Journal of Hydrologic Engineering

This study is the first to investigate and improve a quasi–physically based model, MicroMet, by fitting its adjustment factor ($\chi$) for a large data-scarce mountainous watershed in an arid area in Northwest China. The derived factors for both the whole study area (Derived Factor I) and different elevation classes (Derived Factor II) were determined and compared with the original factor in MicroMet for precipitation estimates, with a training period from 1990 to 2010 and a validation period from 2011 to 2013. Results show that the original factor in MicroMet is more suitable for estimating high precipitation over low-elevation areas (below 2,000 m above sea level), but it is not suitable in high mountainous areas. Both Derived Factors I and II can improve the performance of precipitation estimates, and more reliable adjustment factors could be obtained with more in situ observations. The Barnes objective analysis scheme used in MicroMet is more suitable for interpolating large precipitation events with small variability in humid areas; it requires adjustments for applications in arid areas. Although determined from limited observations in the study area, Derived Factor II performed better than Derived Factor I in winter, indicating that the derived factor for the elevation classes is more suitable for estimating extremely low precipitation with greater variability in data-scarce, high-elevation mountainous watersheds in arid areas.

The high impact of river ice phenomena on the hydrology of cold regions has led to the extensive use of numerical models in simulating and predicting river ice processes. Consequently, there is a need to utilize efficient and robust sensitivity analysis (SA) methods to characterize the role of different parameters on the functioning of these models. To gain greater insight into how the internal parameters affect a river ice model’s behavior, this paper presents a comparative performance investigation of the two global SA methods: (1) the recently proposed variogram analysis of response surfaces (VARS); and (2) the widely used regional sensitivity analysis (RSA). The methods were benchmarked on a one-dimensional hydrodynamic river ice model of the Lower Dauphin River, Manitoba, Canada. Furthermore, using a bootstrapping strategy, a procedure was developed to estimate confidence intervals on the resulting sensitivity indices and evaluate reliability of the inferred parameter rankings. Results show that (1) the water levels simulated by the river ice model are most sensitive to the ice cover characteristics (i.e., porosity and thickness at the ice cover front) and upstream discharge; (2) the hydraulic roughness parameters and slush ice properties (i.e., porosity and thickness of the slush pans) are medium- and low-sensitivity parameters, respectively; (3) the VARS and RSA methods provide contradictory assessments regarding the sensitivity of the model output to variations in the slush ice porosity and ice roughness parameters; and (4) the VARS method appears to be superior to RSA in terms of generating robust estimates of the parameter sensitivity rankings.

The Canadian Precipitation Analysis (CaPA) produces a gridded product by assimilating data from surface stations and radar, using a background field provided by the Global Environmental Multiscale (GEM) model. This study assesses the performance of two satellite-based rainfall estimates for Canada and the results of their assimilation within CaPA. Evaluations are completed on 10 years of rainfall data for June, July, and August. The satellite-based rainfall estimates considered are those from the Climate Prediction Center morphing (CMORPH) method and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN). Relative to the second generation of adjusted daily precipitation for Canada (APC2), it is found that both satellite products possess unique characteristics within central Canada in the form of greater skill on par with the greatest observed rainfall across all of Canada, and an unrivalled ability to detect large rainfall events (between 2 and $50\mathrm{mm}/\mathrm{day}$) within the scope of this study. This exceptional performance is correlated in space and time with the occurrence of convective rainfall events. Comparing the two satellite products’ categorical scores, it was found that CMORPH is better at detecting rainfall events and estimating the events’ magnitude than PERSIANN. Two CaPA configurations are tested: one where CMORPH is used as a background field in place of the GEM model and one where it is used as an additional data source. Combining CMORPH with stations results in an overall increase in skill, except for the detection of light rainfall events less than $1\mathrm{mm}/6\mathrm{h}$. When CMORPH is used alongside stations to adjust the GEM model, the resulting analysis is a combination of the strengths of both gridded datasets used, in that light rainfall events are influenced by the GEM model and larger events by CMORPH. The ability of CMORPH to detect and estimate convective rainfall events within central Canada, specifically events greater than $2\mathrm{mm}/6\mathrm{h}$, has proven particularly useful when assimilated with the GEM model.

The effects of climate change are increasingly considered in conjunction with changes in water demand and reservoir sedimentation in forecasts of water supply vulnerability. Here, the relative effects of these factors are evaluated for the Washington, DC metropolitan area water supply for the near (2010–2039), intermediate (2040–2069), and distant (2070–2099) future by repeated water resources model simulations. This system poses water management challenges because of long water-delivery travel times that increase uncertainty, multiple water jurisdictions that constrain potential decisions, and future scenarios that simultaneously increase demand and decrease water supply during the critical summer period. Adaptation strategies were developed for the system using a multiobjective evolutionary algorithm. Optimized reservoir management policies were compared using six distinct objectives ranging from reservoir storage to environmental and recreational benefits. Simulations of future conditions show water stress increasing with time. Reservoir sedimentation is projected to more than double (114% increase) the severity of reservoir storage failures by 2040. Increases in water demand and climate change are projected to further stress the system, causing longer periods of low flow and a loss of recreational reservoir storage. The adoption of optimized rules mitigates some of these effects, most notably returning simulations of 2070–2099 climate to near historical levels. Modifying the balance between upstream and downstream reservoirs improved storage penalties by 20.7% and flowby penalties by 50%. Changing triggers for shifting load to off-line reservoirs improved flowby (8.3%) and environmental (4.1%) penalties slightly, whereas changing demand restriction triggers provided only moderate improvements, but with few adverse effects.

Most flood risk studies quantify flood likelihood and damage as functions of depth. The length of time that a flood remains above a stage (duration) is also necessary to quantify infrastructure vulnerability. The proposed approach adds a second dimension, time, into flood risk studies by developing a joint distribution of flood stage and event duration. Bivariate flood frequency model metrics are developed; the metrics are illustrated using data from four gauged sites, two riverine and two tidal. It is concluded that a general joint distribution model is applicable to both riverine and tidal sites. A minimum time between events (interevent time) determines the assignment of sequential periods to a single event or multiple events; this measure is shown to affect the statistics of the tidal sites, but not the riverine sites. The flood stage-duration-frequency curves developed in this study demonstrate a new approach that allows stakeholders to consider the coupling between depth and duration, and thereby develop preparedness and response plans that address temporal aspects such as time out of service in addition to depth-dependent damages.

Any change in characteristics of low flows under climate change may have important effects on various socioeconomic, environmental, water resource, and governmental planning aspects. The aim of this research was to use various low-flow indices to investigate low flows for a future period (2015–2044) under the effects of climate change for the case of Sezar basin, southwestern Iran. In this research, outputs of 10 atmosphere–ocean general circulation models (AOGCMs) under an A2 emissions scenario were used to predict temperature and precipitation scenarios under climate change. The low-flow indices used in this study included those extracted from a flow duration curve (discharges exceeded 70, 90, and 95% of the time, i.e., $Q_{70}$, $Q_{90}$, and $Q_{95}$), indices of low-flow frequency analysis (annual minimum 7-day flow with 2, 10, 20, and 100 return periods, i.e., $\mathrm{AM}{7}_{T=2}$, $\mathrm{AM}{7}_{T=10}$, $\mathrm{AM}{7}_{T=20}$, and $\mathrm{AM}{7}_{T=100}$), and flow deficit characteristics (number of dry days in a year, maximum length of dry period, deficit volume, and deficit intensity). The results indicated that the status of low flow in the future will be better than the baseline (1971–2000) in all subbasins, so values of indices including $Q_{70}$, $Q_{90}$, and $Q_{95}$ as well as $\mathrm{AM}{7}_{T=2}$, $\mathrm{AM}{7}_{T=10}$, $\mathrm{AM}{7}_{T=20}$, and $\mathrm{AM}{7}_{T=100}$ will increase and values of indices of deficit volume and intensity will decrease. The results also showed consistency between the indices derived from the flow duration curve and deficit characteristics, whereas indices of frequency analysis were determined to be inherently different.

Design storms are used for the sizing of urban drainage systems and for delineating floodplains. Recorded rainfall extremes are commonly used to develop design storms with the assumption of stationarity. However, the projections provided by climate models indicate a continued increase in the magnitude and frequency of rainfall extremes. Therefore, a continuous revision of design storm distribution is required. The objective of this study is to propose a framework for the development of site-specific temporal rainfall distributions (design storm distributions) from the historical rainfall data, using a clustering technique. The $k$-means for the longitudinal data ($KmL$) package on the R platform was used to cluster the storm distributions prepared from the historical rainfall data. The effects of temporal rainfall patterns and the antecedent moisture condition (AMC) on the peak discharges were assessed by comparing them with the results obtained from the standard Soil Conservation Service (SCS) Type II curve. The proposed framework was successfully implemented using the long-term hourly rainfall data obtained from the Asheville gauging station in North Carolina. It was found that 14, 61, and 25% of the storm distributions were categorized under the lower, middle and upper cluster curves, respectively. The clustered curves, which represent the ranges of severe storm events, can be considered design storm distributions for the study area. For the same hydrologic engineering center-hydrologic modeling system (HEC-HMS) model setup, the difference in peak discharges obtained from the SCS Type II curve and the cluster curves were as high as $179.1{\mathrm{m}}^3/\mathrm{s}$. This result indicates that the use of the SCS Type II curve for estimating the design flood can cause the overdesign of structures. The difference in the peak flows is even more pronounced for wetter initial soil moisture conditions. In consideration of the percentage change in the peak flows and the preceding storms, the antecedent wetness needs to be included in estimating design floods.

Hydrological models must be carefully calibrated to ensure that predictions are scientifically sound and reliable. As a goodness-of-fit measure, an ideal threshold of the Nash-Sutcliffe efficiency coefficient ($E_{ns}$) has been reported to be 0.65, based on yearly and monthly simulation results with the soil and water assessment tool (SWAT). To further explore the use of $E_{ns}$ as a goodness-of-fit measure for daily runoff simulation with SWAT, the authors ran the model with five different parameter values, using Jinjiang River Basin on the southeast coast of China as the study area. In addition, to find how the model versions varied in predicting changes in flood and drought with changes in land use, each of the five model versions were run two times separately: once for the land use condition in 1985, and then again for the land use condition in 2006. The authors investigated the relationships between $E_{ns}$ variation and model accuracy and discussed the differences in predicted runoff due to land use change as caused by the five model versions. The results show that the coefficient of determination ($R^2$) was highly associated with $E_{ns}$, with a high $R^2$ corresponding to a high $E_{ns}$; the $E_{ns}$ value was quite sensitive to simulation results of the maximum one-day (1d) runoff; 0.75; the $E_{ns}$ value was not sensitive to simulation results of the minimum 1d and seven-day (7d) runoff, and unlike the flood runoff, the errors in drought flow increased adversely with increased $E_{ns}$, indicating the poor performance of the SWAT model regarding drought flows; and the variations in runoff simulations, due to the land use change, were sensitive to model version switching and were fundamentally similar to the changes in the error indicators said above. The five SWAT versions varied in performance ($E_{ns}$) in daily runoff simulation and had different measures of goodness-of-fit for various characteristics of the flood and drought processes. As a result, an $E_{ns}$ of 0.75 is suggested as a threshold for satisfactory simulation of maximum 1d discharge in the study watershed.

The spatial variability of surface roughness can affect the overland flow processes. To accurately simulate the overland flow processes of watersheds, many theories have been developed which consider the spatial variability in surface roughness. Several theories claim that the surface roughness has anisotropic attributes, but there is little experimental evidence to support this claim. The authors developed a hydraulic laboratory model to simulate land surfaces, and replicated water flows in different directions over the exact same overland surface. Two typical strip overland surfaces were simulated, and the Manning’s roughness coefficient is calibrated. The results show that in a completely turbulent regime when the angle $\theta$ of the flow direction and the; overland strips direction are 90° and 45°, respectively; the average Manning’s roughness coefficient values $n$ are 0.0438 and 0.0492 over a $60\times 60\mathrm{mm}$ simulation overland surface, and 0.0276 and 0.0382 over a $120\times 60\mathrm{mm}$ simulation overland surface, respectively. Over the exact same strips of surface, there are different Manning’s roughness coefficient values in different flow directions. This proves that the surface roughness of watersheds has anisotropic attributes, which will have effects on overland runoff path and distribution.