基于Lasso函数的分层混合模糊-神经网络及其在遥感影像分类中的应用

本文提出一种新的分层混合模糊-神经网络(HHFNN)算法.在模糊系统中使用Takagi-Sugeno模型和三角波隶属函数.同时,为降低离散输入变量中可能存在的强交瓦作用,采用了系数收缩机制中的Lasso函数.最后,以福建的漳平洛阳—安溪潘田地区LANDSAT ETM+遥感影像数据地物分类为例,应用本文的改进算法与其他神...     

Comparison of static-feedforward and dynamic-feedback neural networks for rainfall–runoff modeling

A systematic comparison of two basic types of neural network, static and dynamic, is presented in this study. Two back-propagation (BP) learning optimization algorithms, the standard BP and conjugate gradient (CG) method, are used for the static network, and the real-time recurrent learning (RTRL) algorithm is used for the dynamic-feedback network. Twenty-three storm-events, about 1632 rainfall and runoff data sets, of the Lan-Yang River in Taiwan are used to demonstrate the efficiency and practicability of the neural networks for one hour ahead streamflow forecasting. In a comparison of searching algorithms for a static network, the results show that the CG method is superior to the standard BP method in terms of the efficiency and effectiveness of the constructed network's performance. For a comparison of the static neural network using the CG algorithm with the dynamic neural network using RTRL, the results show that (1) the static-feedforward neural network could produce satisfactory results only when there is a sufficient and adequate training data set, (2) the dynamic neural network generally could produce better and more stable flow forecasting than the static network, and (3) the RTRL algorithm helps to continually update the dynamic network for learning—this feature is especially important for the extraordinary time-varying characteristics of rainfall–runoff processes.     

致密砂岩气藏动态分类委员会机器测井流体识别方法

致密砂岩流体识别难度大,智能算法能够较好地建立其流体识别模型.相较于单一智能算法,分类委员会机器通过联合多个专家(智能算法)有助于提升智能模型整体性能.而针对分类委员会机器中单个专家性能难以提升的问题,添加门网络构建动态分类委员会机器是一种更有效的模块化学习方式.本研究首先采用门网络将输入数据划分为多个子数据集,然后联合决策树、概率神经网络、贝叶斯分类、BP神经网络、最近邻算法分别训练子数据集得到多个子模型,最后利用组合器最优化子模型组合得到最佳的流体识别模型.针对塔里木盆地库车坳陷大北、克深、博孜地区致密砂岩地层测井数据和测试数据,采用平均影响值法优选敏感测井系列作为输入,构建了动态的测井流体识别模型,其训练、验证准确率分别为96.29%和91.39%.利用此模型以BZ9井为例进行流体类型判别,预测结果与测试结果一致.该方法将无监督与有监督学习相结合,引入门网络提高了数据集利用效率,避免了数据集分布不均衡对模型构建的影响;采用投票机制集成多种专家,建立了子模型与专家的适应关系,流体识别模型预测精度和泛化能力大大提高.     

Optimizing statistical classification accuracy of satellite remotely sensed imagery for supporting fast flood hydrological analysis

The aim of this study is to improve classification results of multispectral satellite imagery for supporting flood risk assessment analysis in a catchment area in Cyprus. For this purpose, precipitation and ground spectroradiometric data have been collected and analyzed with innovative statistical analysis methods. Samples of regolith and construction material were in situ collected and examined in the spectroscopy laboratory for their spectral response under consecutive different conditions of humidity. Moreover, reflectance values were extracted from the same targets using Landsat TM/ETM+ images, for drought and humid time periods, using archived meteorological data. The comparison of the results showed that spectral responses for all the specimens were less correlated in cases of substantial humidity, both in laboratory and satellite images. These results were validated with the application of different classification algorithms (ISODATA, maximum likelihood, object based, maximum entropy) to satellite images acquired during time period when precipitation phenomena had been recorded.     

MODIS Terra Collection 6 fractional snow cover validation in mountainous terrain during spring snowmelt using Landsat TM and ETM+

Daily swath MODIS Terra Collection 6 fractional snow cover (MOD10_L2) estimates were validated with two‐day Landsat TM/ETM + snow‐covered area estimates across central Idaho and southwestern Montana, USA. Snow cover maps during spring snowmelt for 2000, 2001, 2002, 2003, 2005, 2007, and 2009 were compared between MODIS Terra and Landsat TM/ETM + using least‐squared regression. Strong spatial and temporal map agreement was found between MODIS Terra fractional snow cover and Landsat TM/ETM + snow‐covered area, although map disagreement was observed for two validation dates. High‐altitude cirrus cloud contamination during low snow conditions as well as late season transient snowfall resulted in map disagreement. MODIS Terra's spatial resolution limits retrieval of thin‐patchy snow cover, especially during partially cloudy conditions. Landsat's image acquisition frequency can introduce difficulty when discriminating between transient and resident mountain snow cover. Furthermore, transient snowfall later in the snowmelt season, which is a stochastic accumulation event that does not usually persist beyond the daily timescale, will skew decadal snow‐covered area variability if bi‐monthly climate data record development is the objective. As a quality control step, ground‐based daily snow telemetry snow‐water‐equivalent measurements can be used to verify transient snowfall events. Users of daily MODIS Terra fractional snow products should be aware that local solar illumination and sensor viewing geometry might influence fractional snow cover estimation in mountainous terrain. Cross‐sensor interoperability has been confirmed between MODIS Terra and Landsat TM/ETM + when mapping snow from the visible/infrared spectrum. This relationship is strong and supports operational multi‐sensor snow cover mapping, specifically climate data record development to expand cryosphere, climate, and hydrological science applications. Copyright © 2013 John Wiley & Sons, Ltd.     

Pixel‐oriented land use classification in energy balance modelling

Mass and energy transfer between soil, vegetation and atmosphere is the process that allows to maintain an adequate energy and water balance in the earth–atmosphere system. However, the evaluation of the energy balance components, such as the net radiation and the sensible and latent heat fluxes, is characterized by significant uncertainties related to both the dynamic nature of heat transfer processes and surfaces heterogeneity. Therefore, a detailed land use classification and an accurate evaluation of vegetation spatial distribution are required for an accurate estimation of these variables. For this purpose, in the present article, a pixel‐oriented supervised classification was applied to obtain land use maps of the Basilicata region in Southern Italy by processing three Landsat TM and ETM+ satellite images. An accuracy analysis based on the overall accuracy index and the agreement Khat of Cohen coefficient showed a good performance of the applied classification methodology and a good quality of the obtained maps. Subsequently, these maps were used in the application of a simplified two‐source energy balance model for estimating the actual evapotranspiration at a regional scale. The comparison between the simulations made by applying the simplified two‐source energy balance model and the measurements of evapotranspiration at a lysimetric station located in the study area showed the applicability and the validity of the proposed methodology. Copyright © 2012 John Wiley & Sons, Ltd.     

Estimating the distribution of snow water equivalent using remotely sensed snow cover data and a spatially distributed snowmelt model: A multi-resolution,multi-sensor comparison

Time series of fractional snow covered area (SCA) estimates from Landsat Enhanced Thematic Mapper (ETM+), Moderate Resolution Imaging Spectroradiometer (MODIS), and Advanced Very High Resolution Radiometer (AVHRR) data were combined with a spatially distributed snowmelt model to reconstruct snow water equivalent (SWE) in the Rio Grande headwaters (3419 km²). In this reconstruction approach, modeled snowmelt over each pixel is integrated during the period of satellite-observed snow cover to estimate SWE. Due to underestimates in snow cover detection, maximum basin-wide mean SWE using MODIS and AVHRR were, respectively, 45% and 68% lower than SWE estimates obtained using ETM+ data. The mean absolute error (MAE) of SWE estimated at 100-m resolution using ETM+ data was 23% relative to observed SWE from intensive field campaigns. Model performance deteriorated when MODIS (MAE = 50%) and AVHRR (MAE = 89%) SCA data were used. Relative to differences in the SCA products, model output was less sensitive to spatial resolution (MAE = 39% and 73% for ETM+ and MODIS simulations run at 1 km resolution, respectively), indicating that SWE reconstructions at the scale of MODIS acquisitions may be tractable provided the SCA product is improved. When considering tradeoffs between spatial and temporal resolution of different sensors, our results indicate that higher spatial resolution products such as ETM+ remain more accurate despite the lower frequency of acquisition. This motivates continued efforts to improve MODIS snow cover products.     

基于GA-BP理论的储层视裂缝密度地震非线性反演方法

基于GA-BP理论,将自适应遗传算法与人工神经网络技术(BP算法)有机地相结合,形成了一种储层裂缝自适应遗传-神经网络反演方法.这种新的方法是由编码、适应度函数、遗传操作及混合智能学习等组成,即在成像测井裂缝密度数据约束下,通过对目标问题进行编码(称染色体),然后对染色体进行选择、交叉和变异等遗传操作,使染色体不断进化,从而快速获得全局最优解.在反演执行过程中,利用地震数据和成像测井裂缝密度数据之间的非线性映射关系建立训练样本,将GA算法与BP算法有机地结合,优化三层前向网络参数;或将GA与ANFIS相结合,优化ANFIS网络参数.并采用GA算法与TS算法(Tabu Search)相结合的自适应混合学习算法,该学习算法自始至终将GA和BP两种算法按一定的概率比例进行,其概率自适应变化,以达到混合算法的均衡.这种混合算法提高了网络的收敛速度和精度.我们分别利用两个研究地区的6井和1井成像测井裂缝密度数据与地震数据之间的非线性映射关系建立训练样本,对过这两口井的测线的地震数据进行反演,获得了视裂缝密度剖面,视裂缝密度剖面上裂缝分布特征符合沉积相分布特征和岩石力学性质的变化特征.这种视裂缝密度剖面含有储层裂缝的定量信息,其误差可为油气勘探开发实际要求所允许.因此,这种新的方法优于只能作裂缝定性分析的常规裂缝地震预测方法,具有广阔的应用前景.     

A wavelet-neural network hybrid modelling approach for estimating and predicting river monthly flows

Abstract

A wavelet-neural network (WNN) hybrid modelling approach for monthly river flow estimation and prediction is developed. This approach integrates discrete wavelet multi-resolution decomposition and a back-propagation (BP) feed-forward multilayer perceptron (FFML) artificial neural network (ANN). The Levenberg-Marquardt (LM) algorithm and the Bayesian regularization (BR) algorithm were employed to perform the network modelling. Monthly flow data from three gauges in the Weihe River in China were used for network training and testing for 48-month-ahead prediction. The comparison of results of the WNN hybrid model with those of the single ANN model show that the former is able to significantly increase the prediction accuracy.

Editor D. Koutsoyiannis; Associate editor H. Aksoy

Citation Wei, S., Yang, H., Song, J.X., Abbaspour, K., and Xu, Z.X., 2013. A wavelet-neural network hybrid modelling approach for estimating and predicting river monthly flows. Hydrological Sciences Journal, 58 (2), 374–389.     

结合温度因子估算太湖叶绿素a含量的神经网络模型

神经网络方法估算复杂水体水质参数的优越性已经得到证实.基于太湖水体实测叶绿素a浓度,利用MODIS 250m影像和反演得到的水温数据建立了估算太湖水体叶绿素a含量的两个单隐层BP神经网络模型:NN1模型不含温度因子、NN2模型包含温度因子,采用Levenberg-Marquardt算法训练网络,利用初期终止方法提岛网络泛化能力,均取得了较高估算精度,其中包含温度因了的反演模型精度稍有提高,但不显著.     

基于遗传算法优化神经网络权值的大坝结构损伤识别

针对传统 BP 神经网络存在着容易陷入局部极小点、训练时间太长等缺点,本文采用基于浮点编码的遗传算法,对 BP 神经网络的初值空间进行了遗传优化。用基于浮点编码的遗传算法来优化 BP 神经网络的权值,可得到最佳初始权值矩阵,并按误差前向反馈算法,沿负梯度搜索进行网络学习。文中以混凝土重力坝结构作为算例,用结构的模态频率变化作为网络的输入向量,结构的损伤位置作为输出向量,对网络进行了训练。仿真结果表明：遗传 BP 神经网络的收敛和诊断能力优于传统 BP 神经网络,可有效地运用到大坝结构的健康诊断与损伤识别中。     

Comparison of the response of the Landsat 7 Enhanced Thematic Mapper Plus and the Earth Observing-1 Advanced Land Imager over active lava flows

Using near simultaneously acquired Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Earth Observing-1 Advanced Land Imager (ALI) data we assess the relative radiant responses over active lava flows from the Mt. Etna July/August 2001 flank eruption. By assessing the extent of saturation between the two instruments and using the dual-band method of extracting sub pixel thermal information, we show that the ALI represents an improvement over the ETM+ in the present ability to assess temperatures of hot active lava flows for a number of reasons. (1) The extra spectral channels provided by ALI compliment the current SWIR channels on ETM+ by providing a greater number of paired channel combinations for input into the dual-band method. Thus, dual-band temperature solutions can be determined for a greater range of lava flow types than previously possible using the two paired channel combinations available with the ETM+. (2) The ALI instrument is less susceptible than ETM+ to saturation within the SWIR, especially when using channels 5, 5p and 4p at wavelengths of 1.65, 1.25 and 0.87 μm respectively. (3) The greater radiometric sensitivity of the ALI 12 bit electronics coupled with a significantly higher signal to noise ratio aid in obtaining successful dual-band solutions.     

新型Landsat8卫星影像的反射率和地表温度反演

Landsat 8卫星自2013年2月发射以来,其影像的定标参数经过了不断调整和完善,针对Landsat 8开发的各种算法也相继问世.本文采用最新的参数、算法和引入COST算法建立的大气校正模型,对Landsat 8多光谱和热红外波段进行了处理,反演出它们的反射率和地表温度,并与同日的Landsat 7数据和实测地表温度数据进行了对比.结果表明,现有Landsat 8多光谱数据的定标参数和大气顶部反射率反演算法已有很高的精度,本文引入COST算法建立的Landsat 8大气校正模型也与Landsat 7的COST模型所获得的结果几乎相同,相关系数可高达0.99.但是现有针对Landsat 8提出的地表温度反演算法仍不理想,已提出的劈窗算法误差都较大.鉴于TIRS 11热红外波段的定标参数仍不理想,因此在现阶段建议采用单通道算法单独反演TIRS 10波段来求算地表温度,但要注意根据大气水汽含量的情况选用正确的大气参数计算公式.     

Soil erosion assessment using geomorphological remote sensing techniques: an example from southern Italy

The aim of this study is to assess of the distribution and map the geomorphological effects of soil erosion at the basin scale identifying newly‐formed erosional landsurfaces (NeFELs), by means of an integration of Landsat ETM 7+ remotely sensed data and field‐surveyed geomorphological data. The study was performed on a 228·6 km²‐wide area, located in southern Italy. The study area was first characterized from a lithological, pedological, land‐use and morpho‐topographic point of view and thematic maps were created. Then, the georeferenced Landsat ETM 7+ satellite imagery was processed using the RSI ENVI 4.0 software. The processing consisted of contrast stretching, principal component analysis (PCA), decorrelation stretching and RGB false colour compositing. A field survey was conducted to characterize the features detected on the imagery. Particular attention was given to the NeFELs, which were located using a global positioning system (GPS). We then delimited the Regions of Interest (ROI) on the Landsat ETM 7+ imagery, i.e. polygons representing the ‘ground‐truth’, discriminating the NeFELs from the other features occurring in the imagery. A simple statistical analysis was conducted on the digital number (DN) values of the pixels enclosed in the ROI of the NeFELs, with the aim to determine the spectral response pattern of such landsurfaces. The NeFELs were then classified in the entire image using a maximum likelihood classification algorithm. The results of the classification process were checked in the field. Finally, a spatial analysis was performed by converting the detected landsurfaces into vectorial format and importing them into the ESRI ArcViewGIS 9.0 software. Application of these procedures, together with the results of the field survey, highlighted that some ‘objects’ in the classified imagery, even if displaying the same spectral response of NeFELs, were not landsurfaces subject to intense soil erosion, thus confirming the strategic importance of the field‐checking for the automatically produced data. During the production of the map of the NeFELs, which is the final result of the study, these ‘objects’ were eliminated by means of simple, geomorphologically‐coherent intersection procedures in a geographic information system (GIS) environment. The overall surface of the NeFELs had an area of 22·9 km², which was 10% of the total. The spatial analysis showed that the highest frequency of the NeFELs occurred on both south‐facing and southwest‐facing slopes, cut on clayey‐marly deposits, on which fine‐textured and carbonate‐rich Inceptisols were present and displaying slope angle values ranging from 12° to 20°. The comparison of two satellite imageries of different periods highlighted that the NeFELs were most clearly evident immediately after summer tillage operations and not so evident before them, suggesting that these practices could have played an important role in inducing the erosional processes. Copyright © 2009 John Wiley & Sons, Ltd.     

藻类生长预测模型的比较研究——以洪湖水体为例

根据洪湖2014—2019年水质及藻类监测数据,运用综合营养状态指数法评价了丰、平、枯3个时期的营养状态.在此基础上运用逐步回归分析法确定影响藻类生长的显著因子,并根据不同水量不同营养状态细分9种情形对藻类生长做回归预测分析,同时运用BP神经网络模型对回归预测的结果进行比较验证.结果表明:洪湖丰、平水期以蓝藻门为主,枯水期以硅藻门为主;湖泊的营养状态处于中度富营养与轻度富营养之间.分析各时期藻种生物量与影响因子的相关性,发现丰水期控制因子有水温、COD_Mn和透明度;平水期和枯水期控制因子有水温、总氮、总磷.以2014—2018年数据逐步回归分析得出枯水期+中营养和枯水期+轻度富营养决定系数较低,其余7种时期决定系数均在0.5以上,说明逐步回归并不适用于所有时期.使用2014—2018年的数据进行神经网络训练和验证,2019年的数据进行预测,比较BP神经网络与逐步回归的均方根误差发现全年预测时BP神经网络效果更好;枯水期+中营养和枯水期+轻度富营养逐步回归效果较好,逐步回归的均方根误差仅为1600～4000;丰水期和平水期2种方法预测效果相当.合理地选择预测模型能为湖泊水华做出预警,控制显著变量可以达到防治水华污染的效果.     

基于MATLAB的BP预测模型在地震前兆预测中的应用研究

依据神经网络理论,基于MATLAB的神经网络工具箱建立了一个BP神经网络预测模型,并通过对陕西省地震前兆数据的预测分析来检验模型的效果,实验结果证明该模型用于地震预测的可行性,操作简单灵活,直接面向用户。具有很好的应用价值。     

Dynamic neuro-fuzzy local modeling system with a nonlinear feature extraction for the online adaptive warning system of river temperature affected by waste cooling water discharge

This paper proposes a dynamic modeling methodology based on a dynamic neuro-fuzzy local modeling system (DNFLMS) with a nonlinear feature extraction technique for an online dynamic modeling task. Prior to model building, a nonlinear feature extraction technique called the Gamma test (GT) is proposed to compute the lowest mean squared error (MSE) that can be achieved and the quantity of data required to obtain a reliable model. Two different DNFLMS modes are developed: (1) an online one-pass clustering and the extended Kalman filtering algorithm (mode 1); and (2) hybrid learning algorithm (mode 2) of extended Kalman filtering algorithm with a back-propagation algorithm trained to the estimated MSE and number of data points determined by a nonlinear feature extraction technique. The proposed modeling methodology is applied to develop an online dynamic prediction system of river temperature to waste cooling water discharge at 1?km downstream from a thermal power station from real-time to time ahead (2?h) sequentially at the new arrival of each item of river, hydrological, meteorological, power station operational data. It is demonstrated that the DNFLMS modes 1 and 2 shows a better prediction performance and less computation time required, compared to a well-known adaptive neural-fuzzy inference system (ANFIS) and a multi-layer perceptron (MLP) trained with the back propagation (BP) learning algorithm, due to local generalization approach and one-pass learning algorithm implemented in the DNFLMS. It is shown that the DNFLMS mode 1 is that it can be used for an online modeling task without a large amount of training set required by the off-line learning algorithm of MLP-BP and ANFIS. The integration of the DNFLMS mode 2 with a nonlinear feature extraction technique shows that it can improve model generalization capability and reduce model development time by eliminating iterative procedures of model construction using a stopping criterion in training and the quantity of required available data in training given by the GT.     

Evolution of lava flow-fields at Mount Etna, 27–28 October 1999, observed by Landsat 7 ETM+

Landsat 7 Enhanced Thematic Mapper + (ETM+) data are presented which document the thermal characteristics of a series of lava flows emplaced at Mount Etna volcano, Sicily, during 27-28 October 1999. By examining the composition of the short-wave infrared (SWIR) signal emitted from the flow surface, we identified distinctive flow units. The first unit appears to comprise recently active lava flows with relatively cool crusts which, by virtue of the integrity of this crust as determined from the ETM+ data, we infer are stationary or barely moving. The second unit is characterized by much higher levels of SWIR radiance, consistent with a channel-fed active flow unit. Analysis of the SWIR data confirm that this is fed by a lava channel, the properties of which are consistent with vigorously active channels observed on Kilauea, Hawaii. Model predictions of the maximum length that such flows could attain compare favorably with the actual flow lengths observed in the ETM+ data, indicating that the cooler flows had indeed stopped advancing, and may have attained a cooling-limited, rather than volume-limited, maximum length. Our observations and modeling provide a physical corroboration for the supposition made by Wadge (1978) in his analysis of the shape of lava flow fields on Mount Etna, which in the cooling-limited case principal flows are active one after the other and not at the same time.     

基于改进EWT和LogitBoost集成分类器的地震事件分类识别算法

快速、准确地识别天然地震和人工爆破事件是地震台网监测的重要工作之一,也是提高地震观测记录质量、开展地震研究工作的重要基础。针对反向传播神经网络、支持向量机等主流分类识别方法在地震事件分类识别应用上的不足,提出一种基于改进EWT和LogitBoost集成分类器的地震事件分类识别算法。首先,基于S谱能量曲线对传统经验小波变换进行改进,将信号自适应分解为按频率和能量分布的本征模函数;其次,提取P波与S波最大振幅比,前4个本征模函数的香农熵、对数能量熵,以及去噪后重构信号主频等特征;最后,采用基于集成学习LogitBoost的决策树集成分类器进行分类。实验结果表明,所提算法具有较高的鲁棒性,能有效解决样本不足的问题,识别准确率达93.1%以上,比集成学习AdaBoost、反向传播神经网络和支持向量机等方法提高了1%以上,且分类识别效果好。     

Determination of regional land surface heat fluxes over a heterogeneous landscape of the Jiddah area of Saudi Arabia by using Landsat‐7 ETM data

In this study, a parameterization method based on Landsat‐7 Enhanced Thematic Mapper (ETM) data and field observations is presented and tested for deriving the regional land surface variables, vegetation variables and land surface heat fluxes over a heterogeneous landscape. As a case study, the method and two Landsat‐7 ETM images are applied to the Jiddah area of Saudi Arabia. The regional distribution maps of surface reflectance, normalized difference vegetation index, modified soil adjusted vegetation index (MSAVI), vegetation coverage, leaf area index, surface temperature, net radiation flux, soil heat flux, sensible heat flux and latent heat flux have been determined over the Jiddah area. The derived results have been validated by using the ‘ground truth’. The results show that the more reasonable regional distributions of land surface variables (surface reflectance, surface temperature), vegetation variables (MSAVI and vegetation coverage), net radiation, soil heat flux and sensible heat flux can be obtained by using the method proposed in this study. Further improvement of the method is also discussed. Copyright © 2006 John Wiley & Sons, Ltd.