Journal of Hydrologic Engineering

Obtaining appropriate, applicable exponent values of distance and elevation in the modified inverse distance method for hourly rainfall interpolation is difficult because the range of the exponent values has not yet been established at existing studies. In this research, an attempt is made to find the optimal exponent values in the modified inverse distance method based on hourly rainfall data. For the purpose of this study, 28 rainfall observation stations providing the data were classified into four groups, which are located at the Han River upstream, the Han River downstream, the Geum River upstream, and the Nakdong River midstream areas in South Korea. A total of 44 cases consisting of one base rainfall station and four index rainfall stations were analyzed. The optimal exponent values of distance and elevation were calculated by a simple nonlinear optimization technique of the steepest gradient method using hourly rainfall data observed for 10 years (2004–2013) and validated using another data for three years (2014–2016) on a case basis across different groups. The proposed modified inverse distance method was found to be superior to the conventional inverse distance method. Worth noting is that the root mean squared error values obtained by the method of this study were smaller than those of the conventional methods in all groups even though they were small in size hydrometeorologically. Therefore, the method proposed in this research can be considered more advanced than the conventional method. The optimal exponent values of distance and elevation obtained in this research were 3.463 and $-0.017$, 2.559 and 0.741, 2.877 and $-0.681$, and 3.738 and 2.001, respectively, in the four groups, and the total mean values of the two values were 3.159 and 0.511, respectively. The main factor influencing rainfall between stations when applying the modified inverse distance method with the optimal exponents was found to be the distance, and the altitude is the minor factor.

Statistical techniques have been developed for assessing structures to confront extreme events such as floods. Among them are probability distribution functions (PDFs). For example, the waiting time of the first flood exceeding the design flood is geometric, while the number of floods exceeding the design flood in $n$ years is binomial. The expected waiting time (EWT) and hydrologic risk of structures are commonly used metrics developed assuming stationarity and independence. And newer techniques include PDFs and project evaluation metrics applicable for nonstationary conditions. This article focuses on first arrival time of $k$th floods exceeding the design flood and associated metrics. The likelihood of first occurrence of 1, 2, 3, etc. floods exceeding the design flood becomes important, as uncertainties of hydrologic regimes increase due to climatic and anthropogenic drivers. We use stationary and nonstationary negative binomial distribution and develop EWT and risk for the $k$th event. They generalize the traditional return period and risk that refer to the occurrence of the first event, i.e., $k=1$. The newer metrics consider the probability of $k$ or more events. We tested by simulation the concepts and derived equations and apply them using annual floods of Assunpink Creek. The results show that the return period and risk functions developed can be helpful in assessing flood related projects.

Evaporation ($Ep$) has a vital importance for the management and development of water resources projects. In this study two scenarios are considered in prediction of monthly pan evaporation. The first scenario challenges the ability of three artificial intelligence–based models [neural network autoregressive with exogenous input (NNARX), gene expression programming (GEP), and adaptive neuro-fuzzy inference system (ANFIS)]. The second scenario investigates the capability of five different metaheuristic algorithms [particle swarm optimization (PSO), firefly algorithm (FA), artificial bee colony (ABC), continuous ant colony optimization (CACO), and genetic algorithm (GA)] integrated with the ANFIS model in $Ep$ modeling. Meteorological factors (monthly air temperature, solar radiation, relative humidity, and wind speed data) of two stations in Turkey were used as inputs to the models. Various statistic measures [root-mean-square error (RMSE), mean absolute error (MAE), and determination coefficient ($R^2$)] and diagnostic analysis (Taylor diagram) were deployed to evaluate and compare the performance of the models. The results of the first scenario show that the ANFIS model gave better performance in Gaziantep Station, whereas the NNARX model performed better in estimating $Ep$ values in Adiyaman Station. In the second scenario, it was observed that the PSO and GA algorithms performed better in comparison to the other algorithms in Gaziantep and Adiyaman stations, respectively. The non-parametric Kruskal–Wallis test denoted that there is a significant difference (alpha of 0.05) between the observed versus predicted amounts of monthly $Ep$ for the NNARX, ANFIS, and GEP. However, there is no sign of significant difference in predicting monthly $Ep$ between the applied metaheuristic algorithms.

Soil Conservation Service curve number (SCS-CN) and Green-Ampt (GA) infiltration models are probably the most applied equations in practical hydrology to compute the amount of direct runoff from rainfall. Both models are very simple, require few parameters, and present drawbacks and advantages. The empirical CN model concentrates the infiltration effect in the curve number and in the so-called soil hydrological groups, which have been widely characterized for different soil types, land cover, and antecedent soil moisture conditions (ASMCs), although the latter was considered ambiguous, whereas soil hydrological characteristics, including ASMC, are taken into account for the simplified physically based GA model. The main advantage of the GA model is the temporal variation of the rainfall excess intensity, which is not considered in the CN model. In this paper, CN and GA models are jointly used in order to analytically establish relationships linking each other, so that the positive features of both models together can be taken into account for applications. It is shown that the suggested procedure makes it possible to move from one model to the other and vice versa, according to the derived equations linking the relative parameters. Constant rainfall intensity was assumed; therefore, the procedure is better aimed to its design purpose at the small basin scale, rather than to reproducing rainfall-runoff events. A comparison between the results obtained by applying the suggested analytical procedure with those numerically derived by other researchers for SCS temporal storm distribution is performed and discussed.

The increase of urbanization combined with the increase in impervious surfaces has led to an increasing frequency of flooding events in urban catchments. This context highlights the inadequacy of traditional urban drainage systems and the need for alternative solutions. Permeable pavements have proven to be a valuable low-impact development (LID) technique. They are able to retain surface runoff volumes by increasing the infiltration and evaporation processes. Even though their benefits are well known, the lack of adequate modeling tools limits their adoption. This paper presents the development of a physically based model to describe their hydrological behavior. The particularity of this partial differential equation (PDE) model is to base the parameter estimations on basic measurements. The model is calibrated and validated using measurements from a laboratory permeable pavement rig. Results demonstrate a high reliability of the model with a Nash—Sutcliffe model efficiency (NSE) value of 0.969 and 0.891, respectively, for calibration and validation. Finally, a sensitivity analysis was conducted and highlighted the influence of the hydraulic conductivity of the base layer on the performance of the structure.

The Walnut Gulch Experimental Watershed is a semi-arid experimental watershed and long-term agro-ecosystem research (LTAR) site managed by the USDA-Agricultural Research Services (ARS) Southwest Watershed Research Center for which high-resolution, long-term hydroclimatic data are available across its $149\text{-}{\mathrm{km}}^2$ drainage area. Quality control and quality assurance of the massive data set are a major challenge. We present the analysis of 50 years of data sets to develop a strategy to identify errors and inconsistencies in historical rainfall and runoff databases. A multiple regression model was developed to relate rainfall, watershed properties, and the antecedent conditions to runoff characteristics in 12 subwatersheds ranging in area from $0.002–94{\mathrm{km}}^2$. A regression model was developed based on 18 predictor variables, which produced predicted runoff with correlation coefficients ranging from 0.4–0.94 and Nash efficiency coefficients up to 0.76. The model predicted 92% of runoff events and 86% of no-runoff events. The modeling approach is a complement to existing quality assurance and quality control (QAQC) procedures and provides a specific method for ensuring that rainfall and runoff data in the USDA-ARS Walnut Gulch Experimental Watershed database are consistent and contain minimal error. The model has the potential for making runoff predictions in similar hydroclimatic environments with available high-resolution observations.

An understanding of groundwater flow in soil and rocks is necessary for civil and environmental engineering. Real-time monitoring of groundwater flow is a relatively new research area. This paper presents an experimental investigation into the response of geoelectric fields to the steady flow of water through a heterogeneous sand and clay medium. The parameters of the geoelectric field were measured during seepage. The self-potential decreased gradually and then increased rapidly when the infiltration front passed reference electrode N at 3 min. The self-potential decreased rapidly when the infiltration front reached an electrode. Stepwise changes in the current were used to calculate the seepage velocities in different media. The apparent resistivity decreased rapidly to below $600\mathrm{\Omega }·\mathrm{m}$ when the water reached each electrode. Therefore, geoelectric fields can be used to monitor water flow in real time.

Two recharge ways, point inflow and lateral inflow, exist on the conduit network in karst aquifers. Point inflow is commonly included in pipe-flow models, but lateral inflow along the conduit network is often ignored. Few studies have paid attention to the potential influence of different recharge mechanisms on model simulation results. In this study, three models with different ways of recharging the conduit network were established based on a reservoir model and two analytical solutions. These models were applied to a karst spring to investigate the impact of different ways of recharge on simulation results. It is shown that the three models had very similar simulation results for both high and low flows after calibration. A series of synthetic experiments also indicated the model difference caused by the different ways of recharge of the conduit network could always be masked by the adjustment of other parameters in the model during the calibration process. These results imply that spring discharge as a model target alone may not help to identify the recharge way of the conduit network.

The time of concentration is a fundamental parameter used for hydrological analysis in professional and scientific communities. The availability of equations for its estimation leads to significant variability and uncertainty. In this work, the most commonly used 24 equations were selected among the empirically based models to compare their estimates and to quantify the accuracy in predicting the time of concentration observed from 46 river basins in northern Italy. The practical definition for direct time of concentration estimation used in this work computes it as the time from the end of effective rainfall to the end of direct runoff on the recession limb of the total hydrograph. The results show a large discrepancy between the observed values and those estimated with the selected equations: the best performances are achieved by Bransby-Williams, Giandotti, Soil Conservation Service, Témez, and Ferro equations. The catchment classification into classes of basins with homogeneous climatological and physiographic characteristics shows that none of the analyzed equations significantly improves the predictions. Finally, a new equation is proposed for the area of the Upper Po River Basin that can significantly decrease the time-of-concentration estimation error.

Understanding the cumulative impact of a set of stormwater interventions is of paramount importance in urban stormwater planning. A model is introduced to study the effect of 179 stormwater projects in the Cub Run watershed in Fairfax County, Virginia. The results showed gains in environmental benefits when the system was treated as a network of interacting and coevolving measures. The study demonstrates the importance of considering broader system benefits in the planning of stormwater projects instead of point performance. Under different implementation strategies, the dynamics of coevolution reveals a fixed pattern of growth in the number of projects to be implemented. However, their environmental benefits are not necessarily the same. Thus, any decision to advance some projects for early implementation would have a long-term impact on the system’s overall trajectory. Furthermore, at a certain threshold, the environmental benefits from these interventions could diminish since more control does not yield a linear increase in benefits.

Streamflow simulation is of great importance for water engineering design and water resource management. Most existing models simulate streamflow by establishing a quantitative relationship among climate, human activities, and streamflow and assuming the relationship is stationary in the long term. However, in a changing environment, this relationship may vary over time, resulting in the poor performance of many existing streamflow simulation models. In this study, inspired by data stream mining, adapting the gradient-boosted regression tree (XGBoost) to work in an online setting, a new statistically based model, called an online gradient-boosted regression tree (online XGBoost), is proposed to simulate streamflow dynamically in a changing environment. Here, the data of streamflow, climatic variables, and human activities are regarded as a data stream and the change in their relationships is treated as concept drift. The proposed model has two attractive properties. First, it makes it possible to capture the changed relationship between streamflow and its impact factors with the concept of a drift detection algorithm. Second, it can be used to simulate streamflow dynamically by updating models based on the concept drift detection results. Taking the Qingliu River catchment as a case study, the results show that the proposed method achieved good performance on monthly streamflow simulations during 1989 and 2010 with a Nash–Sutcliffe model efficiency coefficient (NSE) of 0.73. Furthermore, it outperformed comparable methods, including four statistically based methods (online support vector regression, online regression tree, online random forest regression, and online boosting tree regression) and four lumped parameter hydrological models (SimHyd, Sacramento, soil moisture accounting and routing, and Tank). The proposed model provides a useful tool for streamflow simulation in a changing environment. Findings will help water resource managers adapt to climate change.

The groundwater level (GWL) is an important subsoil condition that affects water resource management, soil contamination analysis, and infrastructure design. In this study, the spatial and time-dependent variation of the GWL was analyzed and a GWL prediction model based on river stage (RS) is proposed. A series of finite-element analyses were performed for the various geological and hydrological conditions. The results were used to quantify the response of the GWL to the RS for various permeability and RS conditions. A correlation model of the GWL and RS was established and the model parameters were evaluated to yield design equations. Implementation steps for the method were derived, including an adjustment procedure for cases with multiple RS fluctuations. Two case studies were used to verify the validity of the method by comparing its predictions with both the measured values and results from an artificial neural network (ANN). The correlation model produced well-predicted results in both the shape and values compared with the measured GWL.

Droughts result in billions of dollars in annual losses, loss of life, and the displacement of people. The characterization of hydrologic drought enhances the prediction of streamflow statistics that are important for various water resource applications. L-moments were used to examine the goodness of fit of selected 3-parameter probability distributions to the low streamflow series at 704 unregulated gaged rivers in the contiguous United States. Stream gages with and without zero flows were identified, and the use of the inverse transformation of nonzero low flows and censoring of both tails of the distribution were explored. Overall, the log-Pearson type III (LP3), 3-parameter lognormal (LN3), and Pearson type III (PE3) all appear to be acceptable distributions for describing the low streamflow series in the United States at sites with no zero low flows. For the inverse flows, the Generalized Extreme Value (GEV) distribution also performed well, while censoring did not improve the ability to distinguish between distributions.

In the design flood estimation procedure, the areal reduction factor (ARF) is used to convert ground-level point rainfall records into areal design rainfall for a reference area. Practically, the ARF is estimated through the fixed-area (${\mathrm{ARF}}_{\mathrm{fa}}$) scheme but has limitations as a statistical approach based on sparse ground-observation density. The ${\mathrm{ARF}}_{\mathrm{fa}}$ indicates potential biases because of the unsynchronized frequency analysis between point and areal rainfall. The storm-centered ARF (${\mathrm{ARF}}_{\mathrm{sc}}$) is obtained directly from individual storm captured by high resolution radar. In this study, the ${\mathrm{ARF}}_{\mathrm{sc}}$ values were estimated during the monsoon season (June to September) during 2007–2012 that covered the entire nation of South Korea, and then expressed as a function of reference area, duration, and return period. Both the ${\mathrm{ARF}}_{\mathrm{fa}}$ and ${\mathrm{ARF}}_{\mathrm{sc}}$ are proportional to the reference area. However, the most distinct difference is their responses to the return periods. For the fixed specific duration, the ${\mathrm{ARF}}_{\mathrm{fa}}$ indicates insensitive response to the return period over the reference area, whereas the ${\mathrm{ARF}}_{\mathrm{sc}}$ indicates quite varied declining rates according to the return periods. The ${\mathrm{ARF}}_{\mathrm{fa}}$’s invariant characteristics arise from its statistical scheme assuming identical distributions with similar shape parameters for the point and areal annual maximum rainfalls. The results can be used to avoid excessively conservative designs and assist in more economic and reliable uses of practical areal reduction factors.

Estimation of a discharge hydrograph is considered to be a fundamental research inquiry in both computational hydrology and hydraulics. As a tradition, in hydrometric stations, continuous monitoring of the stage is combined with sporadic measurement of discharge to develop the rating curve. To address the drawback in rating curve development, a coupled simulation-optimization methodology for system identification and subsequent discharge hydrograph estimation based on water level hydrograph alone is presented and validated for various synthetic and real case scenarios. For this purpose, a few experiments were designed to explore the extent to which measurement of the velocity field can be eliminated in discharge estimation. In the first numerical experiment, it was assumed that the river corridor was equipped with three hydrometric stations, while in the second one kinematic wave condition was replaced with the downstream hydrometric station. In the third experiment, the synthetic data were replaced by field data (i.e., Tiber River in Italy). The results for the first case show that the proposed methodology is very accurate for identifying the system and subsequently predicting discharge hydrograph. Goodness-of-fit criteria related to the second case are still remarkable but lower than the first case. As for the Tiber River data, the proposed approach managed to reproduce the timing of the discharge hydrograph both in the rising and falling limbs of the hydrograph quite well, but not the peak flow. The study also touches on the variability of channel roughness structure for discharge estimation giving more flexibility to model development. In light of this, a proposal is made to dismiss the traditional calibration and try to go for dynamic calibration as a by-product of discharge estimation.

Indoor experiments consisting of 15 flow discharges, six kinds of roughness, and five different gradients were systematically investigated to discover the evolution laws of rolling waves, flow regimes, and resistance characteristics. The results showed increasingly apparent production of rolling waves due to instability of overland flow on the slope surface when unit discharge increased. However, as discharge increased continuously, rolling waves disappeared. Meanwhile, the defined unit charge regions of rolling-wave flow decreased with increasing roughness. When slope gradient $J<0.1564$, the flow state index ($m$) of overland flow first increased and then decreased with increasing roughness. However, it decreased steadily when $J>0.1564$. Moreover, when $k_s\le 0.08\mathrm{mm}$, $m$ increased as the slope increased and decreased when $k_s\ge 0.08\mathrm{mm}$, with a mean value of 0.543. Thus, the overland flow on a slope is a mixed-flow zone dominated by transitional flow and supplemented by laminar flow. This is consistent with results obtained using the traditional discriminant method. However, the flow regimes existed in the laminar instability zone when a better discriminant method was adopted based on the viscosity-depth ratio. As for flow patterns, when the Froude number discriminant method was adopted, the water flow was supercritical flow. The resistance coefficient was inversely proportional to Reynolds number and was based on frictional resistance, thickness of the viscous sublayer, pressure drag, and roll waves resulting from the increasing-resistance phenomenon of roughness. Finally, a formula for calculating resistance is proposed based on the resistance characteristics [$f=(24/\mathrm{Re})+(\mathrm{14,041}/{\mathrm{Re}}_d^{1.33}{\mathrm{\Lambda }}^{3.11})$, $R^2=0.66$]. These results provide a theoretical basis for the prediction model of soil erosion on the slope and applies the theory of open-channel flow to overland flow.

Simulation based flood risk or rainfall-runoff studies require multiple extreme rainfall sequences as input, for which synthetic hyetographs are generally employed. Temporal distribution for synthetic hyetograph can be generated by deterministic approaches (geometric shapes, standard temporal distribution) or stochastic approaches/weather generators. The pertinent limitations are the missing inherent randomness of rainfall, limited numbers (former), and/or complexity in implementation (latter). Addressing these limitations, a novel data-driven nonparametric approach is proposed in this article for extraction of a huge number of temporal patterns of rainfall from observations. When compared to stochastic approaches, generation of synthetic hyetographs is easier in the proposed approach. Synthetic hyetographs generated in this approach represented the observed storms better than other deterministic approaches from literature.

In this study, an approximate mathematical formulation is derived to delineate a steady-state freshwater–saltwater interface by a horizontal scavenger well system in a confined aquifer that consists of two horizontal pumping wells. One horizontal well is above the freshwater–saltwater interface to recover freshwater. The other horizontal well is below the interface to extract saltwater. A set of dimensionless variables are introduced to facilitate solution derivations. Critical pumping rates that correspond to critical upconing and downconing interfaces are then derived for two different interface extents. Finally, the horizontal scavenger well system is applied to a real-world case, the 1,500-foot sand in Baton Rouge, Louisiana, and is compared to the current vertical scavenger well system, where the Field-Critchley analytical solution is adopted. The real-world case study demonstrates that among many advantages, the horizontal scavenger well system can recover more freshwater and extract less saltwater than the vertical scavenger well system. The study also indicates that the current freshwater and saltwater pumping rates are far below the safe zone and that freshwater is pulled into the saltwater well.