水动力学模型与集合卡尔曼滤波耦合的实时校正多变量分析方法

为减少非恒定水流计算中的不确定性,在水流随机运动系统状态空间模型基础上,应用集合卡尔曼滤波(EnKF)技术建立了非恒定水流分析的实时更新(校正)方法。该方法适用于非线性的随机微分方程,过程和观测噪声可以是非正态分布。同时,为充分利用水位、流量等误差量级相差巨大的观测中所蕴含的有效信息,导出了EnKF多变量分析格式。以明渠单峰洪水过程的合成数据为例,考察了运用建立的实时更新方法分析预报一维洪水演进的性能。重点对比了采用不同精度等级下的水位和流量观测资料进行滤波的效果。在中国现行标准规定的允许观测误差范围内,以水位观测进行一维洪水动力学模型的滤波分析可有效地控制误差、估计流量、识别水流运动系统状态。长江干流清溪场至万县江段实际洪水计算还证实:该方法通过插入即时观测,可实时更新模型状态,给出与实际更为接近的计算。     

BP神经网络洪水预报模型在洪水预报系统中的应用

采用相关分析法,在区域降水、观测断面流量(或水位)因子中识别出影响预报断面径流过程的主要变量,在多个观测断面的数据均为流量情况下,采用基于时延组合的合成流量为影响预报断面径流过程的变量,采用自相关分析法,识别出影响预报断面径流过程的前期流量(或水位),以这些变量为BP神经网络模型的输入,以预报断面的流量(或水位)为模型的输出,在BP神经网络隐层节点数自动优选的基础上,构建了基于BP神经网络的洪水预报模型。将模型载入中国洪水预报系统中,应用结果表明:模型在历史洪水训练样本具有一定代表性的情况下,可获得较高的预报精度。     

长江中游洞庭湖防洪系统水流模拟——Ⅱ．模型实现和率定检验

在长江中游洞庭湖防洪系统水流模拟模型的建模思路和基本算法的基础上,侧重讨论建模过程中遇到的内、外边界条件的特殊处理、洪水实时预报校正模型的应用、资料处理和模型数据的拟定等重要环节,强调了有关处理及其与模型基本算法的磨合在复杂水流模型中的重要作用.采用80年代两场实测洪水对所建模型进行了率定和检验.结果表明,所建模型清晰地重现了长江中游--洞庭湖区的洪水动态演进过程,得到了总体趋势合理的水位分布和流速或流量分布;模拟的控制站水位、流量过程与实况比较,峰谷完全对应,过程吻合良好,精度满足防洪预报和分析的要求,能够较好地模拟洪水波在这样复杂水系中的水流特征.     

双校正模式下的大清河流域陆气耦合洪水预报研究

基于数值大气模式WRF、三维变分数据同化WRF-3DVar、河北雨洪模型以及实时校正模型ARMA,在北方半湿润半干旱地区的大清河流域构建了陆气耦合洪水预报系统,并利用2012、2013年发生的3场降雨洪水,对系统的降雨洪水预报结果进行分析。结果表明:雷达反射率与GTS数据的同时同化,可有效改善数值大气模式对中小尺度流域的降雨预报效果,从而降低系统的洪水预报误差,ARMA模型的应用,能够进一步提升系统的洪水预报精度,随着预见期的延长,系统的预报精度下降,但系统在6h预见期内仍表现出较好的应用效果。因此,在数据同化和实时校正的"双校正"模式下,陆气耦合洪水预报系统在延长洪水预报预见期的同时,具有较高的洪水预报精度,具有一定的应用前景。     

水箱模型在大尺度流域实时洪水预报模型研制中的应用

水箱模型用于实时洪水作业预报的具体成果尚不多，本文研究的目的在于将水箱模型用于大尺度流域实时洪水预报。本文介绍了应用水箱模型建立实时洪水预报模型的方法。按照河段流量传播时间将寸滩以上干、支流划分为若干子河段，各子河段按照计算时段长分成若干单元河段，各单元河段区间降雨径流预报采用水箱模型．河道流量演算采用连续马斯京根法。河系预报模型精度在85％以上，能够满足实时洪水预报的要求。     

结合糙率校正的河网水情数据同化

为解决河网水动力模型重要参数糙率与水力状态量水位、流量的同步校正问题,以糙率和水力状态量作为河网非线性动态系统变量,采用扩展卡尔曼滤波,构建结合糙率动态校正的河网水情数据同化模型.通过算例计算,系统分析了水位动态噪声水平、糙率动态噪声水平、糙率初始值及测站个数对模型校正的影响.结果表明:模型能够有效用于水位状态量的实时校正;靠近测站的糙率校正值趋于真值,远离测站的糙率校正值趋于初始值;通过调整糙率动态噪声水平,可以有效控制糙率的修正量,防止糙率修正过大而引起计算失效问题.     

河道洪水实时预报的半自适应模型研究

提出和讨论了基于马斯京根流量演算河道洪水实时预报的半自适应滤波模型.在该模型中量测误差系列的协方差矩阵可以通过信息更新系列实时估计出来,只有模型误差系列的协方差矩阵需要预先给出.提出了一个处理区间入流较为合理、方便的方法.通过验证和应用说明了该模型的合理性.     

冰盖下矩形河道的综合糙率

综合糙率是采用曼宁公式确定河道水位和流量关系的关键参数。在河道冰封期,冰盖的出现增加了流动的阻力,明流条件下确定的综合糙率不再适用,需要重新估算。基于Einstein阻力划分过流断面的原理,冰盖下矩形河道的过水断面可划分为冰盖区、河床区和边壁区。根据总流的连续性方程,在确定各分区糙率系数、水力半径和断面面积的基础上,提出了冰盖下矩形河道综合糙率的计算公式。采用已有的试验水槽测量数据和天然河道实测资料对公式进行了验证,结果表明:公式计算的综合糙率与实测值吻合较好,与Einstein公式和Sabaneev公式相比,计算精度更高;对于冰封水流,宽浅河道采用分区水深代替水力半径进行简化计算的条件有别于明渠水流,在宽深比大于20时,计算结果才满足精度要求。     

糙率对感潮河段水位预报的影响

感潮河段洪水受上游洪水和下游潮水的共同作用,且河道断面复杂,这些因素加剧了糙率取值的难度.本文通过糙率影响试验,分析了糙率改变对感潮河段计算水位的影响.在具有明显漫滩的断面,将糙率取值分为漫滩糙率和主槽糙率后再进行水位计算.试验结果表明,糙率值的改变对计算水位影响较大,且对感潮河段洪水过程的不同阶段影响程度不同;漫潍糙率的改变对漫滩洪水位影响很大,主槽糙率的改变对低水位影响较大.     

三种基于神经网络的洪水实时预报方案的比较研究

在总结神经网络应用的基础上,归纳了3种基于神经网络的洪水实时预报方案。第一种是神经网络水文模型的模拟模式加模拟误差的自回归校正模型,第二种是权重系数固定的神经网络实时预报方案,第三种是权重系数自动更新的神经网络实时预报方案。采用10个不同流域的日流量资料对这3种方案进行率定和校核。比较这3种方案的实时预报精度。结果发现,第三种方案不仅预报精度要高于其他两种方案,而且比第一种方案少了一个自回归校正模型,结构简洁。本文建议采用第三种洪水实时预报方案。     

Flood forecasts based on multi-model ensemble precipitation forecasting using a coupled atmospheric-hydrological modeling system

The recent improvement of numerical weather prediction (NWP) models has a strong potential for extending the lead time of precipitation and subsequent flooding. However, uncertainties inherent in precipitation outputs from NWP models are propagated into hydrological forecasts and can also be magnified by the scaling process, contributing considerable uncertainties to flood forecasts. In order to address uncertainties in flood forecasting based on single-model precipitation forecasting, a coupled atmospheric-hydrological modeling system based on multi-model ensemble precipitation forecasting is implemented in a configuration for two episodes of intense precipitation affecting the Wangjiaba sub-region in Huaihe River Basin, China. The present study aimed at comparing high-resolution limited-area meteorological model Canadian regional mesoscale compressible community model (MC2) with the multiple linear regression integrated forecast (MLRF), covering short and medium range. The former is a single-model approach; while the latter one is based on NWP models [(MC2, global environmental multiscale model (GEM), T213L31 global spectral model (T213)] integrating by a multiple linear regression method. Both MC2 and MLRF are coupled with Chinese National Flood Forecasting System (NFFS), MC2-NFFS and MLRF-NFFS, to simulate the discharge of the Wangjiaba sub-basin. The evaluation of the flood forecasts is performed both from a meteorological perspective and in terms of discharge prediction. The encouraging results obtained in this study demonstrate that the coupled system based on multi-model ensemble precipitation forecasting has a promising potential of increasing discharge accuracy and modeling stability in terms of precipitation amount and timing, along with reducing uncertainties in flood forecasts and models. Moreover, the precipitation distribution of MC2 is more problematic in finer temporal and spatial scales, even for the high resolution simulation, which requests further research on storm-scale data assimilation, sub-grid-scale parameterization of clouds and other small-scale atmospheric dynamics.     

一个稳健的河网水情数据同化耦合模型

为解决河网水情动态系统的数据同化问题,将水情量测数据作为内边界条件,建立了一个基于误差最小二乘原理的河网水情数据同化耦合模型.同时,为了有效控制水情状态变量的校正程度,避免因校正过度而破坏河网系统的水量平衡关系,构建了一个能有效控制及平衡各量测变量校正程度的权重矩阵.通过模拟算例和实例应用,系统检验了模型在河网水情仿真与预报中的实时校正能力,并得到了理想的结果.结果表明:模型理论完备、构建简便、数据同化可控性好,能够有效地进行河网水情动态系统的数据同化,提高河网水情仿真与预报的精度.     

考虑降水预报的三峡入库洪水集合概率预报方法比较

为了考虑预见期内降水预报的不确定性对洪水预报的影响,采用中国气象局、美国环境预测中心和欧洲中期天气预报中心的TIGGE(THORPEX Interactive Grand Global Ensemble)降水预报数据驱动GR4J水文模型,开展三峡入库洪水集合概率预报,分析比较BMA、Copula-BMA、EMOS、M-BMA 4种统计后处理方法的有效性。结果表明:4种统计后处理方法均能提供一个合理可靠的预报置信区间;其期望值预报精度相较于确定性预报有所提高,尤其是水量误差显著减小;M-BMA方法概率预报效果最佳,它能够考虑预报分布的异方差性,不需要进行正态变换,结构简单,应用灵活。     

Evaluation of the real-time TRMM-based multi-satellite precipitation analysis for an operational flood prediction system in Nzoia Basin,Lake Victoria,Africa

Many researchers seek to take advantage of the recently available and virtually uninterrupted supply of satellite-based rainfall information as an alternative and supplement to the ground-based observations in order to implement a cost-effective flood prediction in many under-gauged regions around the world. Recently, NASA Applied Science Program has partnered with USAID and African-RCMRD to implement an operational water-hazard warning system, SERVIR-Africa. The ultimate goal of the project is to build up disaster management capacity in East Africa by providing local governmental officials and international aid organizations a practical decision-support tool in order to better assess emerging flood impacts and to quantify spatial extent of flood risk, as well as to respond to such flood emergencies more expediently. The objective of this article is to evaluate the applicability of integrating NASA’s standard satellite precipitation product with a flood prediction model for disaster management in Nzoia, sub-basin of Lake Victoria, Africa. This research first evaluated the TMPA real-time rainfall data against gauged rainfall data from the year 2002 through 2006. Then, the gridded Xinanjiang Model was calibrated to Nzoia basin for period of 1985–2006. Benchmark streamflow simulations were produced with the calibrated hydrological model using the rain gauge and observed streamflow data. Afterward, continuous discharge predictions forced by TMPA 3B42RT real-time data from 2002 through 2006 were simulated, and acceptable results were obtained in comparison with the benchmark performance according to the designated statistic indices such as bias ratio (20%) and NSCE (0.67). Moreover, it is identified that the flood prediction results were improved with systematically bias-corrected TMPA rainfall data with less bias (3.6%) and higher NSCE (0.71). Although the results justify to suggest to us that TMPA real-time data can be acceptably used to drive hydrological models for flood prediction purpose in Nzoia basin, continuous progress in space-borne rainfall estimation technology toward higher accuracy and higher spatial resolution is highly appreciated. Finally, it is also highly recommended that to increase flood forecasting lead time, more reliable and more accurate short- or medium-range quantitative precipitation forecasts is a must.     

嘉陵江流域三江汇流洪水联机实时预报研究

不直接描述嘉陵江流域三洪汇流存在的洪水顶托等复杂的物理过程，而是采用时变参数线性模型，由作者修正的一种自适应滤波水文实时预报方法进行参数在线识别，并分别研究集总输入和多输入两种情况，实现了洪水联机实时预报，结果表明，预报精度高，且模型简单，滤波参数稳定，便于推广使用。     

Dynamic routing modeling for flash flood forecast in river system

A real-time flood-forecasting method coupled with the one-dimensional unsteady flow model was developed for the Danshuei River system in northern Taiwan. Based on the flow at current time, the flow at new time is calculated to provide the water stage forecasting during typhoons. Data, from two typhoons in 2000: Bilis and Nari, were used to validate and evaluate the model capability. First, the developed model was applied to validate and evaluate with and without discharge corrections at the Hsin-Hai Bridge in Tahan Stream, Chung-Cheng Bridge in Hsintien Stream, and Sir-Ho Bridge in the Keelung River. The results indicate that the calculated water stage profiles approach the observed data. Moreover, the water stage forecasting hydrograph with discharge correction is close to the observed water stage hydrograph and yields a better prediction than that without discharge correction. The model was then used to quantify the difference in prediction between different methods of real-time water stage correction. The model results reveal that water stages using the 1–6 h forecast with real-time stage correction exhibits the best lead times. The accuracy for 1–3 h lead time is higher than that for 4–6 h lead time, suggesting that the flash flood forecast in the river system is reasonably accurate for 1–3 h lead time only. The method developed is effective for flash flood forecasting and can be adopted for flood forecasting in complicated river systems.     

基于分布式水文模拟的三峡区间洪水预报(Ⅰ)——模型构建及验证

长江三峡区间因暴雨形成的洪水峰高量大,对三峡水库的防洪安全和运行调度的影响很大.本论文依据三峡地区的地形地貌特征,采用基于GIS的机理性分布式水文模型,来模拟三峡区间入库洪水,以尽量减少洪水预报中的不确定性.利用近期建成的78个自动雨量站网监测的小时降雨信息作为模型的输入,对模型参数进行了率定和验证,结果表明:大多数洪水过程的模拟精度较好,但也有的模拟结果较差,其中降雨信息缺失是洪水预报不确定性的主要来源.     

Probabilistic forecasting of landslide displacement accounting for epistemic uncertainty: a case study in the Three Gorges Reservoir area,China

Accurate and reliable displacement forecasting plays a key role in landslide early warning. However, due to the epistemic uncertainties associated with landslide systems, errors are unavoidable and sometimes significant in traditional methods of deterministic point forecasting. Transforming traditional point forecasting into probabilistic forecasting is essential for quantifying the associated uncertainties and improving the reliability of landslide displacement forecasting. This paper proposes a hybrid approach based on bootstrap, extreme learning machine (ELM), and artificial neural network (ANN) methods to quantify the associated uncertainties via probabilistic forecasting. The hybrid approach consists of two steps. First, a bootstrap-based ELM is applied to estimate the true regression mean of landslide displacement and the corresponding variance of model uncertainties. Second, an ANN is used to estimate the variance of noise. Reliable prediction intervals (PIs) can be computed by combining the true regression mean, variance of model uncertainty, and variance of noise. The performance of the proposed hybrid approach was validated using monitoring data from the Shuping landslide, Three Gorges Reservoir area, China. The obtained results suggest that the Bootstrap-ELM-ANN approach can be used to perform probabilistic forecasting in the medium term and long term and to quantify the uncertainties associated with landslide displacement forecasting for colluvial landslides with step-like deformation in the Three Gorges Reservoir area.     

洪水预报产流误差的动态系统响应曲线修正方法

为提高实时洪水预报精度,提出了一种基于动态系统响应曲线洪水预报误差修正新方法。该方法将动态系统响应曲线引入洪水预报误差修正中,建立一种向误差源头追溯的动态反馈修正模型。此修正方法将新安江模型产流以下的部分作为响应系统, 用线性差分近似代替非线性系统响应函数的偏导数值,得到时段产流量所对应的系统响应曲线。用实测流量和计算流量之间的差值作为信息,使用最小二乘估计原理,对产流量进行修正, 再用修正后的产流量重新计算出流过程。该修正方法分别用理想案例和王家坝流域进行检验,结果证明此方法效果比传统二阶自回归模型有明显提高。