Spatial Sampling Designs - Sampling Spatial Units for Agricultural Surveys - page 147

Chapter 7
Spatial Sampling Designs

7.1 Introduction

Geographically distributed observations have characteristics and peculiarities that
should be appropriately considered when we are designing a sample. In fact,
traditional sampling designs (see Chap. 6 ) may be inadequate when investigating
geo-coded data, because they do not capture any spatial homogeneity that may be
present. The presence of this spatial effect may be inherent to the phenomenon
under investigation, so it is desirable and appropriate that we consider this infor-
mation in the sampling design.
Surveys are routinely used to gather primary data for agricultural and environ-
mental research. The observed units are often randomly selected from a finite
geo-referenced population. Thus, the spatial distribution of this population can be
used as key information when designing the sample. In this chapter, we will focus
on probability samples that are well-spread over the population, in every dimen-
sion. These particular samples are denoted as spatially balanced samples. Over the
last few decades, this characteristic has become so important that several new
sampling algorithms have been developed (Christman 2000 ; Wang et al. 2012 ;
Delmelle 2013 ).
Surprisingly, this criterion for sample selection is mainly based on intuitive
considerations, and it is not clear when, and to what extent, it could have an impact
on the efficiency of the estimates. Besides, it is also useful to consider that this
characteristic of sampling plans is not properly defined.
The main reason that the geographical position of statistical units in the random
selection of samples is important in agricultural and environmental surveys is that
the statistical units themselves are defined using purely spatial criteria (see
Chap. 5 ) . Another motivation is that recently many NSIs geo-reference their
sampling frames according to geographical nomenclature codes and by adding
information regarding the exact or estimated position of each record.