Spatial variability is all that stands between hydrology and science, forcing us to deal in probabilities and averages. Because of scale, we can not consider forces on individual soil particles, water molecules and solute ions when addressing human size problems. We must therefore look at aggregate properties and mean values for parameters and inputs in computer modeling of hydrologic phenomena. This research explores the impact of spatially variable inputs on the Water Erosion Prediction Project soil erosion computer program. Distributions of input variables are generated and assigned randomly to a grid of homogeneous rangeland hillslope elements. Values for runoff volume and sediment loss from each flow path are recorded and averaged to provide a distribution of outputs in the form of a sensitivity analysis. Variabilities of slope, slope length, soil textures, soil characteristics, terrain, convex and concave slopes, soil saturation, rainfall amount and vegetation were examined. Results show that use of mean inputs values in the WEPP representative hillslope model yields very similar outputs to the spatially variable research model using a distribution of inputs in all simulations in the case of totally random bare rangeland soils. When a decreasing trend in soil clay content is introduced in the variable model, the hillslope model using average values as inputs no longer provides a good estimate of the sediment loss. When random vegetation is generated and added to the simulation, runoff volume continues to be similar between the two models, but the sediment loss is much higher in the spatially variable model. In addition, the results of the standard hillslope model are much less responsive to changes in slope than those of the spatially variable model. It is concluded that spatial variability of soils must be considered when there is a linear change in input values with slope position. Likewise spatial variability of vegetation needs to be addressed in order to accurately estimate erosion on the rangeland watersheds considered in this dissertation. It is also found that this type of simulation provides a model for sensitivity analysis of a complex computer programs. Physically related inputs can be generated in such a way as to preserve the desired interrationships and distributions of inputs can be directly compared to generated distributions of outputs.
| Identifer | oai:union.ndltd.org:arizona.edu/oai:arizona.openrepository.com:10150/191165 |
| Date | January 1991 |
| Creators | Parker, Ronald Dean,1948- |
| Contributors | Lane, Leonard J., Fogel, Martin M., Hawkins, Richard H., Post, Donald F., Huete, Alfredo R. |
| Publisher | The University of Arizona. |
| Source Sets | University of Arizona |
| Language | English |
| Detected Language | English |
| Type | Dissertation-Reproduction (electronic), text |
| Rights | Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author. |
Page generated in 0.0017 seconds