Overview and Brief History - Sampling Spatial Units for Agricultural Surveys - page 45

investigation. The sampling units for the second phase are based on a preliminary
identification of zones of interest for analyzing the different levels of administrative
division: province, region, and entire national territory. Finally, the yield estimates
in the third sampling phase are calculated for the same geographical

levels

described above, but only for 12 different crops.
The land use archive for agricultural surveys may be derived from three different
sources: a ground survey, interpretation of digital orto-photos, and remotely sensed
image classification. AGRIT uses photo interpretation of remotely sensed images
acquired from aircraft. This activity, which formed the hierarchical nomenclature
of reference (on two levels, for a total of 25 labels), was developed within the
Permanent Observed POints for Land Use Statistics (POPOLUS) project. The
25 POPOLUS classes were aggregated into six groups, and only four were sampled.
The first step when preparing the archive was the stratification of the first phase
units. The points were stratified using 103 Italian provincial codes and 6 classes of
land use, obtaining 618 non-empty strata. The strata codes for land uses are: arable
land, permanent crops, permanent fodder land (altitude 1,200 m), permanent

fodder land (altitude 1,200 m), wooded land, isolated trees and agricultural

buildings, and other (i.e., artificial surfaces, water, non-vegetated natural surfaces).
The target population for the selection of the second phase sample is composed of
the points that display an agricultural activity. The criterion for the definition of the
strata is conservative: all the points that have a potential agricultural interest are
considered. Therefore, a random second phase sub-sample was extracted from the
strata codes that are considered useful for AGRIT 2005 and each of the 103 Italian
provinces. In total, approximately 150,000 points must be collected.
AGRIT 2005 was a multipurpose survey, and was defined to produce estimates
with a predetermined sample error for a set of variables (i.e., areas of different
crops). Bethel

>

s algorithm ( 1989 ) was used as the allocation method in the sampling
procedure. It is a generalization of Neyman

'

s classic formulas for calculating the
optimal sample size, in the case of stratified sampling (see Chap. 8 ) .
Finally, the third phase sample produces the yield estimates. Bethel

'

s algorithm
was again used to determine the sub-sample size, using the variances of the yield of
the previous AGRIT rounds. The sample contained approximately 60,000 points,
located in different regions of Italy.
The survey has remained essentially unchanged across the years.

'

2.3.4 TER-UTI

The knowledge and monitoring of land use and land cover are old issues in France.
The Utilisation du territoire (TER-UTI) survey is annually conducted by the
statistical services of the French Ministry of Agriculture and Fishing. It collects
data on land use throughout the whole continental territory, using a set of points
constituting a representative sample of the territory.