Key concepts 1 GIS - Spatial Data Analysis: An Introduction for GIS Users - page 15

low frequency. To capture spatial variation in the case of a mountainous terrain, a i ner
sampling grid would be needed than would be the case for the river l ood plain. If the
sample spacing is too large, then important information may be lost. If it is too small,
then some of then ef ort expended in sampling will have been wasted—some of the
data will be redundant and add little information. It is therefore important to consider
how the available sample meets the requirements of the analysis and Section 2.8.1
discusses this issue further.

Spatial data collection

2.8

h is section introduces the basic principles of some widely used means of spatial data
collection. Spatial data may derive from secondary sources (e.g. paper maps) or data
may be collected by, or for, a particular user with a particular purpose in mind (termed
'primary data'). h e i rst subsection deals with the topic of spatial sampling. Paper-
based secondary data sources are the focus of Section 2.6.2. Spatial data can be divided
into those collected by remote sensing and those collected by ground survey, and
Sections 2.6.3 and 2.6.4 introduce these two modes of data collection.

2.8.1 Spatial sampling

Any spatial data set is a sample. A remotely sensed image may cover the entire area of
interest, but there are limits to the spatial resolution of such imagery. In any applica-
tion making use of spatial data, it is necessary to i nd a balance between the amount of
information required and the number and spatial positioning of observations. In
terms of measurements on the ground, perhaps using GPS or some more traditional
survey technology, making observations on a i ne grid over the entire study area, will
allow detailed characterization of the particular property of interest. However, such
a strategy will be wasteful of ef ort and money if many neighbouring observations
have similar characteristics, and thus contain very similar information (see the previ-
ous section for a related discussion). h e objective of sampling design is, therefore,
to design a sampling scheme whereby the maximum possible information is acquired
for the minimum ef ort.
Various commonly used strategies for sampling exist. h ese may be based on mak-
ing observations at locations with particular characteristics. Following the example of
mapping elevations, making measurements at locations where there is a break of slope
would be sensible. Other strategies are based on random selection of sampling loca-
tions, with a key objective being to minimize bias. Note that many routines exist for
randomly selecting locations or values from a list. Using such approaches, a particular
number of locations might be randomly selected from across the whole study area
and measurements could then be made at these locations. Other strategies are based
on, for example, random selection of locations within areas (a stratii ed sampling
scheme), thus ensuring that particular areas are represented, but that the locations