Setting Up the Frame - Sampling Spatial Units for Agricultural Surveys - page 98

corresponding frame unit have been transferred. The selected sample segments are
located on appropriate aerial photo enlargements used to control field data
collection.
In the case of spatial frames with regular polygons or with segments that
coincide with the land of agricultural holdings, the frame construction is simpler
than in the previous case.
For frames with segments that coincide with the land of agricultural holdings, a
grid is overlaid on the strata and a sample of points is selected. Then, the points are
identified on the ground and the corresponding holdings form the sample.
The construction of point frames is basically simpler as the generation of
coordinates is made in an automatic way. The grid, produced by software, is then
overlaid with a point to polygon map intersection in order to attribute to each point
one and only one territorial nomenclature code, based on administrative divisions
(e.g. NUTS) excluding points outside the study region (e.g. water, foreign States,
etc.). Then each point is overlaid on aerial photos or low-resolution satellite images
so that a photo-interpreter can assign a stratum code.
Moreover, the frame is often overlaid with a digital elevation model to assign to
each point, the elevation, which could be useful (see Sect. 6.7 ) for cut-off sampling
(see Sect. 6.11 ) or stratification (see Sect. 6.5 ) .
Finally point to polygon intersection (see Sect. 3.3 ) can be used to overlay this
frame to other ancillary data as available land use map, cadastral map, other
administrative data source (Carfagna and Carfagna 2010 ).
Conclusions
This chapter has been devoted to the description of the main characteristics of
frames. The reduction of non-sampling errors is a strategy often crucial for
the success of a survey and of these errors probably the most influential are
those deriving from the list used for the representation of the target
population.
Any difference between the frame and the target population implies a
possible bias in the results whose weight will grow with the increase of the
distance between the two populations.
A list of farms, whose construction is usually based on the results of a
census, has high costs due to the need to integrate it with data from admin-
istrative sources, and to update it periodically to prevent that it deviates too
much from the target population. Avoiding any possible duplication, omis-
sion and improper entries, a list of units based on a suitable partition of the
study region could significantly reduce the non-sampling errors.
Such a list could also complement and not necessarily replace the list
frame, it could be also used to update existing lists, to use auxiliary informa-
tion available only on a geographical basis (such as remotely sensed data),
and could facilitate certain aspects of quality controls, of the definition of
(continued)