Journal of Hydrologic Engineering

Data assimilation is a useful tool in hydrologic and agricultural application studies because of its ability to produce predicted results with high accuracy. However, different data-assimilation methods have different performances for a given application. Although the popular ensemble Kalman filter (EnKF) performs well with Gaussian distribution, the error is difficult to conform to the Gaussian distribution. To take advantage of the EnKF, this study presents a new data-assimilation method, ensemble particle filter (EnPF), which is an integration of the EnKF and the particle filter (PF). This new method was evaluated in comparison with two existing methods (EnKF and PF) through soil temperature predictions. The simple biosphere model (SiB2) and the filters were assessed with observations from the Wudaogou experimental area in the Huaihe River basin, China. Results show that when the time interval increases adequately, all the simulated or assimilated results improve significantly. All of these filters tend to be more stable when the number of particles reaches a certain amount (e.g., 60) or the variance is small (e.g., less than 0.6) in the study. When the number of particles is less than a threshold value (e.g., 30), the advantage among these three methods is not appreciable. The error obtained by EnPF is smaller than that by EnKF and PF; this means that EnPF performs better than EnKF and PF.

This study evaluates the implicit safety level in hydrologic infrastructure design, performed in common practice with the calculation of the peak-flow frequency curve for high return periods ($100<\mathrm{return}\mathrm{period}<\mathrm{10,000}\mathrm{years}$) through use of single-peak hyetographs in a semidistributed hydro-meteorological event-based model. Through consideration of two basins located in mainland Spain, the paper studies the effect on the peak-flow frequency curve of (1) the shape of the design hyetograph defined by the alternating blocks (AB) method or through applying a Monte Carlo (MC) procedure; (2) the catchment area (ranging from 140 to $\mathrm{1,800}{\mathrm{km}}^2$); and (3) the extent of disaggregation in the hydro-meteorological model (varying from 14 to 66 subbasins). The effect on hydrologic safety was measured through using the ratio $R=\mathrm{QmaxAB}/\mathrm{QmaxMC}$, with QmaxAB, the peak flows obtained with the AB method and QmaxMC, the application of MC. The paper finds the following: (1) that the degree of safety decreases as the basin area increases, regardless of the extent of disaggregation in the model used; (2) a greater disaggregation in semidistributed hydrologic models leads to greater water safety, mainly for smaller basins (in this study, $150{\mathrm{km}}^2$); and (3) at the level of subbasin, the degree of safety decreases as the time of concentration and area increase.

Because natural flow (NF) values are either not directly measured or have the potential to contain considerable error, when deemed necessary, the reconstruction of a reliable NF series is ostensibly important. Selecting the appropriate method depends on available data. For a time period before reservoir construction (pre-reservoir construction period), the only available data for ungauged basins came from the neighboring basins and simulated flow used in a rainfall-runoff model. A new Kalman-based method developed in this paper looks to reconstruct the NF series using the state fusion technique, which is then compared with the area ratio method, the maintenance of variance (Move) type III method, and the multivariable regression method using different quality indexes (QIs). In the perspective of the post-reservoir construction period, when hydrometric data (i.e., turbine flow, water level in the reservoir, and discharged flow) is collected in an ungauged basin (with no flow measurements), a new water balance equation (WBE)-based method is recommended for reconstructing and filtering the NF data using an optimization technique that would then be compared with the classic WBE that implements different QIs.

During the prereservoir years, when limited data (neighboring basin flow and simulated flow) are available for ungauged basins, simple methods are relevant to reconstruct natural flow (NF). On the other hand, for the postreservoir years, when hydrometric data (water level in the reservoir, turbine flow, and discharged flow) are available for ungauged basins, the classic water balance equation (WBE) method is usually employed to determine NF. However, applying classic methods still produce noisy and unreliable flow data series. In an attempt to produce more reliable NF data, new methods are proposed and described in Part 1 of this paper: the state fusion Kalman filter method (prereservoir construction period) and the deterministic WBE method and the stochastic WBE method (postreservoir construction period). This paper that is presented as Part 2 deals with the execution of these new methods for three subbasins of different sizes in Quebec, Canada. Obtained results are then compared against those produced by the use of the current methods of NF reconstruction: area ratio method, maintenance of variance (Move) type III method, and multivariable regression method (prereservoir construction period), and the classic WBE model (postreservoir construction period). This comparative analysis shows the capability of proposed methods for improving the constructed data series with improved results that do not contain any negative flow. They are less noisy, perfectly matched with regional flows, and are reliable enough for frequency analyses. Because these NF values include evaporation, infiltration, and other losing terms of a closed hydraulic system, with more studies needed in order to define the contribution of each of them.

Determination of water-table elevation corresponding to aquifer recharge or discharge is an important issue in sustainable groundwater management. This approach necessitates the application of numerical simulation models that may require substantial amounts of input data related to aquifer parameters and specifications, even for modeling only part of the aquifer, which makes the calculations expensive. Moreover, comprehensive aquifer modeling is a time-consuming and computationally intensive process. Artificial intelligence tools can replace simulation models and decrease computational efforts by using input and output data sets without considering complex relations of the system to be modeled. This paper employs an adaptive neural fuzzy inference system (ANFIS) and genetic programming (GP) as artificial intelligence tools to extract governing groundwater flow equations in Ghaen and Karaj aquifers in Iran. For both aquifers, several input-output data sets, for both training and testing data sets, are determined by using a developed numerical simulation model [iterative alternating direction implicit method (IADIM)]. In addition, the water table elevation at each cell in the model is considered to be a function of aquifer recharge and discharge at the current period as well as water table elevation at the previous period. Application of ANFIS and GP models in these case studies illustrates the superior flexibility of GP over ANFIS in time series modeling. In fact, GP provides water-table elevation results with less root mean squared error (RMSE) as the error criterion, especially in the testing data set. Thus, GP is a good candidate for use in groundwater modeling.

A three-point approximation method, the perturbance moments method (PMM), is applied to quantify the uncertainty of predicted dissolved oxygen (DO) concentrations in the Buffalo River that arise from data variability and parameter uncertainty. A risk assessment is then conducted to systematically compute the probability of exceeding a predetermined water quality standard. Numerical investigations reveal that the overall uncertainty of the output DO concentrations does not necessarily equal the sum of the uncertainties of all individual variables or parameters of interest. Therefore, different sources of uncertainty can have compensative or additive effects on the overall uncertainty of output DO concentrations. In a particular case study of the Buffalo River, the variability of measured concentrations was demonstrated to have the largest influence on the overall uncertainty of predicted DO concentrations. Notably, data with a higher collection frequency should have a smaller uncertainty, and therefore a smaller variability in the expected values, and so contribute less to the magnitude of the standard deviation of the output DO concentrations. The effective risk that DO concentrations exceed the selected standard over time is evaluated. The effective risk that the DO concentration drops below the selected standard is relatively high in a higher-temperature season.

The present paper addresses critical issues that describe the transient transfer of stored rock-matrix flow into high-permeable fractures and rate-limited diffusive solute flux into low-permeable rock matrix using a typical dual-porosity approach. An improved mathematical model is suggested that better describes fluid flow through a coupled fracture-matrix system using a dual-porosity approach. The suggested model differs from a conventional model as the fracture flow equation contains a hyperbolic term in addition to the conventional dispersive term. The matrix flow equation contains the coupling term that controls the transient nature of fluid exchange from the stored rock matrix into the hydraulic conductors. The Langmuir sorption isotherm is suggested to describe the limited sorption sites available on fracture walls, while the Freundlich sorption isotherm is recommended to describe the unlimited sorption sites available within the rock matrix. The dispersion mechanism in a coupled fracture-matrix dual-porosity system becomes more complex as the convective longitudinal dispersion coefficient diverges resulting from huge variations in mean velocities of streamlines between the fracture and rock matrix.

The energy and mass balance model for terrestrial ice and snow is an essential tool for the future projection of interannual snow and ice storage including glaciers. However, the snow models in the hydrologic engineering field have barely considered the long-term behavior of the snow and ice storage because the time scale of glacier dynamics is much longer than those of river flow and seasonal snowmelt. This paper proposes an appropriate treatment for inland glaciers as systems in dynamic equilibrium that stay constant under a static climate condition. It is conjectured that the snow and ice vertical movement from high-elevation areas to valleys (lower elevation areas) by means of wind redistribution, avalanches, and glacial motion may be considered as an equilibrator of the snow and ice storage system because it stimulates snow and ice ablation. In order to demonstrate this concept, a simple dynamic equilibrium model for regional to global scales is proposed and applied to the Pamirs. The interannual snow and ice storage in the Pamir Mountains may lose about half of their volume by the end of 21st century by the anticipated global warming based on nine selected general circulation model (GCM) projections.

The curve number (CN) method is widely used for estimating direct runoff depth from rainstorms. The procedure is on the basis of the parameter CN, a lumped expression of basin absorption and runoff potential, and a second parameter, initial abstraction (IA), which represents the interception, infiltration, and surface depression during the early part of a storm. The evaluation of CN in Sicily at a basin scale from rainfall-runoff multiday events is done using rainfall-runoff observations during the period 1940–1997 (mean record length of 20 years) in 61 Sicilian watersheds using three different methods: (1) the national engineering handbook, section 4 hydrology (NEH4) method (NEH4M) (the median CN for the annual flood events); (2) asymptotic fitting of ordered and natural data; and (3) least-squares method using rainfall-runoff using both ordered and natural data. Asymptotic fitting showed a major occurrence of the standard CN response (43 basins), with a lesser complacent response (10 basins), and a few cases of violent behavior (three basins). For five basins, the data quality and/or small sample size did not allow the identification of a CN. The least-squares and asymptotic fittings furnished similar results but showed that the original assumption of the initial abstraction ratio ($\mathrm{IA}/S$ or $\lambda$) of 0.20 in watersheds with a standard and violent CN response is unusually high. Using both natural and ordered rainfall-runoff data, a median $\lambda$ value of 0 and 0.05 was found for natural data and ordered data, respectively. The CNs found by the recommended handbook method (NEH4 method) were higher than those found from the handbook tables. The asymptotic fitting and least-squares fitting methods were all lower than the table values. Comparisons were made between the CNs estimated using hydrologic soil-cover complexes on the basis of Soil Conservation Service (SCS)–National Resources Conservation Service (NRCS) handbook tables and the CN evaluated from rainfall-runoff multiday events in 36 of the basins. Overall, although comparisons were poor, the NEH4 method and asymptotic fitting method that used all of the ordered data were closest to the CNs from the handbook tables. The link between the CN estimated with the asymptotic fitting method (AFM) (T-series, ordered data) and the watersheds’ morphoclimatic and hydrologic characteristics was also investigated, and a useful relationship for basins with a larger drainage area ($S_{\mathrm{DA}}>160{\mathrm{km}}^2$) was found.

Modern dams are overwhelmingly designed under the assumption of climatic stationarity by using a static design value known as probable maximum precipitation (PMP). Therefore, it is worthwhile to explore the impact of relaxing the assumption of stationarity and recalculating design PMP values by using currently practiced procedures enhanced by numerical modeling or observational climate trends. This study reports the findings of nonstationary PMP recalculations at three large dam sites in the United States (South Holston Dam in Tennessee, Folsom Dam in California, and Owyhee Dam in Oregon). The results indicate that currently accepted PMP values are significantly increased when future changes in dew points from observational trends or numerical models are taken into account. It is plausible that such future changes in these meteorological thresholds, had they been known among the engineering community when PMPs were designed, would have received the necessary attention regarding the future uncertainty of stationary PMP values as a dam ages.