Setting Up the Frame - Sampling Spatial Units for Agricultural Surveys - page 93

The first category consists mainly of agricultural holdings (or farms), house-
holds, and businesses. Agricultural surveys are generally based on farms, but it is
important to underline that for many other aspects related to agriculture also the
other two kinds of units are frequently used. For example for milk statistics, the
statistical unit is an establishment, i.e., a dairy or the farm itself, while for the
composition of agricultural income the household is usually surveyed.
Moreover, in many developing countries the household is the unit to be
interviewed as it corresponds to the farm. From here on, only agricultural holdings
or farms will be discussed as representatives of this type of unit.
The spatial units are segments of land areas and may be polygons (regular or
irregular), points, and lines. In agricultural surveys only the first two kinds are used
while the last is mainly interest to environmental studies.
The segments can be specified in three different ways:

1. They can be set by identifiable physical boundaries (such as rivers, roads) that
are easily found and supply a precise identification of the segment (irregular
polygons). The theoretical definition concerns an irregular and varying dimen-
sion polygons while the operational definition concerns a irregular portion of
land identifiable by physical boundaries (type 1).
2. Through the use of a squared grid of map coordinates (regular polygons). The
theoretical definition specifies regular dimension (the units have all the same
size) and the operational definition a square portion of land (type 2).
3. Superimposing their limits to the boundaries of the land of the agricultural
holding (points). In this case point sampling procedures are used. The theoretical
definition is “no dimension but it may be defined as having a certain size for
coherence with the observation rules or the location accuracy that can be
achieved” while from an operational point of view it is a circular portion of
land (type 3, FAO 1996 ; Ferraz 2013 ).
A segment is subdivided into non-overlapping tracts. One tract is the part of a
holding that is found within the limits of a segment, or a piece of land that does not
belong to any holding. A tract is determined by the limits of the segment and by the
holdings with land in the segment. Note that its definition is not necessary for
spatial sampling designs in which the segments are of the third type above. A
holding comprises at least one tract. Tracts are observational units.
With spatial units of type 1, limits should follow physical boundaries thus it
should be easier to identify the statistical unit than with the other two kinds. In the
process of constructing a new spatial frame (in this case a spatial area frame), all
land has to be partitioned into primary sampling units (PSUs).
PSUs, with the aid of an aerial photo (to recognize physical boundaries), have to
be further divided in segments of similar size. This is a very difficult task. For this
reason only a sample of PSU is selected, and then segments for these PSUs are
delineated. In this way, it is not needed to divide an entire frame into segments,
saving a large amount in time and labour costs (Cotter et al. 2010 ). Final segments
rarely have the same size, and so a ps sampling scheme is always adopted (see

Sect. 6.4 ). However, the spatial frame construction and the sample selection with

π