Journal of Hydrologic Engineering

Preferential flow may significantly affect the acceleration of contaminant transport, which can be described by the use of nonequilibrium models such as dual-porosity and dual-permeability models. Its modeling requires many parameters that are generally difficult to measure. The determination of the parameters in these nonequilibrium models from experimental measurements is a challenging task and is usually carried out using parameter optimization methods. In this study, prior to the identification process, an estimability analysis was carried out in order to determine the set of the most estimable parameters in the dual-porosity model from the available experimental data collected from the field lysimeter (pressure heads, water contents, cumulative percolation, and evaporation). The most estimable parameters were then identified, and the less estimable ones were fixed from previous studies, literature, or additional measures. The results obtained from the estimability analysis showed that five out of the eight parameters were estimable from the combination of daily pressure heads and water contents. The five estimable parameters were then identified using the Levenberg-Marquardt method implemented within one-dimensional finite-element model software. Saturated water contents in the mobile and immobile regions were impossible to estimate simultaneously due to their high correlation. When the water transfer coefficient was included in the identification procedure, the optimization results worsened due to its low estimability from pressure heads and water contents. Removing saturated water content in the immobile region and saturated hydraulic conductivity, which had a strong correlation with more estimable parameters, led to the improvement of the optimization results. As a result, the water transfer parameter was estimated with reliability.

This research aims at improving the performance of regional groundwater models by incorporating high-resolution elevation data into the head-dependent packages of MODFLOW. Model code specific to the evapotranspiration package (EVT) of MODFLOW was modified to account for the variability in elevation data and to effectively delineate the evapotranspiration (ET) simulated region at user-specified digital elevation model (DEM) resolution. The regional groundwater model of the Rincon Valley–Mesilla Basins (NMOSE-2007 flow model) was improved and considered to evaluate and validate the developed code. The base DEM of the study area is smoothened and aggregated to various resampled resolutions that are integer divisors of NMOSE-2007 flow model resolution for use with ET simulation. A gradual decrease in ET outflow is observed when the variability in elevation is eliminated across the grid cell. Also, changes in cumulative ET outflow (as a fraction of total outflow) at different resampled grids followed a similar trend during the simulation. The computational cost is high for the models simulated at fine resolution, whereas the simulation accuracy is low for the models simulated at coarse resolution. To select the optimum resampled DEM resolution to simulate the ET component of groundwater for use with the NMOSE-2007 flow model, a chi-square test of homogeneity was performed at 5 and 10% significance levels by considering computational cost and simulation accuracy as the base parameters. Results of the statistical analysis concluded that simulating ET component at 80.467-m resolution and integrating the outflows to model cell (402.336 m) resolution would significantly improve the performance of the flow model without compromising on the computational cost.

When applying a back-propagation neural network (BPNN) model in hydrological simulation, researchers generally face three problems. The first one is that real-time correction mode must be adopted when forecasting basin outlet flow, i.e., observed antecedent outlet flows must be utilized as part of the inputs of the BPNN model. Under this mode, outlet flow can only be forecasted one time step ahead, i.e., continuous simulation cannot be implemented. The second one is that topology, weights, and biases of BPNN cannot be optimized simultaneously by traditional training methods. Topology designed by the trial-and-error method and weights and biases trained by back-propagation (BP) algorithm are not always global optimal and the optimizations are experience-based. The third one is that simulation accuracy for the validation period is usually much lower than that for the calibration period, i.e., generalization property of BPNN is not good. To solve these problems, a novel coupled black-box model named BK (BP-KNN) and a new methodology of calibration are proposed in this paper. The BK model was developed by coupling BPNN model with $K$-nearest neighbor (KNN) algorithm. Unlike the traditional BPNN model previously reported, the BK model implemented continuous simulation under nonreal-time correction mode. Observed antecedent outlet flows were substituted by simulated values. The simulated values were calculated by the BPNN model first and then corrected based on the KNN algorithm, historical simulation error, and other relevant factors. According to the calculation process, parameters of the BK model were divided into three hierarchies and each hierarchy was calibrated respectively by the NSGA-II multiobjective optimization algorithm. This new methodology of calibration ensured higher accuracy and efficiency, and enhanced the generalization property of the BPNN. The accuracy of flow concentration module of Xinanjiang model is not always high enough, in order to combine advantages of conceptual and black-box models, XBK and XSBK models were proposed. The XBK model was constituted by coupling runoff generation module of Xinanjiang model with BK flow concentration model and the XSBK model was constituted by coupling runoff generation and separation module of Xinanjiang model with BK flow concentration model. BK, XBK, XSBK, and Xinanjiang models were applied in Chengcun, Dongwan, and Dage watersheds. The simulation results indicated that improved models obtained higher accuracies than Xinanjiang model and can overcame limitations of traditional BPNN model.

A common approach to multivariate flood frequency analysis is to apply discharge time series records with the same length (i.e., one upstream and one downstream discharge gauge in the same river for bivariate flood frequency analysis, and multivariate flood frequency analysis at the confluence of river systems). However, in reality the gauged discharge time series records may have different lengths due to different activation times of the gauges. In addition, there exists one common assumption for flood frequency analysis, that is, the discharge time series may be considered as a stationary signal. However, due to land-use and land-cover (LULC) and climate changes, the stationary assumption may need to be justified. To answer the above questions, this paper investigates (1) the full-length discharge record at each discharge gauge; (2) the dependence structure of bivariate and multivariate discharge time series with different lengths using the copula theory; (3) employment of the vine copula for multivariate flood frequency analysis (i.e., $d\ge 3$); and (4) the validation of the proposed method and comparison of its performance with asymmetric, symmetric Archimedean, and meta-elliptical copulas using the discharge time series from the Cuyahoga River basin, Ohio, as a case study.

Rainfall forecast errors are the key sources of uncertainty in flood forecasting. To quantify this uncertainty operational flood forecasting centers make use of rainfall forecasts obtained by ensemble predicting systems (EPS). The EPS forecasts are generated by perturbing the initial conditions of numerical weather prediction models. Question, however, remains whether these EPS cover the real forecast uncertainty range and whether the EPS-based uncertainty estimates are similar to the ones obtained by statistical methods. Both questions are addressed in this research based on data of a flood forecasting system in Belgium. Comparison is made between the uncertainty bounds generated by EPS and by a Monte Carlo-based statistical method after historical forecasted rainfall uncertainty analysis. The latter analysis calculates the error between forecasted and observed catchment rainfall, taking into account the dependency on lead time and rainfall depth. The forecasted rainfall errors are described by truncated normal distributions, which allow to calculate the full uncertainty distribution on a deterministic rainfall forecast. It is concluded that the EPS may underestimate the influence of the total forecasted rainfall uncertainty. For the Belgian case study of the Rivierbeek, the forecasted rainfall uncertainty explains 30% of the total river flow forecast uncertainty.

A case study examining aspects of nonstationarity in the estimation of flood magnitudes in the Southeast and mid-Atlantic United States is presented here. The focus is on challenges posed by changes in the number and location of stream gauges over time to assess climate variability and land-use land-cover (LULC) change impacts. The study shows that statistical outliers are spatially aligned with LULC change from urbanization independently of time period and record length across the region with substantial increases ($>30\%$) in the magnitude of the 25-year and 100-year events since 1950. Large amplification ($>10$) of the ratio of flood magnitude to the mean annual flood was detected in expanding urban areas, but only systematically for the 100-year event, suggesting that existing infrastructure constrains the sensitivity of flood response for events below applied engineering design criteria. Overall, annual streamflows do not exhibit statistically significant trends (98% cfl), except for positive trends in urban areas ($<7\%$), and negative trends in regulated rivers and regions of karst processes ($<12\%$). The critical importance of long-term, high-density observing networks as the basis to assess and interpret climate change and LULC impacts is stressed.

Small dams and associated reservoirs have notable effects on soil and water dynamics in prairie streams. In this study, we developed a simulation module of small dams in the soil and water assessment tool (SWAT) to evaluate their long-term effects on stream flow and water quality at a watershed scale. To overcome the challenges in characterizing small storage and short retention time in small reservoir routing, concepts of equivalent reservoir storage and equivalent reservoir discharge are applied by which the average daily storage and daily discharge of the small reservoirs are calculated. Accordingly, the sediment deposition and nutrient abatement within the reservoir are computed using available SWAT routines. The effects of small dams in reduction of daily peak flow, sediment, and nutrient loads at the watershed outlet are obtained by summing the effects of all small dams within the watershed considering both reservoir and channel processes. The model is applied to the $74.6\text{-}{\mathrm{km}}^2$ South Tobacco Creek watershed located in Southern Manitoba of Canada. A total of 26 small dams exist in the watershed with surface area ranging from 0.002 to $0.492{\mathrm{km}}^2$ and storage capacity from 3,380 to $\mathrm{642,000}{\mathrm{m}}^3$. The simulation results show that the combined effect of these small dams can reduce daily peak flow by 0–14% at the watershed outlet depending on climate and initial reservoir storage conditions. The estimated average annual sediment, total nitrogen and total phosphorus reductions at the watershed outlet are about 4.51, 3.59, and 2.96%, respectively. However, the on-site effects of individual small dams are much higher depending on its size, location, shape, drainage area, and land use compositions in the contribution area. The simulation results also show that snowmelt-flooding events have higher reduction amounts but lower relative reduction rates compared to rainfall storm events and the back-flood small dams have greater impact on sediment, nitrogen, and phosphorous abatement followed by multipurpose small dams and dry dams in the study watershed.