Journal of Hydrologic Engineering

This study reports on the performance of two medium-range streamflow forecast models: (1) a multilayer feed-forward artificial neural network; and (2) a distributed hydrologic model. Quantitative precipitation forecasts were used as input to both models. The Furnas Reservoir on the Rio Grande River was selected as a case study, primarily because of the availability of quantitative precipitation forecasts from the Brazilian Center for Weather Prediction and Climate Studies and due to its importance in the Brazilian hydropower generating system. Streamflow forecasts were calculated for a drainage area of about $\mathrm{51,900}{\mathrm{km}}^2$ , with lead times up to $12\text{days}$ , at daily intervals. The Nash–Sutcliffe efficiency index, the root-mean-square error, the mean absolute error, and the mean relative error were used to assess the relative performance of the models. Results showed that the performance of streamflow forecasts was strongly dependent on the quality of quantitative precipitation forecasts used. The artificial neural network (ANN) method seemed to be less sensitive to precipitation forecast error relative to the distributed hydrological model. Hence, the latter presented a better skill in flow forecasting when using the more accurate perfect precipitation forecast. The conceptual hydrological model also demonstrates better forecast skill than ANN models for longer lead times, when the representation of the rainfall-runoff process and of the water storage in the watershed becomes more important than the flow routing along the drainage network. In addition, results obtained by incorporating a quantitative precipitation forecast in both models performed better than the current streamflow obtained by the Brazilian national electric system operator using statistical models which do not utilize information on precipitation, whether observed or forecast.

In this study an effective procedure is developed to evaluate the aquifer parameters from the late time-drawdown data observed both at pumping and observation wells in confined aquifers. For this aim, the time-drawdown data graph is considered by means of the MS Excel spreadsheets and later, its regression line and equation for the late time-drawdown data are obtained. The data after the minimum pumping duration criterion valid for both pumping and observation wells, in the case of flow toward a large-diameter well, by M. Çimen in 2006 and after the some early drawdown for observation wells, in the absence of the pumping well data, are considered in the calculations. It is observed that this late time-drawdown data yield the straight line in the logarithmic function. The transmissivity of the confined aquifer is then obtained from the slope value of the regression equation derived for late time data. The storage coefficient is also calculated with the Cooper–Jacob equation. In the Cooper–Jacob equation, the time intercepting zero drawdown is easily found from this regression equation. Finally, the application of the methodology has been tested for some data sets of the pumping and observation wells and the well storage effect has been taken, in some applications, into account.

In the core of an integrated watershed model there is coupling between surface water and subsurface water flows. Recently, interest in hydrology literature, regarding the fully coupled approach for surface and subsurface water interactions, has increased. For example, the assumption of a gradient-type flux equation, based on Darcy’s law and the numerical solution of all governing equations in a single global matrix, has been reported. This paper argues that this “fully coupled approach” is only a special case of all possible coupling combinations and, if not applied with caution, the nonphysics interface parameter becomes a calibration tool. Generally, there are two cases of surface/subsurface coupling based on the physical nature of the interface: continuous or discontinuous assumption; when a sediment layer exists at the interface, the discontinuous assumption may be justified. As for numerical schemes, there are three cases: time lagged, iterative, and simultaneous solutions. Since modelers often resort to the simplest, fastest schemes in practical applications, it is desirable to quantify potential errors and the performance specific to each coupling scheme. This paper evaluates these coupling schemes in a watershed model, WASH123D, with numerical experiments. They are designed to compare the performance of each coupling approach for different types of surface water and subsurface interactions. These experiments are evaluated in terms of surface water and subsurface water solutions, along with exchange flux (e.g. infiltration/seepage rate).

The genetic algorithm (GA) technique is applied to obtain optimal parameter values of the standard rating curve model (RCM) for predicting, in real time, event-based flow discharge hydrographs at sites receiving significant lateral inflows. The standard RCM uses the information of discharge and effective cross-sectional flow area at an upstream station and effective cross-sectional flow area wave travel time later at a downstream station to predict the flow rate at this last site. The GA technique obtains the optimal parameter values of the model, here defined as the GA-RCM model, by minimizing the mean absolute error objective function. The GA-RCM model was tested to predict hydrographs at three different stations, located on the Upper Tiber River in central Italy. The wave travel times characterizing the three selected river branches are, on the average, 4, 8, and $12\mathrm{h}$ . For each river reach, seven events were employed, four for the model parameters’ calibration and three for model testing. The GA approach, employing 100 chromosomes in the initial gene pool, 75% crossover rate, 5% mutation rate, and 10,000 iterations, made the GA-RCM model successfully simulate the hydrographs observed at each downstream section closely capturing the trend, time to peak, and peak rates with, on the average, less than 5% error. The model performance was also tested against the standard RCM model, which uses, on the contrary to the GA-RCM model, different values for the model parameters and wave travel time for each event, thus, making the application of the standard RCM for real time discharge monitoring inhibited. The comparative results revealed that the RCM model improved its performance by using the GA technique in estimating parameters. The sensitivity analysis results revealed that at most two events would be sufficient for the GA-RCM model to obtain the optimal values of the model parameters. A lower peak hydrograph can also be employed in the calibration to predict a higher peak hydrograph. Similarly, a shorter travel time hydrograph can be used in GA to obtain optimal model parameters that can be used to simulate floods characterized by longer travel time. For its characteristics, the GA-RCM model is suitable for the monitoring of discharge in real time, at river sites where only water levels are observed.

Most operational flood alert systems (FASs) are based on real-time observations of rainfall, runoff, and simultaneous simulation of hydrologic and hydraulic models that are often augmented with weather forecast data. However, implementation of such a system can be cost prohibitive. Alternatively, a model FAS (MFAS) can effectively be developed for a channel section using reporting stream gauges measured against frequency-based design storms and unsteady-state hydraulic models. A MFAS is based on the principle of stage-to-stage crest relationship of river forecasting and in principle is similar to model floodplain maps produced on the basis of frequency-based design storms with the applications of steady-state hydraulic models. A MFAS is based on the predictable nature of the hydraulic response of channel cross sections and the hypothesis that sufficient warning time can be derived from quantitative estimation of flood wave propagation time from upstream to downstream. Based on these concepts, a model, yet real-time, FAS is developed specifically for the Houston Theater District, where heavy flooding during Tropical Storm Allison in 2001 caused significant structural, equipment, and records damage, as well as loss of lives.

Because of differences in roughness, flow depth, and slope, the timing of runoff from most parts of a watershed differs from that along the principal flow path, which is generally used to compute times of concentration. Thus, times of concentration computed from principal flow path information do not accurately reflect the timing of runoff from most parts of a watershed. Specifically, travel times based on the principal flow path will be shorter than times based on the entire watershed, which could result in inaccurate computed peak discharge rates. This potential bias in watershed runoff timing was investigated using both measured rainfall—runoff data from 27 watersheds and computer simulation. Multiple definitions of the time of concentration were compared, with the expected uncertainty in the alternative definitions evaluated. The difference in timing between the centers of mass of rainfall excess and direct runoff was found to be the least uncertain timing parameter while the traditional definition of time of concentration, i.e., the difference between the end of rainfall excess and the inflection point, was found to be highly uncertain. An adjustment factor that can be applied to times of concentration computed solely from the principal flow path is provided to eliminate the bias inherent to times computed from information obtained from the principal flow path.

The mixing of water, solutes, and sediments from two or more tributaries produces streamflow possessing unique biogeochemical, ecological, hydrological, and morphological signatures. To understand and predict these signatures, confluence locations within river networks are commonly used to divide a watershed into individual modeling units, within these units hydrologic characteristics are spatially averaged and used to parameterize process models. In this context, model parameters from an individual unit directly influence the simulated hydrograph at its corresponding downstream junction and reach. Previous researchers have shown complex changes in flow hydraulics, stream geometry, and channel morphology in the downstream direction of a drainage network. However, the uncertainty of model predictions at and between junction locations due to parameter variability within individual modeling units is poorly understood. This paper presents a methodology for identifying stream junctions and reaches that are potentially vulnerable to increased model uncertainty due to the structural arrangement of their upstream drainage network and model parameter variability. A case study, using coastal southern California, is presented to illustrate the concepts and explore the relationships between changes in simulated peak discharge at junctions and along reaches due to expected model parameter variability for varying drainage network characteristics. Based on simulation results for 300 different drainage network configurations, eight model parameter scenarios, and three design storms (7,200 simulations), the uncertainty in simulated peak discharge within a reach is related to the drainage network characteristics. In the study region, 18% of the reaches are shown to be especially vulnerable to model parameter uncertainty, with 32% of those researches crossed by a major road and/or railroad. This approach is intended to improve risk assessment in sensitive hydrologic applications, such as the design of bridges, culverts, and channels to convey flood flows.

Evapotranspiration (ET) is likely to be greatly affected by global warming because of the dependence of ET on surface temperature. The increasing atmospheric concentration of carbon dioxide $\left(\mathrm{C}{\mathrm{O}}_2\right)$ and other greenhouse gases is expected to increase precipitation and evaporation proportionally. However, a few studies have shown a decreasing trend for evaporation over the last $50\text{years}$ globally. In India, earlier works showed that there was a significant increasing temporal trend in surface temperature and a decreasing trend in grass reference ET (ETo). To study the temporal trend of ETo along with its regionwise spatial variation, $32\text{years}$ (1971–2002) monthly meteorological data were collected for 133 selected stations evenly distributed over different agro-ecological regions (AERs) of India. ETo was estimated by the globally accepted Food and Agriculture Organization (FAO) Penman Monteith (PM) method (FAO-56 PM). These ETo values were then analyzed by a nonparametric Mann–Kendall (MK) test (with modified effective sample size approach for serially correlated data) and Sen slope to determine the existence and magnitude of any statistically significant trend over the time period considered in this study. The same analysis was also performed on governing meteorological variables to identify the cause of existence of such trend in ETo. A significant decreasing trend was found in ETo all over India during the study period, which was mainly caused by a significant increase in the relative humidity and a consistent significant decrease in the wind speed throughout the country. However, a general increase in rainfall was not found in recent years.

This paper proposes a stochastic procedure for generating seasonal flows at multiple locations simultaneously. The proposed procedure uses the technique of singular value decomposition (SVD) to transform a historical flow-data matrix into its standardized principal components (SPCs), and then applies the simple Kaiser’s cut-off rule to retain only the significant SPCs (SSPCs) for further fitting the multivariate autoregressive scheme of order one [MAR(1)] to them. It is shown that the Kaiser’s rule is proper for giving the size of the SSPCs of multisite seasonal-flow records because application of the rule yields only a few more SSPCs for each additional site, and its dimension is reasonably definitive. The comparison of the proposed SVD/MAR(1) procedure with the existing one has demonstrated that the proposed procedure has less parameters than the existing one due to taking into consideration the SSPCs only. In addition, the ability of the parsimonious procedure to reproduce adequately the historical key statistics, storage-, and drought-related characteristics at considered monthly and successive aggregation levels, still remains the same as the existing candidate.

To design urban storm water infrastructures, hydrologists apply the SCS Types I, II, and III $24\text{-}\mathrm{h}$ rainfall distribution curves to simulate the design rainfall hyetographs by which storm runoff can be predicted accordingly. Although SCS $24\text{-}\mathrm{h}$ rainfall curves are recommended by the Natural Resources Conservation Service and have been widely adopted for many cosmopolitan areas in the United States, it is not enough to understand how these curves compare with observed data. In this study, a $57\text{-}\text{year}$ hourly rainfall data recorded in Denver was analyzed and then compared with the SCS $24\text{-}\mathrm{h}$ rainfall curves. It is concluded that these SCS $24\text{-}\mathrm{h}$ Types I and II curves do not represent the statistical average values. Rather, they represent the worst time distribution to form a severe storm. Using the concept of enveloping curves, conservative rainfall distributions can be derived using a similar approach as revealed by the SCS $24\text{-}\mathrm{h}$ curves. The approach suggested in this study requires much less rainfall data than the statistical average.