Journal of Hydrologic Engineering

The Gezhouba Reservoir (GR) and Three Gorges Reservoir (TGR) are the two largest water-storage projects in the middle reach of the Yangtze River. The impoundments of the reservoirs have changed natural flow-regimes in the river. This paper uses the Yichang Hydrological Station as a proxy for water-storage projects along the Yangtze River to assess alterations in river-flow conditions in three separate periods. The periods coincide with the start of water storage in the GR and TGR. The ecohydrological conditions in the river system are assessed based on annual discharge, water level, water temperature, and suspended-sediment concentration. More importantly, the entire analysis is done in relation to the Chinese sturgeon species, which face the danger of extinction in the Yangtze River. Based on the research reported in this paper, the impoundment of the GR and TGR increases downstream average annual water-temperature. In contrast, it decreases downstream average annual runoff, water level, and suspended-sediment concentration. The change of hydrological conditions negatively influences the spawning and hatching conditions of Chinese sturgeon. This change is the key reason that causes the fertilized rate of Chinese sturgeon eggs to decline.

The runoff coefficient, $C$, of the rational method is an expression of rate proportionality between rainfall intensity and peak discharge. Values of $C$ were derived for 80 developed and undeveloped watersheds in Texas using two distinct methods. First, the rate-based runoff coefficient, $C_{\text{rate}}$, was estimated for each of about 1,500 rainfall-runoff events. Second, the frequency-matching approach was used to derive a runoff coefficient, $C_r$, for each watershed. Published $C$ values, $C_{\text{lit}}$, or literature-based runoff coefficients were compared to those obtained from the methods investigated here. Using the 80 Texas watersheds, comparison of the two methods shows that about 75% of literature-based runoff coefficients are greater than $C_r$ and the watershed-median $C_{\text{rate}}$, but for developed watersheds with more impervious cover, literature-based runoff coefficients are less than $C_r$ and $C_{\text{rate}}$. An equation applicable to many Texas watersheds is proposed to estimate $C$ as a function of impervious area.

This study considers the problem of quantifying stream depletion from pumping test data. Bayesian inference is used to quantify the posterior uncertainty of parameters for a simple vertically heterogeneous aquifer model, in which the pumped semiconfined aquifer is separated by an aquiclude from a phreatic aquifer hydraulically connected to a stream. This study investigates the effects of using different data sets and shows that a single pumping test is generally not sufficient to determine stream depletion within reasonable limits. However, uncertainty quantification conducted within a Bayesian context reveals that by judicious design of aquifer tests, stream depletion can be accurately determined from data.

Analysis of the dependence between the main stream and its upper tributaries is important for hydraulic design, flood prevention, and risk control. The concept of total correlation, computed by the copula-entropy method, was applied to measure the dependence. This method only needs to calculate the copula entropy instead of the marginal or joint entropy, which estimates the total correlation more directly and avoids the accumulation of systematic bias. To that end, bivariate and multivariate Archimedean and metaelliptical copulas were employed, and multiple-integration and Monte Carlo methods were used to calculate the copula entropy. The methodology was applied to the upper Yangtze River reach in China, which has five major tributaries: Jinsha, Min, Tuo, Jialing, and Wu. Results showed that the selected copulas fitted the empirical probability distributions satisfactorily. There was a significant difference in total correlation values, when different copula functions were used. The copula entropy, calculated using the multiple-integration and Monte Carlo methods, led to similar results. The total correlation among the rivers was not high, and the one between Min and Tuo Rivers was the largest. There was some dependence among Jinsha, Min, and Tuo rivers, which constitutes a threat to flood control by the Three Gorges Dam (TGD). The flows of the Jinsha, Jialing, Min, and Tuo rivers significantly influence the flood occurrence in the Yangtze River.

Iraq is experiencing water-resource shortages which are expected to become more severe in the future. It is believed that rainwater harvesting will be one of the solutions to overcome this problem. In this paper, rainwater harvest modeling techniques were applied to the Sinjar area of northwest Iraq for agricultural purposes. A watershed modeling system (WMS) and linear programming (LP) optimization techniques were applied to maximize the irrigated area, which could be supplied by each selected reservoir for the period 1990–2009. This technique proved to be efficient for solving large-scale water supply problems with multiple parameters and constraints, including the required input data for the model. Two scenarios of operation were considered for each main basin. In the first, each reservoir was operated as a separate unit while, in the second, all reservoirs within the basin were operated as one system. Both scenarios gave encouraging results but Scenario 2 provided better results.

The knowledge of hydrologic variables and related functions at a given river cross section, as well as its time series of daily mean flows, supports the planning, design, and operation of facilities for the use of water resources in gauged river basins. In basins lacking systematic records of water levels and discharges, however, it becomes necessary to develop methodologies that enable the transfer of hydrologic information that is available in other basins. The present study proposes and evaluates a method for the automatic calibration of the parameters of a rainfall-runoff model in ungauged basins, using synthetic long-term flow duration curves (FDCs) estimated by a regional statistical model as the paradigm for the hydrologic processes. In this context, the long-term flow duration curves are used as an instrument to calibrate the parameters of a conceptual rainfall-runoff model. Once the parameters that control the hydrologic synthesis by the rainfall-runoff model are estimated, the continuous simulation of discharges over a given period of time enables various hydrologic evaluations such as the frequency analysis of rare events, the water balance of reservoirs, flow routing, and water availability studies. The methodology shown in here is composed, basically, of two parts. In the first part, a method for the regionalization of long-term flow duration curves is proposed, thus enabling the transfer of these hydrologic functions to ungauged watersheds, provided these are located in the same homogenous region as that of the basins for which data are available. In the second stage, the calibration of the parameters of the RIO GRANDE rainfall-runoff model is performed, having as a target to simulate the flow duration curves synthesized by the regional statistical model. The method was applied to a number of watersheds located in southeastern Brazil. In both phases of the proposed method, some performance indices are calculated in order to evaluate the accuracy of the procedures involved.

In the coal-mining region of the Appalachians, the Forestry Reclamation Approach (FRA) is currently promoted by the U.S. Dept. of Interior, Office of Surface Mining (OSM). FRA’s goal in establishing native forest cover on reclamation sites requires the use of loose spoils rather than traditionally compacted spoils during hillslope reconstruction. Loose spoils improve tree planting survival and enhance overall forest growth. Although FRA has been shown to be successful on low-gradient slopes without excessive runoff leading to erosion, rainfall-runoff relationships have not been studied on steep slopes ($>20°$) where most surface-mine reclamation typically occurs. A curve number (CN) range representing low-compaction steep-sloped conditions was needed for Appalachian FRA sites. Three active surface-mine sites in East Tennessee were monitored during a 1-year period (June 2009–July 2010) for rainfall (5-min intervals) and runoff hydrology (discharge volume, peak discharge) using a unique Pinson-type collection system. CNs were estimated by the traditional NEH-4 (Part 630) method and two asymptotic, frequency-matching techniques. It was found that an initial abstraction ($\lambda$) of 0.2 best fits the data. CNs generated by the NEH-4 method among all three mining sites were not significantly different suggesting common shale-based loose spoils on steep slopes have similar runoff behavior. The CN determined by the asymptotic method provided an estimate with a practical range between 58.5 and 60. This CN range is lower than one would expect from surface-mine sites, suggesting infiltration influenced the measured rainfall-runoff relationships. Further research is needed to investigate the influence of slope and infiltration on runoff behavior at surface-mine sites applying FRA.

The Lavras Simulation of Hydrology (LASH) hydrological model is classified as a deterministic, semiconceptual, and spatially distributed model. This model was developed to make available a tool for hydrological analysis in watersheds with a limited database, and it has presented good streamflow predictions. The aim of this study was to calibrate and validate the LASH model for hydrological simulation of the following watersheds which make up the headwaters of the Grande River basin (southern Minas Gerais state, Brazil): the Aiuruoca, Grande, Sapucaí, and Verde river basins. The LASH model simulated adequately the hydrological regimes of the previously mentioned watersheds, presenting satisfactory performances for all the watersheds, which can be emphasized by the Nash-Sutclifffe coefficient ($C_{NS}$) values greater than 0.70. The analysis of the main hydrological processes simulated by the LASH model confirmed that the results were consistent with the hydrological characteristics observed on the respective watersheds. The simulated daily flow-duration curves demonstrated good agreement in relation to the observed curves, indicating only a few errors in estimation of extreme streamflow events. It was found that the LASH model can be applied for hydrological simulation of headwater watersheds situated in the Grande River basin under land-use scenarios and meteorological and environmental conditions other than the ones used in this study. Therefore, it can be considered an important tool for planning and management of water resources in this region.

The identification of seasonal and annual trends in precipitation and streamflow volumes at the regional scale contributes to the understanding of global climate change and variation and is essential to the development of hydrologic models, hydrologic forecasting, and water resources planning and management for basins. This study examines such trends over the past 60 years (1950–2009) for the Nueces River basin, the San Antonio River basin, and the Guadalupe River basin, all in south Texas. The correlation of precipitation and streamflow with the El Niño Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) climatic indices is also evaluated, with a view toward incorporating these two relatively predictable cyclical patterns into modeling and forecasting of precipitation and streamflow. Trends and correlations were measured using the Kendall and Mann-Kendall test statistics, while the more commonly used Pearson $r$ coefficient was shown not to be consistently appropriate for the data analyses. The influence of PDO on annual precipitation is substantially less in these basins than it was found to be in a basin approximately 1,000 km northwest of them. The majority of trends and correlations were found to be statistically significant at the level of 0.05, and all such relationships were positive (or increasing). The analyses suggest that forecasting and management practices require a deterministic hydrologic model capable of accommodating natural features, management practices, and meteorological input for estimating evapotranspiration potential. Calibration and uncertainty analysis must be of a sophistication to accommodate nonnormal residuals, so that confidence intervals may be reliably used in forecasting and in management decisions.

The soil and water assessment tool (SWAT) is a distributed basin model that includes the option of defining spatial discretization in terms of terrain slope. Influence of terrain slope in runoff results from mountain basins is a determining factor in its simulation results; however, its use as a criterion for basin discretization and for the parameter calibration has not yet been analyzed. In this study, this influence is analyzed for calibrations using two different cases. Ten discretization cases were carried out to evaluate the relative importance of slope discretization compared with other discretization criteria. Data from 1999–2005 were used for model calibration, and those from 2006–2009, for model validation. Parameter identification and specification were performed with the combined latin hypercube and one-factor-at-a-time (LH-OAT) and the shuffled complex evolution-uncertainty analysis methods (SCE-UA), respectively. All cases resulted in very good statistical values, with the Nash-Sutcliffe efficiency coefficient of 0.82–0.85, a bias of 2–10%, and the observations standard deviation ratio of 0.4–0.3. More realistic calibrated parameters were found when terrain slope variation was not included in the spatial discretization criteria. The inclusion of slope did not significantly improve simulations results when a good set of parameters was used, but it did enhance the calibration when a reduced number of subbasins was used.

Reanalysis of the records, upon which the 120-year-old Rational Formula is based, reveals that the published linear relationship between rainfall duration and the ratio of peak discharge to the product of rainfall intensity and watershed area can also be replaced by a logarithmic regression. Such a regression is found, with reasonable correlation, suitable for two-dimensional laboratory watersheds, for four experimental urban watersheds, and for five small watersheds in Israel. Combining this regression with a power-law regression between rainfall intensity and duration, associated with any given recurrence interval, yields a general relationship between the variables. From this and through simple mathematical operations, the maximum discharge associated with that recurrence interval is determined. The variables in the new equation are watershed area and the empirical parameters of the regressions. Application of this equation bypasses determination of the highly influential time of concentration, which is an immeasurable variable and for which varied surrogates are used. It also saves determination of the coefficient $C$ of the Rational Formula. Yet, it requires preparation of regional models for estimation parameter values of the logarithmic regression. Application of the proposed method on the outdoor watersheds, with intensity-duration-frequency curves for representative meteorological stations, yields realistic results. The application for one of those watersheds, with an observational period of 40 years, reveals a wide difference between the prediction through the proposed method and that through statistics of recorded peak discharges. The difference is attributable to association of recurrence intervals of rainfall intensities with peak runoff discharges. The difference is reduced by use of statistics for rainfall intensities during the subseason of high discharges rather than during the entire rainy season.

The performance of the ensemble Kalman filter (EnKF) and the particle filter (PF) is assessed for short-term streamflow forecasting with a distributed hydrologic model, namely, the water and energy transfer processes (WEP) model. To mitigate the drawbacks of conventional filters, the ensemble square root filter (EnSRF) and the regularized particle filter (RPF) are implemented. For both the EnSRF and the RPF, sequential data assimilation is performed within a lag-time window to consider the response times of internal hydrologic processes. The proposed methods are applied to two catchments in Japan and Korea to assess their performance. The model ensembles are perturbed by the noise of the soil moisture content and are assimilated with streamflow observations. The forecasting accuracy of both the EnSRF and the RPF is improved when sufficient lag-time windows are provided. The EnSRF is sensitive to the length of the lag-time window and has a limited ability to forecast within short lead times, whereas the RPF has a more stable forecasting capability for the entire range of lead times. Filtering with a limited number of ensembles also yields improved performance using a lag-time window.

In this paper, a simple hydrologic example is employed to illustrate the important features of reservoir regulated flood frequency. Despite its practical significance, the estimation of reservoir regulated flood frequency is largely dominated by empirical methodologies containing assumptions that could lead to incorrect results. The goal of this paper is to show by means of a continuous rainfall-runoff simulation how several reservoir variables, including the reservoir storage capacity, the size of release structures, operation rules, and the statistical variability of inflows to the reservoir, quantitatively control the regulated flood frequency. Although the example presented in this paper does not encompass the full complexity of the problem, it reveals important features of the regulated flood frequency. The study also highlights how specific assumptions in the traditional and widely used inflow volume-duration-frequency (VDF)–based methodology could lead to underestimation of flood risk for locations downstream from reservoirs. This paper, therefore, provides insight into the steps necessary to move away from the VDF-based empirical analysis towards a more realistic framework for estimating regulated flood frequencies. Simultaneously, the difficulties in fully addressing this problem are acknowledged and, consequently, continued examination of this important subject by the hydrological-science and engineering community is advocated.

Wetlands are rich ecosystems recognized for ameliorating floods, improving water quality, and providing other ecosystem benefits. This part of a two-paper series presents a relatively detailed process-based model for nitrogen and phosphorus retention, cycling, and removal in flooded wetlands. The model captures salient features of nutrient dynamics and accounts for complex interactions among various physical, biogeochemical, and physiological processes. The model simulates oxygen dynamics and the impact of oxidizing and reducing conditions on nitrogen transformation and removal, and approximates phosphorus precipitation and releases into soluble forms under aerobic and anaerobic conditions, respectively. Nitrogen loss pathways of volatilization and denitrification are explicitly accounted for on a physical basis. Processes in surface water and the bottom-active soil layer are described by a system of coupled ordinary differential equations. A finite-difference numerical scheme is implemented to solve the coupled system of ordinary differential equations for various multiphase constituents’ concentrations in the water column and wetland soil. The numerical solution algorithm is verified against analytical solutions obtained for simplified transport and fate scenarios. Quantitative global sensitivity analysis revealed consistent model performance with respect to critical parameters and dominant nutrient processes. A hypothetical phosphorus loading scenario shows that the model is capable of capturing the phenomenon of phosphorus precipitation and release under oxic and anoxic conditions, respectively.

In this paper, the authors applied and evaluated the wetland nutrient model that was described in Paper I. Hydrologic and water quality data from a small restored wetland located on Kent Island, Maryland, which is part of the Delmarva Peninsula on the eastern shores of the Chesapeake Bay, was used for this purpose. The model was assessed through various methods against the observed data in simulating nitrogen (N), phosphorus (P), and total suspended sediment (TSS) dynamics. Time series plots of observed and simulated concentrations and loads generally compared well; better performance was demonstrated with dissolved forms of nitrogen, i.e., ammonia and nitrate. Through qualitative and quantitative sensitivity analysis, dominant processes in the study wetland were scrutinized. Nitrification, plant uptake, and mineralization were the most important processes affecting ammonia. Denitrification in the sediment layer and diffusion to bottom sediments were identified as key processes for nitrate. Settling and resuspension were the most important processes for particulate matter (organic N, sediment) and sediment-bound phosphate (inorganic P). Order of parameter sensitivities and dominant processes exhibited seasonality. Uncertainty bands created from Monte Carlo simulations showed that parameter uncertainty is relatively small; however, uncertainty in the wetland inflow rates and loading concentrations have much more bearing on model predictive uncertainty. N, P, and TSS mass balance analysis showed that the wetland removed approximately 23, 33, and 46%, respectively, of the incoming load (runoff + atmospheric deposition) over the two-year period, with more removal in year 1 (34, 43, and 55%, respectively), which had a long stretch of a dry period. The developed model can be employed for exploring wetland response to various climatic and input conditions, and for deeper understanding of key processes in wetlands.

Rain gardens have increasingly been used to control the adverse effects of urbanization on storm-water quantity and quality. The ratio or percentage of storm water generated from the contributing area of a rain garden that is captured by the surface depression of the rain garden is known as its storm-water capture efficiency. This capture efficiency is an important indicator of a rain garden’s performance for storm-water management. Based on the probability distributions of local rainfall event characteristics and the hydrologic operation of rain gardens, an explicit analytical equation is derived for estimating the long-term average storm-water capture efficiency of a rain garden. The validity of the analytical equation is demonstrated by comparing its outcomes with results from a series of continuous simulations. Example applications of this equation are made for two locations.

In hydrology, one often needs to use goodness-of-fit tests to describe how well a statistical frequency model fits a set of observations. The two-parameter kappa (KAP2) distribution is a frequency model that has already shown useful hydrological applications. The Cramér-von Mises and the Anderson-Darling tests are among the most frequently used goodness-of-fit tests in statistical practice. To apply these tests to the KAP2 distribution, statistical tables of critical values are needed. Such tables already exist for various other statistical distributions, but have not yet been developed for KAP2. This paper fills the gap by developing such tables for the KAP2 distribution. These tables are based on large sample theory, but correction formulas are also provided that make them applicable to small samples. KAP2 and the generalized Pareto distribution (GPD) are two subfamilies of the four-parameter kappa distribution that have been used in extreme value hydrological applications in the past. A detailed example is given that shows the usefulness of the KAP2 model for fitting some types of extreme hydrological data, and another example is provided to show how KAP2 may be used as an alternative to the GPD in certain extreme value applications.

A new index called a modified relative strength effect (RSE) is developed for determining the influence of an input in an artificial neural network (ANN) model. This modified RSE, which is an improvement over the usual RSE, indicates the influence of each input on the target output in different ranges and can be estimated easily during ANN model development by a user. The potential of this modified RSE was examined using ANN models developed for finding atypical bacteria counts and daily flow. The modified RSE and traditional RSE were used to identify the essential inputs for the ANN models in this research. The results show an improvement of 10% in the mean square error when the modified RSE is used for input selection instead of the traditional RSE. This study also explores the usefulness of the modified RSE in identifying the required input lags while developing an ANN time-series model. A time-series model was constructed using ANN, and the modified RSE values of inputs were compared to the partial autocorrelation coefficients (PCCs) of the first 12 lags. The PCCs were calculated using standard procedure. Perfect matching of the modified RSE with the PCC was observed. The results indicate that the RSE can be used like the PCC while developing ANN-based time-series models.

The development of watershed models with minimal quantified uncertainty under nonstationary conditions is a major challenge in the field of hydrology. This is especially problematic in data-poor areas where values for model inputs are lacking or measured on temporally and/or spatially sparse scales. The objective of this work is to conduct a global sensitivity and uncertainty analysis (GSA/UA) of the Pitman semidistributed hydrologic model for the data-poor Okavango Basin in southern Africa under both stationary and climate change scenarios. The Morris GSA method allowed qualitative ranking of important model inputs whereas the variance-based Fourier amplitude sensitivity test (FAST) method quantitatively identified the parametric uncertainty and sensitivity to these inputs. Results showed that the most important model inputs determining mean annual flow and model fit to observed data were the infiltration rate and the temporal rainfall distribution. In addition, the wetter western headwaters region was shown to be the most important region in determining the flow at the outlet of the basin. Parameter equifinality was significant in this study, and hence the evaluation of the relationships between mechanisms was not straightforward. Analysis of model results under climate change scenarios showed that a hot and wet scenario introduced more change in mean annual flow than a hot and dry scenario. The climate change scenarios also altered model sensitivity. For example, the parameter that controls the rate of infiltration decreased in importance and the parameter that controls soil moisture storage gained importance under the dry scenario. These results are useful when determining the applicability of model predictions under stationary and nonstationary conditions and when focusing watershed monitoring efforts.

The population has increased at a fast pace in the last two decades in Arkansas. One direct result from this process is urbanization, especially in northwest Arkansas. Land use and land cover (LULC) had been changed from forests, pastures, and farmland to urbanized areas over a short time span. Increased flood discharge and decreased water quality have been negative side effects of urbanization on the environment. In this study, five watersheds were selected as representative of the entire state. These watersheds were selected to examine the hydrological response to land use change over the last two decades in Arkansas. The curve number, a term used to describe the runoff depth, was calculated for the five watersheds. Out of the five watersheds, the Illinois River watershed experienced the greatest change in land use from 1993–2006, whereas the other four watersheds experienced slight land use change. The land transformation model (LTM) was used to forecast the land use of the Illinois River watershed in the year 2019. Simulated urbanized area in the Illinois River watershed from 1993–2019 increased from 3.7 to 23.6%. The curve number of the Illinois River watershed would increase from 68.9 in 1993 to 72.1 in 2019. The Spring Creek-Osage Creek watershed, a subwatershed in the Illinois River watershed, was selected for detailed evaluation of the hydrologic response attributable to land use change. The storm water management model (SWMM) was used to calculate the flood discharges resulting from different storm events. The simulation results showed that during 1993–2019, the peak flood discharge of the Spring Creek-Osage Creek watershed for the storm return periods 2, 5, 10, 25, 50, and 100 years increased 230, 180, 160, 140, 130, and 110%, respectively, whereas the total flood volume increased 270, 220, 200, 170, 160, and 140%. The results in this study might suggest to small cities like Bethel Heights that it is better to plan the drainage systems for the coming decades before such cities have already sprawled. This study also applies a combined methodology of a regional scale urbanization-flood model, which could be used to improve flood frequency analysis.

A challenging problem in basin-scale hydrology is the ability to represent and quantify the nonlinearity of the output hydrograph or response. An important aspect of this challenge is to clearly distinguish among different sources of nonlinearity. To address the previous, a scale-dependent network instantaneous response function (IRF) is obtained by using the theory of transport by travel times, relationships of basin geomorphology, hydraulic geometry, and the inverse Gaussian (IG) distribution. Inverse Gaussian is used to represent the network response because of its connection with diffusion processes. It is shown that both geometric and hydraulic regularity in the stream network, together with the scaling property of IG, can result in the network response being approximately scaling, implying in turn that the peak flow and time to peak of the response are scaling as well. The derived IRF has the advantage over previous formulations of explicitly distinguishing between nonlinearity and scaling. In addition, it provides an explanation for the scaling in terms of well-known hydrogeomorphic relationships. The scaling is found to arise from the dependence of the network length (i.e., the mean path length) and hydraulic geometry, specifically the mean velocity and the cross-sectional area, on the basin size. Moreover, after dividing the network travel times by the mean travel time, a response that is independent of scale is obtained. This response can be seen as analogous to the instantaneous unit hydrograph but representing instead the output associated with a unit travel time. Furthermore, the derived IRF may be useful in design applications as it provides a way for estimating the response across basin sizes by using readily available data sets. For the latter, additional comparison against observed data is necessary to better understand the applicability of the proposed scaling under a variety of climatic and terrain scenarios.

The nonlinear Muskingum model has three parameters (storage parameter, weighting parameter, and exponent parameter) that are assumed in model estimation to be constant. The exponent parameter, which has no physical meaning, represents the average nonlinear behavior of the flood during the entire routing period. To address the variations of nonlinearity during the routing period, this paper considers a variable exponent parameter that varies with the inflow level. The boundaries of the inflow levels are considered to be dimensionless parameters. The problem is formulated as a mathematical optimization model that minimizes the sum of the squared (SSQ) or absolute deviations between the observed and estimated outflows. An efficient spreadsheet-based software is implemented. The proposed model was applied by using three examples involving single peak, multipeak, and nonsmooth hydrographs. The results show that the range of the optimal exponent parameters is small, yet the improvement in the fit of the nonlinear Muskingum model is substantial; the SSQ reduction reaches 35%, compared with the case of a constant exponent parameter. The proposed model should be of interest to researchers and engineers working in the area of flood management.