Artificial neural network models for forecasting monthly precipitation in Jordan

Forecasting precipitation in arid and semi-arid regions, in Jordan in the Middle East for example, has particular importance since precipitation is the unique source of water in such regions. In this study, 1-month ahead precipitation forecasts are made using artificial neural network (ANN) models. Feed forward back propagation (FFBP), radial basis function (RBF) and generalized regression type ANNs are used and compared with a simple multiple linear regression (MLR) model. The models are tested on monthly total precipitation recorded at three meteorological stations (Baqura, Amman and Safawi) from different climatological regions in Jordan. For the three stations, it is found that the best calibrated model is FFBP with respect to all performance criteria used in the study, including determination coefficient, mean square error, mean absolute error, the slope and the intercept in the best-fit linear line of the scatter diagram. In the validation stage, FFBP is again the best model in Baqura and Amman. However, in Safawi, the driest station, not only FFBP but also RBF and MLR perform equally well depending on the performance criterion under consideration.     

A combined rotated general regression neural network method for river flow forecasting

ABSTRACT

This study focused on the performance of the rotated general regression neural network (RGRNN), as an enhancement of the general regression neural network (GRNN), in monthly-mean river flow forecasting. The study of forecasting of monthly mean river flows in Heihe River, China, was divided into two steps: first, the performance of the RGRNN model was compared with the GRNN model, the feed-forward error back-propagation (FFBP) model and the soil moisture accounting and routing (SMAR) model in their initial model forms; then, by incorporating the corresponding outputs of the SMAR model as an extra input, the combined RGRNN model was compared with the combined FFBP and combined GRNN models. In terms of model efficiency index, R², and normalized root mean squared error, NRMSE, the performances of all three combined models were generally better than those of the four initial models, and the RGRNN model performed better than the GRNN model in both steps, while the FFBP and the SMAR were consistently the worst two models. The results indicate that the combined RGRNN model could be a useful river flow forecasting tool for the chosen arid and semi-arid region in China.

Editor D. Koutsoyiannis; Associate editor not assigned     

A simple formula for shape and depth determination from residual gravity anomalies

This paper presents a simple method for shape and depth determination of a buried structure from residual gravity anomalies along profile. The method utilizes the anomaly values of the origin and characteristic points of the profile to construct a relationship between the shape factor and depth of the causative source. For fixed points, the depth is determined for each shape factor. The computed depths are then plotted against the shape factor representing a continuous monotonically increasing curve. The solution for the shape and depth of the buried structure is then read at the common intersection point of the depth curves. This method is applied to synthetic data with and without random errors. Finally, the validity of the method is tested on two field examples from the USA.     

Short note: Source geometry identification by simultaneous use of structural index and shape factor

A cross-plot of the shape factors and the structural indices, determined from gravity anomalies over various idealized sources, namely horizontal/vertical lines and vertical ribbons with various strike lengths and depth extents, forms a closed loop. Different segments of this loop, termed the source geometry identification loop (SGIL), correspond to different source geometries. Combined use of the structural index and the shape factor determined from an isolated gravity anomaly reduces the ambiguity in characterizing the source geometry. A simulated example and three field examples, namely a Cuban chromite anomaly, an Indian example over manganese ore and a sulphide ore from Quebec, have been analysed by the proposed method in order to identify their respective source geometries.     

Self-potential data interpretation using standard deviations of depths computed from moving-average residual anomalies

We have developed a least‐squares minimization approach to determine simultaneously the shape (shape factor) and the depth of a buried structure from self‐potential (SP) data. The method is based on computing the standard deviation of the depths determined from all moving‐average residual anomalies obtained from SP data, using filters of successive window lengths for each shape factor. The standard deviation may generally be considered a criterion for determining the correct depth and shape factor of the buried structure. When the correct shape factor is used, the standard deviation of the depths is less than the standard deviations computed using incorrect shape factors. This method is applied to synthetic data with and without random errors, complicated regionals and interference from neighbouring sources, and is tested on a known field example from Turkey. In all cases, the shape and depth solutions obtained are in a good agreement with the actual values.     

人工神经网络法在烃源岩测井评价中的应用

运用有机地球化学方法分析岩芯、岩屑样品的有机炭含量存在着昂贵、时且不准确等问题。利用测井方法的优点是经济、准确。在测井评价中使用人工神经网络法具有极大的优越性和适用性。本文结合Kohonen和BP网络方法，完成了塔里木台盆区19口井的寒武、奥陶系烃源岩层段的识别与评价，并通过测井资料处理成果和岩芯有机地化资料、地质录井情况的相互检验，证实，其本上能够满足评价烃源岩的需要，从而为利用测井资料进行烃源岩评价做出了新的尝试。     

Inversion of elongated magnetic anomalies using magnitude transforms

The calculable magnitudes of the anomalous magnetic field from simple 2D sources and their gradients and Laplacians appear as ratios that can be synthesized in functional forms, corresponding to the different source shapes. Field components and first‐order derivatives are involved in the inversion procedures presented. The structural index and source depth are estimated independently of each other. The applied functions allow magnetic profiles and magnetic maps to be shape‐ and depth‐converted with immediate imaging of the inversion results. The contours of these functions outline elongated loops around the 2.5D anomaly axis on magnetic maps. The width of the loops reflects the depth and structural index N of the source in the scale units of the inverted map. Model and field tests illustrate the effectiveness of this approach for fast automatic inversion of large sets of magnetic data for depth, shape, length and location of simple sources.     

地震油气储层的小样本卷积神经网络学习与预测

地震储层预测是油气勘探的重要组成部分,但完成该项工作往往需要经历多个环节,而多工序或长周期的研究分析降低了勘探效率.基于油气藏分布规律及其在地震响应上所具有的特点,本文引入卷积神经网络深度学习方法,用于智能提取、分类并识别地震油气特征.卷积神经网络所具有的强适用性、强泛化能力,使之可以在小样本条件下,对未解释地震数据体进行全局优化提取特征并加以分类,即利用有限的已知含油气井段信息构建卷积核,以地震数据为驱动,借助卷积神经网络提取、识别蕴藏其中的地震油气特征.将本方案应用于模型数据及实际数据的验算,取得了预期效果.通过与实际钻井信息及基于多波地震数据机器学习所预测结果对比,本方案利用实际数据所演算结果与实际情况有较高的吻合度.表明本方案具有一定的可行性,为缩短相关环节的周期提供了一种新的途径.     

Depth and shape estimates from simultaneous inversion of magnetic fields and their gradient components using differential similarity transforms

The differential similarity transform of a magnetic anomaly is a linear combination of its intensity and gradient components. This transform is sensitive to the distance between a chosen central point of similarity and the source and depends on the degree of homogeneity of the field. Taking advantage of this property, a new field inversion method resulting in the evaluation of source position and shape type is proposed and implemented. The field gradient components are measured directly in magnetic gradiometry, or they can be calculated from the measured field data. Regional and local linear backgrounds are accounted for by the method. The method can be applied on either regularly or irregularly-spaced data sets, on even or uneven surfaces of observation. The solving of the systems of equations is not necessary. A semi-automated inversion for both location and shape of the sources is implemented. Model and field tests illustrate the effectiveness of the proposed inversion technique for depth and shape estimates.     

NEAR-SOURCE PULSE PROPAGATION: APPLICATION TO Q-DETERMINATION1

Among the approaches generally used to measure attenuation from field data, the study of the first pulse broadening appears to be one of the more promising methods to estimate the quality factor Q for different geological formations including soils. Using a numerical scheme, we studied the evolution of the pulse shape in the neighbourhood of the source in order to establish the limits of the method. It was found that the pulse width variations depend strongly upon source depth. At short distances from the source, the pulse shape is controlled mainly by the near-field terms and/or the onset of surface waves. The investigations proved that the pulse-broadening method is reliable for distances greater than about 1.2 wavelengths. From numerical experiments, the maximum error in Q-determination is found to be 10% in the half-space case.     

考虑实际投影成像平面位置的扇束滤波反投影(FFBP)算法

扇束滤波反投影(Fan-beam Filtered Back Projection- FFBP)算法理论公式中,投影成像平面位于旋转中心;而实际CT扫描系统中,投影成像平面与旋转中心都存在一定距离.考虑到实际投影成像平面位置,本文推导了它的FFBP算法(即IFFBP-Improved Fan-beam Filtered Back Projection算法),比较了此两种情况下(即忽略成像平面与旋转中心距离)FFBP、IFFBP算法重建质量.计算机模拟和实验结果证明了IFFBP算法的正确性.     

Quantitative Interpretation of Self-Potential Anomalies of Some Simple Geometric Bodies

We have developed a new numerical method to determine the shape (shape factor), depth, polarization angle, and electric dipole moment of a buried structure from residual self-potential (SP) anomalies. The method is based on defining the anomaly value at the origin and four characteristic points and their corresponding distances on the anomaly profile. The problem of shape determination from residual SP anomaly has been transformed into the problem of finding a solution to a nonlinear equation of the form q = f (q). Knowing the shape, the depth, polarization angle and the electric dipole moment are determined individually using three linear equations. Formulas have been derived for spheres and cylinders. By using all possible combinations of the four characteristic points and their corresponding distances, a procedure is developed for automated determination of the best-fit-model parameters of the buried structure from SP anomalies. The method was applied to synthetic data with 5% random errors and tested on a field example from Colorado. In both cases, the model parameters obtained by the present method, particularly the shape and depth of the buried structures are found in good agreement with the actual ones. The present method has the capability of avoiding highly noisy data points and enforcing the incorporation of points of the least random errors to enhance the interpretation results.     

基于神经网络的时间域直升机电磁数据电导率深度成像

采用G-S变换以及高斯数值积分法,形成了时间域直升机的航空电磁响应正演样本集,分析了飞机测量过程中吊舱高度变化对电磁响应的影响,并将吊舱高度的变化等效成电导率为零的假层厚度的变化,以去除高度计等的影响.以假层半空间模型为基础,研究了基于人工神经网络的电导率深度成像算法,通过分析两个三层模型的电导率深度成像结果得出,神经网络方法计算时间域航空电磁探测的视电导率精度较高,特别是对高阻层的视电导率计算.     

A new least-squares minimization approach to depth and shape determination from magnetic data

We have developed a least‐squares minimization approach to depth determination using numerical second horizontal derivative anomalies obtained from magnetic data with filters of successive window lengths (graticule spacings). The problem of depth determination from second‐derivative magnetic anomalies has been transformed into finding a solution to a non‐linear equation of the form, f(z) = 0. Formulae have been derived for a sphere, a horizontal cylinder, a dike and a geological contact. Procedures are also formulated to estimate the magnetic angle and the amplitude coefficient. We have also developed a simple method to define simultaneously the shape (shape factor) and the depth of a buried structure from magnetic data. The method is based on computing the variance of depths determined from all second‐derivative anomaly profiles using the above method. The variance is considered a criterion for determining the correct shape and depth of the buried structure. When the correct shape factor is used, the variance of depths is less than the variances computed using incorrect shape factors. The method is applied to synthetic data with and without random errors, complicated regionals, and interference from neighbouring magnetic rocks. Finally, the method is tested on a field example from India. In all the cases examined, the depth and the shape parameters are found to be in good agreement with the actual parameters.     

Classified real-time flood forecasting by coupling fuzzy clustering and neural network

This paper presented a new classified real-time flood forecasting framework by integrating a fuzzy clustering model and neural network with a conceptual hydrological model. A fuzzy clustering model was used to classify historical floods in terms of flood peak and runoff depth, and the conceptual hydrological model was calibrated for each class of floods. A back-propagation (BP) neural network was trained by using real-time rainfall data and outputs from the fuzzy clustering model. BP neural network provided a rapid on-line classification for real-time flood events. Based on the on-line classification, an appropriate parameter set of hydrological model was automatically chosen to produce real-time flood forecasting. Different parameter sets was continuously used in the flood forecasting process because of the changes of real-time rainfall data and on-line classification results. The proposed methodology was applied to a large catchment in Liaoning province, China. Results show that the classified framework provided a more accurate prediction than the traditional non-classified method. Furthermore, the effects of different index weights in fuzzy clustering were also discussed.     

Neural estimation of strong ground motion duration

This paper presents and discusses the use of neural networks to determine strong ground motion duration. Accelerometric data recorded in the Mexican cities of Puebla and Oaxaca are used to develop a neural model that predicts this duration in terms of the magnitude, epicenter distance, focal depth, soil characterization and azimuth. According to the above the neural model considers the effect of the seismogenic zone and the contribution of soil type to the duration of strong ground motion. The final scheme permits a direct estimation of the duration since it requires easy-to-obtain variables and does not have restrictive hypothesis. The results presented in this paper indicate that the soft computing alternative, via the neural model, is a reliable recording-based approach to explore and to quantify the effect of seismic and site conditions on duration estimation. An essential and significant aspect of this new model is that, while being extremely simple, it also provides estimates of strong ground motions duration with remarkable accuracy. Additional but important side benefits arising from the model’s simplicity are the natural separation of source, path, and site effects and the accompanying computational efficiency.     

Integrating facies-based Bayesian inversion and supervised machine learning for petro-facies characterization in the Snadd Formation of the Goliat Field,south-western Barents Sea

Seismic petro-facies characterization in low net-to-gross reservoirs with poor reservoir properties such as the Snadd Formation in the Goliat field requires a multidisciplinary approach. This is especially important when the elastic properties of the desired petro-facies significantly overlap. Pore fluid corrected endmember sand and shale depth trends have been used to generate stochastic forward models for different lithology and fluid combinations in order to assess the degree of separation of different petro-facies. Subsequently, a spectral decomposition and blending of selected frequency volumes reveal some seismic fluvial geomorphological features. We then jointly inverted for impedance and facies within a Bayesian framework using facies-dependent rock physics depth trends as input. The results from the inversion are then integrated into a supervised machine learning neural network for effective porosity discrimination. Probability density functions derived from stochastic forward modelling of endmember depth trends show a decreasing seismic fluid discrimination with depth. Spectral decomposition and blending of selected frequencies reveal a dominant NNE trend compared to the regional SE–NW pro-gradational trend, and a local E–W trend potentially related to fault activity at branches of the Troms-Finnmark Fault Complex. The facies-based inversion captures the main reservoir facies within the limits of the seismic bandwidth. Meanwhile the effective porosity predictions from the multilayer feed forward neural network are consistent with the inverted facies model, and can be used to qualitatively highlight the cleanest regions within the inverted facies model. A combination of facies-based inversion and neural network improves the seismic reservoir delineation of the Snadd Formation in the Goliat Field.     

Effective prediction of scour downstream of ski-jump buckets using artificial neural networks

The main objective of this research was to analyze and quantify the uncertainty of artificial neural network in prediction of scour downstream ski-jump buckets. Hence, at first, three artificial neural network models were developed to predict depth, length, and width of scour hole. Then, Monte-Carlo simulation was applied in the estimates of artificial neural network modeling procedure. The uncertainties were quantified by means of two criteria: 95 percent prediction uncertainty and d-factor. The results of the artificial neural network models showed superior performance of it in comparison with some empirical formulas because of higher correlation coefficient (R ² > 0.95) and lower error (RMSE < 1.63). The obtained result from uncertainty analysis of the models revealed the satisfactory performance of them. In this procedure it was clarified that the artificial neural network model for length prediction was more reliable than the others with d-factor and 95 percent prediction uncertainty equal to 2.53 and 92, respectively.