Including spatial distribution in a data‐driven rainfall‐runoff model to improve reservoir inflow forecasting in Taiwan

Multi‐step ahead inflow forecasting has a critical role to play in reservoir operation and management in Taiwan during typhoons as statutory legislation requires a minimum of 3‐h warning to be issued before any reservoir releases are made. However, the complex spatial and temporal heterogeneity of typhoon rainfall, coupled with a remote and mountainous physiographic context, makes the development of real‐time rainfall‐runoff models that can accurately predict reservoir inflow several hours ahead of time challenging. Consequently, there is an urgent, operational requirement for models that can enhance reservoir inflow prediction at forecast horizons of more than 3 h. In this paper, we develop a novel semi‐distributed, data‐driven, rainfall‐runoff model for the Shihmen catchment, north Taiwan. A suite of Adaptive Network‐based Fuzzy Inference System solutions is created using various combinations of autoregressive, spatially lumped radar and point‐based rain gauge predictors. Different levels of spatially aggregated radar‐derived rainfall data are used to generate 4, 8 and 12 sub‐catchment input drivers. In general, the semi‐distributed radar rainfall models outperform their less complex counterparts in predictions of reservoir inflow at lead times greater than 3 h. Performance is found to be optimal when spatial aggregation is restricted to four sub‐catchments, with up to 30% improvements in the performance over lumped and point‐based models being evident at 5‐h lead times. The potential benefits of applying semi‐distributed, data‐driven models in reservoir inflow modelling specifically, and hydrological modelling more generally, are thus demonstrated. Copyright © 2012 John Wiley & Sons, Ltd.     

Real-time assimilation of streamflow observations into a hydrological routing model: effects of model structures and updating methods

This paper comparatively assesses the performance of five data assimilation techniques for three-parameter Muskingum routing with a spatially lumped or distributed model structure. The assimilation techniques used include direct insertion (DI), nudging scheme (NS), Kalman filter (KF), ensemble Kalman filter (EnKF) and asynchronous ensemble Kalman filter (AEnKF), which are applied to river reaches in Texas and Louisiana, USA. For both lumped and distributed routing, results from KF, EnKF and AEnKF are sensitive to the error specification. As expected, DI outperformed the other models in the case of lumped modelling, while in distributed routing, KF approaches, particularly AEnKF and EnKF, performed better than DI or nudging, reflecting the benefit of updating distributed states through error covariance modelling in KF approaches. The results of this work would be useful in setting up data assimilation systems that employ increasingly abundant real-time observations using distributed hydrological routing models.     

Integration of gauge and radar rainfall to enable best simulation of hydrological parameters

This study is about use of spatially distributed rain in physically based hydrological models. In recent years, spatially distributed radar rainfall data have become available. The distributed radar rain is used to precisely model hydrologic processes and it is more realistic than the past practice of distribution methods like Thiessen polygons. Radar provides a highly accurate spatial distribution of rainfall and greatly improves the basin average rainfall estimates. However, quantification of the exact amount of rainfall from radar observation is relatively difficult. Thus, the fundamental idea of this study is to apply hourly gauge and radar rainfall data in a distributed hydrological model to simulate hydrological parameters. Hence the comparison is made between the outcomes of the WetSpa model from radar rainfall distribution and gauge rainfall distributed by the Thiessen polygon technique. The comparative plots of the hydrograph and the results of hydrological components such as evapotranspiration, surface runoff, soil moisture, recharge and interflow, reflect the spatially distributed radar input performing well for model outflow simulation.

EDITOR D. Koutsoyiannis

ASSOCIATE EDITOR F. Pappenberger     

A spatially distributed Clark’s unit hydrograph based hybrid hydrologic model (Distributed-Clark)

ABSTRACT

A hybrid hydrologic model (Distributed-Clark), which is a lumped conceptual and distributed feature model, was developed based on the combined concept of Clark’s unit hydrograph and its spatial decomposition methods, incorporating refined spatially variable flow dynamics to implement hydrological simulation for spatially distributed rainfall–runoff flow. In Distributed-Clark, the Soil Conservation Service (SCS) curve number method is utilized to estimate spatially distributed runoff depth and a set of separated unit hydrographs is used for runoff routing to obtain a direct runoff flow hydrograph. Case studies (four watersheds in the central part of the USA) using spatially distributed (Thiessen polygon-based) rainfall data of storm events were used to evaluate the model performance. Results demonstrate relatively good fit to observed streamflow, with a Nash-Sutcliffe efficiency (E_NS) of 0.84 and coefficient of determination (R²) of 0.86, as well as a better fit in comparison with outputs of spatially averaged rainfall data simulations for two models including HEC-HMS.     

Investigating the use of spatial discretization of hydrological processes in conceptual rainfall runoff modelling: a case study for the meso‐scale

In this study a simple modelling approach was applied to identify the need for spatial complexity in representing hydrological processes and their variability over different scales. A data set of 18 basins was used, ranging between 8 and 4011 km² in area, located in the Nahe basin (Germany), with daily discharge values for over 30 years. Two different parsimoniously structured models were applied in lumped as well as in spatially distributed according to two distribution classifications: (1) a simple classification based on the lithology expressed in three permeability types and (2) a more complex classification based on seven dominating runoff production processes. The objective of the study was to compare the performances of the models on a local and on a regional scale as well as between the models with a view to identifying the accuracy in capturing the spatial variability of the rainfall‐runoff relationships. It was shown that the presence of a specific basin characteristic or process of the distribution classification was not related with higher model performance; only a larger basin size promoted higher model performance. The results of this study also indicated that the permeability generally contained more useful information on the spatial heterogeneity of the hydrological behaviour of the natural system than did a more detailed classification on dominating runoff generation processes. Although model performance was slightly lower for the model that used permeability as a distribution classification, consistency in its parameter values was found, which was lacking with the more complex distribution classification. The latter distribution classification had a higher flexibility to optimize towards the variability of the runoff, which resulted in higher performance, however, process representation was applied inconsistently. Copyright © 2007 John Wiley & Sons, Ltd.     

Assessment on the function of reservoirs for flood control during typhoon seasons based on a distributed hydrological model

A grid‐based distributed hydrological model, PDTank model, is used to simulate hydrological processes in the upper Tone River catchment. The Tone River catchment often suffers from heavy rainfall events during the typhoon seasons. The reservoirs located in the catchment play an important role in flood regulation. Through the coupling of the PDTank model and a reservoir module that combines the storage function and operation function, the PDTank model is used for flood forecasting in this study. By comparing the hydrographs simulated using gauging and radar rainfall data, it is found that the spatial variability of rainfall is an important factor for flood simulation and the accuracy of the hydrographs simulated using radar rainfall data is slightly improved. The simulation of the typhoon flood event numbered No. 9 shows that the reservoirs in the catchment attenuate the peak flood discharge by 423·3 m³/s and validates the potential applicability of the distributed hydrological model on the assessment of function of reservoirs for flood control during typhoon seasons. Copyright © 2011 John Wiley & Sons, Ltd.     

Incorporating the spatio-temporal distribution of rainfall and basin geomorphology into nonlinear analyses of streamflow dynamics

Many recent studies have been devoted to the investigation of the nonlinear dynamics of rainfall or streamflow series based on methods of dynamical systems theory. Although finding evidence for the existence of a low-dimensional deterministic component in rainfall or streamflow is of much interest, not much attention has been given to the nonlinear dependencies of the two and especially on how the spatio-temporal distribution of rainfall affects the nonlinear dynamics of streamflow at flood time scales. In this paper, a methodology is presented which simultaneously considers streamflow series, spatio-temporal structure of precipitation and catchment geomorphology into a nonlinear analysis of streamflow dynamics. The proposed framework is based on “hydrologically-relevant” rainfall-runoff phase-space reconstruction acknowledging the fact that rainfall-runoff is a stochastic spatially extended system rather than a deterministic multivariate one. The methodology is applied to two basins in Central North America using 6-hour streamflow data and radar images for a period of 5 years. The proposed methodology is used to: (a) quantify the nonlinear dependencies between streamflow dynamics and the spatio-temporal dynamics of precipitation; (b) study how streamflow predictability is affected by the trade-offs between the level of detail necessary to explain the spatial variability of rainfall and the reduction of complexity due to the smoothing effect of the basin; and (c) explore the possibility of incorporating process-specific information (in terms of catchment geomorphology and an a priori chosen uncertainty model) into nonlinear prediction. Preliminary results are encouraging and indicate the potential of using the proposed methodology to understand via nonlinear analysis of observations (i.e., not based on a particular rainfall-runoff model) streamflow predictability and limits to prediction as a function of the complexity of spatio-temporal forcing relative to basin geomorphology.     

A test of TOPMODEL'a ability to predict spatially distributed groundwater levels

TOPMODEL was calibrated to a small catchment using precipitation and runoff data. Acceptable fits of simulated and observed runoff were obtained during both the calibration and validation periods. Predictions of groundwater levels using this calibration did not agree well with observations at the 37 points within the catchment where groundwater levels were measured, including three locations with continuous recordings. Groundwater level observations at one single point in time, however, sufficed to calibrate new topographic–soil indices that improved the prediction of the local groundwater levels at the observed tubes. This suggests that spatially distributed calibration data are necessary to exploit reliably TOPMODEL's ability to predict spatially distributed hydrology. The mean or recalibrated transmissivity values at these 37 points differed from the catchment mean as determined by the precipitation–runoff calibration. Thus, while groundwater information can help in predicting groundwater levels at specific locations, increasing the number of local groundwater level measurements is not sufficient to improve the spatially distributed representation of subsurface flow by TOPMODEL for the catchment as a whole, as long as the groundwater information is not integrated in the precipitation–runoff calibration. © 1997 John Wiley & Sons, Ltd.     

Observational and predictive uncertainties for multiple variables in a spatially distributed hydrological model

In this study, uncertainty in model input data (precipitation) and parameters is propagated through a physically based, spatially distributed hydrological model based on the MIKE SHE code. Precipitation uncertainty is accounted for using an ensemble of daily rainfall fields that incorporate four different sources of uncertainty, whereas parameter uncertainty is considered using Latin hypercube sampling. Model predictive uncertainty is assessed for multiple simulated hydrological variables (discharge, groundwater head, evapotranspiration, and soil moisture). Utilizing an extensive set of observational data, effective observational uncertainties for each hydrological variable are assessed. Considering not only model predictive uncertainty but also effective observational uncertainty leads to a notable increase in the number of instances, for which model simulation and observations are in good agreement (e.g., 47% vs. 91% for discharge and 0% vs. 98% for soil moisture). Effective observational uncertainty is in several cases larger than model predictive uncertainty. We conclude that the use of precipitation uncertainty with a realistic spatio‐temporal correlation structure, analyses of multiple variables with different spatial support, and the consideration of observational uncertainty are crucial for adequately evaluating the performance of physically based, spatially distributed hydrological models.     

Cokriging for enhanced spatial interpolation of rainfall in two Australian catchments

Rainfall data in continuous space provide an essential input for most hydrological and water resources planning studies. Spatial distribution of rainfall is usually estimated using ground‐based point rainfall data from sparsely positioned rain‐gauge stations in a rain‐gauge network. Kriging has become a widely used interpolation method to estimate the spatial distribution of climate variables including rainfall. The objective of this study is to evaluate three geostatistical (ordinary kriging [OK], ordinary cokriging [OCK], kriging with an external drift [KED]), and two deterministic (inverse distance weighting, radial basis function) interpolation methods for enhanced spatial interpolation of monthly rainfall in the Middle Yarra River catchment and the Ovens River catchment in Victoria, Australia. Historical rainfall records from existing rain‐gauge stations of the catchments during 1980–2012 period are used for the analysis. A digital elevation model of each catchment is used as the supplementary information in addition to rainfall for the OCK and kriging with an external drift methods. The prediction performance of the adopted interpolation methods is assessed through cross‐validation. Results indicate that the geostatistical methods outperform the deterministic methods for spatial interpolation of rainfall. Results also indicate that among the geostatistical methods, the OCK method is found to be the best interpolator for estimating spatial rainfall distribution in both the catchments with the lowest prediction error between the observed and estimated monthly rainfall. Thus, this study demonstrates that the use of elevation as an auxiliary variable in addition to rainfall data in the geostatistical framework can significantly enhance the estimation of rainfall over a catchment.     

MODELLING RUNOFF GENERATION ON SMALL AGRICULTURAL CATCHMENTS: CAN REAL WORLD RUNOFF RESPONSES BE CAPTURED?

This paper focuses on the problem of quantifying real world catchment response using a distributed model and discusses the ability of the model to capture that response. The rainfall–runoff responses of seven small agricultural catchments in the eastern wheatbelt region of south-western Australia are examined. The variability in runoff generation and the factors that contribute to that variability (i.e. rainfall intensity, soil properties and topography) are investigated to determine if their influence can be captured in a mathematical model. The spatially distributed rainfall–runoff model used in this study is based on the TOPMODEL concepts of Beven and Kirkby (1979), and simulates runoff generation by both the infiltration excess and saturation excess mechanisms. Simulations with the model revealed the highly complex nature of catchment response to rainfall events. Runoff generation was highly heterogeneous in both space and time, with the runoff response being governed by the spatial variability of soil properties and topography, and by the temporal variation in rainfall intensity. Although the model proved capable of simulating catchment response for many events, the investigation has demonstrated that not all aspects of the variability associated with agricultural catchments (particularly the effects of land management) can be captured using this relatively simple model. © 1997 by John Wiley & Sons, Ltd     

On the use of multi‐algorithm,genetically adaptive multi‐objective method for multi‐site calibration of the SWAT model

With the availability of spatially distributed data, distributed hydrologic models are increasingly used for simulation of spatially varied hydrologic processes to understand and manage natural and human activities that affect watershed systems. Multi‐objective optimization methods have been applied to calibrate distributed hydrologic models using observed data from multiple sites. As the time consumed by running these complex models is increasing substantially, selecting efficient and effective multi‐objective optimization algorithms is becoming a nontrivial issue. In this study, we evaluated a multi‐algorithm, genetically adaptive multi‐objective method (AMALGAM) for multi‐site calibration of a distributed hydrologic model—Soil and Water Assessment Tool (SWAT), and compared its performance with two widely used evolutionary multi‐objective optimization (EMO) algorithms (i.e. Strength Pareto Evolutionary Algorithm 2 (SPEA2) and Non‐dominated Sorted Genetic Algorithm II (NSGA‐II)). In order to provide insights into each method's overall performance, these three methods were tested in four watersheds with various characteristics. The test results indicate that the AMALGAM can consistently provide competitive or superior results compared with the other two methods. The multi‐method search framework of AMALGAM, which can flexibly and adaptively utilize multiple optimization algorithms, makes it a promising tool for multi‐site calibration of the distributed SWAT. For practical use of AMALGAM, it is suggested to implement this method in multiple trials with relatively small number of model runs rather than run it once with long iterations. In addition, incorporating different multi‐objective optimization algorithms and multi‐mode search operators into AMALGAM deserves further research. Copyright © 2009 John Wiley & Sons, Ltd.     

Modelling long‐term contaminant migration in a catchment at fine spatial and temporal scales using the UP system

A method has been developed to simulate the long‐term migration of radionuclides in the near‐surface of a river catchment, following their release from a deep underground repository for radioactive waste. Previous (30‐year) simulations, conducted using the SHETRAN physically based modelling system, showed that long‐term (many decades) simulations are required to allow the system to reach steady state. Physically based, distributed models, such as SHETRAN, tend to be too computationally expensive for this task. Traditional lumped catchment‐scale models, on the other hand, do not give sufficiently detailed spatially distributed results. An intermediate approach to modelling has therefore been developed which allows flow and transport processes to be simulated with the spatial resolution normally associated with distributed models, whilst being computationally efficient.The approach involves constructing a lumped model in which the catchment is represented by a number of conceptual water storage compartments. The flow rates to and from these compartments are prescribed by functions that summarize the results from physically based distributed models run for a range of characteristic flow regimes. The physically based models used were, SHETRAN for the subsurface compartments, a particle tracking model for overland flow and an analytical model for channel routing. One important advantage of the method used in constructing the lumped model is that it makes down scaling possible, in the sense that fine‐scale information on the distributed hydrological regime, as simulated by the physically based distributed models, can be inferred from the variables in the lumped model that describe the hydrology at the catchment scale. A 250‐year flow simulation has been run and the down scaling process used to infer a 250‐year time‐series of three‐dimensional velocity fields for the subsurface of the catchment. This series was then used to drive a particle tracking simulation of contaminant migration. The concentration and spatial distribution of contaminants simulated by this model for the first 30 years were in close agreement with SHETRAN results. The remaining 220 years highlighted the fact that some of the most important transport pathways to the surface carry contaminants only very slowly so both the magnitude and spatial distribution of concentration in surface soils are not apparent over the shorter SHETRAN simulations. Copyright © 1999 John Wiley & Sons, Ltd.     

Effect of rainfall uncertainty on the performance of physically based rainfall–runoff models

This paper analyses the effect of rain data uncertainty on the performance of two hydrological models with different spatial structures: a semidistributed and a fully distributed model. The study is performed on a small catchment of 19.6 km² located in the north‐west of Spain, where the arrival of low pressure fronts from the Atlantic Ocean causes highly variable rainfall events. The rainfall fields in this catchment during a series of storm events are estimated using rainfall point measurements. The uncertainty of the estimated fields is quantified using a conditional simulation technique. Discharge and rain data, including the uncertainty of the estimated rainfall fields, are then used to calibrate and validate both hydrological models following the generalized likelihood uncertainty estimation (GLUE) methodology. In the storm events analysed, the two models show similar performance. In all cases, results show that the calibrated distribution of the input parameters narrows when the rain uncertainty is included in the analysis. Otherwise, when rain uncertainty is not considered, the calibration of the input parameters must account for all uncertainty in the rainfall–runoff transformation process. Also, in both models, the uncertainty of the predicted discharges increase in similar magnitude when the uncertainty of rainfall input increase.     

Internal catchment process simulation in a snow‐dominated basin: performance evaluation with spatiotemporally variable runoff generation and groundwater dynamics

Hydrologic models have increasingly been used in forest hydrology to overcome the limitations of paired watershed experiments, where vegetative recovery and natural variability obscure the inferences and conclusions that can be drawn from such studies. Models are also plagued by uncertainty, however, and parameter equifinality is a common concern. Physically‐based, spatially‐distributed hydrologic models must therefore be tested with high‐quality experimental data describing a multitude of concurrent internal catchment processes under a range of hydrologic regimes. This study takes a novel approach by not only examining the ability of a pre‐calibrated model to realistically simulate watershed outlet flows over a four year period, but a multitude of spatially‐extensive, internal catchment process observations not previously evaluated, including: continuous groundwater dynamics, instantaneous stream and road network flows, and accumulation and melt period spatial snow distributions. Many hydrologic model evaluations are only on the comparison of predicted and observed discharge at a catchment outlet and remain in the ‘infant stage’ in terms of model testing. This study, on the other hand, tests the internal spatial predictions of a distributed model with a range of field observations over a wide range of hydroclimatic conditions. Nash‐Sutcliffe model efficiency was improved over prior evaluations due to continuing efforts in improving the quality of meteorological data collection. Road and stream network flows were generally well simulated for a range of hydrologic conditions, and snowpack spatial distributions were well simulated for one of two years examined. The spatial variability of groundwater dynamics was effectively simulated, except at locations where strong stream–groundwater interactions exist. Model simulations overall were quite successful in realistically simulating the spatiotemporal variability of internal catchment processes in the watershed, but the premature onset of simulated snowmelt for one of the simulation years has prompted further work in model development. Copyright © 2011 John Wiley & Sons, Ltd.     

SWAT模型在斯里兰卡河流径流预测中的运用

本文运用SWAT模型和新安江模型对斯里兰卡卡鲁河流域上游地区日径流进行了预测.卡鲁河是斯里兰卡的第二大河,由于流域的降雨量很大,上游地区河流沿峡谷流下,中下游平原地区河床平坦.卡鲁河流域的洪水变的很正常.应用SWAT模型来对卡鲁河的日径流量进行预测,并同应用新安江模型所得到的结果做对比.研究表明,新安江模型要比SWAT (分布式水文模型)模型在卡鲁河日径流量预测上稍微好一些.实际上,或许数据质量不高或不恰当是部分原因,因为SWAT的输出成果严格取决于其输入的数据质量.此外,在斯里兰卡,许多人的日常用水是靠井水.当把流域看作一个整体,通常都是一个很大的范围,那样的话就不可能详尽的记录所有各个小规模的水利用,例如:小灌溉、小规模的家畜管理和工业水利用.这些水利用累积起来或许就很可观.这些数据的缺失对分布式水文模型在水平衡的应用有着独特的影响.但是概念水文模型(如新安江模型)可以根据实际情况在校正中调节它的参数,因为这些参数并没有实质的物理含义.因此,在流域特征和模型输入数据有限或不完整的情况下,概念水文模型比分布式水文模型更具优势.     

Comparative modelling of two catchments in Taiwan and England

A comparative modelling of two catchments of similar sizes in Taiwan and England is described. In the study, despite its success in many Taiwanese catchments, including the Yan‐Shui Creek catchment in this study, the distributed model GBDM was initially found unsuitable when applied to the Brue catchment in South West England. However, the simulations are much better after revising the infiltration capacity. Further exploration reveals several interesting findings. (1) The infiltration computation based on soil characteristics and classifications is unreliable in the model. Other factors, such as climate, farming practice and vegetation cover, could have a much more significant impact. (2) The application of the GBDM far away from its ‘home country’ unveils a possible weakness of such a model for being ‘underfitting’. The fact that an ‘adjustment factor’ added in the model could improve both its calibration and validation may indicate that there is a room to improve the GBDM structure for catchments outside Taiwan. (3) The study illustrates the difficulty in creating a universal distributed model that could suit all possible hydrological environments, under the constraint of model parameter parsimony to minimize the ‘equifinality’ problem. Copyright © 2006 John Wiley & Sons, Ltd.     

Development and application of a spatially‐distributed Arctic hydrological and thermal process model (ARHYTHM)

A process‐based, spatially distributed hydrological model was developed to quantitatively simulate the energy and mass transfer processes and their interactions within arctic regions (arctic hydrological and thermal model, ARHYTHM). The model first determines the flow direction in each element, the channel drainage network and the drainage area based upon the digital elevation data. Then it simulates various physical processes: including snow ablation, subsurface flow, overland flow and channel flow routing, soil thawing and evapotranspiration. The kinematic wave method is used for conducting overland flow and channel flow routing. The subsurface flow is simulated using the Darcian approach. The energy balance scheme was the primary approach used in energy‐related process simulations (snowmelt and evapotranspiration), although there are options to model snowmelt by the degree‐day method and evapotranspiration by the Priestley–Taylor equation. This hydrological model simulates the dynamic interactions of each of these processes and can predict spatially distributed snowmelt, soil moisture and evapotranspiration over a watershed at each time step as well as discharge in any specified channel(s). The model was applied to Imnavait watershed (about 2·2 km²) and the Upper Kuparuk River basin (about 146 km²) in northern Alaska. Simulated results of spatially distributed soil moisture content, discharge at gauging stations, snowpack ablations curves and other results yield reasonable agreement, both spatially and temporally, with available data sets such as SAR imagery‐generated soil moisture data and field measurements of snowpack ablation, and discharge data at selected points. The initial timing of simulated discharge does not compare well with the measured data during snowmelt periods mainly because the effect of snow damming on runoff was not considered in the model. Results from the application of this model demonstrate that spatially distributed models have the potential for improving our understanding of hydrology for certain settings. Finally, a critical component that led to the performance of this modelling is the coupling of the mass and energy processes. Copyright © 2000 John Wiley & Sons, Ltd.