太湖蓝藻生境对气候变化的响应

近年来太湖蓝藻暴发已成为重要水污染事件,是太湖面临的重大水安全问题,气候变化因素加剧了其严重性。为了预测未来由于气候变化对太湖蓝藻暴发的影响,并提出应对气候变化的策略,开展了蓝藻生境对气候变化响应关系的研究。基于大量实测气象资料的统计分析与气候变化情景计算,总结分析了近50年来太湖流域气候变化呈现出气温增高、风速略有下降、日照减少、降雨增多、湿度降低等趋势。气候变化是太湖蓝藻水华暴发的重要影响因素,其中气温与风速变化是影响太湖藻类生长的敏感气象因子。气温升高导致的蓝藻暴发风险平均10年将增加约2%,风速降低导致蓝藻水华暴发的风险平均每5年增加约3.5%。结合蓝藻对敏感因子响应关系的试验结果,提出了截污减排、适当清除底泥、打捞蓝藻、调水引流、修复生态等应对措施。     

Modeling the Role of Zebra Mussels in the Proliferation of Blue-green Algae in Saginaw Bay,Lake Huron

Between 1991 and 1993, Saginaw Bay experienced an invasion by zebra mussels, Dreissena polymorpha, which caused a significant perturbation to the ecosystem. Blooms of Microcystis, a toxin-producing blue-green alga, became re-established in the bay after the zebra mussel invasion. Microcystis blooms had all but been eliminated in the early 1980s with controls on external phosphorus loadings, but have re-occurred in the bay most summers since 1992. An apparent paradox is that these recent Microcystis blooms have not been accompanied by increases in external phosphorus loadings. An ecosystem model was used to investigate whether the re-occurrence of Microcystis could be due to changes caused by zebra mussels that impacted phytoplankton community structure and/or internal phosphorus dynamics. The model was first used to establish baseline conditions in Saginaw Bay for 1991, before zebra mussels significantly impacted the system. The baseline model was then used to investigate: (1) the composite impacts of zebra mussels with average 1991–1995 densities; (2) sensitivity to changes in zebra mussel densities and external phosphorus loadings; and (3) three hypotheses on potential causative factors for proliferation of blue-green algae. Under the model assumptions, selective rejection of blue-green algae by zebra mussels appears to be a necessary factor in the enhancement of blue-green production in the presence of zebra mussels. Enhancement also appears to depend on the increased sediment-water phosphorus flux associated with the presence of zebra mussels, the magnitude of zebra mussel densities, and the distribution of zebra mussel densities among different age groups.     

Effect of selenium on the growth of some selected green and blue-green algae

An algal assay procedure using microplate technique was carried out to assess the effect of selenium on the growth of some green and blue-green algae. Sodium selenite pentahydrate (Na₂SeO₃.5H₂O) and selendioxide (Se₂O₄) were tested as selenium. The test algae were Selenastrum capricornutum, Scenedesmus obliquus, Chlorella sp., Monoraphidium convolutum, Monoraphidium contortum, Monoraphidium griffithii, Anabaena flos-aquae, Micro-cystis aeruginosa, and Oscillatoria agardhii. The blue-green algae are toxin-producing strains. Cell number and in vivo chlorophyll fluorescence were the growth parameters of green and blue-green algae respectively. Dose-response curves quantify the selenium toxicity in terms of EC₅₀. The lowest selenium concentrations giving no detectable growth (EC₁₀₀) were visually inspected. The test algae showed distinctly different responses to various selenium treatments. Selenium at concentrations < 0.1 mg L⁻¹ stimulates, to varying degrees, the growth of all the green algae with one exception: S. capricornutum was dramatically inhibited at all treatments. O. agardhii, A. flos-aquae and M. aeruginosa showed marginal growth stimulation up to 0.1 mg Se L⁻¹, 1.0 mg Se L⁻¹ and 3.2 mg Se L⁻¹ respectively. In terms of EC₅₀, the compound selendioxide was more toxic to greens than sodium selenite. The reverse was indicated for the blue-greens. In general, the EC₅₀ presented a wide range (0.08–7.2 mg Se L⁻¹), with the lowest values reported for S. obliquus and the highest for M. aeruginosa. All the test algae, except S. obliquus, maintained observable growths at higher selenium levels (EC₁₀₀>1.0<100mg Se L⁻¹).     

基于SWAT模型的湟水流域蓝绿水与不同土地利用类型的绿水差异研究

蓝绿水的时空分布直接关系到一个地区的水资源与粮食安全。本文利用SWAT分析了1981-2000年间蓝绿水随时间的变化规律以及不同土地利用类型绿水分异规律。研究表明:1981-2000年间湟水流域蓝绿水均来源于降水,绿水资源约为蓝水的3.8倍;年均绿水流在不同的土地利用类型中差异明显,农业用地最大,居民用地最小;绿水贮存量较为丰富的是湿地和中低覆盖度的草甸;选取1981-1985年和1996-2000年分别为研究的前期和后期,得到农业用地的绿水流、绿水贮存变化量与作物生长周期和管理条件密切相关。     

基于遗传优化神经网络的多源遥感数据反演土壤水分

为快速反演较高精度土壤水分,提出用遗传算法优化后的神经网络辅以多源遥感数据的方法进行地表土壤水分反演。首先建立4层神经网络并用遗传算法优化此网络,之后以雷达数据不同极化(VV、VH、VH/VV)的后向散射系数、雷达入射角、光学数据的归一化植被指数(NDVI)、以及高程数据作为网络的输入,土壤水分数据为输出,对网络进行训练与仿真,再运用地表实际测量数据与反演数据做对比验证。结果表明:反演结果与实际测量数据相关性良好,R2可达0. 79。采用遗传算法对神经网络优化的土壤水分反演方法可行,且添加光学数据等辅助数据后土壤水分反演效果更优,为多源遥感土壤水分的协同反演研究提供新思路。     

秃尾河流域年径流变化特性分析

对流域径流的变化规律进行准确地分析及合理地预测对流域水资源的合理开发、水利工程的建设以及社会经济的发展具有重要的指导意义。利用Mann-Kendall秩次相关检验法和小波分析理论对秃尾河流域的径流变化规律进行分析研究,并建立BP神经网络模型对径流变化进行预测分析。结果表明:秃尾河流域年径流量变化总体上有明显的下降趋势;从小波系数图可以看出年径流过程主要存在2年、8年和19年左右的变化周期,其中19年左右时间尺度为第一主周期,同时发现目前年径流处在枯水期后期,水量有转向增加的趋势;采用BP神经网络法对秃尾河流域高家川站年径流量进行预测,预测结果相对误差仅为5.92%,说明所建立的BP神经网络模型用于该流域的年径流预测得精度较高,是一种有效地年径流预测方法。     

RBF神经网络模型在砂土液化判别中的应用研究

以时松孝次收集的砂土液化数据为研究对象,选取黏粒含量ρ_c、相对密实度D_r、临界深度d_s、竖向有效应力σ′、地下水位d_w、地震震级M、最大地面水平加速度α_max和标准贯入次数SPT-N等8个砂土液化的主要影响因素作为RBF神经网络的输入参数,利用MATLAB7.0中的神经网络工具箱,对部分样本数据进行训练和测试。并利用建立的RBF神经网络模型分析了各因素对砂土液化的影响规律。结果表明：砂土液化判别指标随α_max的增加而增大,随SPT-N和d_w的增加而减小。研究成果表明,建立的RBF网络模型完全满足砂土液化判别的精度要求,能够精确模拟输入和输出之间复杂的非线性映射关系,具有较高的预测精度,具有重要的工程应用价值。     

Elman与GRNN神经网络模型在水环境承载力评价中的应用——以文山州区域水环境承载力评价为例

利用层次分析法构建符合区域水环境承载力的评价指标体系和分级标准,基于Elman神经网络与广义回归神经网络(GRNN)算法原理,提出Elman与GRNN神经网络水环境承载力评价模型,采用内插法构造网络训练样本,将水环境承载力分级评价标准阈值样本进行评价,将结果作为区域水环境承载力等级评价的划分依据,对文山州不同规划水平年水环境承载力进行评价。结果表明:文山州不同规划水平年水环境承载力处于绝对可承载与基本可承载之间,客观反映了区域水环境现状及规划期望效果,可为区域水环境承载力评价和研究提供参考。Elman与GRNN神经网络模型评价结果基本相同,表明研究建立的区域水环境承载力评价模型和评价方法均是合理可行的,二者均可作为区域水环境承载力评价的选用模型。     

基于碾压波速的堆石坝压实质量实时监测指标的研究

已有的连续压实质量评价指标在评估堆石坝料的压实质量时仍存在评价精度低、表征压实效果复杂以及结果易受压实材料属性影响等缺点.为给堆石坝施工质量的连续控制提供有效指标,本文采用数据延拓式相关的相位差求解方法来间接获取碾压波速(VR),提出了以实时监测的VR作为堆石坝料压实状态的表征指标.从定性分析角度考虑碾压参数对VR的影...     

基于遗传算法与神经网络的水库水沙联合优化调度模型

针对一维泥沙数学模型维数高、求解耗时长以及水沙联合调度模型多目标难以求解的问题,结合遗传算法与神经网络的特性,以发电量最大和有效库容最大为基本目标,构建了水库水沙联合优化调度模型。利用约束法和权重法,将多目标模型转化为单目标模型,采用加速遗传算法进行求解,其中泥沙冲淤量使用自适应BP神经网络进行拟合预测。三峡水库实例计算结果表明:运行20 a,与原设计运行方式相比,采用该优化调度模型优化运行年均发电量增加7.732%,泥沙淤积量增加0.044%,在淤积量增加很小的情况下能大幅度增加发电量,模型能较好地解决水库水沙联合调度问题,在工程实际中是有效可行的。     

基于互信息与神经网络的天山西部山区融雪径流中长期水文预报

为提高天山西部山区融雪径流的预报精度,更好地指导所在区域的工农业生产发展,针对影响预报精度的关键问题(预报因子的选择),基于互信息法、相关系数法、主成分分析法对研究区的预报因子进行优选,采用RBF神经网络以及组合小波BP神经网络模型进行径流预报研究,并进行不同方案的比较。结果表明:①互信息法优选出的预报因子作为模型输入可以提高预报精度;②采用不同优选预报因子作为RBF神经网络以及组合小波BP神经网络模型的输入变量,结果表明RBF神经网络模型的预测精度要好于组合小波BP神经网络模型;③以相对误差作为评价模型精确度的标准,预测效果最好的是基于互信息方法挑选出的预报因子作为RBF神经模型输入数据的模型预测结果。     

双隐含层BP神经网络模型在老哈河水质预测中的应用

为了快速准确预测老哈河水质,采用老哈河2011-2015年水质监测数据,运用拉格朗日插值法补充缺失值,分别对化学需氧量、生化需氧量、高锰酸盐指数和总磷浓度建立Levenberg-Marquardt优化的双隐含层BP神经网络模型,利用2011-2014的数据建立训练网络,以2015年的数据进行验证与测试。结果表明:五日生化需氧量预测模型,第一隐含层节点数为4,第二隐含层节点数为12时,决定系数0.751 6(P=0.000 3),平均相对误差25.73%;化学需氧量预测模型,第一隐含层节点数为12,第二隐含层节点数为10时,决定系数0.887 5(P0.000 1),平均相对误差27.69%;高锰酸盐预测模型,第一隐含层节点数为6,第二隐含层节点数为3时,决定系数0.854 7(P0.000 1),平均相对误差28.90%;总磷预测模型,第一隐含层节点数为12,第二隐含层节点数为12时,决定系数0.889 2(P0.000 1),平均相对误差17.94%。应用拉格朗日插值法对缺失数据进行补充后建立的双隐含层BP神经网络模型相对误差均小于28.90%,模型的预测效果较好,其中总磷浓度预测效果最好。通过拉格朗日插值,可以建立老哈河赤峰段甸子点位污染指标的双隐含层人工神经网络模型进行水质预测。     

ARIMA与ANN组合预测模型在中长期径流预报中的应用

基于时间序列预测模型及BP神经网络,提出了新的组合预测方法.该方法采用三层结构的BP神经网络来构造组合预测模型,运用时间序列模型预测方法得出的预测结果,采用历史滚动法将前5年的预测结果数据作为BP网络的输入,以当前年份的预测结果为网络期望输入,建立了ARIMA-ANN组合预报模型.利用Matlab7神经网络工具箱对塔里木河上游源流卡群水文站的年径流量进行了预报及验证.结果表明:组合模型的预报结果精度高,容错能力强,是中长期径流预报的有效方法.     

新安江模型和改进BP神经网络模型在闽江水文预报中的应用

精确的水文预报是防洪减灾中重要的非工程措施,水文模型是开展水文预报最有力的工具。采用LM算法改进了的BP神经网络水文预报模型,以闽江富屯溪流域为例,进行了BP模型和新安江模型在日流量模拟预报中的应用比较。结果表明:两个模型总体均达到水文预报的精度要求,水文预报合格率可达到90%以上;新安江模型在丰水年模拟效果较好,相比而言,BP神经网络模型的模拟精度更高一些;两个模型均可用于闽江流域的水文预报研究。     

Modeling of Phytoplankton-Nutrient Dynamics in Saginaw Bay,Lake Huron

A multi-class, phytoplankton simulation model was developed and calibrated to an extensive set of field data acquired on Saginaw Bay, Lake Huron, during 1974. Phytoplankton biomass was partitioned into five functional groups: diatoms, greens, non-N₂-fixing blue-greens, N₂-fixing blue-greens, and “others”. Nutrients included in the model were phosphorus, nitrogen, and silicon. The model was applied to a single spatial segment encompassing the inner portion of Saginaw Bay.Process level analyses were conducted with the calibrated model to determine the relative importance of various factors affecting phytoplankton and nutrient dynamics. The concept of a single limiting factor for phytoplankton growth was found to be overly simplistic. Results indicated that temperature and light were relatively more growth rate limiting than nutrients on an annual average basis. However, as a consequence of nutrient depletion, nutrients became relatively more important at the times of peak phytoplankton crops. Nitrogen was found to be relatively more growth rate limiting than phosphorus to the total phytoplankton crop, although important differences occurred among the individual functional groups. At various times, and for various groups, all three nutrients were important in limiting either the rates of growth and/or the maximum sizes of the phytoplankton crops. Results were consistent with the hypothesis that while nitrogen and silicon were important in phytoplankton-nutrient dynamics, the supply of phosphorus would ultimately determine the size of the blue-green component of the total crop because N₂-fixing blue-greens do not have absolute requirements for dissolved available nitrogen or silicon. Results indicated that phosphorus requirements of spring and fall diatom crops were satisfied primarily by external loadings. Phosphorus requirements of summer blue-green crops were satisfied primarily by recycle processes within the water column. Upon cell death, direct nutrient recycle to the available nutrient compartments in the water column from excess internal phytoplankton stores was found to be important for both phosphorus and nitrogen. Phytoplankton production was found to be extremely sensitive to variations in the light extinction coefficient in the water column, and relatively insensitive to variations in incident solar radiation.     

基于多尺度卷积神经网络的管道泄漏检测模型研究

如何有效检测管道泄漏是节水型社会建设迫切需要解决的关键和热点问题之一。近年来基于深度学习的管道泄漏检测方法发展迅速,本文针对传统单尺度卷积神经网络对泄漏特征提取不充分的问题,提出一种基于多尺度一维卷积神经网络(MS1DCNN)的管道系统泄漏检测模型。该方法利用多个不同卷积尺度的卷积通路并行提取管道泄漏的特征并进行泄漏信息的分类预测。基于经典的管道系统布置,利用瞬变流模型生成管道泄漏工况下的三个水压数据集对模型进行验证,三个数据集分别用于预测管道的泄漏位置、泄漏量和非恒定摩阻系数,对应样本数为39601、3980、4900,并将预测结果与其他深度学习方法和传统的机器学习方法进行对比分析。结果表明：MS1DCNN模型对数据集样本下泄漏位置、泄漏量、非恒定摩阻系数的分类准确率达到99.96%、98.48%、100%,三者平均预测精度比传统一维卷积神经网络(1DCNN)、BP神经网络、支持向量机(SVM)和k近邻算法(KNN)提高0.31%、2%、1.27%、22.8%;MS1DCNN在信噪比为-4～12 dB的噪声环境下各数据集的平均F₁分数分别为99.2%、97.02...     

考虑气象因子不确定性的参考作物蒸散量预报方法

参考作物蒸散量(ET0)的准确预测预报对于制定作物灌溉制度与实时灌溉调度具有重要意义,然而气象因子的不确定性极大的影响着ET0的预测精度.因此本研究采用马尔科夫蒙特卡罗模拟与自适应采样算法相结合的方法(AM-MCMC)对气象因子的不确定性进行修正,以气象站实测ET0作为标准值,利用径向基神经网络(RBF)模型建立气象因...     

基于灰色系统与RBF神经网络的中长期水文预报

及时、准确的中长期水文预报能有效促进水库管理优化。以非汛期各月径流量为预报因子,通过计算所需预报年份与已有径流资料历史年份的预报因子之间的灰色关联度,遴选出与该年灰色关联度较大的年份作为代表年份。采用MATLAB数学软件构建RBF神经网络预报模型,利用选定的代表年份径流量对目标年份汛期径流量进行预报。以清河水库为例,用该模型预报汛期径流量。结果表明,模型简单可操作、运行速度快、预报效果好。     

西藏拉萨河径流预测方法研究

本文介绍了人工神经网络模型、分期平稳自回归模型、一阶季节性自回归模型这三种径流预测模型的基本原理,并且利用西藏拉萨河拉萨站1956至1968年,1973年至2000年41年的月平均流量资料对月径流进行预测和比较,得出BP-人工神经网络模型是相对于其它两种方法更适合对拉萨河径流进行预测的方法。     

Circulation and mixing in Sodus Bay due to water exchange with Lake Ontario

Embayments on large lakes may be affected by water exchange with the lake that, in turn, impact water quality in the embayment. In this study we examine the influence of hydrodynamic factors that may play a role in controlling water quality in Sodus Bay, the largest enclosed embayment on the U.S. shoreline of Lake Ontario. The motivation for this study was the occurrence of a blue-green algal (cyanobacterial) bloom in 2010, and the need to understand the factors that influence this and other water quality issues. A hydrodynamic model with high spatial and temporal resolution was applied and calibrated to field data collected in a detailed sampling program in 2013. The model, along with field data collected over several additional years, was then used to develop a comprehensive understanding of the hydrodynamic impacts on physical conditions in the bay. A primary result of this process is a determination of the importance of flow exchange between Lake Ontario and Sodus Bay, particularly during lake upwelling periods that cause colder water to enter the bay as an underflow. Hydrodynamic features identified to have played a potential role in the bloom formation of 2010 include a lake upwelling event, strong stratification, and relatively warm conditions throughout much of the summer. Lack of continuous monitoring in 2010 precludes a specific comparison of the model and data when the bloom formed, but the model clearly shows conditions that would have led to a bloom, assuming other preconditions were in place. The customized hydrodynamic model provides opportunities for future ecological modeling, hypothesis development, and what-if scenario testing. This study reinforces the importance of hydrodynamic interactions between lakes and embayments, and their impacts on water quality.