|
GIS Guide to Good Practice |
|
4.4 Derived Data
You will often be using data derived from other sources when
creating or managing a GIS data set. There are often important considerations
in documenting derived data sets, as discussed in Section 5.
When deriving data from another source, or when making use
of derived data, it is the responsibility of the data user
to ensure that any intellectual property rights belonging to the
initial data creator(s) are respected. In some cases this may
simply be a requirement to acknowledge the originating source,
in other cases a royalty payment may be due for some part of the
data to be used. Be sure to check out the situation in advance. One of the core components within many GIS databases is the Digital
Elevation Model or DEM.
This sub-section will look briefly at the principal pathways you can take to
create a DEM, special issues that practitioners should be aware of relating to
accuracy and integrity, and the specialised process-related metadata that should
be recorded. DEM vs DTM. There is considerable confusion
between these two terms - which many people use interchangeably.
Some people also use the term Digital Surface Model (DSM). DSM appears to be a synomym for DEM, but with the added possibility
of being a component of a stack of surface models, for example
modelling atmospheric or subsurface layers. DSMs are not yet in
common usage. Strictly speaking, the term DTM,
Digital Terrain Model, should be reserved for those models of reality which include
information relating to surface texture, etc., in addition to
information regarding elevation. The term Digital Elevation Model,
DEM, should be reserved for
representations of altitude alone. There are few genuine DTMs
around yet - the concept is established but the tools to display, or visualise,
information of this nature are not yet fully/widely available. A DEM normally consists
of a regular matrix of elevation values, from which altitude functions
such as slope and aspect can be calculated, and which may be rendered
for visualisation as isolines (contours), perspective or panoramic
views, etc. A DEM is typically described in terms of its horizontal
resolution. Resolution, for a DEM, defines the horizontal and
vertical precision by which the information is recorded. A typical
example might define the horizontal grid to be fifty metres
- by which it is meant that the information in the DEM is arranged
with one value every fifty metres in each of the co-ordinate directions.
The resolution of the vertical element of the information will
indicate whether the value represents the computed average elevation
for that, say, fifty metre cell, of the elevation at the mid point as
well as defining the range of accuracy to be expected of the elevation
values, possibly plus or minus two or three metres. Sometimes the surface elevation information does not form a regular matrix,
but rather comprises a collection of measured locations with altitude.
A DEM constructed from such data may be interpolated to
form a regular matrix, or the surface may be represented by linking
the measured points within a Triangulated Irregular Network (TIN).
The TIN structure, of triangular facets each with a measureable
slope and direction, is an efficient storage mechanism from which
a regular matrix can readily be derived when necessary. TINs have another useful property: they can be stacked. This
means that they can be used to represent layers of information,
for example atmospheric layers, subsurface archaeological or geological
stratigraphy, etc. Resolution, in the context of a TIN, is
a simpler concept than for DEMs formed from a regular matrix
of values. The vertical resolution remains the same but the horizontal
resolution is a function of the precision of the co-ordinates defining
the data points and the number and distribution of the data points
with respect to the surface morphology. A primarily flat surface
will require fewer points than a rugged or undulating surface. Contour lines are not a good form in which to store elevation
data. Contours are derived data, data interpolated from information
of altitude at known points, and in themselves offer no information
about the surface morphology between them. Contours may
be an effective way of illustrating the third dimension (altitude
or depth) on two dimensional paper, but are a poor technology
for storing altitude information that may be used analytically. A DEM held
as a regular matrix suffers the same disadvantages
due to size as does the raster data model. Although the co-ordinates
of each cell in the data set can be derived from the co-ordinates
of the origin and from the number of cells in each direction and
their separation, a value must still be stored for every cell.
The computer file size for such a data set is thus the product
of the number of rows and the number of columns. For example,
an OS LandForm
Panorama tile, 20km square and with a 50m
cell separation, holds 401 rows and columns - 160,801 cells. The
same area in the LandForm Profile data set, with a cell separation
of 10m, comprises 402,002,500 cells. In computer storage terms,
the coarser data set would require around 629Kbytes; the finer
resolution data set requires some 1.57Gbytes (note difference in
units). If the elevation data is held in floating point form (fractional
numbers) then the storage requirements may be doubled,
depending upon the individual computer system. It is less easy
to give guidelines on the space requirements of TIN data sets ... a rough guide might be (number of points times 6) times
4 bytes. Thus a TIN with 900 data points would require 21600 bytes
(21Kbytes). This figure may vary according to the implementation
of the TIN structures. There are several potential sources of elevation data:
Digital photogrammetry uses stereo digital images and typically produces a regular matrix of values, a DEM
Several times in the above discussion the terms interpolation or interpolated have been used. In the context of DEMs these refer to the technique used to approximate the altitude of points for which there is no measured data. The purpose of interpolation is to attempt to regain a representation of the actual surface morphology. There are perhaps three techniques commonly used in GIS to perform such a task: each has specific capabilities dependent upon the nature of the data. These three techniques are Linear interpolation, which effectively runs a straight line between the points with altitude values, the Cubic Spline, which interpolates a smooth curve through the given data points, and Statistical interpolation using Kriging semi-variograms. For further discussion see Chapter 8 of Burrough (1986). |
The right of Mark Gillings, Peter Halls, Gary Lock, Paul Miller, Greg Phillips, Nick Ryan, David Wheatley, and Alicia Wise to be identified as the Authors of this Work has been asserted by them in accordance with the Copyright, Designs and Patents Act 1988. All material supplied via the Arts and Humanities Data Service is protected by copyright, and duplication or sale of all or part of any of it is not permitted, except that material may be duplicated by you for your personal research use or educational purposes in electronic or print form. Permission for any other use must be obtained from the Arts and Humanities Data Service(info@ahds.ac.uk). Electronic or print copies may not be offered, whether for sale or otherwise, to any third party.
|