Geoscience Reference
In-Depth Information
calculated. The main issue in the construction of a TIN is the best discretisation or
tessellation
of the
space to represent the topographic form of a catchment most efficiently. The study by Nelson
et al.
(1999) provides a technique for automatically generating a TIN representation of the topography from
elevation points or a vector DEM. Once a TIN is defined, algorithms are also available for automatically
delineating the river network and catchment area for any point on the network (Palacios-Velez and
Cuevas-Renaud, 1986; Jones
et al.
, 1990). A number of distributed rainfall-runoff models have been
based on a TIN representation of topography (such as the tRIBS model of Ivanov
et al.
, 2004, 2008, and
the Pennsylvania Integrated Hydrological Model of Qu and Duffy, 2007).
The idea of analysing the topography of the catchment to give an indication of flow pathways is
clearly an attractive one and can result in some attractive computer graphics when model predictions
are superimposed back onto a three-dimensional picture of the topography. However, there are some
limitations to such analyses that the user must be aware of. Regardless of the algorithms or type of DEM
used, all DEM analyses depend crucially on the assumption that the flow pathways will be controlled
predominantly by the topography of the catchment. This will only be a good assumption for catchments
with relatively shallow soils underlain by impermeable or near impermeable bedrock. If there are deeper
flow pathways they may deviate significantly from those suggested by an analysis of the surface topog-
raphy. Recent work has also shown that even in shallow systems, the bedrock topography may have a
greater control over downslope saturated flow than the surface topography, at least in some catchments
(McDonnell
et al.
, 1996). Finally, it is worth repeating that, even under ideal hydrological circumstances,
an adequate representation of the flow pathways will require a DEM of fine enough resolution to define
the shapes of the hillslopes. Analyses suggest that, for raster data, resolutions coarser than 100 m will
not suffice.
3.8 Geographical Information and Data Management Systems
Topographic data are just one type of distributed data that are becoming more readily available to the
hydrological modeller in the form of digital geographical information systems (GIS). These are software
packages that allow different types of spatial data to be overlain and manipulated. Most GIS do not
easily handle data variables that change over time but some, such as the Institute of Hydrology Water
Information System (WIS), have been specifically designed for time variable data. Several have facilities
for the analysis of flow directions, although they are generally limited to single flow direction algorithms,
such as in the ARC-VIEW or GRASS packages. There are now many hydrological modelling packages
with links to GIS packages such as ArcHydro (Maidment, 2002), Green Kenue from Environment Canada
and the open source RHydro.
Other variables that can be easily stored and manipulated within a GIS are maps of vegetation type,
soil type and geology. The data can then be used in different ways. The characteristics of each element
of a raster grid of arbitrary size could be derived, for example. Or, by overlaying different layers of
information, irregular
hydrological response units
(HRUs) of different characteristics could be identified
(see Section 6.6). A map could be displayed of all grid elements or HRUs having similar characteristics
and so on (see, for example, Figure 2.7).
The problem for the hydrological modeller is that the types of information that are generally available
in GIS form are only indirectly relevant to the rainfall-runoff processes. Knowing the soil and vegetation
type classifications of an HRU is certainly informative, but what parameter values should be used for
each classification? In principle, such parameter values could be stored directly within a GIS but only
if the values are known. Soil type, as mapped by the soil scientist, for example, may not be the same as
that required by the hydrologist. In the UK, a hydrological classification of soil types is now available
in GIS form (the HOST classification, Boorman
et al.
, 1995) but this is based on expectations of soil
hydrological behaviour and does not directly give parameter values required in a hydrological model. In
