Journal of Hydrologic Engineering

Hourly rainfall depths are probably the most important data input for the simulation of watershed, instream, and water quality processes such as when using the Hydrological Simulation Program FORTRAN model. However, the density of stations measuring precipitation data on an hourly basis is low, so the disaggregation of daily rainfall values or the transfer of information from nearby stations is commonly applied. This investigation evaluated the accuracy of four methods for estimating hourly precipitation depths and assessed whether or not the accuracy of the estimates was a function of the distance between stations. Hourly rainfall records from 74 gauges located within or near the Chesapeake Bay watershed were used to evaluate the biases and accuracies of procedures for disaggregating measured daily depths into hourly amounts or transferring rainfall information for nearby gauges. None of the four methods provided accurate predictions of hourly rainfall, which suggests that it is difficult to get reliable estimates of hourly rainfall intensities by disaggregating daily totals or transferring hourly values from nearby gauges. However, the transfer of daily totals is more accurate than the estimation of hourly depths.

A multiple linear regression (MLR) model, in conjunction with Geographic Information Systems technology, was used to derive the relationship between annual precipitation and elevation, longitude, and latitude. The island of Crete, in Greece, was used as the case study. A multiscale precipitation analysis was performed on areas ranging from large areas (the whole island and the northern, southern, and eastern parts of the island), to medium areas (watersheds), to small areas (sub-basins). While the MLR annual precipitation estimates (which used elevation, latitude, and longitude information) were found to be more reasonable than estimates obtained using elevation only when applied to the whole island, the difference between the MLR estimates and the elevation-only estimates was smaller when applied to individual watersheds. The MLR provides realistic estimates for mean areal precipitation for the island of Crete: 700±100, 950±150, and 1,300±200 mm for dry, average, and wet years, respectively. Elevation-rainfall gradients are: 0.45–0.6, 0.6–0.9, and 0.9–1.3 mm/m for dry, average, and wet years, respectively. Of this, 44% falls on the northern, 33% on the southern, and 23% on the eastern parts of the island for a typical average year.

Currently, the river forecasting system deployed in each of 13 River Forecast Centers of the National Weather Service primarily uses lumped parameter models to generate hydrologic simulations. With the deployment of the weather surveillance radar 1988 Doppler radars, more and more precipitation data with high spatial and temporal resolution have become available for hydrologic modeling. Hydrologists inside and outside the National Weather Service are now investigating how to effectively use these data to enhance river-forecasting capabilities. In this paper, six years of continuously simulated hydrographs from an eight-subbasin model are compared to those from a single-basin (or lumped) model, both applied to the Blue River basin (1,232 km²) in Oklahoma. The Sacramento soil moisture accounting model is used to generate runoff in all cases. Synthetic unit hydrographs for each subbasin convey the water to the outlet of the basin without explicit flow routing. Subdividing the basin into eight subbasins captures spatially variable rainfall reflected in the next generation weather radar products and produces improved results without greatly increasing the computational and data requirements. Strategies for calibrating the hydrologic model parameters for multiple subbasins are explored.

Information regarding topographic, meteorologic, geologic, and geomorphic characteristics is increasingly available in spatially explicit digital formats. Of interest is whether enhanced spatial processing of newly available digital grids can lead to new estimators of watershed characteristics which may in turn, improve our ability to predict extreme hydrologic events. Regional hydrologic models of low-flow processes often produce estimators with unacceptably large errors. Using a continuous digital elevation model (DEM) of the conterminous United States, watershed boundaries were developed for the streamflow gauges of the USGS’s Hydro-Climatic Data Network. Using these watershed boundaries, many watershed characteristics were developed from digital grids, including: the original DEM, the USDA’s State Soil Geographic grids, and the Spatial Climate Analysis Service’s orographically weighted precipitation and temperature grids of varying spatial and temporal resolution. Digital processing of grids leads to improvements in estimation and reproducibility of spatial statistics over traditional manual processing approaches. Low-flow regional regression models were developed for regions across the conterminous United States. Inclusion of the new watershed characteristics led to improvements in regional regression models for all regions. The inclusion of hydrogeologic indices, in particular a new smoothed baseflow recession constant estimator, led to dramatic improvements in low-flow prediction.

As a part of a study examining alternatives for flow augmentation for fall run Chinook salmon migration on the Cosumnes River, California, two models were constructed to simulate channel flow combined with seepage losses. The first model is a numerical, one-dimensional diffusion wave approximation of the St. Venant equations coupled to a seepage routine that uses either a Green and Ampt infiltration routine or a head-based seepage routine that can handle seepage to or from the channel. The second model uses an analytical solution to the diffusion wave approximation of the St. Venant equations and also incorporates the Green and Ampt method to represent losses due to seepage for the simulation of flow over a dry riverbed. The two models are used to simulate low flow periods in the Michigan Bar to McConnell reach of the Cosumnes River in order to determine flows associated with a minimum depth necessary for fall run Chinook salmon migration. These computed flows were compared against historical data in order to determine volumes of flow augmentation for the time period October through December which corresponds to the spawning period of fall run Chinook salmon. Computed flow augmentation volumes were compared with available reservoir storage in the watershed in order to determine if such flow augmentation would be possible. Both models proved capable of simulating low flow periods on the Cosumnes River. During dry years, the computed flow augmentation volumes represent a significant percentage of the available reservoir storage, indicating that augmentation flows may not always be possible for the entire time period from this source alone.

A sample-data system discretization of the continuous Kalinin–Milyukov–Nash cascade is performed in a state-space analysis framework allowing for stream–aquifer interactions that include bank storage during flood events and groundwater discharge to the stream under low-flow conditions. These interactions generally result in faster attenuation of propagating flood waves and in elevated streamflow levels during drought conditions. An example is given that demonstrates how accounting for these processes ensures more reliable streamflow forecasts. The model is based on a simplified physical description of the processes both in the stream and across the stream–aquifer boundary.

Pumping from a well beside a spring also drains water from the spring in a process that is termed spring depletion. A solution is obtained for this problem by assuming that the spring partially penetrates an aquitard that overlies the pumped aquifer. The Boulton solution for delayed-yield flow to a well and the Duhamel superposition theorem are used to obtain a Volterra integral equation of the second kind, and the Laplace transform is used to solve this integral equation. Numerical values computed from this solution are checked with a direct numerical solution of the integral equation. Flow depleted from the spring depends upon dimensionless time and four dimensionless parameters, and the solution is illustrated by showing in plots the effect of changing time and one parameter while holding all other parameters constant. The plotted results appear physically reasonable.

The focus of this paper is the optimum use of mulch (biodegradable erosion control blankets and waste ballast) on steep slopes (batters) to aid the establishment of grasses for erosion control. Biodegradable blankets are commonly used on steep slopes as mulch to minimize the risk of washout of soil, seeds, and ameliorants during grass establishment to control erosion. However, the blankets can be costly when laid on all sections of the batters. In order to reduce costs, it is suggested to lay the blankets on critical sections such as the outer verge (top) and the bottom. Waste ballast spread on a railway batter provides a degree of permanent protection against erosion, and is useful where fire is a potential hazard. Runoff and soil loss were monitored on experimental plots of a railway embankment batter between December 2000 and May 2001. The cost-effective erosion control strategies on the plots were centerd on control (doing nothing after surface preparation), limited use of erosion control blankets, and variation of waste ballast density on some seeded plots. It was observed that waste ballast and erosion control blankets laid on the outer verge have the potential to control incipient rills by spreading runoff. Erosion control blankets placed at the bottom section of the embankment have the potential to induce deposition of eroded material. Runoff production and soil loss decreases with an increase in the application density of waste ballast, up to a certain level. Waste ballast spread on railway batters enhances grass germination and sustains grass growth.