基于多重网格的地表水文与二维水动力动态双向耦合模型研究

针对流域洪涝模拟模型的计算精度、格式稳定性及计算效率等问题,本文提出基于多重网格技术的地表水文与二维水动力动态双向耦合模型(M-DBCM)。地表水文模型采用非线性水库法模拟降雨产流和径流;二维水动力模型采用浅水方程模拟洪水演进过程。采用不同分辨率的网格划分计算区域,在粗网格区域采用地表水文模型模拟降雨径流过程;在细网格区域采用二维水动力模型模拟洪涝积水区的水流运动。地表水文和二维水动力模型通过内部耦合移动界面(Coupling Moving Interface, CMI)实现无缝连接,保证通过CMI的水量和动量等通量守恒,提高模型的模拟精度。采用时间显式格式同时求解地表水文和水动力模型,在不同区域采用不同的计算时间步长,以提高模型的计算效率。通过典型案例验证本文构建的耦合模型的性能,结果表明本文提出的动态双向耦合模型能够在保证模拟精度的同时提高计算效率。     

A Genetic Programming Approach to System Identification of Rainfall-Runoff Models

Advancements in data acquisition, storage and retrieval are progressing at an extraordinary rate, whereas the same in the field of knowledge extraction from data is yet to be accomplished. The challenges associated with hydrological datasets, including complexity, non-linearity and multicollinearity, motivate the use of machine learning to build hydrological models. Increasing global climate change and urbanization call for better understanding of altered rainfall-runoff processes. There is a requirement that models are intelligible estimates of underlying physics, coupling explanatory and predictive components, maintaining parsimony and accuracy. Genetic Programming, an evolutionary computation technique has been used for short-term prediction and forecast in the field of hydrology. Advancing data science in hydrology can be achieved by tapping the full potential of GP in defining an evolutionary flexible modelling framework that balances prior information, simulation accuracy and strategy for future uncertainty. As a preliminary step, GP is used in conjunction with a conceptual rainfall-runoff model to solve model configuration problem. Two datasets belonging to a tropical catchment of Singapore and a temperate catchment of South Island, New Zealand with contrasting characteristics are analyzed in this study. The results indicate that proposed approach successfully combines the merits of evolutionary algorithm and conceptual knowledge in the generation of optimal model structure and associated parameters to capture runoff dynamics of catchments.     

Surrogate-Based Stochastic Multiobjective Optimization for Coastal Aquifer Management under Parameter Uncertainty

Linked simulation-optimization (S/O) approaches have been extensively used as tools in coastal aquifer management. However, parameter uncertainties in seawater intrusion (SI) simulation models often undermine the reliability of the derived solutions. In this study, a stochastic S/O framework is presented and applied to a real-world case of the Longkou coastal aquifer in China. The three conflicting objectives of maximizing the total pumping rate, minimizing the total injection rate, and minimizing the solute mass increase are considered in the optimization model. The uncertain parameters are contained in both the constraints and the objective functions. A multiple realization approach is utilized to address the uncertainty in the model parameters, and a new multiobjective evolutionary algorithm (EN-NSGA2) is proposed to solve the optimization model. EN-NSGA2 overcomes some inherent limitations in the traditional nondominated sorting genetic algorithm-II (NSGA-II) by introducing information entropy theory. The comparison results indicate that EN-NSGA2 can effectively ameliorate the diversity in Pareto-optimal solutions. For the computational challenge in the stochastic S/O process, a surrogate model based on the multigene genetic programming (MGGP) method is developed to substitute for the numerical simulation model. The results show that the MGGP surrogate model can tremendously reduce the computational burden while ensuring an acceptable level of accuracy.

     

基于长短时记忆神经网络（LSTM）的降雨径流模拟及预报

长短时记忆神经网络(LSTM)具有很强的时间序列关系拟合能力,非常适用于模拟及预报流域产汇流这一复杂的时间序列过程。基于LSTM针对不同预见期的径流预报分别建立了流域降雨径流模型,以探讨LSTM在水文预报当中的应用。模型采用流域降雨、气象及水文数据作为输入,不同预见期后的径流过程作为输出,率定期为14a,验证期为2a。结果显示,在预见期为0～2d时LSTM预报精度很高,在预见期为3d时预报精度较差,但仍优于新安江模型。隐藏层神经元数量作为神经网络复杂程度的代表既会影响模型预报精度,也会影响模型训练速度。而输入数据长度则仅会在预见期为0的条件下影响模型预报效果。     

Monthly Rainfall Forecasting Using Echo State Networks Coupled with Data Preprocessing Methods

In this paper, two novel methods, echo state networks (ESN) and multi-gene genetic programming (MGGP), are proposed for forecasting monthly rainfall. Support vector regression (SVR) was taken as a reference to compare with these methods. To improve the accuracy of predictions, data preprocessing methods were adopted to decompose the raw rainfall data into subseries. Here, wavelet transform (WT), singular spectrum analysis (SSA) and ensemble empirical mode decomposition (EEMD) were applied as data preprocessing methods, and the performances of these methods were compared. Predictive performance of the models was evaluated based on multiple criteria. The results indicate that ESN is the most favorable method among the three evaluated, which makes it a promising alternative method for forecasting monthly rainfall. Although the performances of MGGP and SVR are less favorable, they are nevertheless good forecasting methods. Furthermore, in most cases, MGGP is inferior to SVR in monthly rainfall forecasting. WT and SSA are both favorable data preprocessing methods. WT is preferable for short-term forecasting, whereas SSA is excellent for long-term forecasting. However, EEMD tends to show inferior performance in monthly rainfall forecasting.     

Emotional ANN (EANN) and Wavelet-ANN (WANN) Approaches for Markovian and Seasonal Based Modeling of Rainfall-Runoff Process

The hydrological time series have three principle components (autoregressive, seasonality and trend) and the performance of the models is strongly related to the nature of these components. The current research examines the accuracy of two Artificial Neural Network (ANN) based approaches for rainfall-runoff (r-r) modeling of two catchments with different geomorphological conditions at monthly and daily time scales. The techniques proposed here are hybrid wavelet-ANN (WANN) model, as a multi-resolution forecasting tool and Emotional Artificial Neural Network (EANN) (a new generation of ANN based models) which serves artificial emotional factors as well as classic bias and weights parameters. The obtained results for monthly modeling show that WANN could perform better than the simple feed forward neural network (FFNN) model up to 40% and 35% in terms of verification and training efficiency criteria due to significant seasonality involved in the monthly time series of the process. On the other hand, the obtained results for daily modeling via FFNN and EANN, both as Markovian models, indicates the superiority of EANN over FFNN because of EANN capability to better learning of extraordinary and extreme conditions of the process in the training phase.     

Solving Complex Rainfall-Runoff Processes in Semi-Arid Regions Using Hybrid Heuristic Model

In the current research, a hybrid model was proposed to solve the complexity of rainfall-runoff models in semi-arid regions. The proposed hybrid model structure consists of linking two data mining models, namely, Group Method of Data Handling (GMDH) and Generalized Linear Model (GLM). The proposed hybrid model structure consists of two phases. The GMDH model was used in the first phase of the hybrid model to predict daily streamflow. The first phase consists of two sections. In the first section a predictive model is developed using the time series of the daily streamflow. In the second section the rainfall-runoff model was developed. The outputs of the first phase of the hybrid model are used as inputs to the second phase of the hybrid model. The second phase of the hybrid model was developed using the GLM model. The Gomel River in Iraq was selected as a case study. The daily rainfall data and daily streamflow data for the period from January 1, 2004 to December 19, 2016 were used to train and validate the model. The results proved the accuracy of the proposed hybrid model in estimating the daily streamflow of the study area, where the value of R² was 0.92 in the training period and 0.88 in the validation period of the model.

     

Potential implications of pre-storm soil moisture on hydrological prediction

Numerous hydrological models with various complexities, strengths, and weaknesses are available. Despite technological development, the association of runoff accuracy with the underlying model's parameters in watersheds with limited data remains elusive. Evaluating the soil moisture impacts at the watershed scale is often a difficult task, but it can be vital to optimally managing water resources. Incorporating pre-storm soil moisture accounting (PSMA) procedures into hydrologic models affects the watershed response to generate runoff from storm rainfall. This study demonstrated the impact of pre-storm and post-storm soil moisture in order to circumvent major obstacles in accurate runoff estimation from watersheds employing the conventional curve number (CN) model. The proposed hydrological lumped model was tested on a data set (1,804 rainfall-runoff events) from 39 watersheds in South Korea. Its superior performance indicates that the reconciliation of pre- and post-storm conceptualization has the potential to be a solution for efficient hydrological predictions and to demonstrate the complex and dynamic nature of tractable hydrological processes. The statistically significant results reveal that the proposed model can more effectively predict runoff from watersheds in the study area than the conventional CN model and its previously proposed modifications.     

Green-Ampt Infiltration Models for Varied Field Conditions: A Revisit

The Green-Ampt (GA) infiltration model is a simplified version of the physically based full hydrodynamic model, known as the Richards equation. The simplicity and accuracy of this model facilitates for its use in many field problems, such as, infiltration computation in rainfall-runoff modelling, effluent transport in groundwater modelling studies, irrigation management studies including drainage systems etc. The numerous infiltration models based on the Green-Ampt approach have been widely investigated for their applicability in various scenarios of homogeneous soils. However, recent advances in physically based distributed rainfall-runoff modeling demands for the use of improved infiltration models for layered soils with non-uniform initial moisture conditions under varying rainfall patterns to capture the actual infiltration process that exists in nature. The difficulty that modelers are facing now-a-days includes the estimation of time of ponding and the application of the infiltration model to unsteady rainfall events occurring in heterogeneous soil conditions. The investigation in this direction exhibits that only few infiltration models can handle these situations. Hence, it is of vital importance to analyze the usefulness of different variants of the Green-Ampt infiltration models in terms of their degree of accuracy, complexity and applicability limits. This paper provides a brief review of these infiltration models to bring out their usefulness in the rainfall-runoff and irrigation modeling studies as well as the drawbacks associated with these models.     

Efficient Identification of Unknown Groundwater Pollution Sources Using Linked Simulation-Optimization Incorporating Monitoring Location Impact Factor and Frequency Factor

This study aims to improve the accuracy of groundwater pollution source identification using concentration measurements from a heuristically designed optimal monitoring network. The designed network is constrained by the maximum number of permissible monitoring locations. The designed monitoring network improves the results of source identification by choosing monitoring locations that reduces the possibility of missing a pollution source, at the same time decreasing the degree of non uniqueness in the set of possible aquifer responses to subjected geo-chemical stresses. The proposed methodology combines the capability of Genetic Programming (GP), and linked simulation-optimization for recreating the flux history of the unknown conservative pollutant sources with limited number of spatiotemporal pollution concentration measurements. The GP models are trained using large number of simulated realizations of the pollutant plumes for varying input flux scenarios. A selected subset of GP models are used to compute the impact factor and frequency factor of pollutant source fluxes, at candidate monitoring locations, which in turn is used to find the best monitoring locations. The potential application of the developed methodology is demonstrated by evaluating its performance for an illustrative study area. These performance evaluation results show the efficiency in source identification when concentration measurements from the designed monitoring network are utilized.     

A software-based sensor for combined sewer overflows

A new methodology for online estimation of excess flow from combined sewer overflow (CSO) structures based on simulation models is presented. If sufficient flow and water level data from the sewer system is available, no rainfall data are needed to run the model. An inverse rainfall-runoff model was developed to simulate net rainfall based on flow and water level data. Excess flow at all CSO structures in a catchment can then be simulated with a rainfall-runoff model. The method is applied to a case study and results show that the inverse rainfall-runoff model can be used instead of missing rain gauges. Online operation is ensured by software providing an interface to the SCADA-system of the operator and controlling the model. A water quality model could be included to simulate also pollutant concentrations in the excess flow.     

Influence of Time Step Synchronization on Urban Rainfall-Runoff Simulation in a Hybrid CPU/GPU 1D-2D Coupled Model

The 1D sewer - 2D surface coupled hydrodynamic model has increasingly become an essential tool for simulating and predicting the flood process and is widely used in the study of urban rainfall-runoff simulation. The current method of using the smaller time step of the sub model in the coupled model as the synchronization time greatly limits the computational efficiency, especially in the case of the large data amount or models executed in different platforms and in various types of codes. To evaluate the impact of time synchronization on the rainfall-runoff process in a coupled hydrodynamic model, a new model that couples the 2D GPU accelerated shallow water model and the 1D SWMM is applied to two urban catchments to simulate the rainfall-runoff-drainage processes, the fixed time step (5 s, 10 s, 30 s, 60 s, 120 s, 180 s and 300 s) is adopted to ensure the calculation efficiency and precision of the model. The results show that the time computational efficiency can be improved by 7.27%–27.37% in different scenarios compared with the method applying 2D model time step as the synchronization time; the surface runoff process is hardly affected as the synchronization time changes; and the relative error of the drainage process is less than 2.5% when the synchronization time is less than 60 s. Therefore, the fixed synchronization time method is recommended in the 1D-2D coupled model to improve the computational efficiency for flood and inundation simulation. Based on the advantage that the fixed synchronization time is easy to realize in the programming of the model and the high efficiency of the fixed synchronization time method concluded above, this work is expected to provide a reference for model coupling applications.

     

Development and Implementation of Support Vector Machine Regression Surrogate Models for Predicting Groundwater Pumping-Induced Saltwater Intrusion into Coastal Aquifers

Predicting the extent of saltwater intrusion (SWI) into coastal aquifers in response to changing pumping patterns is a prerequisite of any groundwater management framework. This study investigates the feasibility of using support vector machine regression (SVMr), an innovative artificial intelligence-based machine learning algorithm for predicting salinity concentrations at selected monitoring wells in an illustrative aquifer under variable groundwater pumping conditions. For evaluation purpose, the prediction results of SVMr are compared with well-established genetic programming (GP) based surrogate models. SVMr and GP models are trained and validated using identical sets of input (pumping) and output (salinity concentration) datasets. The trained and validated models are then used to predict salinity concentrations at specified monitoring wells in response to new pumping datasets. Prediction capabilities of the two learning machines are evaluated using different proficiency measures to ensure their practicality and generalisation ability. The performance evaluation results suggest that the prediction capability of SVMr is superior to GP models. Also, a sensitivity analysis methodology is proposed for assessing the impact of pumping rates on salt concentrations at monitoring locations. This sensitivity analysis provides a subset of most influential pumping rates, which is used to construct new SVMr surrogate models with improved predictive capabilities. The improved prediction capability and the generalisation ability of the SVMr models together with the ability to improve the accuracy of prediction by refining the input set for training makes the use of proposed SVMr models more attractive. Prediction models with more accurate prediction capability makes it potentially very useful for designing large scale coastal aquifer management strategies.     

Estimation of hydraulic jump on corrugated bed using artificial neural networks and genetic programming

Artificial neural networks (ANNs) and genetic programming (GP) have recently been used for the estimation of hydraulic data. In this study, they were used as alternative tools to estimate the characteristics of hydraulic jumps, such as the free surface location and energy dissipation. The dimensionless hydraulic parameters, including jump depth, jump length, and energy dissipation, were determined as functions of the Froude number and the height and length of corrugations. The estimations of the ANN and GP models were found to be in good agreement with the measured data. The results of the ANN model were compared with those of the GP model, showing that the proposed ANN models are much more accurate than the GP models.     

An investigation of the instrumental variable-approximate maximum likelihood method of modeling and forecasting daily flows

The instrumental variable-approximate maximum likelihood (IV0AML) method provides a technique to develop better models for short-time increment hydrologic data. In this method, a recursive input-output model, which consists of a deterministic model and a stochastic noise model are used. These models handle the system and measurement noise separately. The instrumental variable method has been developed to eliminate the bias in parameter estimates.The IV-AML method is investigated in the present study. Parameters of daily rainfall-runoff models are estimated by the IV-AML and by least squares methods and compared. The effects of a rainfall filter on parameter estimates are also investigated. Forecast accuracies of models whose parameters are estimated by IV-AML and least squares methods are compared.The results indicate that the forecast accuracy of models whose parameters are estimated by least squares method is comparable to that of models whose parameters are estimated by IV-AML method. The rainfall filter, on the other hand, reduces the parameter variation and improves forecasts.     

基于遗传规划的地下水盐分动态模拟

地下水盐分动态研究对认识地下水盐化规律以及合理规划、利用和管理地下水资源具有重要的意义。本文以地下水盐分动态时间序列数据为基础,利用自回归综合移动平均(ARIMA)模型确定输入向量,建立了地下水盐分动态遗传规划模型。以黄河三角洲为例,使用该遗传规划模型对地下水盐分动态进行模拟。为检验GP模型的有效性,与ARIMA模型进行对比。结果表明:地下水盐分动态遗传规划模型适于地下水盐分动态模拟研究,而且模拟精度比单纯使用ARIMA模型有显著提高。     

Prediction of Surface Flow by Forcing of Climate Forecast System Reanalysis Data

Meteorological data are key variables for hydrologists to simulate the rainfall-runoff process using hydrological models. The collection of meteorological variables is sophisticated, especially in arid and semi-arid climates where observed time series are often scarce. Climate Forecast System Reanalysis (CFSR) Data have been used to validate and evaluate hydrological modeling throughout the world. This paper presents a comprehensive application of the Soil and Water Assessment Tool (SWAT) hydrologic simulator, incorporating CFSR daily rainfall-runoff data at the Roodan study site in southern Iran. The developed SWAT model including CFSR data (CFSR model) was calibrated using the Sequential Uncertainty Fitting 2 algorithm (SUFI-2). To validate the model, the calibrated SWAT model (CFSR model) was compared with the observed daily rainfall-runoff data. To have a better assessment, terrestrial meteorological gauge stations were incorporated with the SWAT model (Terrestrial model). Visualization of the simulated flows showed that both CFSR and terrestrial models have satisfactory correlations with the observed data. However, the CFSR model generated better estimates regarding the simulation of low flows (near zero). The results of the uncertainty analysis showed that the CFSR model predicted the validation period more efficiently. This might be related with better prediction of low flows and closer distribution to observed flows. The Nash-Sutcliffe (NS) coefficient provided good- and fair-quality modeling for calibration and validation periods for both models. Overall, it can be concluded that CFSR data might be promising for use in the development of hydrological simulations in arid climates, such as southern Iran, where there are shortages of data and a lack of accessibility to the data.     

Modeling Evaporation-Seepage Losses for Reservoir Water Balance in Semi-arid Regions

In the water balance of reservoir system, evaporation plays a crucial role particularly so for the reservoir systems of smaller size located in the semi-arid or arid regions. Such regions are most often characterized by significant seepage losses from reservoirs, besides evaporation losses. Usually, in the optimization of a reservoir system, it is a common practice to assume evaporation loss either as some constant value or as negligible. Such assumptions, however, may affect the results of reservoir optimization. This is demonstrated in this study by a case study in the optimal scheduling of Pilavakkal reservoir system in Vaipar basin of Tamilnadu, India. For modeling reservoir losses, many models are available, of which, Penman combination model is most commonly used. In this study, an alternative approach based on Genetic Programming (GP) is proposed. The results of GP and Penman model for both evaporation loss estimation and reservoir scheduling are compared. It is found that while GP and Penman combination model performs equally well for estimating evaporation losses, GP is also able to model seepage losses (or other losses from reservoir) to a much better degree. It is also shown the reservoir scheduling does get influenced based on how the reservoir losses are modeled in the reservoir water balance equation.     

Soft Computing Techniques for Rainfall-Runoff Simulation: Local Non–Parametric Paradigm vs. Model Classification Methods

Accurate simulation of rainfall-runoff process is of great importance in hydrology and water resources management. Rainfall–runoff modeling is a non-linear process and highly affected by the inputs to the simulation model. In this study, three kinds of soft computing methods, namely artificial neural networks (ANNs), model tree (MT) and multivariate adaptive regression splines (MARS), have been employed and compared for rainfall-runoff process simulation. Moreover, this study investigates the effect of input size, including number of input variables and number of data time series on runoff simulation by the developed models. Inputs to the simulation models for calibration and validation purposes consist two parts: I1: five variables, including daily rainfall and runoff time series (30 years) with lag times, and I2: twelve variables, including daily rainfall and runoff time series (10 years). To increase the model performances, optimal number and type for input variables are identified. The efficiency of the training and testing performances using the ANNs, MT and MARS models is then evaluated using several evaluation criteria. To implement the methodology, Tajan catchment in the northern part of Iran is selected. Based on the results, it was found that using I1 as input to the developed models results in higher simulation performance. The results also provided evidence that MT (R = 0.897, RMSE = 6.70, RSE = 0.33) with set I2 is capable of reliable model for rainfall-runoff process compared with MARS (R = 0.892, RMSE = 7.47, RSE = 0.83) and ANNs (R = 0.884, RMSE = 7.40, RSE = 0.43) models. Therefore, size (length of data time series) and type of input variables have significant effects on the modeling results.     

Modelling Rainfall Runoff Relations Using HEC-HMS and IHACRES for a Single Rain Event in an Arid Region of Jordan

The HEC-HMS and IHACRES rainfall runoff models were applied to simulate a single streamflow event in Wadi Dhuliel arid catchment that occurred on 30–31/01/2008. Streamflow estimation was performed on the basis of an hourly scale. The aim of this study was to develop a new framework of rainfall-runoff model applications in arid catchments by integrating a re-adjusted satellite-derived rainfall dataset (GSMaP_MVK+) to determine the location of the rainfall storm. The HEC-HMS model was applied using the HEC-GeoHMS extension in ArcView 3.3 while the IHACRES is Java-based version model. The HEC-HMS model input data include soil type, land use/land cover, and slope. By contrast, the lumped model IHACRES was also applied, based on hourly rainfall and temperature data. Both models were calibrated and validated using the observed streamflow data set collected at Al-Za’atari discharge station. The performance of IHACRES showed some weaknesses, while the flow comparison between the calibrated streamflow results fits well with the observed streamflow data in HEC-HMS. The Nash-Sutcliffe efficiency (E_f) for the two models was 0.51 and 0.88 respectively.