Newsletter and Technical Publications
Freshwater Management Series No. 5
Guidelines for the Integrated Management of
- Phytotechnology and Ecohydrology -
D. Sources of spatial data
to a GIS program is the variety of data types that they can store and process.
These can be vector data, raster data (including remote sensing data), and
attributes recorded in data bases. This feature is often called data integration.
The value of a GIS data base depends on following criteria:
- spatial coverage of existing digital maps,
- numbers of thematic layers,
- data age,
- data spatial resolution and accuracy.
In many countries, an important source of digital
thematic layers for GIS analysis is cartographic agencies. These include
national agencies such as the British Ordnance Survey and the United States
Geological Survey (USGS) Mapping Division. An important issue remains that of
standardisation of the data collected. This is one of the subjects being
considered in institutions like the U.S. National Center for Geographic
Information and Analysis (NCGIA), Rosgeoinform in Russia, the British Natural
Environment Research Council (NERC), and the Conseil National de l'Information
Geographique (CNIG) in France. There are also agencies which provide
information on existing sources of digital maps, a good example being the
National Geospatial Data Clearinghouse in United States.
For research purposes in the earth sciences, the
construction of a GIS data base starts with the defining model of the natural
environment which then is converted into a digital form. This defining model is
most often expressed as an analogue thematic map. In the next step, digitising
devices are used to perform quantification and compilation of data in a vector
or raster format. In the first case, this is accomplished by digitisation of
analogue maps using a digitiser or by the vectorisation of a raster image
displayed on a screen. In both methods, there is a need to register the map to
define the relationship between the digitiser and screen co-ordinates or
cartographic co-ordinates. Most often a linear transformation equation is used,
having the form:
x = au + bv + e
y = cu + dv + f
||(x, y) - cartographic co-ordinates,
||(v, u) - digitiser or screen co-ordinates,
||a...f - coefficients.
Registration of the map is done using control
points. These points are used to solve the above equations and determine the
coefficients. Using these equations, it is possible to calculate cartographic
co-ordinates from the registered analogue map on a digitiser or displayed on a
A scanner is a device for obtaining raster data. Digital images are registered
using a set of CCD (charge coupled device) elements. This can be done using a
table or drum scanner, as well as by scanner installed on an aircraft or
satellite. The scanner works on the same principle as a digital camera. The
resolution of scanners used to quantify printed analogue documents is very
high, reaching 0.025 to 0.050 mm. The parameter used to describe scanner
resolution is the number of dots per inch (dpi), which ranges up to a value of
1200, and, by interpolation, to 9600 dpi. A scanner records all elements of the
analogue picture, and, in this regard, it is similar to remote sensing scanning
Satellite images and aerial photographs, as a type
of raster data, contain a lot of information that has to be first extracted by
image processing and interpretation. The spatial resolution of aerial
photographs depends on their scale, and is in the range of 2 to 200 m. Data
from these sources can be used in GIS programs within thematic layers such as
land use, vegetation cover type, and patterns of settlement. Digital camera
technologies open new possibilities for aerial photography, and are likely to
improve the spectral resolution and range of this type of image. To use an
aerial photograph in GIS, it is often necessary to rectify the image, or to
convert the image registered in the central projection (principle of optical
objective) to an image in an orthogonal projection. Such
a transformation allows the user to
develop an "orthophoto" map from aerial photography. Satellite remote sensing
images are most often obtained by recording a wide spectrum of radiation
reflected from the ground. The spatial resolution of such a multispectral
satellite images ranges from a few kilometres (meteorological satellites) to a
Raster data can be also created by other programs.
For example, spatial interpolation programs or computer models with distributed
parameters can generate raster data. Interpolation programs use randomly
distributed observation points to calculate new values located within a regular
grid. Interpolated values can be visualised by assigning them pixel values and
then displaying the values in the form of a raster image. Because raster data
come from various sources, and may have different resolutions, a resampling
procedure is used to make them compatible within the GIS program. This is
commonly done by changing the position of the pixels and the pixel size. The
commonly used formats for raster images (for instance, TIF, BMP, PCX files) do
not contain information on the scale and cartographic co-ordinates of image.
This creates a problem in transferring raster data between the GIS programs,
since many GIS systems have their own standards for coding and compressing
raster images registered to cartographic co-ordinates. Likewise, there are
similar concerns with regard to satellite images, with each satellite source
having its own characteristic data format
(e.g., the Landsat MSS uses ACRES or EROS format, the Landsat TM uses
the AMIRA format, and Spot uses the Image format).
Editing the geometry of raster data is called calibration (also known as
"rubbersheeting". The parameters of the transformation equations are
calculated by the least squares method. The following transformation methods
are widely used:
- linear isotropic (Helmert transformation),
- linear anisotropic (affine transformation),
of raster models is used to correct distortions in aerial photographs and
satellite images. It is used also for facilitating comparisons with old topographic
maps (Figure 3.1).
|Fig. 3.1. Sequence of topographic maps showing the evolution of
channel forms in the Vistula River near Plock (in central Poland) using raster
image calibration. This makes it possible to estimate the age of alluvial
deposits at the sampling sites|
Both vector and raster data have positive and
negative properties. Which properties prevail depend on how closely the data
model fits the properties of the object being modelled. When we want to depict
the spatial variability of an environment, a raster model is generally better.
Raster models correspond well to the continuity property of landscape units.
When dealing with objects having sharp or well-defined boundaries, vector
models are generally better. For purposes of using stored data in calculations,
raster data are simpler because pixel values are stored in a matrix format.
Many GIS programs use both types of data, and work with so-called "hybrid graphics"
Both raster and vector layers should be registered in the same co-ordinate
system, however, for these hybrid systems to function. During registration, a
relationship between the raster co-ordinates (i - row, j - column) and
cartographic co-ordinates (x - eastings, y - northings) is calculated. Within
the hybrid graphics environment, it is possible to perform conversions between
data types by means of vectorisation or rasterisation. Vectorisation is more
difficult, since precise vector data are being generated from less accurate
raster data. In a raster data set, there is no information on topology, which
has to be included in a vector data set. Automatic methods of vectorisation
work on similar principles to those used in OCR (optical character recognition)
programs. The raster layers have to be specially prepared before vectorisation.
Specific concerns include the number of thematic layers visible, the continuity
of lines, and width.
is much simpler, since topological information does not need to be stored. The
vector layer is converted to a raster image at a declared resolution (pixel size).