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.     

GRACE terrestrial water storage data assimilation based on the ensemble four-dimensional variational method PODEn4DVar:Method and validation

Seasonal and interannual changes in the Earth's gravity field are mainly due to mass exchange among the atmosphere,ocean,and continental water sources.The terrestrial water storage changes,detected as gravity changes by the Gravity Recovery and Climate Experiment(GRACE) satellites,are mainly caused by precipitation,evapotranspiration,river transportation and downward infiltration processes.In this study,a land data assimilation system LDAS-G was developed to assimilate the GRACE terrestrial water storage(TWS) data into the Community Land Model(CLM3.5) using the POD-based ensemble four-dimensional variational assimilation method PODEn4 DVar,disaggregating the GRACE large-scale terrestrial water storage changes vertically and in time,and placing constraints on the simulation of vertical hydrological variables to improve land surface hydrological simulations.The ideal experiments conducted at a single point and assimilation experiments carried out over China by the LDAS-G data assimilation system showed that the system developed in this study improved the simulation of land surface hydrological variables,indicating the potential of GRACE data assimilation in large-scale land surface hydrological research and applications.     

Direct and Inverse Problems in a Variational Concept of Environmental Modeling

A concept of environmental forecasting based on a variational approach is discussed. The basic idea is to augment the existing technology of modeling by a combination of direct and inverse methods. By this means, the scope of environmental studies can be substantially enlarged. In the concept, mathematical models of processes and observation data subject to some uncertainties are considered. The modeling system is derived from a specially formulated weak-constraint variational principle. A set of algorithms for implementing the concept is presented. These are: algorithms for the solution of direct, adjoint, and inverse problems; adjoint sensitivity algorithms; data assimilation procedures; etc. Methods of quantitative estimations of uncertainty are of particular interest since uncertainty functions play a fundamental role for data assimilation, assessment of model quality, and inverse problem solving. A scenario approach is an essential part of the concept. Some methods of orthogonal decomposition of multi-dimensional phase spaces are used to reconstruct the hydrodynamic background fields from available data and to include climatic data into long-term prognostic scenarios. Subspaces with informative bases are constructed to use in deterministic or stochastic-deterministic scenarios for forecasting air quality and risk assessment. The results of implementing example scenarios for the Siberian regions are presented.     

Test of three methods to detect the overbank flow from water level time‐series analysis

In this study, we propose to identify morphological and hydraulic characteristics related to overbank flows in the water level time‐series available at many gauged stations. The results obtained at 13 river stations (the catchment sizes vary from 10 to 1700 km² with contrasted geology, land use and rainfalls regime) show that overflow mechanisms at the river‐reach scale can be systematically identified in the water level frequency distribution estimated with the peaks‐over‐threshold (POT) method. A first level (Lts1) was in the range of the incipient flooding onto the floodplain. Even if the definition of this level is variable in terms of flooded area at the reach scale, this method can be useful in providing a first estimation of the bankfull level for many gauged stations, without complex and costly field surveys. A second level (Lts2) was systematically detected on average 38 cm above the topographic flat floodplain elevation. The Lts2 inflection in the water‐level frequency distribution is assumed to reflect a composite effect of catchment hydrology and local hydraulics and channel geometry, without possibility to make a clear distinction between both processes at the moment. The local or reach scale effect would probably play an essential role in the frequency distribution as flood attenuation at catchment scale may explain the inflection Lts2 at only three sites. In light of the knowledge acquired in laboratory flumes with compound channels, most of the time Lts2 level would correspond to the level of resumption of flow both in the main channel and the floodplain. Once this method is validated in various physiographic contexts, it should apply to many hydrometric stations for both synchronic (e.g. regional analysis) and diachronic analysis (e.g. evolution over time of the bankfull discharge) to evaluate anthropic impacts on river morphology. Copyright © 2010 John Wiley & Sons, Ltd.     

Snow water equivalent prediction using Bayesian data assimilation methods

Using the U.S. National Weather Service’s SNOW-17 model, this study compares common sequential data assimilation methods, the ensemble Kalman filter (EnKF), the ensemble square root filter (EnSRF), and four variants of the particle filter (PF), to predict seasonal snow water equivalent (SWE) within a small watershed near Lake Tahoe, California. In addition to SWE estimation, the various data assimilation methods are used to estimate five of the most sensitive parameters of SNOW-17 by allowing them to evolve with the dynamical system. Unlike Kalman filters, particle filters do not require Gaussian assumptions for the posterior distribution of the state variables. However, the likelihood function used to scale particle weights is often assumed to be Gaussian. This study evaluates the use of an empirical cumulative distribution function (ECDF) based on the Kaplan–Meier survival probability method to compute particle weights. These weights are then used in different particle filter resampling schemes. Detailed analyses are conducted for synthetic and real data assimilation and an assessment of the procedures is made. The results suggest that the particle filter, especially the empirical likelihood variant, is superior to the ensemble Kalman filter based methods for predicting model states, as well as model parameters.     

Hydrological information for sustainable development

Abstract

This paper presents the development of a framework for data collection network design that considers sustainable development goals. The proposed framework adopts sustainable development principles and incorporates and revises traditional methodologies used in data collection network design. Important components of the framework include a focus on hydrological information, the preservation of long term gauging stations, and the adoption of integrated ecosystem monitoring. The eventual adoption of the framework should ensure that hydrological information required for effective decision making in water resources is available even as hydrometric data collection networks are rationalized.     

PUBLICATIONS OF THE IAHS

Abstract

Gauging stations where the stage—discharge relationship is affected by hysteresis due to unsteady flow represent a challenge in hydrometry. In such situations, the standard hydrometric practice of fitting a single-valued rating curve to the available stage—discharge measurements is inappropriate. As a solution to this problem, this study provides a method based on the Jones formula and nonlinear regression, which requires no further data beyond the available stage—discharge measurements, given that either the stages before and after each measurement are known along with the duration of each measurement, or a stage hydrograph is available. The regression model based on the Jones formula rating curve is developed by applying the monoclinal rising wave approximation and the generalized friction law for uniform flow, along with simplifying assumptions about the hydraulic and geometric properties of the river channel in conjunction with the gauging station. Methods for obtaining the nonlinear least-squares rating-curve estimates, while factoring in approximated uncertainty, are discussed. The broad practical applicability and appropriateness of the method are demonstrated by applying the model to: (a) an accurate, comprehensive and detailed database from a hydropower-generated highly dynamic flow in the Chattahoochee River, Georgia, USA; and (b) data from gauging stations in two large rivers in the USA affected by hysteresis. It is also shown that the model is especially suitable for post-modelling hydraulic and statistical validation and assessment.     

Improvement of Global Hydrological Models Using GRACE Data

After about 6 years of GRACE (Gravity Recovery and Climate Experiment) satellite mission operation, an unprecedented global data set on the spatio-temporal variations of the Earth’s water storage is available. The data allow for a better understanding of the water cycle at the global scale and for large river basins. This review summarizes the experiences that have been made when comparing GRACE data with simulation results of global hydrological models and it points out the prerequisites and perspectives for model improvements by combination with GRACE data. When evaluated qualitatively at the global scale, water storage variations on the continents from GRACE agreed reasonably well with model predictions in terms of their general seasonal dynamics and continental-scale spatial patterns. Differences in amplitudes and phases of water storage dynamics revealed in more detailed analyses were mainly attributed to deficiencies in the meteorological model forcing data, to missing water storage compartments in the model, but also to limitations and errors of the GRACE data. Studies that transformed previously identified model deficiencies into adequate modifications of the model structure or parameters are still rare. Prerequisites for a comprehensive improvement of large-scale hydrological models are in particular the consistency of GRACE observation and model variables in terms of filtering, reliable error estimates, and a full assessment of the water balance. Using improvements in GRACE processing techniques, complementary observation data, multi-model evaluations and advanced methods of multi-objective calibration and data assimilation, considerable progress in large-scale hydrological modelling by integration of GRACE data can be expected.     

River flow modelling: comparison of performance and evaluation of uncertainty using data-driven models and conceptual hydrological model

Hydrological and statistical models are playing an increasing role in hydrological forecasting, particularly for river basins with data of different temporal scales. In this study, statistical models, e.g. artificial neural networks, adaptive network-based fuzzy inference system, genetic programming, least squares support vector machine, multiple linear regression, were developed, based on parametric optimization methods such as particle swarm optimization (PSO), genetic algorithm (GA), and data-preprocessing techniques such as wavelet decomposition (WD) for river flow modelling using daily streamflow data from four hydrological stations for a period of 1954–2009. These models were used for 1-, 3- and 5-day streamflow forecasting and the better model was used for uncertainty evaluation using bootstrap resampling method. Meanwhile, a simple conceptual hydrological model GR4J was used to evaluate parametric uncertainty based on generalized likelihood uncertainty estimation method. Results indicated that: (1) GA and PSO did not help improve the forecast performance of the model. However, the hybrid model with WD significantly improved the forecast performance; (2) the hybrid model with WD as a data preprocessing procedure can clarify hydrological effects of water reservoirs and can capture peak high/low flow changes; (3) Forecast accuracy of data-driven models is significantly influenced by the availability of streamflow data. More human interferences from the upper to the lower East River basin can help to introduce greater uncertainty in streamflow forecasts; (4) The structure of GR4J may introduce larger parametric uncertainty at the Longchuan station than at the Boluo station in the East river basin. This study provides a theoretical background for data-driven model-based streamflow forecasting and a comprehensive view about data and parametric uncertainty in data-scarce river basins.     

A data assimilation approach to discharge estimation from space

River discharge is currently monitored by a diminishing network of gauges, which provide a spatially incomplete picture of global discharges. This study assimilated water level information derived from a fused satellite Synthetic Aperture Radar (SAR) image and digital terrain model (DTM) with simulations from a coupled hydrological and hydrodynamic model to estimate discharge in an un‐gauged basin scenario. Assimilating water level measurements led to a 79% reduction in ensemble discharge uncertainty over the coupled hydrological hydrodynamic model alone. Measurement bias was evident, but the method still provided a means of improving estimates of discharge for high flows. The study demonstrates the potential of currently available synthetic aperture radar imagery to reduce discharge uncertainty in un‐gauged basins when combined with model simulations in a data assimilation framework, where sufficient topographic data are available. The work is timely because in the near future the launch of satellite radar missions will lead to a significant increase in the volume of data available for space‐borne discharge estimation. Copyright © 2009 John Wiley & Sons, Ltd.     

Estimation of peak streamflows through channel geometry / Estimation de pics de débit fluviatiles à l'aide de la géométrie des cours d'eau

Abstract

Knowledge of peak discharge is essential for safe and economical planning and design of hydraulic structures. In India, as in most developing countries, the majority of river basins are either sparsely gauged or not gauged at all. The gauged records are also of short length (generally 15–30 years), therefore development of robust models is necessary for estimation of streamflows. Various studies reveal that flood estimation through channel geometry is an alternative method for ungauged catchments. It is appropriate for use where flow characteristics are poorly related to catchment area and other catchment characteristics. In the present study, stream geometry parameters for 42 river sites in central-south India were used; calibration equations were developed with data for 35 stations and tested on data for the remaining seven stations. The relationships developed between mean discharge and channel geometry parameters provide an alternative technique for estimation of mean annual channel discharge.     

Flow dynamics at the continental scale: Streamflow correlation and hydrological similarity

Streamflow variability in space and time critically affects anthropic water uses and ecosystem services. Unfortunately, spatiotemporal patterns of flow regimes are often unknown, as discharge measurements are usually recorded at a limited number of hydrometric stations unevenly distributed along river networks. Advances in understanding the physical processes that control the spatial patterns of river flows are therefore necessary to predict water availability at ungauged locations or to extrapolate pointwise streamflow observations. This work explores the use of the spatial correlation of river flows as a metric to quantify the similarity between hydrological responses of two catchments. Following a stochastic framework, 340,000 cross‐correlations between pairs of daily streamflows time series are predicted at a seasonal timescale across the contiguous United States using 413 catchments of the MOPEX dataset. Model predictions of streamflow correlation obtained in absence of run‐off information are successfully used to identify catchment outlets sharing similar discharge dynamics and flow regimes across a broad range of geomorphoclimatic conditions, without relying on calibration. The selection of reference streamgauges based on predicted streamflow correlation generally outperforms the selection based on spatial proximity, especially as the density of available gauged sections decreases. Interestingly, correlated outlets share a broad spectrum of hydrological signatures (mean discharge, flow variability, and recession properties), suggesting that catchments forced by analogous frequency and intensity of effective rainfall events might exhibit common geomorphoecological traits leading to similar hydrological responses. The proposed framework provides a physical basis to assist the regionalization of flow dynamics and to interpret the spatial variability of flow regimes along stream networks.     

Neural networks in auroral data assimilation

Data assimilation is an essential step for improving space weather forecasting by means of a weighted combination between observational data and data from a mathematical model. In the present work data assimilation methods based on Kalman filter (KF) and artificial neural networks are applied to a three-wave model of auroral radio emissions. A novel data assimilation method is presented, whereby a multilayer perceptron neural network is trained to emulate a KF for data assimilation by using cross-validation. The results obtained render support for the use of neural networks as an assimilation technique for space weather prediction.     

集合资料同化中方差滤波技术研究及试验

本文基于YH4DVAR业务系统构建了集合资料同化试验平台,利用10个集合样本统计得到的流依赖背景误差能显著改进业务应用中背景误差方差的结构和大小.但是受样本数的限制,背景误差方差的集合估计值中引入了大量的随机取样噪声.为了降低噪声对估计值的影响,本文采用谱滤波方法,根据信号和噪声尺度的统计特征构造一个低通滤波器来滤除背景误差方差估计值中的大部分随机取样噪声.在2013年第九号台风"飞燕"的集合方差滤波试验中,10个样本的滤波结果优于30个样本的集合估计值.谱滤波方法的成功应用有效降低了集合资料同化系统对集合样本数的要求,将是集合资料同化系统未来业务化运行的一项不可或缺的关键技术.     

The channel-geometry method: Guidelines and applications

All river engineering schemes require flood discharge estimates as part of the design and appraisal process. Unfortunately, continuous measurement of flood discharges is limited to those river sites with instrumented gauging stations, which constitute only a small proportion of channel reaches where information is required. Therefore, considerable research effort has been devoted to the development of reliable indirect techniques of flood discharge estimation. Research on the interrelationship of stream channel geometry and river discharge has provided the basis for an indirect method of flood estimation – the channel-geometry method – which employs river channel dimensions alone to estimate discharge characteristics at ungauged river sites. Channel-geometry equations are developed empirically by relating streamflow data from gauging stations and channel dimensions measured from natural river reaches in the vicinity of the gauge, and take the form of power function relations. Once regional channel-geometry equations have been defined, a channel width or channel capacity measurement is the only variable needed to estimate the flood flow characteristics at a specified river site. The method is useful as an alternative to traditional catchment-based approaches or as a rapid reconnaissance technique. In addition to the application for flood discharge prediction, channel-geometry equations could prove helpful in the management of river channels, first, by providing a basis for assessing local deviations in the channel form–discharge relation, deviations which could be employed as indicators of the sensitivity of particular stretches of river channel to change, and secondly, in the computation of natural channel dimensions for use in river channel design and river restoration.     

"THREE DIMENSIONAL MATHEMATICAL MODEL OF SEDIMENT TRANSPORT IN ESTUARINE REGIONS-- A CASE STUDY OF THE HAIHE RIVER MOUTH" Yuchuan Bai,Huanting SHEN and Shixiong HU

The paper presents the 3D finite element simulation of tidal flow and Sediment transport in the estuarine region of the Haihe river. The proposed model adopts sigma-transformation of the hydrodynamic and sediment transport equations. The hydrodynamic and sediment transport models are verified in case of a simple test problem for which analytical solutions are available. Finally the models are applied to muddy Haihe river estuary of North China and it is claimed that hydrodynamic and sediment transport models give a reliable comparison with the observed field data. However, there are certain discrepancies, and some reasonable questions regarding the present state-of-art, in the modeling of three-dimensional multilevel hydrodynamics and sediment transport, which are provided below for answer.     

A comparative study of three different learning algorithms applied to ANFIS for predicting daily suspended sediment concentration

The modeling and prediction of suspended sediment in a river are key elements in global water recourses and environment policy and management. In the present study, an Adaptive Neuro-Fuzzy Inference System model trained with the Levenberg-Marquardt learning algorithm is considered for time series modeling of suspended sediment concentration in a river. The model is trained and validated using daily river discharge and suspended sediment concentration data from the Schuylkill River in the United States. The results of the proposed method are evaluated and compared with similar networks trained with the common Hybrid and Back-Propagation algorithms, which are widely used in the literature for prediction of suspended sediment concentration. Obtained results demonstrate that models trained with the Hybrid and Levenberg-Marquardt algorithms are comparable in terms of prediction accuracy. However, the networks trained with the Levenberg-Marquardt algorithm perform better than those trained with the Hybrid approach.