A comparison of performance of several artificial intelligence methods for forecasting monthly discharge time series

Developing a hydrological forecasting model based on past records is crucial to effective hydropower reservoir management and scheduling. Traditionally, time series analysis and modeling is used for building mathematical models to generate hydrologic records in hydrology and water resources. Artificial intelligence (AI), as a branch of computer science, is capable of analyzing long-series and large-scale hydrological data. In recent years, it is one of front issues to apply AI technology to the hydrological forecasting modeling. In this paper, autoregressive moving-average (ARMA) models, artificial neural networks (ANNs) approaches, adaptive neural-based fuzzy inference system (ANFIS) techniques, genetic programming (GP) models and support vector machine (SVM) method are examined using the long-term observations of monthly river flow discharges. The four quantitative standard statistical performance evaluation measures, the coefficient of correlation (R), Nash–Sutcliffe efficiency coefficient (E), root mean squared error (RMSE), mean absolute percentage error (MAPE), are employed to evaluate the performances of various models developed. Two case study river sites are also provided to illustrate their respective performances. The results indicate that the best performance can be obtained by ANFIS, GP and SVM, in terms of different evaluation criteria during the training and validation phases.     

顾及次优噪声模型的GNSS坐标时间序列噪声模型选择

在全球导航卫星系统（Global Navigation Satellite System,GNSS）坐标时间序列噪声建模领域,选择合适的噪声模型对GNSS坐标时间序列的速度信号提取具有重要价值.长期以来,研究者专注于建立最优噪声模型来描述GNSS坐标时间序列中的噪声成分,却往往直接忽略与最优噪声模型性能接近的次优噪声模型.本文选用8种常见的噪声模型,使用极大似然估计法（Maximum Likelihood Estimate,MLE）估计模型参数,以赤池信息量准则（Akaike information criterion,AIC）作为评价准则来选取噪声模型.实验数据采用中国区域12个国际GNSS服务（International GNSS Service,IGS）基准站36个坐标分量及全球范围内50个IGS基准站150个坐标分量的坐标时间序列.结果表明,超过半数的测站坐标分量上存在与最优噪声模型性能接近的次优噪声模型.在此基础上中国区域的IGS站,在东方向有14%的次优噪声模型拥有比最优噪声模型数值更大的速度不确定度,在北方向和垂直方向,这一值分别为33%和63%.而对于全球范围的IGS站,这一值分别为31%、39%和48%.因此在GNSS坐标时间序列噪声模型选择过程中,充分考虑与最优噪声模型性能接近的次优噪声模型有助于获得到更保守的估计结果.

     

A hidden Markov model segmentation procedure for hydrological and environmental time series

In this paper we present a procedure for the segmentation of hydrological and enviromental time series. We consider the segmentation problem from a purely computational point of view which involves the minimization of Huberts segmentation cost; in addition this least squares segmentation is equivalent to Maximum Likelihood segmentation. Our segmentation procedure maximizes Likelihood and minimizes Huberts least squares criterion using a hidden Markov model (HMM) segmentation algorithm. This algorithm is guaranteed to achieve a local maximum of the Likelihood. We evaluate the segmentation procedure with numerical experiments which involve artificial, temperature and river discharge time series. In all experiments, the procedure actually achieves the global minimum of the Likelihood; furthermore execution time is only a few seconds, even for time series with over a thousand terms.     

Comparison of forecasting performance of nonlinear models of hydrological time series

Mean monthly flows of the Tatry alpine mountain region in Slovakia are predominantly fed by snowmelt in the spring and convective precipitation in the summer. Therefore their regime properties exhibit clear seasonal patterns. Positive deviations from these trends have substantially different features than the negative ones. This provides intuitive justification for the application of nonlinear two-regime models for modelling and forecasting of these time series. Nonlinear time series structures often have lead to good fitting performances, however these do not guarantee an equally good forecasting performance. In this paper therefore the forecasting performance of several nonlinear time series models is compared with respect to their capabilities of forecasting monthly and seasonal flows in the Tatry region. A new type of regime-switching models is also proposed and tested. The best predictive performance was achieved for a new model subclass involving aggregation operators.     

Singular spectrum analysis and forecasting of hydrological time series

The singular spectrum analysis (SSA) technique is applied to some hydrological univariate time series to assess its ability to uncover important information from those series, and also its forecast skill. The SSA is carried out on annual precipitation, monthly runoff, and hourly water temperature time series. Information is obtained by extracting important components or, when possible, the whole signal from the time series. The extracted components are then subject to forecast by the SSA algorithm. It is illustrated the SSA ability to extract a slowly varying component (i.e. the trend) from the precipitation time series, the trend and oscillatory components from the runoff time series, and the whole signal from the water temperature time series. The SSA was also able to accurately forecast the extracted components of these time series.     

顾及有色噪声的GNSS时间序列时域信号提取

随着大型地震的发生,GNSS时间序列中除线性趋势和周期信号外,还存在大量震后瞬态,准确地提取各项时域信号是运用GNSS时间序列进行地学研究的关键.为此,本文提出了顾及有色噪声的GNSS时间序列时域信号提取法.该方法首先基于白噪声（White Noise,WN）+闪烁噪声（Flicker Noise,FN）模型,使用最大似然估计（Maximum Likelihood Estimation,MLE）对震前GNSS时间序列进行参数估计,并根据参数估值来去除震后时间序列中的震前信号,以此获取残差序列;然后将残差序列作为求解特征时间尺度的观测量,WN+FN模型作为观测量的随机模型,并采取非线性最小二乘法（Non-linear Least Squares,NLS）法估计特征时间尺度;最后利用估计的特征时间尺度构建GNSS时间序列函数模型,并采用MLE估计其未知参数,进而实现时域信号的提取.经模拟数据分析,考虑有色噪声时,特征时间尺度估计算法的收敛性提高了25%,各项未知参数的标准差（Standard Deviation,STD）显著下降.最后,将该算法应用于日本区域实测数据,并与传统方法进行了对比分析.

     

Analysis and interpretation of patterns within and between hydroclimatological time series in an alpine glacier basin

This paper explores patterns within and between climatological and hydrological time series from an alpine glacier basin. Time series recorded in the basin of the Haut Glacier d'Arolla over the 1989 ablation season are subdivided into five subperiods. Box-Jenkins ARIMA (AutoRegressive Integrated Moving Average) and TFN (Transfer Function-Noise) models are estimated for each of the five subperiods and differences between the models are interpreted in the context of changing glacier hydrology, particularly the changing nature and extent of the glacier drainage network.     

The analysis of turbulence time series: Statistical and correlation approaches using the minitab package

The theoretical basis and use of statistical and correlation approaches for the analysis of turbulence time series are described. In particular, the analysis of data from high frequency current meter measurements using MINITAB is presented and the data files and analytical routines are included on the software disk.     

A two-stage time series model for monthly hydrological projections under climate change in the Lim River basin (southeast Europe)

ABSTRACT

Climate change projections of precipitation and temperature suggest that Serbia could be one of the most affected regions in southeastern Europe. To prepare adaptation measures, the impact of climate changes on water resources needs to be assessed. Pilot research is carried out for the Lim River basin, in southeastern Europe, to predict monthly flows under different climate scenarios. For estimation of future water availability, an alternative approach of developing a deterministic-stochastic time series model is chosen. The proposed two-stage time series model consists of several components: trend, long-term periodicity, seasonality and the stochastic component. The latter is based on a transfer function model with two input variables, precipitation and temperature, as climatic drivers. The Nash-Sutcliffe model efficiency for the observed period 1950–2012 is 0.829. The model is applied for the long-term hydrological prediction under the representative concentration pathway (RCP) emissions scenarios for the future time frame 2013–2070.     

An improved hemispherical photography model for stream surface shortwave radiation estimations in a central U.S. hardwood forest

Hemispherical photographs of forest canopies can be used to develop sophisticated models that predict incident below canopy shortwave radiation on the surface of interest (i.e. soil and water). Hemispherical photographs were collected on eight dates over the course of a growing season to estimate leaf area index and to quantify solar radiation incident on the surface of two stream reaches based on output from Gap Light Analyser and Hemisfer software. Stream reaches were shaded by a mixed‐deciduous Ozark border forested riparian canopy. Hemispherical photo model results were compared to observed solar radiation sensed at climate stations adjacent to each stream reach for the entire 2010 water year. Modeled stream‐incident shortwave radiation was validated with above‐stream pyranometers for the month of September. On average, the best hemispherical photo models underestimated daily averages of solar radiation by approximately 14% and 12% for E–W and N–S flowing stream reaches, respectively (44.7 W/m² measured vs 38.4 W/m² modeled E–W, 46.8 W/m² vs. 41.1 W/m²N–S). The best hemispherical photo models overestimated solar radiation relative to in–Stream pyranometers placed in the center of each stream reach by approximately 7% and 17% for E–W and N–S stream reaches respectively (31.3 W/m² measured vs 33.5 W/m² modeled E–W, 31.5 W/m² vs. 37.1 W/m²N–S). The model provides a geographically transferable means for quantifying changes in the solar radiation regime at a stream surface due to changes in canopy density through a growing season, thus providing a relatively simple method for estimating surface and water heating in canopy altered environments (e.g. forest harvest). Copyright © 2012 John Wiley & Sons, Ltd.     

GRACE月时变模型缺失值的填补方法与精度分析

长时间序列的GRACE时变重力场对研究全球地表质量变化具有重要的意义.部分月份的GRACE卫星观测数据质量不佳导致了相应月份的时变重力场模型缺失,为了保持时变重力场模型的连续性,可采用一定的插值方法填补.本文以ITSG-Grace2016时变重力场模型序列为研究对象,详细分析了三次样条插值、三次埃尔米特插值和三次多项式插值等3种方法用于填补GRACE时变模型序列的精度,实验结果表明:(1)利用3种插值方法获取空缺1个月或连续空缺2个月的时变重力场模型时,插值时变模型与实测时变模型比较,阶误差均较小,且三次埃尔米特插值的精度稍好;(2)利用插值时变模型分析区域质量变化时,在空缺1个月数据的情况下,插值时变模型与实测时变模型符合度较高,但在连续空缺2个月数据的情况下,插值时变模型与实测时变模型的计算结果差异较大,说明利用阶误差评定模型精度具有一定局限性;(3)对区域质量变化的趋势项进行分析时,区域质量变化的复杂程度决定了模型内插的精度,当时间序列的长度在3年或3年以上时,插值时变模型的精度对区域质量变化分析的影响较小.在分析区域质量变化时,三次样条插值方法的插值结果与实测结果更为接近,建议...     

Neural network time series prediction of environmental variables in a small upland headwater in NE Scotland

Short‐term prediction of environmental variables such as stream flow rate is useful to members of the general public and environmental scientists alike, providing the ability to predict environmental disasters or scientifically interesting events. Here, a neural‐network based method is presented, which is capable of providing advance flood warnings or the prediction of high stream flow events for research purposes in a small upland headwater in NE Scotland. This method relies on training from past time series data acquired in the field, and provides the ability to predict a range of hydrological and meteorological variables up to 24 h ahead using feedback of predicted values at time t as new inputs for the next time step t + 1. The system is rapid and effective, relies on standard neural network training methods, and has the potential to be implemented in a web‐based monitoring and prediction package. The model design could be implemented at any study site where time series data has been gathered, and is sufficiently flexible to accept whatever data is available. Copyright © 2007 John Wiley & Sons, Ltd.     

Application of spatial EOF and multivariate time series model for evaluating agricultural drought vulnerability in Korea

This study proposed a methodology using the empirical orthogonal function (EOF) and multivariate time series model for the analysis of drought both in time and space. The methodology proposed was then applied to evaluate the vulnerability of agricultural drought of major river basins in Korea. First, the three-month SPI data from 59 rain gauge stations over the Korean Peninsula were analyzed by deriving and spatially characterizing the EOFs. The shapes of major estimated EOFs were found to well reflect the observed spatial pattern of droughts. Second, the coefficient time series of estimated EOFs were then fitted by a multivariate time series model to generate the SPI data for 10,000 years, which were used to derive the annual maxima series of areal average drought severity over the Korean Peninsula. These annual maxima series were then analyzed to determine the mean drought severity for given return periods. Four typical spatial patterns of drought severity could also be selected for those return periods considered. This result shows that the southern part of the Korean Peninsula is most vulnerable to drought than the other parts. Finally, the agricultural drought vulnerability was evaluated by considering the potential water supply from dams. In an ideal case, when all the maximum dam storage was assumed to be assigned to agriculture, all river basins in Korea were found to have the potential to overcome a 30-year drought. However, under more realistic conditions considering average dam storage and water allocation priorities, most of the river basins could not overcome a 30-year drought.     

基于复杂度理论的地磁变化场时间序列特征分析

复杂度理论为研究地磁变化场提供了新思路和新途径.地磁变化场从物理起源上可以看作是一种复杂系统,因此,本文从军事工程应用需求出发,基于复杂度理论,提出运用样本熵、多尺度熵及滑动窗样本熵方法对不同磁扰程度下的地磁变化场时间序列进行复杂度特征分析,结果表明：（1）样本熵和多尺度熵能够很好地表征地磁扰动强度及演化特征,启发我们可设计一种新的“熵指数”来衡量地磁扰动大小;（2）滑动窗样本熵能够准确定位地磁扰动时间段,对于磁暴识别与预测、空间灾害性天气预警等有一定的参考价值;（3）Mackey-Glass时滞混沌方程随时间的演化呈现出与地磁场日变曲线非常相似的形态,因此,地磁变化场或许可用时滞混沌方程来表示,对于我们更好地认识地磁变化场物理机理、建模预测与地磁寻的等军事工程应用都有一定的意义.     

Seemingly unrelated intervention time series models for effectiveness evaluation of large scale environmental remediation

Large scale environmental remediation projects applied to sea water always involve large amount of capital investments. Rigorous effectiveness evaluations of such projects are, therefore, necessary and essential for policy review and future planning. This study aims at investigating effectiveness of environmental remediation using three different Seemingly Unrelated Regression (SUR) time series models with intervention effects, including Model (1) assuming no correlation within and across variables, Model (2) assuming no correlation across variable but allowing correlations within variable across different sites, and Model (3) allowing all possible correlations among variables (i.e., an unrestricted model). The results suggested that the unrestricted SUR model is the most reliable one, consistently having smallest variations of the estimated model parameters. We discussed our results with reference to marine water quality management in Hong Kong while bringing managerial issues into consideration.     

基于同步大地电磁时间序列依赖关系的噪声处理

本文从信号与系统的角度讨论了同步大地电磁时间序列信号之间的依赖关系,选取高信噪比的时间序列信号作为先验数据,用最小二乘法估算依赖关系;结合参考道的数据,合成本地道含噪声时段的数据,最后用合成数据替代噪声段数据,组成新数据,从而在时域中去除大地电磁噪声.西藏地区高信噪比实测数据的试算结果表明,无论电场还是磁场,信号之间的依赖关系是相对稳定的,只与先验数据的长度有关,与时间无关;虽然不同参考点之间的依赖关系不同,但都可以精确合成本地点数据,与参考点地下电性结构和参考距离无关.仿真实验显示,去噪后的信号与原始信号基本一致.实测数据处理结果表明,该方法可以有效去除强噪声干扰,抑制中高频段的近场源效应,同时保留了微弱的有效信号,保证了处理结果的正确性.最后针对方差比方法无法识别的方波噪声,提出了一种简单的平移方法,成功去除了持续时间大于窗口长度的方波噪声;将该方法与远参考技术结合,可以有效抑制近场源噪声干扰,获得光滑连续并且可信的测深资料.     

Modelling the impact of recent land‐cover changes on the stream flows in northeastern Puerto Rico

We investigated the influence of recent and future land‐cover changes on stream flow of a watershed northeastern Puerto Rico using hydrological modeling and simulation analysis. Monthly and average annual stream flows were compared between an agricultural period (1973–1980) and an urbanized/reforested period (1988–1995) using the revised Generalized Watershed Loading Function model. Our validated results show that a smaller proportion of rainfall became stream flows in the urbanized/forested period compared with the agricultural period, apparently because of reforestation. Sensitivity analysis of the model showed that evapotranspiration, precipitation, and curve number were the most significant factors influencing stream flow. Simulations of projected land‐cover scenarios indicate that annual stream flows would increase by 9·6% in a total urbanization scenario, decrease by 3·6% in a total reforestation scenario, and decrease by 1·1% if both reforestation and urbanization continue at their current rates to 2020. An imposed hurricane event that was similar in scale to the largest recent event on the three land‐cover scenarios would increase the daily stream flow by 62·1%, 68·4% and 67·1% respectively. Owing to the environmental setting of eastern Puerto Rico, where sea breezes caused by temperature differences between land surface and the ocean dominate the local climate, we suggest that managing local land‐cover changes can have important consequences for water management. Copyright © 2007 John Wiley & Sons, Ltd.     

Unsaturated flows in soils with self-similar hydraulic conductivity distribution

In this article, we are concerned with the statistics of steady unsaturated flow in soils with a fractal hydraulic conductivity distribution. It is assumed that the spatial distribution of log hydraulic conductivity can be described as an isotropic stochastic fractal process. The impact of the fractal dimension of this process, the soil pore-size distribution parameter, and the characteristic length scale on the variances of tension head and the effective conductivity is investigated. Results are obtained for one-dimensional and three-dimensional flows. Our results indicate that the tension head variance is scale-dependent for fractal distribution of hydraulic conductivity. Both tension head variance and effective hydraulic conductivity depend strongly on the fractal dimension. The soil pore-size distribution parameter is important in reducing the variability of the unsaturated hydraulic conductivity and of the fluxes.     

foF2月中值太阳循环变化及单站谱模型研究

利用我国9个电离层观测站第21和22太阳周大约20年的foF2月中值数据,分析太阳活动和地磁活动对电离层foF2的影响,结果显示白天和夏季夜间foF2和太阳黑子数R之间存在着明显非线性关系,并且随着纬度的降低逐渐增强.当回归分析加入地磁Ap指数时,多重回归模型与实测值误差进一步减小,说明同时考虑太阳活动和地磁活动的非线性影响能够更好地描述foF2的变化.基于foF2与太阳黑子数R及地磁指数Ap之间的非线性统计关系,利用Fourier级数建立9个单站谱模型,并与国际参考电离层IRI进行了比较,精度有一定提高.