Essential Statistical Concepts, Definitions, and Terminology - Sampling Spatial Units for Agricultural Surveys - page 34

Conclusions
The particular nature of geo-referenced data has influenced the definition of
models for spatial data analysis. However, it is surprising that there has not
been a similar research effort into the issue of spatial data collection. The data
collection phase is crucial, and has a remarkable influence on the data
analysis that follows.
The definition and analysis of appropriate methods for spatial sampling
represent a huge challenge for statisticians and researchers who work with
geographical data.
The main objective of spatial sampling is to make inferences about some
parameters of a geo-referenced target population, using a sample of units
selected from that population. Spatial sampling is used, for example, when
the population is so large that a census would be impractical. Furthermore, we
may have infinitely many locations where measurements could be observed.
The topic of spatial sampling has not been deeply analyzed in the thematic
literature of spatial statistics. The reference topics do little to address this
issue, and if they do, it is only marginally mentioned. The only exception is
Haining ( 2003 ), who included a section (see p. 93 and the following text)
completely devoted to the problem of spatial sampling.
Haining ( 2003 ) underlined that appropriate spatial sampling decisions are
needed when making inferences about a geographically distributed popula-
tion. We should define the variable to be estimated, the sample size, and the
corresponding required level of precision. However, when dealing with
spatial data, we must also define a proximity criterion and a homogeneity
measure, so that we avoid duplications of information. This may happen if the
sample consists of similar units.
In agricultural surveys, the parameter of interest is often geographical in
nature. In other words, it pertains to specific locations. This type of spatial
data requires us to simultaneously consider both the location and attribute
information. It is well known that spatial data are not typically independent,
unlike non-spatial data. This consideration leads to a definition of an appro-
priate sampling plan that cannot be considered a simple extension of the
design-based and model-based sampling approaches.
This topic is designed to try to address these questions and problems. In
fact, our main aim is to show the connections between the topics of sampling
and spatial statistics, which are often not considered to be linked.
In this first chapter, we have outlined the basic methodological back-
ground for describing the problem of spatial sampling of agricultural data.
After some more introductory chapters (that discuss, for example, GIS and
remote sensing), the remainder of the topic attempts to fill a gap in the
literature, by providing an in-depth description of the issue of sampling for
spatial data.