Journal of Hydrologic Engineering

Results from diagnostic analyses to understand the seasonal evolution of the large-scale climatic state responsible for the development and melt of the winter snowpack, and spring–early summer precipitation in the Churchill Falls region on the Québec-Labrador Peninsula, Canada, are presented in the context of the development of an empirical model for seasonal to annual streamflow forecasting, with a special emphasis on the May–July spring freshet. Teleconnection indices and gridded global measures of atmospheric circulation inferred from the National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis are used as climatic indicators. Composite and correlation analyses are applied to the climatic indicators conditioned on the spring streamflow for identification of potential predictors. Meridional and zonal atmospheric fluxes over the Atlantic and the Pacific Oceans emanating from regionally persistent sea surface temperature/sea level pressure modes are identified as potential carriers of information. We speculate on the ocean-atmosphere and regional hydrologic mechanisms that may be involved in lending multiseasonal predictability to streamflows in the region.

The performance of different models and procedures for forecasting aggregated May–July streamflow for the Churchill Falls basin on the Québec-Labrador peninsula is compared. The models compared have different lead times and include an autoregressive model using only past streamflow data, an autoregressive with exogenous input model utilizing both past streamflow and precipitation, and a linear regression model using the principal components of exogenous measures of atmospheric circulation inferred from the National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis project. The forecast skills of the different approaches are compared using a variety of measures of performance. The results indicate that relatively accurate forecasts using only measures of atmospheric circulation can be issued as early as in December of the prior year. A multimodel combination approach is found to be more effective than the use of a single forecast model. In addition, it is concluded that forecasting models utilizing atmospheric circulation data are useful, especially for basins where hydroclimatic observations are scarce and for basins where flows and other hydroclimatic variables are not strongly autocorrelated (do not depend on their past).

Precipitation areal reduction factors (ARFs) for the $\mathrm{685,000}{\mathrm{km}}^2$ of Texas were calculated using next generation radar (NEXRAD) rainfall estimates. The study was based on 18,531 storms of different durations that took place in different seasons and regions of Texas. The storms were assumed of elliptical shape. It was found that, in addition to the storm duration and area, other factors such as the season, region, and precipitation depth (i.e., storm total rainfall accumulation for the given duration) have a statistically significant effect on the ARFs. Elongated ellipses and orientation angles somewhat parallel to the Texas gulf coast were found more frequent in winter, when warm and cold fronts produce frontal storms, than in summer. The effect of the precipitation depth on the ARFs was found to be stronger in summer than in winter. Even though part of the ARF variability could be explained by seasonality, regionality, and precipitation depth, the uniqueness of each storm event appears to be an important cause of this variability. Lower ARF values were observed compared to previous studies.

The determination of the annual exceedence probability (AEP) of extreme water levels in complex estuarine systems is an important and challenging issue in flood management. Extreme estuarine levels are caused by the combined effects of river flows, local winds, and coastal ocean levels. This paper describes a stochastic, event-based approach for generating concurrent hourly mean sea level pressure (MSLP) and wind speed, which are needed in a larger project to derive stochastic hourly river flows, winds, and coastal ocean levels to drive a hydrodynamic model for estuarine flood level simulation. The minimum MSLP versus rainfall and maximum wind speed versus rainfall relationships for the extreme flood events at the daily time scale are first used to generate minimum daily MSLP and maximum daily wind speed using a nonparametric resampling method. The generated minimum daily MSLP and maximum daily wind speed are then used to scale concurrent hourly MSLP and wind speed sequences resampled from the historical record. This approach is novel because it is simple, generic, and computationally efficient for generating stochastic cross-correlated forcing time series for any estuarine hydrodynamic and flood risk study. The approach is tested using 50 years of forcing data from the Gippsland Lakes system in southeast Australia. The results indicate that the approach produces realistic stochastic concurrent hourly sequences of MSLP and wind speed that preserve and distinguish the MSLP-rainfall and wind speed-rainfall relationships for different synoptic weather conditions and times of the year. The approach also produces hourly and daily minimum MSLP and maximum wind speed with similar AEP characteristics as the historical data.

Active microwave remote sensing data (e.g., radar) can be used for estimation of soil moisture beneath the ground surfaces up to 0 to $5\mathrm{cm}$ depths. A number of analytical and empirical models are available that relate the synthetic aperture radar (SAR) data to the surface soil moisture. Each of these models has its own merits and demerits. In this paper, results obtained from a study on the application of artificial neural networks (ANN) for soil moisture retrieval from ERS-2 SAR data over bare and vegetated surfaces is presented. SAR images of three dates (i.e., July 28, 2003, March 29, 2004, and May 3, 2004) were acquired over a portion of the Solani river catchment in India. A feed forward back-propagation neural network was used for establishing the relationship between surface soil moisture and terrain as well as sensor variables. A maximum of seven input variables were considered for training the ANN. These are: Digital number (DN) or backscatter coefficient ( $\sigma$ deg) (as the case may be) of each pixel of SAR image selected individually, incidence angle of radar beam, land cover, surface roughness height, terrain height, leaf area index, and plant water content. Several ANN experiments were conducted and the coefficient of determination $\left(R^2\right)$ and root-mean-square error (RMSE) were determined between the ANN derived soil moisture and the observed soil moisture through in-situ measurements taken concurrently with the satellite pass. It has been observed that it is the backscattering coefficient $\sigma$ deg rather than DN, used as one of the inputs that produced high $R^2$ $(\sim 0.9)$ and low RMSE. The results also indicate that only a few number of input variables may be sufficient to retrieve the soil moisture with high accuracies using the ANN. The leaf area index has been found to be as good as the single bulk variable representing the vegetation characteristics in the study area. In addition, the ANN technique has yielded more accurate results than the traditional statistical regression, indicating its usefulness for soil moisture estimation from microwave remote sensing data.

A method to disaggregate daily rainfall into hourly precipitation is evaluated across Texas. Based on measured precipitation data, the method generates disaggregated hourly hyetographs that match measured daily totals by selecting storm intensity patterns from measured hourly databases using a one parameter selection method. The model is applied across Texas using historic data measured at hourly precipitation gauges; performance is evaluated by the model’s ability to reproduce the measured hourly rainfall statistics at each gauge. Initially, each hourly precipitation gauge is evaluated in its performance as a disaggregation database. Based on a cluster analysis of the results to determine which precipitation databases should be used for general disaggregation, no preferences in space or among gauge characteristics (e.g., period of record, precipitation statistics) were identified. As a result, a Texas state database that contains all of the measured hourly data in Texas is proposed for use by the disaggregation algorithm. The state database is verified for a selection of gauges across Texas and performed as well at a given station as using that station’s measured rainfall for the disaggregation. The method is further applied to estimate intensity-duration curves, which show that the disaggregation algorithm captures the majority of the storm intensities well and diverges by less than 17% from the measured intensities for the extreme runoff-generating events.

This study investigates the effect of local systematic and random errors of the commonly used tipping-bucket (TB) rain gauges on the accuracy of runoff predictions. A physically-based rainfall-runoff model is applied to analyze several storms in a humid midsize watershed in south Louisiana. Two types of systematic TB errors are considered, wind-induced losses and underestimation of high rainfall intensities due to the lack of gauge dynamic calibration. The effect of the TB errors is assessed by comparing hydrographs simulated using uncorrected and corrected rainfall input. The results indicate that wind and dynamic calibration effects can cause differences in estimating runoff volumes and peaks on the order of 5 to 15%. We also analyze the effect of random TB errors caused by the discrete gauge sampling mechanism. The computed runoff differences caused by the TB random errors were dependent on the magnitude of the runoff discharge, and on the temporal resolution of the rainfall input. However, the effect of random errors was found to be rather small, especially for large discharge.

Soil erosion occurring at highway construction sites can contribute large amounts of sediment to the local stream network and degrades overall water quality. The environmental impacts of the sediment-laden stormwater runoff can be mitigated by employing appropriate best management practices (BMPs). Silt fence is one of the most widely used BMPs at highway construction sites. Numerous field studies have shown that silt fence installations at highway sites have a high degree of failure due to flow around the end of the fence, and due to concentrated erosion along the toe of the fence. To address these problems, erosion control manuals often recommend installing a tieback (also known as “j-hook”) at the down slope end of the fence. However, no one has systematically studied the sediment removal efficiency of tieback designs to develop a rational procedure for field installations. In this research, we quantified the effectiveness of silt fence systems installed with tiebacks by comparing its efficiency against systems installed without tiebacks. We completed this comparison study using an intermediate-scale laboratory model that was designed to study the erosion processes occurring at highway construction sites. The model allowed us to observe various erosion and sediment transport processes in a controlled setting. The results show that silt fence installations utilizing tiebacks remove sediment primarily through the process of sedimentation, not by filtration. Installation of a tieback improves the ability of the silt fence system to act as a temporary detention basin and greatly enhances the system’s ability to remove suspended particles by sedimentation. Our experimental data shows that a well designed silt fence with tiebacks can remove up to 90% of solids transported by the runoff water.

Longitudinal solute transport modeling requires knowledge of the transport parameters involved. At best, these parameters have been estimated from stream tracer experiments, usually conducted, however, at flow rates different from those characterizing the intended application. Thus, the mathematical simulation of breakthrough curves invariably rests on a transfer of parameters, resulting in errors of an a priori unknown magnitude. In the research reported here, data from nine stream tracer experiments performed on the same channel reach have been used to find out, up to which ratio of respective flow rates transport parameter transfer will still yield acceptable modeling accuracy. The results of the study (obtained with the transient storage model of longitudinal solute transport) suggest that parameter transfer across a range of flow rates not exceeding a ratio of 1:2 will be associated with errors in peak concentrations below the 40% mark.

A numerical simulation for a longitudinally one-dimensional upscaled solute transport model is developed in order to predict the mean solute concentration in natural streams subject to irregular variations in various flow and transport parameters and forcing conditions. A solute transport equation at the local scale of a river cross section is deterministic, but the equation at the river reach scale is stochastic due to various uncertainties in flow and transport parameters in every reach. The upscaling transformation makes a transport equation at the river reach scale deterministic. Compared with the transport equation at local scale, the upscaled equation at the river reach scale has seven covariance integrals, which incorporate the influence of the stream variabilities on the mean value of solute concentration. One of these integrals provides the explicit model of the macrodispersion coefficient as it varies in time and space under varying flow conditions. Lagrangian trajectories of flow are simulated to compute the time-space covariance integrals (including macrodispersion) of the flow properties. The Monte Carlo simulation of the stochastic solute transport is implemented to verify the upscaling methodology. The validation exercise confirms that the upscaled model is accurate, feasible, and physically meaningful. The Monte Carlo simulation with the stochastic Saint-Venant flow model also supplies all the random flow information used in both the predictive transport model and the validation transport model.

The discrete linear cascade model (DLCM) was applied for historical flow routing along the Nebraska City–Rulo section of the Missouri River in southeastern Nebraska. With the help of optimized model parameters it has been possible to identify the triggering mechanism responsible for historical changes in the stage-discharge relationship at Rulo over the 1952–2006 period, following construction of Gavins Point dam on the Missouri above Sioux City, Iowa. It was found that in the second part of the past century flood celerity in the study reach slowed by about 25%, most likely caused by an increase in the Manning roughness coefficient as a result of a large increase (almost a doubling) in the number of wing-dykes constructed over the reach within the same period. The ease of application and minimal data requirement (only discharge values at regular intervals) makes the DLCM a practical tool for stream-flow analysis. It also can serve as a preliminary investigative tool for more advanced and detailed hydraulic approaches that typically require a data-rich environment and significantly greater development time.