Survey Data Collection and Processing - Sampling Spatial Units for Agricultural Surveys - page 230

contains information on the regional/central offices, the external quality control,
and Eurostat, and their individual roles within the data quality control procedures.
To provide good quality statistics on land cover/land use, Eurostat devoted
special attention to the production process to ensure that the resulting data satisfy
customer requirements in a systematic and reliable way (Eurostat 2009b ). They
implemented actions that concerned different levels to respect quality assurance
standards. They set up a five level hierarchical structure (surveyor, supervisor,
central office, external quality control, and Eurostat). They also introduced various
training steps, continuous monitoring, independent data quality checks, and stan-
dardization and computerization of the main data management phases.
It is also very important to compare the data with information available from
other sources. This kind of action is the main form of validation carried out for data
dissemination. It seeks to exclude the possibility that derived information has
quality problems that can be revealed by comparing it with any available statistics.
It does not try to identify the causes of the errors, but only considers whether the
errors exceed defined limits. This can affect the release of the data. The internal
consistency of the data is used to assess the quality by calculating some index of
recognized stability, using estimates of the quantities produced by the survey.
Additionally, the produced statistics can be compared with those obtained in
previous surveys or with estimates from independent sources. In both cases, special
attention must be paid to the elements of variability introduced by the different
definitions and operating conditions used by the different sources.
The most expensive approach to quality control is to estimate the main compo-
nents of the error using appropriate quality measurement and control surveys. This
assesses the quality of the data using a direct estimate of the components of the error
profile.
It consists of obtaining (typically using ad hoc surveys on specific sources of
error) auxiliary information that is used to estimate specific components of the total
error. For example, to investigate any variability caused by respondents not accu-
rately providing the requested information we can re-interview a sample of respon-
dents. The questions are repeated to analyze any variability in the responses
provided by the same individual on two different occasions.
This kind of study is expensive and complex (Forsman and Shreiner 1991 ), so it
necessarily has an episodic character and covers a very specific analysis of the more
relevant sources of error.
The accuracy assessment phase is commonly carried out after the observation/
interviewing phase of the land cover/land use survey (Lunetta et al. 1991 ). It is
based on comparisons of the collected and reference data, for a subsample of units.
The reference data are mainly gathered using ground visits or telephone inter-
views to the landowner, and are assumed to be more accurate than the observed
information. In reference data collection, ground visits tend to be preferable to
telephone interviews, as the latter may lead to questionable results.
There are three major components to a land cover/land use survey quality
assessment: the sampling design, the response or measurement design, and the
estimation and analysis protocol.