适于城市交通时序空间数据预测的层叠泛化模型

受地理过程的动态随机性影响,时序空间数据建模普遍存在先验知识缺乏与模型参数设置问题,导致单一模型难以有效地反映地理系统的整体运行状态。本文提出了一个普适性的集成学习框架,设计了基于异态集成学习的层叠泛化模型,按照组合模型最大化减小原始预测误差原则改进了层叠泛化模型平均输出的混合策略,并基于error-ambiguity decomposition对提出的层叠泛化模型的有效性进行了数学证明。基于北京市交通路网通行状态数据的实验结果表明,层叠泛化模型的均方根误差与平均绝对误差均小于单一模型;平均绝对误差方差均小于基于数理统计的混合模型,验证了层叠泛化模型在时序空间数据预测方面的优越性。     

基于EEMD的明代北京地区干旱灾害特征分析

集合经验模态分解（EEMD）是一种适用于非线性、非平稳序列的信号分析方法,将EEMD应用于气候要素时间序列,可提取真实可靠的气候变化信号。根据北京地区历史时期干旱灾害资料,采用EEMD分解方法对明代（1368—1644年）北京地区干旱灾害等级序列进行多时间尺度的分析,获得简洁且平稳性较好的固有模态函数分量,并与所统计的明代北京地区干旱灾害频次多项式拟合曲线进行对比。结果表明：将EEMD应用于干旱灾害等级序列,可以提取干旱灾害中各个尺度的变化,对明代北京地区干旱灾害进行多尺度分析。明代北京地区干旱灾害存在着2.8年、6.3年的年际周期,11.5年、26.6年、53.6年的年代际周期和118.7年、299.5年的世纪周期。北京地区干旱灾害在明代整个时间跨度上呈现着先增加后微减的变化趋势,总体而言明代中期以后旱灾明显增多。     

Seasonal Prediction for the Indian Monsoon Region with FSU Ocean-atmosphere Coupled Model: Model Mean and 2002 Anomalous Drought

Seasonal climate prediction for the Indian summer monsoon season is critical for strategic planning of the region. The mean features of the Indian summer monsoon and its variability, produced by versions of the ‘Florida State University Coupled Ocean-Atmosphere General Circulation Model’ (FSUCGCM) hindcasts, are investigated for the period 1987 to 2002. The coupled system has full global ocean and atmospheric models with coupled assimilation. Four member models were created by choosing different combinations of parameterizations of the physical processes in the atmospheric model component. Lower level wind flow patterns and rainfall associated with the summer monsoon season are examined from this fully coupled model seasonal integrations. By comparing with observations, the mean monsoon condition simulated by this coupled model for the June, July and August periods is seen to be reasonably realistic. The overall spatial low-level wind flow patterns and the precipitation distributions over the Indian continent and adjoining oceanic regions are comparable with the respective analyses. The anomalous below normal large-scale precipitation and the associated anomalous low-level wind circulation pattern for the summer monsoon season of 2002 was predicted by the model three months in advance. For the Indian summer monsoon, the ensemble mean is able to reproduce the mean features better compared to individual member models.     

一种基于预报集合的降维资料同化方法的数值试验研究

对Qiu和Chou(2006)提出的一种基于预报集合的降维资料同化方法(4DSVD)给出了可行的实施方案，利用中尺度模式MM5产生的模拟资料进行数值试验并将其与MM5/3DVAR的同化结果进行比较，分析了不同的观测误差和观测点密度对同化结果的影响。试验表明:(1)和3DVAR相比该方法能更好地从有观测的变量推断无观测的变量(从温度的观测推测风和比湿)；(2)该方法可以相当有效地滤除观测噪音；(3)该方法具有更好的将观测信息扩展到资料空缺地区的能力。     

地质流体热力学性质的蒙特卡罗模拟

地质流体的热力学性质对研究它们在各种地质作用中的地球化学行为起着十分重要的作用。研究地质流体的传统方法有实验方法和经验、半经验模型方法。这些方法的应用范围一般只能涵盖整个地球温压条件中极为有限的区间。新兴的计算机模拟提供了研究地质流体的第三条途径 ,即通过“计算机实验”的方法 ,模拟极限温压条件下地质流体系统的热力学特性和微观性质。与传统方法相比它有优越性 ,且越来越受到地球化学家们的重视。蒙特卡罗法是流体计算机模拟的主要方法和手段之一 ,吉布斯系综法的提出 ,又显示出用蒙特卡罗法在研究物质相平衡和相变特性方面特有的能力。近来 ,人们把蒙特卡罗法应用于地质流体的研究 ,已取得了一些可喜的成果。但是 ,目前在地质研究中蒙特卡罗模拟大部分还局限于相对简单的系统 ,对于成分较复杂的系统 (特别是含水混合物 )的研究还有许多工作要做     

遗传算法优化的BP神经网络在地面温度多模式集成预报的应用研究

基于TIGGE资料集中的ECMWF、CMA和JMA的数值预报产品,利用加权集成、回归集成和消除偏差集成等线性集成方式与遗传算法优化的BP神经网络(GABP)集成,对我国大部开展地面2 m温度在24 h、48 h和72 h预报时效的多模式集成预报试验。通过对2013年1—6月的预报检验,结果表明:GABP集成预报效果有较大提升,均方误差明显小于各单一模式预报。GABP集成的误差分布在新疆和华北均方误差较大,但是在预报效果改进上GABP集成在西部地区相对单一模式的误差减小更加明显。在进行几种多模式集成方式时,GABP集成相比线性方法预报结果更加精准。对于天气过程个例的预报,GABP集成预报出预报量的变化趋势,预报效果优于单一模式和线性集成预报。无论是较长时间段还是短时间的天气过程,在改进预报效果上GABP集成都起到了最佳的作用。     

一种根据湿对流区域时空分布调整的BGM扰动改进方案

在增长模繁殖法(Breeding of the Growing Mode,BGM)的动态繁殖过程中,尺度化因子的选择极为关键。利用WRF(Weather Research and Forecasting)模式,在分析飑线系统数值模拟误差增长机制的基础上,根据飑线发展过程中湿对流区域误差更容易快速增长的特点,提出了一种根据湿对流区域时空分布调整的BGM初始扰动改进方案。该方案通过在动态繁殖过程中对小扰动的水平结构进行调整,加强湿对流区域扰动,捕获到增长最快方向上的小扰动并将其作为初始扰动。试验结果表明:根据降水量调整的改进方案相比其他方案扰动能量较大,各集合成员之间差异也较大,集合平均预报误差较小;对强降水范围的模拟相对理想,暴雨的降水评分较高;对风廓线及水汽场的预报更接近于实况,较好地改善了集合预报效果。     

Nonlinear measurement function in the ensemble Kalman filter

ABSTRACT The optimal Kalman gain was analyzed in a rigorous statistical framework. Emphasis was placed on a comprehensive understanding and interpretation of the current algorithm, especially when the measurement function is nonlinear. It is argued that when the measurement function is nonlinear, the current ensemble Kalman Filter algorithm seems to contain implicit assumptions： the forecast of the measurement function is unbiased or the nonlinear measurement function is linearized. While the forecast of the model state is assumed to be unbiased, the two assumptions are actually equivalent. On the above basis, we present two modified Kalman gain algorithms. Compared to the current Kalman gain algorithm, the modified ones remove the above assumptions, thereby leading to smaller estimated errors. This outcome was confirmed experimentally, in which we used the simple Lorenz 3-component model as the test-bed. It was found that in such a simple nonlinear dynamical system, the modified Kalman gain can perform better than the current one. However, the application of the modified schemes to realistic models involving nonlinear measurement functions needs to be further investigated.     

The Persistent Heavy Rainfall over Southern China in June 2010：Evolution of Synoptic Systems and the Effects of the Tibetan Plateau Heating

This study investigates influencing weather systems for and the effect of Tibetan Plateau （TP）’s surface heating on the heavy rainfall over southern China in June 2010, focusing on the four persistent heavy rainfall events during 14-24 June 2010. The ma jor weather systems include the South Asian high, midlatitude trough and ridge, western Pacific subtropical high in the middle troposphere, and shear lines and eastward-moving vortices in the lower troposphere. An ensemble of convection-permitting simulations （CTL） is carried out with the WRF model for these rainfall events, which successfully reproduce the observed evolution of precipitation and weather systems. Another ensemble of simulations （SEN） with the surface albedo over the TP and its southern slope changed artificially to one, i.e., the surface does not absorb any solar heating, otherwise it is identical to CTL, is also performed. Comparison between CTL and SEN suggests that the surface sensible heating of TP in CTL significantly affects the temperature distributions over the plateau and its surroundings, and the thermal wind adjustment consequently changes atmospheric circulations and properties of the synoptic systems, leading to intensified precipitation over southern China. Specifically, at 200 hPa, anticyclonic and cyclonic anomalies form over the western and eastern plateau, respectively, which enhances the southward cold air intrusion along the eastern TP and the divergence over southern China;at 500 hPa, the ridge over the northern plateau and the trough over eastern China are strengthened, the southwesterly flows along the northwestern side of the subtropical high are intensified, and the positive vorticity propagation from the plateau to its downstream is also enhanced significantly;at 850 hPa, the low-pressure vortices strongly develop and move eastward while the southwesterly low-level jet over southern China strengthens in CTL, leading to increased water vapor convergence and upward motion over the precipitation region.     

The REMO Ocean Data Assimilation System into HYCOM （RODAS_H）： General Description and Preliminary Results

The first version of the Brazilian Oceano- graphic Modeling and Observation Network （REMO） ocean data assimilation system into the Hybrid Coordi- nate Ocean Model （HYCOM） （RODAS H） has recently been constructed for research and operational purposes. The system is based on a multivariate Ensemble Optimal Interpolation （EnOI） scheme and considers the high fre- quency variability of the model error co-variance matrix. The EnOl can assimilate sea surface temperature （SST）, satellite along-track and gridded sea level anomalies （SLA）, and vertical profiles of temperature （T） and salinity （S） from Argo. The first observing system experiment was carried out over the Atlantic Ocean （78°S-50°N, 100°W-20°E） with HYCOM forced with atmospheric reanalysis from 1 January to 30 June 2010. Five integra- tions were performed, including the control run without assimilation. In the other four, different observations were assimilated： SST only （A SST）; Argo T-S profiles only （AArgo）; along-track SLA only （A_SLA）; and all data employed in the previous runs （A_All）. The A_SST, A_Argo, and A_SLA runs were very effective in improv- ing the representation of the assimilated variables, but they had relatively little impact on the variables that were not assimilated. In particular, only the assimilation of S was able to reduce the deviation of S with respect to ob- servations. Overall, the A_All run produced a good analy- sis by reducing the deviation of SST, T, and S with respect to the control run by 39%, 18%, and 30%, respectively, and by increasing the correlation of SLA by 81%.