特征作为模型输入,强化关键特征的影响程度;在此基础上,构建多个基预测器预测风电功率,并利用结合器将预测结果融合,建立Stacking集成预测模型。算例分析表明,以加权特征作为输入的Stacking集成预测模型具有更高的短期风电功率预测精度。     

一种用于方向预测的集成学习算法

对用方向向量标识示例的学习问题,以预测方向与实际方向之间的方向误差最小化为目标,提出了一种可用于方向预测的集成学习算法,详细分析了构造多个预测函数以及组合各个预测函数以实现方向的最优化预测方法. 提出的算法具有广泛的应用特性：当用不同的轴向来标识类别时,可简化得到多分类连续AdaBoost算法,其能确保训练错误率随分类器个数增加而降低;用错分代价组成的向量来标识示例时,可简化得到一种平均错分代价最小化的集成学习算法. 理论分析和实验结果均表明了算法的合理性和有效性.     

Introduction and systematic assessment for IAP numerical annual climate prediction system

The IAP numerical annual climate prediction system has been presented in this paper. In order to evaluate this annual prediction system, annual ensemble hindcast experiments over a 21-year period from 1980 to 2000 have been done. Systematic assessment shows that this annual prediction system has higher predictability for summer climate in tropic than in extra-tropic area, and higher predictabilities over ocean than over land for the fields of precipitation, sea level pressure and surface air temperature; for 500 hPa geopotential height field, the predictability assuming a zonal distribution decreases from tropic to middle-high latitudes, and in China it is the highest among those of all fields. Correlation analysis shows that the prediction ability of IAP annual prediction system to summer temperature is higher than that to precipitation, and the prediction skill can be remarkably improved by the correction system. Furthermore, the comparison between annual and extraseasonal hindcasts indicates that precipitation hindcasted extraseasonally is better than that done annually, and the major discrepancy exists in middle-high latitudes.     

基于集成分类器的用户属性预测研究

用户属性在个性化服务中具有重要的作用,利用手机数据进行用户属性预测逐渐成为新方向.利用手机应用类别均使用时长和应用类别个数,提出了基本属性与辅助属性的概念.首先对所有未标注样本的辅助属性离散化,将辅助属性基于类别的海灵格距离作为基本属性的特征权重,将基本属性与权重的乘积作为特征训练集成分类器中的各个基分类器,并引入随机森林中的带外样本准确率作为基分类器的权重,得到最终的分类结果.实验结果表明,本文所给出的集成分类器框架能够提高用户属性预测的效果.     

一种新的遗传策略在含“开关”过程的集合预报中的有效应用

针对初值及模式的不确定性,进行了初值扰动集合预报和模式扰动集合预报。在初值扰动集合预报中,将一种新的遗传策略用于四维变分资料同化生成集合预报的分析场,求解条件非线-}生最优扰动,并结合第二、第三主奇异向量（SVs）生成集合预报的初始扰动。为了检验该方法的有效性,采用一个含“开关”过程的偏微分方程的预报模式,设计了3种比较数值试验方案。结果表明：采用第3方案的集合预报在预报技巧上明显高于其他两种方案。第2方案和第1方案相比,由于“开关”的影响,集合预报技巧提高并不明显。在模式扰动集合预报数值实验中,为了模拟模式的不确定性,在控制方程右端添加6个随机的误差项模拟由于物理参数化方案的不同而带来的模式扰动,采用新的遗传策略在扰动模式中同化出6个对应于扰动模式的分析场后进行集合预报,并与基于伴随技术的方法进行比较。结果表明,基于遗传算法的扰动模式集合预报的预报技巧明显优于伴随方法,且这种优势随着预报时间的增加愈发明显。     

Predictor-based error correction method in short-term climate prediction

In terms of the basic idea of combining dynamical and statistical methods in short-term climate prediction, a new prediction method of predictor-based error correction (PREC) is put forward in order to effectively use statistical experiences in dynamical prediction. Analyses show that the PREC can reasonably utilize the significant correlations between predictors and model prediction errors and correct prediction errors by establishing statistical prediction model. Besides, the PREC is further applied to the cross-validation experiments of dynamical seasonal prediction on the operational atmosphere-ocean coupled general circulation model of China Meteorological Administration/ National Climate Center by selecting the sea surface temperature index in Niño3 region as the physical predictor that represents the prevailing ENSO-cycle mode of interannual variability in climate system. It is shown from the prediction results of summer mean circulation and total precipitation that the PREC can improve predictive skills to some extent. Thus the PREC provides a new approach for improving short-term climate prediction.     

热带气旋与台风气候变化研究进展

 近年来,引起严重灾害的热带气旋和台风气候变化研究均有新的进展：近60年观测资料比对表明,由于海上观测手段不足,造成前期资料可靠性较低,近30~40年资料较为可靠,观测资料的可靠性随热带气旋强度的增强而增加;全球6个洋区观测资料计算分析显示,热带气旋存在多年代际变率,约自1970年以来,强和超强热带气旋活动有增强趋势;统计和动力降尺度方法模拟热带气旋频数和强度与观测检验证实,这些方法具有一定的模拟热带气旋的能力,但尚存比较大的不确定性;利用统计和动力降尺度模型及模式考虑21世纪人类排放增加全球变暖,较为一致地预估西北太平洋和北大西洋强台风（飓风）强度和频数都有可能增强。热带气旋和台风均属于给人类带来巨大灾难的极端气候事件,利用目前的统计和动力降尺度方法很难进行较为准确的未来年以上时间尺度的气候预测和预估,尚需对影响因子和预测与预估方法进行更加深入的研究,进一步减小预测和预估的不确定性。     

Thermal Insulation Distribution Pattern of Layered Clothing Ensemble

Clothedbodyincludesclothingandhumanbodytwo parts.Clothingperformance,likeitsthermalinsulation, isinfluencedbyclothingformanddressingway.Because theairlayeroverskinsurfaceanditsstatearedetermined bytheairlayerunderclothing,clothingcoveringarea,airlayerdistribution,clothingoverlap,clothingweightandsoon,thethermalinsulationpropertiesatdifferent partschangewithclothingoverlaplayer,airgapsbetweenclothinglayersandmaterialcombinations[1].In thisresearchtherelationshipbetweenthethermal insulationofmul…     

基于完全集成经验模态分解和模糊熵分频的短期风电功率预测

随着风电接入电力系统的比例日益增大,准确的风电功率预测显得愈发重要。为此,提出了一种基于模糊熵和完全集成经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)的短期风电功率预测模型。采用完全集成经验模态分解将原始风电功率序列进行分解,得到一系列不同频率的子序列。再使用模糊熵(Fuzzy Entropy,FE)算法识别各频率分量特征,将子序列分量分为高、中频分量类和趋势项。趋势项为低频分量,具有较为平稳,波动性小的特点,采用麻雀算法(sparrowSsearch algorithm,SSA)优化支持向量回归(support vector regression,SVR)进行预测；高、中频分量的波动性大且特点较为复杂,则采用SSA优化长短期记忆神经网络(Long Short-Term Memory,LSTM),同时引入注意力机制(Attention Mechanism,AM)对重要信息进行更好的权值分配。最后,经过实验结果分析表明,该模型具有更高的风电功率预测精度。     

基于集成学习的用户信用卡违约预测模型研究

用户信用卡违约预测任务有助于银行等金融机构平衡经济风险与经济利益,对于银行信用卡业务的风险管控具有重要作用。针对用户信用卡违约预测问题,提出了一种基于集成学习的预测模型,有异于传统集成学习中的弱学习器。该模型采用集成模型和神经网络模型作为基学习器,从而提升模型整体的预测效果。首先通过预处理提取用户信用卡数据集的相关特征,然后分别采用优化后的决策树、随机森林、GBDT、XGBoost、CatBoost和SPE六种机器学习模型与神经网络模型进行并行训练和预测,最后通过加权软投票法集成基学习器结果并输出最终预测结果。结果表明,相对于基学习器,该模型在各项评估指标上均有所提升,且拥有更好的模型泛化能力。     

Predictor-based error correction method in short-term climate prediction

In terms of the basic idea of combining dynamical and statistical methods in short-term climate prediction, a new prediction method of predictor-based error correction （PREC） is put forward in order to effectively use statistical experiences in dynamical prediction. Analyses show that the PREC can reasonably utilize the significant correlations between predictors and model prediction errors and correct prediction errors by establishing statistical prediction model. Besides, the PREC is further applied to the cross-validation experiments of dynamical seasonal prediction on the operational atmosphere-ocean coupled general circulation model of China Meteorological Administration/National Climate Center by selecting the sea surface temperature index in Nifio3 region as the physical predictor that represents the prevailing ENSO-cycle mode of interannual variability in climate system. It is shown from the prediction results of summer mean circulation and total precipitation that the PREC can improve predictive skills to some extent. Thus the PREC provides a new approach for improving short-term climate prediction.     

基于集成特征选择的盗窃案件预测方法

盗窃类案件是公安机关较为棘手的一类犯罪,呈现高发低破态势.提前预测发案情况是预防该类型犯罪的有效途径,因此对预测盗窃犯罪提出了一种以Bagging方法为基础、基于特征选择准确度和差异性双重考量的集成学习算法,根据集成学习器好而不同的原则,构造由异质基学习器集成的特征选择器,实现对影响盗窃犯罪发生因子的有效选择,使用更少维度的特征数据集提升犯罪预测的效率和准确度.实验结果表明,提出的SEFV_Bagging算法具有较好的泛化能力和稳定性,在测试数据上表现出的预测准确度也较为理想,且算法无需根据先验知识设置所选特征子集维数,在盗窃犯罪数据分析预测领域应用中有较为明显优势.      

A new approach to the generation of initial perturbations for ensemble prediction: Conditional nonlinear optimal perturbation

Conditional nonlinear optimal perturbation （CNOP）, which is a natural extension of singular vector （SV） into the nonlinear regime, is applied to ensemble prediction study by using a quasi-geostrophic model under the perfect model assumption. SVs and CNOPs have been utilized to generate the initial perturbations for ensemble prediction experiments. The results are compared for forecast lengths of up to 14 d. It is found that the forecast skill of samples, in which the first SV is replaced by CNOP, is comparatively higher than that of samples composed of only SVs in the medium range （day 6-day 14）. This conclusion is valid under the condition that analysis error is a kind of fast-growing ones regardless of its magnitude, whose nonlinear growth is faster than that of SV in the later part of the forecast. Furthermore, similarity Index and empirical orthogonal function （EOF） analysis are performed to explain the above numerical results.     

Use of the advanced microwave sounding unit data to improve typhoon prediction

The effects of incorporating the advanced microwave sounding unit(AMSU-A)data with a modified Zhu-Zhang-Weng vortexbogussing algorithm on typhoon prediction are examined through the use of the PSU/NCAR Mesoscale Model version 5(MM5).The AMSU-A data contain the vertical distribution of the retrieved temperature from satellite brightness temperature, with the geopotential height and wind fields derived through a series of statistical and diagnostic calculations.The advantages of the modified vortexbogussing algorithm include the incorporation of realistic asymmetric typhoon structures,the balanced dynamics with the background field,the easiness to implement and the efficient computations.To test the efficiency of this vortex-bogussing algorithm,the Typhoon Dan event in 1999 is simulated by incorporating the derived AMSU-A fields into the initial conditions of the MM5 modeling system.Results show significant improvements in the track and intensity of the storm, as compared to the simulation without the AMSU-A data.Therefore,this modified vortex-bogussing algorithm can be easily implemented on any typhoon modeling system,which will improve the real-time forecast of tropical cyclones.