Geoscience Reference
In-Depth Information
raster DEMs have been built by interpolating from digitised contours (Figure 3.3b) and, as a result, may,
in places, be subject to significant error, particular where in flat topography where there are few contours
(particularly where, in some cases, databases store only integer elevation values) or where there are short
steep slopes. Topography can also be efficiently represented as a
triangular irregular network
(TIN,
Figure 3.3c). There is potential for developing good resolution topographic data from photogrammetric
analysis of aircraft- or satellite-derived stereo images or directly from aircraft-borne laser altimetry (e.g.
Weltz
et al.
, 1994) or synthetic aperture radar (SAR). Aircraft platform techniques can give elevations
on a grid of down to better than 2 m with elevations of an accuracy of 0.1 m, although it must be pointed
out that these will be elevations of the surface seen by the sensor, which is not necessarily the same as
the ground surface where there is significant tall vegetation cover or buildings.
NASA's Shuttle Radar Topography Mission (SRTM) provided a digital elevation model derived from
SAR technology at a global resolution of 90 m and a refined resolution of 30 m in the USA. Since 2009,
this has been superceded by the Advanced Spaceborne Thermal Emission and Reflection Radiometer
(ASTER) mission global DEM which provides a wider coverage at a resolution of 30 m between latitudes
of 83 degrees north and south. This was calculated from 1.3 million near-infrared images using stereo-
scopic methods. It has been suggested that there are some artifacts in the ASTER DEM and that it is inferior
to the SRTM data at the same resolution. It is worth remembering that, as hydrologists, we are generally
interested in relative elevations and gradients to derive flow pathways but this is a difference operation
that is sensitive to uncertainties in the data (e.g. Endreny and Wood, 2001). Thus, it will be worth in-
vestigating the data retrieved from these types of database for unusual features that might be artifacts of
the way they have been produced. Topographic data are also useful, however, in modelling patterns of
evapotranspiration and snowmelt (e.g. Kafle and Yamaguchi, 2009)
Water does have a tendency to flow downhill, at least for shallow hydrological systems, so that knowing
something about the form of the topography should have some utility in hydrological modelling. Dis-
tributed models can clearly use this type of data directly and there are also models, such as TOPMODEL,
that are based on a prior analysis of the catchment topography (see Section 6.3 and Box 6.1). Resolution
is clearly an issue here. Coarse resolution DEMs will not be able to provide an adequate description of
hillslope flow pathways; distributed models may not be able to use all the information in a fine scale
DEM because of computational constraints. Variables derived from topographic data, calibrated param-
eter values, and model predictions within distributed models based on DEMs are known to be sensitive
to grid resolution (e.g. Zhang and Montgomery, 1994; Bruneau
et al.
, 1995; Quinn
et al.
, 1995; Saulnier
et al.
, 1997a).
The analysis of a DEM to derive apparent flow pathways has been an interesting topic of research in
itself. The methods available depend on whether a raster or vector DEM is available. For raster DEMs,
a comparison of methods was published by Tarboton (1997). For each grid cell, there are eight possible
flow directions. There may be several surrounding grid elements with elevations lower than the cell being
considered. The problem is how to distribute the potential flow to these different possible pathways.
Some inaccuracy is clearly inevitable, but the methods that give the best results, at least visually, appear
to be the multiple flow direction algorithm of Quinn
et al.
(1995) (Figure 3.4a) and the resultant vector
method of Tarboton (1997) (Figure 3.4b). An analysis program based on the multiple direction algorithm
is available as freeware (see Appendix A).
With vector data, the problem is how to derive the lines of greatest slope or
streamlines
for flow on the
hillslopes. The idea is that water will follow the same direction of flow as a ball running down the same
(smoothed!!) surface topography. Since, under this assumption, water should not cross a streamline, it
may then be possible to represent the flow between two streamlines, in a
stream tube
, as a one-dimensional
flow of varying width in the downslope direction (two dimensions if the vertical is taken into account,
as in Figure 3.5). This is the basis for distributed models such as TOPOG (Vertessy
et al.
, 1993) and the
Institute of Hydrology Distributed Model (Calver and Wood, 1995). The streamlines should always be
at right angles (orthogonal) to the contours. If the contours are available in digital form, calculating the
