A hybrid Bayesian back‐propagation neural network approach to multivariate modelling

There is growing interest in the use of back‐propagation neural networks to model non‐linear multivariate problems in geotehnical engineering. To overcome the shortcomings of the conventional back‐propagation neural network, such as overfitting, where the neural network learns the spurious details and noise in the training examples, a hybrid back‐propagation algorithm has been developed. The method utilizes the genetic algorithms search technique and the Bayesian neural network methodology. The genetic algorithms enhance the stochastic search to locate the global minima for the neural network model. The Bayesian inference procedures essentially provide better generalization and a statistical approach to deal with data uncertainty in comparison with the conventional back‐propagation. The uncertainty of data can be indicated using error bars. Two examples are presented to demonstrate the convergence and generalization capabilities of this hybrid algorithm. Copyright © 2003 John Wiley & Sons, Ltd.     

Bayesian Discovery Process Modeling of the Lower and Middle Jurassic Play of the Halten Terrace,Offshore Norway,as Compared with the Previous Modeling

A Bayesian version of the discovery process model was applied to the pre-rift Lower and Middle Jurassic play of the Halten Terrace, Mid-Norway. The Bayesian approach estimates the lognormal parameters, the discoverability parameter, and the distribution of sizes of the undiscovered fields as well as the play potential, conditioned on a discovery sequence averaging for all possible prior choices weighted by their likelihood. This approach avoids the problem of having to make arbitrary choices for the parameters. The estimates of parameters and play potential based upon the present methodology compares well with previous estimates, if the play is divided into two sub-plays representing the overpressured and normally pressured zones. These sub-plays have been estimated independently and aggregated in order to get the total undiscovered resource potential. This study estimates that the expected remaining play potential is 100 × 10⁶Sm³ o.e., about 9% of the total resources in the play. There is however a 90% chance that the remaining potential ranges from 13 to 282× 10⁶ Sm³ o.e. and a 5% possibility of exceeding this value.     

Probabilistic Precipitation Forecasting Based on Ensemble Output Using Generalized Additive Models and Bayesian Model Averaging

A probabilistic precipitation forecasting model using generalized additive models (GAMs) and Bayesian model averaging (BMA) was proposed in this paper.GAMs were used to fit the spatial-temporal precipi...     

长江流域夏季降水的贝叶斯多模式集合概率预测

A Bayesian probabilistic prediction scheme of the Yangtze River Valley (YRV) summer rainfall is proposed to combine forecast information from multi-model ensemble dataset provided by ENSEMBLES project.Due to the low forecast skill of rainfall in dynamic models,the time series of regressed YRV summer rainfall are selected as ensemble members in the new scheme,instead of commonly-used YRV summer rainfall simulated by models.Each time series of regressed YRV summer rainfall is derived from a simple linear regression.The predictor in each simple linear regression is the skillfully simulated circulation or surface temperature factor which is highly linear with the observed YRV summer rainfall in the training set.The high correlation between the ensemble mean of these regressed YRV summer rainfall and observation benefit extracting more sample information from the ensemble system.The results show that the cross-validated skill of the new scheme over the period of 1960 to 2002 is much higher than equally-weighted ensemble,multiple linear regression,and Bayesian ensemble with simulated YRV summer rainfall as ensemble members.In addition,the new scheme is also more skillful than reference forecasts (random forecast at a 0.01 significance level for ensemble mean and climatology forecast for probability density function).     

基于贝叶斯网络分类的遥感影像变化检测

遥感成像过程中,地面、大气等诸多要素的不确定性和波段之间的相关性等原因影响了分类精度,导致变化检测的不准确性。为了提高分类精度往往需要引入先验知识。贝叶斯网络是一种新的数据表达和推理模型,对数据没有严格的正态分布前提要求,通过动态地调整先验概率密度,能有效提高分类精度。以北京通州地区1996-05-29和2001-05-19两个时相的陆地卫星Landsat TM遥感影像为例,介绍了基于贝叶斯网络的分类算法,并在此基础上实现了两个时相遥感影像的变化检测。实验结果表明:基于贝叶斯网络分类算法的后分类比较变化检测方法是遥感影像变化检测的一种新的有效方法。     

Soil erosion and sediment transport in Tanzania: Part I – sediment source tracing in three neighbouring river catchments

Water bodies in Tanzania are experiencing increased siltation, which is threatening water quality, ecosystem health, and livelihood security in the region. This phenomenon is caused by increasing rates of upstream soil erosion and downstream sediment transport. However, a lack of knowledge on the contributions from different catchment zones, land-use types, and dominant erosion processes, to the transported sediment is undermining the mitigation of soil degradation at the source of the problem. In this context, complementary sediment source tracing techniques were applied in three Tanzanian river systems to further the understanding of the complex dynamics of soil erosion and sediment transport in the region. Analysis of the geochemical and biochemical fingerprints revealed a highly complex and variable soil system that could be grouped in distinct classes. These soil classes were unmixed against riverine sediment fingerprints using the Bayesian MixSIAR model, yielding proportionate source contributions for each catchment. This sediment source tracing indicated that hillslope erosion on the open rangelands and maize croplands in the mid-zone contributed over 75% of the transported sediment load in all three river systems during the sampling time-period. By integrating geochemical and biochemical fingerprints in sediment source tracing techniques, this study demonstrated links between land use, soil erosion and downstream sediment transport in Tanzania. This evidence can guide land managers in designing targeted interventions that safeguard both soil health and water quality.     

High spatial-resolution monitoring to investigate nitrate export and its drivers in a mesoscale river catchment along an anthropogenic land-cover gradient

Nitrate monitoring is commonly conducted with low-spatial resolution, only at the outlet or at a small number of selected locations. As a result, the information about spatial variations in nitrate export and its drivers is scarce. In this study, we present results of high-spatial resolution monitoring conducted between 2012 and 2017 in 65 sub-catchments in an Alpine mesoscale river catchment characterized by a land-use gradient. We combined stable isotope techniques with Bayesian mixing models and geostatistical methods to investigate nitrate export and its main drivers, namely, microbial N turnover processes, land use and hydrological conditions. In the investigated sub-catchments, mean values of NO₃⁻ concentrations and its isotope signatures (δ¹⁵N_NO3 and δ¹⁸O_NO3) varied from 2.6 to 26.7 mg L⁻¹, from −1.3‰ to 13.1‰, and from −0.4‰ to 10.1‰, respectively. In this study, land use was an important driver for nitrate export. Very strong and strong positive correlations were found between percentages of agricultural land cover and δ¹⁵N_NO3, and NO₃⁻ concentration, respectively. Mean proportional contributions of NO₃⁻ sources varied spatially and seasonally, and followed land-use patterns. The mean contribution of manure and sewage was much higher in the catchments characterized by a high percentage of agricultural and urban land cover comparing to forested sub-catchments. Specific NO₃⁻ loads were strongly correlated with specific discharge and moderately correlated with NO₃⁻ concentrations. The nitrate isotope and concentration analysis results suggest that nitrate from external sources is stored and accumulated in soil storage pools. Nitrification of reduced nitrogen species in those pools plays the most important role for the N-dynamics in the Erlauf river catchment. Consequently, nitrification of reduced N sources was the main nitrate source except for a number of sub-catchments dominated by agricultural land use. In the Erlauf catchment, denitrification plays only a minor role in controlling NO₃⁻ export on a regional scale.     

2013年芦山Ms7.0地震前重力场源变化分析

陆地重力观测获取的时变重力信号是研究地壳内部孕震区介质变化的重要依据。利用贝叶斯平差方法处理了川西地区2010—2012年的流动重力观测资料,发现相对重力仪的格值系数在两年内变化可达2×10^-4,而且仪器存在较小的非线性漂移率;重力场累积动态变化图像显示,2013年芦山MS7.0地震发生在重力梯度带零值线与龙门山断裂带交会部位附近。进一步采用等效源方法划定了测网内以雅安、康定和石棉为中心的场源分辨能力较高的区域,并反演了地震前的地壳视密度变化。场源视密度动态变化结果表明:2013年芦山MS7.0地震前2年,在龙门山断裂带的南侧和鲜水河断裂南段的东侧区域形成了范围较大的等效视密度增加区,而在小金至康定一带和都江堰周边地区也存在小规模的视密度增加。本文基于陆地时变重力测量结果反演得到芦山地震前的地壳内部等效视密度变化,可为研究壳内深部孕震环境变化特征提供参考。     

改进的贝叶斯迭代反演方法及其在白云岩致密储层识别的应用

由于低孔隙度和低渗透率的白云岩致密油储集层的纵波阻抗与其围岩差异非常小,利用叠后反演技术难以有效预测储层,而可以提取丰富弹性信息的叠前反演是解决此问题的有效手段,但是由于噪声等问题,叠前反演方程有较强的不适定性,笔者在贝叶斯框架下引入了改进的多变量柯西分布和改进的低频约束因子,重新推导了反演方程,获得了新的目标函数,有效地减少了反演的不适定性,从而提高了反演的稳定性,并结合迭代的思想来不断更新反演求解过程中的背景纵横波速度比值,从而增加了反演结果的精度。模型数据测试和实际资料应用都证明了该方法的稳定性和适用性。统计表明,利用提出的反演方法,目的层段内优质储层厚度预测吻合率高达89.75%。因此,此方法对类似硅质致密储层的勘探有重要的借鉴意义。     

LSQR法在位场反演中的分析与评价

LSQR法具有计算效率高、对计算机内存要求低的优点,适合于大规模问题的求解。为探讨其应用于位场反演的稳定性和可靠性,笔者以加入不同噪声的两个合成模型数据为实验对象,比较分析了Tikhonov正则化与LSQR法求解结果,显示直接利用LSQR法求解位场反问题能够得到满意的正则化解,其解模型相对Tikhonov正则化,最大相对误差仅为0.36%,说明直接利用LSQR法求解位场反问题是可行的。将其应用于四川盆地雅安地区重力三维反演,极大地降低计算成本,获取了区内沉积盆及主要断裂分布情况,为页岩气靶区优选提供了有力支撑。