Journal of Hydrologic Engineering

The National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center recently began issuing new multiple long-lead outlooks of meteorological probabilities. Operational hydrology approaches for generating probabilistic hydrological outlooks must be compatible with these meteorological outlooks yet preserve spatial and temporal relationships observed in past meteorology. Many approaches, however, either limit the use of historical data to be compatible with meteorological outlooks or limit compatibility with the outlooks to allow fuller use of historical data. An operational hydrology approach that uses all historical data while remaining compatible with many of the new long-lead outlooks, in order of user priority, is described here. The approach builds a hypothetical very large structured set of possible future scenarios, to be treated as a “sample” from which to estimate outlook probabilities and other parameters. The use of this hypothetical set corresponds to the weighted use of a scenario set based on historical data. The determination of weights becomes an optimization problem for the general case. An example illustrates the concepts and method.

A diffusion wave model of distributed catchment dynamics is presented. The effects of catchment topography and river network structure on storm-flow response are incorporated by routing surface runoff in cascade throughout a digital elevation model (DEM) based conceptual transport network, where the Muskingum-Cunge scheme with variable parameters is used to describe surface runoff dynamics. Dynamic scaling of hydraulic geometry is also incorporated in the model by using the “at-a-station” and “downstream” relationships by Leopold and Maddock. Numerical experiments indicate that the model is more than 98% mass conservative for possible slope and roughness configurations, which may occur for hillslopes in a natural catchment. Fluctuations in the simulated discharge may occur in response to discontinuities in rainfall excess representation if Courant number Cu during the simulation exceeds a threshold of about 3. Catchment scale simulations with different temporal resolution show that the model response is independent of structural parameters (model consistency). Also, the overall accuracy is preserved for computationally inexpensive space-time discretizations (for which Cu > 3) because fluctuations that may occur at the local scale are dampened when propagating downstream. Comparison of model results with observed outlet hydrographs of the Rio Missiaga experimental catchment (Eastern Italian Alps) show this approach to be capable of describing both overland and channel phases of surface runoff in mountainous catchments.

Transforms reduce the distributions of surface-runoff and sediment-transport variables to near-exponential distributions typical of rainfall duration and intensity. Two transform parameters are determined from field measurements using either a graphical method or the method of moments. The transforms were tested on both small experimental plots and very large watersheds for the distributions of rainfall depth, runoff volume and discharge, sediment concentration and sediment discharge, as well as concentration and transport of chemicals from surface runoff. As expected from deterministic relationships, the inverse transform exponent $\stackrel{\mathit{ˆ}}{\mathit{b}}$ for point rainfall is close to unity, increases to about 1.5 for runoff discharge, and varies between 1.5 and 3 for sediment and chemical transport. Several examples show that the transforms are useful to determine exceedance probability, flow and sediment duration curves, expected sediment load, and sediment-rating curves for poorly correlated concentration and discharge measurements.

Soils are often layered at scales smaller than the block size used in numerical and conceptual models of variably saturated flow. Consequently, the small-scale variability in water content within each block must be homogenized (upscaled). Laboratory results have shown that a linear volume average (LVA) of water content at a uniform suction is a good approximation to measured water contents in heterogeneous cores. Here, we upscale water contents using van Genuchten's function for both the local and upscaled soil-water-retention characteristics. The van Genuchten (vG) function compares favorably with LVA results, laboratory experiments under hydrostatic conditions in 3-cm cores, and numerical simulations of large-scale gravity drainage. Our method yields upscaled vG parameter values by fitting the vG curve to the LVA of water contents at various suction values. In practice, it is more efficient to compute direct averages of the local vG parameter values. Nonlinear power averages quantify a feasible range of values for each upscaled vG shape parameter; upscaled values of N are consistently less than the harmonic means, reflecting broad pore-size distributions of the upscaled soils. The vG function is useful for modeling soil-water retention at large scales, and these results provide guidance for its application.

The spatial variability of soil hydraulic properties that govern infiltration were investigated at two sites in the Konza Prairie Research Area in Kansas. One site was a pristine prairie site with native grass, and the other was a continuously cropped agricultural site from which wheat had been recently harvested. Soil hydraulic properties studied were saturated hydraulic conductivity, matrix flux potential, soil sorptivity, and pore-size distribution parameter. Spatial variability was assessed through semivariograms that were constructed based on 37 measurements at each site using the Guelph Permeameter. Correlation lengths of the infiltration properties varied from 1 to 4 m. Sorptivity was found to be lognormally distributed. The spatial structure of infiltration properties at both sites exhibited pseudoperiodicity and a hole-effect variogram model was used. Infiltration properties at the agricultural site indicated multiple peaks in the variograms. Implications of these results on computing areally averaged infiltration at the two sites are discussed.