Journal of Hydrologic Engineering

Intermittent streamflow estimates are important for water quality management, planning water supplies, hydropower, and irrigation systems. This paper proposes the application of a conjunction model (neurowavelet) for forecasting daily intermittent streamflow. The neurowavelet conjunction model is improved by combining two methods, discrete wavelet transform and artificial neural networks (ANN), for $1\text{day}$ ahead streamflow forecasting and results are compared with those of the single ANN model. Intermittent streamflow data from two stations in the Thrace Region, the European part of Turkey, in the northwest part of the country are used in the study. The comparison results revealed that the suggested model could significantly increase the forecast accuracy of single ANN in forecasting daily intermittent streamflows. The neurowavelet conjunction model reduced the prediction root mean square errors and mean absolute errors with respect to the single ANN model by 74–65% and 43–12%, and increased the determination coefficient by 47–11%, respectively.

The main goal of this study was to determine regional probability distributions for the annual maximum flood data observed at 45 streamflow gauging sites in the Kucuk and Buyuk River Basins in Turkey using index flood L moments. Mann–Kendall trend analysis was previously applied to the streamflow gauging sites’s observed data, and dam effects were sought at all stations. Seven sites were omitted from the analysis because they were not unregulated or trends were present in the data. A regional analysis was performed on the remaining 36 sites, which satisfied the screening criteria of having a minimum record length of $10\text{years}$ , representing unregulated natural flows, and passing the Mann–Kendall test at the 5% significance level. Data from the 36 streamflow gauging sites in the basins were screened using the discordancy measure $\left(D_i\right)$ . Homogeneity of the region was then tested using the L-moments heterogeneity measure by carrying out 500 simulations using the four parameter kappa distribution. Based on these tests, the streamflow gauging sites were found to constitute two homogeneous subregions, namely the upper- and lower-Menderes subregions. Seven comparative regional flood frequency analysis studies were carried out for each subregion using various L-moment-based frequency distributions: the generalized logistic, general extreme value, generalized normal, Pearson Type III, generalized Pareto, kappa, and the five-parameter Wakeby. Based on the L-moment goodness of fit statistic criterion, the generalized normal extreme value distribution has been identified as the best-fit distribution for the upper- and lower-Menderes subregions. A Monte Carlo simulation was used to evaluate the results on the basis of the relative root mean square error relative bias. For estimation of floods of various return periods for ungauged catchments, a regional relationship between mean annual peak flood and catchment area was developed and coupled with the regional flood frequency relationship derived for gauged catchments.

Droughts are complex events which may impair social, economic, agricultural, and other activities of a society. The present hydroclimatological study comprises three stages. First, a Markov chain model based on annual flows at a gauging station located on the Göksu River (Turkey) is utilized. Second, a critical drought analysis is conducted. Third, the influences of the North Atlantic Oscillation (NAO) on the probability distribution functions (PDFs) of critical droughts have been documented. Drought duration is used as a key parameter in the Markov chain model. As the model degree should be defined prior to applying a Markov chain model to a series of observations, a model degree of first order has been selected according to the results of the Akaike information criteria and Bayesian information criteria. The exact PDFs of the critical drought duration in a finite sample that follows the first-order Markov chain have been determined by the enumeration technique. The critical drought duration is the possible maximum duration likely to occur over the economic life of any water resources system. The expectations of the critical drought duration are provided for different transitional probabilities. The influences of NAO on the PDFs of critical droughts have been examined for two opposite cases; namely, the period of negative NAO index values (1936–1971) and the period of positive NAO index values (1972–2000). The results of this analysis clearly indicate that NAO series have quantifiable influences on the transition probabilities and expectation of the critical drought duration. As a result, the probability distribution functions obtained for various critical drought durations in the observations of the Göksu River may be implemented as a robust tool to determine a design value with the concept of risk.

Evaporation plays a key role in water resources management in arid and semiarid climatic regions. This is the first time that an artificial neural network (ANN) model is applied to estimate evaporation in a hot and dry region (BWh climate by the Köppen classification). It has been found that ANN works very well at the study site and, further, an integrated ANN and autoregressive with exogeneous inputs can have an improved performance over the traditional ANN. Both models significantly outperformed the two empirical methods. It has been demonstrated that the important weather factors to be included in the model inputs are wind speed, saturation vapor pressure deficit, and relative humidity. This result is different from all those reported in the literature and is interestingly linked with a 1936 study by Anderson, who emphasized the importance of saturation vapor pressure deficit. As evaporation is a nonlinear dynamic process, the selection of suitable input weather variables has been a complicated and time-consuming task for modelers. In this study, a new data analysis tool called the gamma test has been used to identify the best combination of model inputs prior to model construction and evaluation.

This study presents gene-expression programming (GEP), which is an extension to genetic programming, as an alternative approach to modeling stage–discharge relationship. The results obtained are compared to more conventional methods, stage rating curve and multiple linear regression techniques. Statistical measures such as average, standard deviation, minimum and maximum values, as well as criteria such as mean square error and determination coefficient, the coefficient of efficiency, and the adjusted coefficient of efficiency are used to measure the performance of the models developed by employing GEP. Also, the explicit formulations of the developed GEP models are presented. Statistics and scatter plots indicate that the proposed equations produce quite satisfactory results and perform superior to conventional models.

Numerical experiments were performed to investigate how the coastal boundary condition could be approximated in a groundwater flow model to yield accurate values for hydraulic heads and fluxes in the freshwater part of a coastal aquifer. These experiments consisted of obtaining steady-state solutions for hydraulic heads in a vertical cross section using the groundwater flow code MODFLOW and comparing the results to a steady-state solution in a similar cross section for equivalent freshwater heads obtained using the variable-density flow and transport code SEAWAT, which was considered to be the accurate solution to the problem. Six different boundary conditions that have been used to approximate the coastal boundary in groundwater flow models were tested, and simulations were run for both specified-flux and specified-head boundary conditions at the upstream freshwater boundary. For both upstream boundary conditions, the MODFLOW solution that best matched the SEAWAT solution for hydraulic heads and fluxes in the freshwater part of the aquifer was the solution in which equivalent freshwater heads were specified over the full thickness of the aquifer at the coastal boundary.

This paper presents an analytical solution of the linearized Boussinesq equation characterizing the unsteady flow in an unconfined sloping aquifer that is in contact with a constant piezometric head at one end and a stream, whose level is rising at a constant rate, at the other end. Considering the idealized one-dimensional region as homogeneous, isotropic, incompressible, and an unconfined aquifer, the analytical expressions for the transient water table and the flow rate at both interfaces are obtained. The governing equations are solved by applying the Laplace transform technique and the results show a considerable change in aquifer’s water head profile and flow rate at the interfaces due to the presence of bed slope. The solution presented here is capable of handling the transient condition, as well as a very slow rise and a sudden rise in stream water for downward and upward sloping bedrocks after employing an asymptotic approach to corresponding scenarios.

Storm water management models using the kinematic wave (KW) approach require the conversion of a natural watershed to its equivalent rectangular cascading plane. AKW cascading plane is specified by the plane’s area, width, and slope. Without on-site calibration, there is no guidance on how to translate the irregularity of a natural watershed into a rectangular shape that has an equivalent hydrologic response. As a result, the reliability of a KW model depends on an engineer’s experience. Often the use of a computer model to simulate stormwater runoff is the only option because there are not field data at the project site to use statistical methods. The need for calibration at each site defeats the purpose for using computer simulations. This long existing problem has been documented in the EPA SWMM4 User Manual, which continued into the EPA SWMM5 user guidance. In this study, the continuity and energy principles are creatively interpreted as the preservations of watershed area and vertical fall over the receiving waterway’s length. The dimensional analysis suggests that the watershed shape factor can be used to correlate the waterway length in the natural watershed to the KW plane width. A parabolic one-to-one single valued function was derived to translate the watershed shape into its equivalent KW plane width, length, and slope. After numerous tests, it is confirmed that the watershed and KW shape factors provide a consistent and stable basis for watershed geometric conversion. It is believed that the procedure outlined in this paper can be very useful to the KW surface runoff modeling to represent urban stormwater predictions when an on-site calibration is not readily available.

The Hydrologic Simulation Program–Fortran (HSPF) model was used to simulate streamflow for the Distributed Model Intercomparison Project (DMIP). This research describes how to integrate a next generation radar, NEXRAD data set into HSPF model and how to generate hydrologic runoff associated with the inputs of the spatial rainfalls, and discusses the challenges of HSPF during automatic calibration processes. The model performance of simulated streamflows with calibration and without calibration was also evaluated to assess the sensitivity of final estimated parameters through calibration procedures to initial parameters derived from physical watershed configuration. An automatic calibration scheme was adopted to optimize objectives that the writer specified in terms of statistical measures. Parameter estimation, a model-independent parameter estimator, was used as an automatic calibration tool in the hydrologic calibration of HSPF. Overall, the calibrated simulations outperformed the uncalibrated simulations for DMIP basins, and the writer anticipates that HSPF may be a potential alternative model to reproduce the hourly flows for streamflow forecasts.

A local perturbation approach is presented for the sensitivity analysis of the one-dimensional shallow water equations. Under steady state conditions the sensitivity of the water depth to the Strickler coefficient is shown to obey a first-order ordinary differential equation that becomes linear under uniform conditions. This allows a “half-distance” to be defined, which indicates the size of the region over which the Strickler coefficient can be calibrated from stage measurements. This region extends downstream of the stage measurement point under subcritical conditions and upstream under supercritical conditions. The validity of the approach is checked against typical backwater curve calculations. The sensitivity obtained by the local perturbation approach is very close to the empirical sensitivity, even when the Strickler coefficient is perturbed by $\pm 25\%$ , which validates the use of the approach in practical applications.

This case study describes the installation and performance of two experimental low-head weirs and infiltration areas (Weir Facility) located in the 183 ha Fox Hollow Watershed, an urbanizing watershed underlain by karst bedrock in central Pennsylvania. Soil/rock corings and infiltration tests conducted prior to installation indicated that the area was suitable for development as an infiltration field. Low-head weirs were constructed to cause temporary, shallow water retention to promote infiltration and groundwater recharge. A monitoring program assessed the performance of the weirs and allowed mass balance calculations. During 2003 through 2008, which included two of the wettest years on record, few events (less than 30%) resulted in significant outflow from the weirs and associated infiltration areas. For smaller rainfall events ( $<1.9\mathrm{cm}$ of rain) almost all (89–100%) of the runoff delivered to the Weir Facility infiltrated; whereas for larger events ( $⩾1.9\mathrm{cm}$ rain) generally over 50% of the runoff volume infiltrated, and peak rates and runoff volumes were substantially reduced. These data support the hypothesis that, in karst-underlain watersheds undergoing substantial development, it is imperative to identify and preserve critical recharge areas, and that simple low-head weirs with corresponding infiltration areas enhance and sustain recharge, thereby reducing potential storm-water runoff.

Kriging is a powerful tool for interpolation and network design. A necessary prerequisite for a kriging application is the variogram that describes the variation of the semivariance with distance. Variograms are routinely derived from the measurements of the concerned variable. Presented herein is a model of the variogram of groundwater hydraulic heads that permits the estimation of the semivariance corresponding to any distance in terms of the transmissivity $\left(T\right)$ and standard deviation $\left({\sigma }_W\right)$ of the flux. The model thus evolves the variogram from the aquifer attributes rather than from the measured heads. Treating the flux as a normally distributed random variable, with zero mean and an assigned standard deviation, synthetic discrete steady-state stationary head fields are generated by the Monte Carlo simulation. The fields so generated are employed to evolve experimental variograms, and hence to estimate the parameters of the Gaussian model for various values of $T$ and ${\sigma }_W$ . Closed form functional relations between the Gaussian parameters and the independent variables viz. $T$ and ${\sigma }_W$ are arrived at by regression. This provides a new model capable of yielding variograms of groundwater hydraulic heads without invoking the head measurements.

Wet mountainous watersheds are characterized by variable source area runoff, the runoff generated from surface saturated source areas being the prominent component in most cases. However, often subsurface stormflow, generated by various different mechanisms, also contributes significantly to stream quick flow. The model presented in the present study has been developed to simulate runoff processes in such watersheds, considered to have two quick flow and a delayed flow components. The lumped watershed model used has been developed in stages, starting from the curve-number method, by drawing inferences at every stage and changing the model structure in accordance with the inferences. The data on streamflow and rainfall from five catchments in the Western Ghat regions of Karnataka, in South India, have been used in the study. The simulations of the model have been interpreted to infer that the subsurface pipe flow component forms a very significant part of quick flow in the region.

One-dimensional infiltration tests were conducted to investigate the effects of rainfall intensity and initial water content on the transient infiltration pattern of unsaturated silty sand. For a constant rainfall intensity greater than the saturated permeability but smaller than the maximum infiltration capacity of soil, the infiltration curve is divided into preponding and postponding stages. The time to ponding reduces with increasing rainfall intensity and initial water content. After ponding, the infiltration rate reduces with time and its reducing rate increases with rainfall intensity but does not vary significantly with initial water content. The infiltration pattern can be simulated qualitatively by solving Richard’s equation numerically with the use of the flux and head type boundary conditions at the soil surface in the preponding and postponding stages, respectively.