Topographic parameterization in regional and continental scale hydrology
Armstrong, Robert N.
The focus of this thesis is on the effects of spatial scale on the parameterization of regional and continental scale drainage basins for surface hydrologic applications. Specifically, this thesis investigates the impact of grid size or resolution of digital elevation models (DEMs) on basin and channel network delineation, and derived topographic variables. The viability of substituting coarse resolution grids for finer grids, and appropriate grid sizes for hydrologic applications are also considered. The digital landscape analysis tool TOPAZ was used for processing of a raster DEM and segmentation and parameterization of the basin. Basin and network properties are extracted from DEMs of varying grid size or resolution, using the same parameter values to generate the drainage network at each scale. Results of an application using this technique are presented for two continental scale basins in North America: the Mackenzie River basin, located in northwestern Canada; and the Missouri River basin, located in the northwestern United States. TOPAZ output, which includes spatial data (raster maps) and tabular data (statistics), is interfaced with a GIS for further processing. The effects of grid size are evaluated by considering changes in the spatial distribution, statistical properties, and derived topographic variables of the basin and the network. Basin and drainage network properties obtained from DEMs with an initial grid resolution of 1 km are used as the standard or reference. Variations in basin and network properties obtained from the coarse resolution DEMs are compared to those obtained from the 1 km DEMs. Increasing the grid size from 1 to 64 km showed several trends. First, the spatial distribution of the basin and network changes with increased grid size, with greater change in the spatial distribution, in general, observed for grid sizes larger than 8 km. Aggregation or averaging of the elevation data results in a much smoother representation of the topography, and basin extents that remain similar or show considerable change has an effect on the systematic smoothing of elevation. Averaging also reduces the value of most variables (e.g., basin area, number of sub-basins, mean elevation, slope, total channel length, mean channel link length, mean distance to nearest channel, mean distance to outlet, mean drop to nearest channel, and mean distance to outlet). Variables remain relatively constant to 8 km resolution (3 aggregations) and tend to depart considerably from the values obtained at 1 km resolution for grids larger than 8 km. Results of an analysis of hydraulic slope suggest that a mean hydraulic slope of about 1% may be used to determine the coarsest grid size for reproducing drainage features (e.g., basin and network) and variable values that are similar to those obtained from the reference DEM. General rules for determining appropriate grid sizes, from previous research, are also compared with findings of this research. Results of the comparison suggest that the application of general rules for determining appropriate grid sizes are limited to the scale at which the general rules are derived. The substitution of coarse resolution DEMs in continental hydrology is constrained by the redirection of flow across large flat areas, and elevation errors produced by averaging of the DEM. Elevation errors produced along valley bottoms result in the blockage of drainage. As a result of general changes in basin and network configuration, and derived topographic variables for grid sizes larger than 8 km, it is suggested that a grid size of approximately 10 km or less be used for continental hydrologic applications.