Terrestrial Hydrology Model (THM)

Objectives

• Homogeneity - A unit hydrograph is generally applied to a basin or sub-basin, where it is assumed that the relevant hydrologic parameters are homogeneous throughout. For this reason, sub-basins or hydrologic response units should be chosen such that land use, soils, and slopes are relatively constant throughout. This is often times a very difficult task, as soils and the landscape in general change from one location to another.
• Linearity - Unit hydrograph procedures operate on the assumption of linearity. This concept is best illustrated by considering that the unit hydrograph is essentially a mathematical filter. For a given input precipitation profile, the output hydrograph is the mathematical convolution of the input profile with the unit hydrograph. Accordingly, the model's output hydrograph will respond to the input precipitation in essentially a linear fashion. Although a real watershed will respond differently to storms of various magnitudes, this varying response will almost certainly not be linear.
• Timing Parameter - A timing parameter, whether it be time of concentration, lag time, or time to peak, is required in synthetic unit hydrograph procedures. This timing parameter is at best difficult to estimate and often makes little or no reference to the physical properties of the watershed. The timing parameter effectively controls the timing and peak of a synthetic unit hydrograph.

Results

Because the THM is capable of simulating a surface response only, a series of small summer storms which produced rather small, fast peaking hydrographs was selected in the expectation that these small storms would be comprised of a limited or negligible amount of return flow from the saturated zone. The Mahantango Creek Watershed (MCW), a USDA-ARS research water shed in east-central PA, was chosen as the test basin. Instrumentation in the MCW includes several meteorological stations, groundwater monitoring stations, and a multistage weir at the outlet. For the initial testing, a total of 6 storms was chosen for each of which all 4 of the available precipitation gauges were active during the event. For these tests the SCS curve number method was used to estimate the hydrologic abstractions. Comparisons were made utilizing all 4 rain gauges as opposed to the temporal average precipitation of the 4 gauges. The motivation for this was the planned linkage of the THM and the MM. Since the MM yields only the average precipitation over the grid cell, it is desirable to understand how the output hydrograph is affected by assuming uniform precipitation over the grid cell instead of the actual, spatially varying, precipitation. As testing began, it was discovered that the initial abstraction term in the SCS method was quite sensitive. With this in mind, the testing evolved in a slightly different manner.

Two versions of the experiment were run. In the first, the average precipitation for the entire test basin was used to calibrate by comparing the predicted runoff hydrograph to the observed runoff hydrograph. The literature typically reports values of to be approximately 0.2; however, lower values are often commonplace. An , was calibrated in this manner for each of the 6 test storms. The calibrated was then used in investigating the use of the 4 individual gauges (herein referred to as the spatially varied case) as opposed to an average. In the second version of the experiment, was calibrated for the spatially varied case and used for the average precipitation case. Both versions gave similar results: for a given initial abstraction, the spatially varied rainfall of the 4 gauges produced a higher peak than did the average rainfall of the same 4 gauges. This is undoubtedly due to the "smoothing" effect of mathematical averaging.

In the formulation of the kinematic wave approximation for overland flow, the Manning's "n" or overland flow roughness factor plays an important role. The more rough the surface, the more the flow is retarded and hence, the slower the overland velocity. This essentially slows the travel time of the kinematic wave as it travels down slope. In the current model, all input describing the earth's surface comes in the form of a gridded or raster data set. In an effort to avoid an additional data layer describing the overland flow roughness, a series of "n" values based on the flow accumulation value of the cell were chosen. The cells with very low flow accumulation (i.e. the ridge tops) were given a roughness value of 0.4. Cells with a very high flow accumulation value were assumed to be the agricultural lands and were given an "n" value of 0.12 and finally, the cells in the middle were given a Manning's "n" value of 0.20 to 0.25.

In an effort to test the sensitivity of this parameter, a series of model simulations were made with Manning's "n" values. The Manning's "n" values ranged between 0.05 and 0.8. For the test runs, all cells were assigned the same roughness value. The comparisons were made on the peak flow versus the actual peak as recorded for the event. This series of simulations indicated that the roughness factor has very little sensitivity above values of approximately 0.5 with the most sensitive values being 0.1 and below. The range of recommended roughness values for the conditions found in the experimental area is 0.05 to 0.4 with most of the basin being be tween 0.05 and 0.2. The "n" value had considerable impact on the peak flow within the recommended range; however, when a homogeneous "n" value anywhere between 0.05 and 0.2 was applied to the basin, the resulting simulated peak flow was consistently within approximately 25% of the actual recorded peak flow. Interestingly, the "n" value had limited effect on the timing of the predicted runoff hydrograph.

Overall, the model is able to simulate the 6 storms rather well. The curve number method has proven to be the most applicable and produces good results. This is most likely due to the difficulty in estimating and the sensitivity of the required parameters for the Green-Ampt routine. Investigation of the Green-Ampt method is continuing.

Future Plans

• A groundwater or subsurface routine -- It is absolutely essential to provide a subsurface flow component to the existing model. This component will enable the analysis and simulation of events and basins which have substantial return flow components in the hydrographs. This component is currently under development.

• A reservoir delineation and operation routine -- Because there are several reservoirs (both controlled and uncontrolled) in the Susquehanna River basin, it is necessary to model the effects of these reservoirs if a viable model is to be developed. Current work includes investigation and development of algorithms to identify reservoirs and other water bodies through the use of remotely sensed and other spatial data structures. Additionally, it is intended that the model will recognize reservoirs and that reservoir operation procedures will become part of the model.

• Snow melt component -- Although snow melt is normally not a major consideration in the Susquehanna River basin, the winter of 1993-1994 saw some regions of the basin receiving their largest snowfalls in history. As a result of this large snowpack, the soil was saturated for most of the spring melt and heavy rains during this time resulted in stream flooding. Thus, snowfall had a significant impact on the basin. A snow melt component is currently being investigated for addition to the model.

• Additional infiltration routines -- Because of problems encountered in the existing infiltration routines, additional methods are being considered.

Table of Contents

Previous Section

Next Section

Download Report