Survey Data Collection and Processing - Sampling Spatial Units for Agricultural Surveys - page 231

The sampling design does not necessarily follow the same rules and principles
used in designing the survey. This is because the domain of interest, the stratifica-
tion, and the selection probabilities are more related to the organizational structure
of the data collection and the probability of error than to the efficiency of the
estimates of aggregates.
As well as some basic approaches based on stratified random sampling (Scepan
1999 ) and two-stage cluster sampling (Nusser and Klaas 2003 ), alternative sam-
pling schemes have been developed to assess the accuracy of land cover/land use
data. For example, Carfagna and Marzialetti ( 2009a ) proposed a sequential sam-
pling design for both quality control and validation of land cover databases.
Different strata were identified according to the land cover type and the size of
polygons, and the sampling units were selected within each stratum according to the
permanent random number method.
The number of selected sampling units may not be predetermined as in sequen-
tial acceptance sampling. The use of this approach to control the quality of land
cover databases was discussed in Carfagna and Marzialetti ( 2009b ). In sequential
acceptance sampling, a sequence of samples is selected. Then, at each stage of the
quality control process, we use the previous results to decide whether to terminate
the inspection or select a further sample.
The validation of land cover/land use data involves assessing both positional and
thematic accuracy (Lunetta et al. 1991 ; Foody 2002 ). Positional accuracy is the
accuracy of the location of a unit in the map, relative to its location in the reference
data. The positional deviation of a selected control unit relative to a reference unit is
traditionally measured in terms of the root mean square error (Lunetta et al. 1991 ).
Positional accuracy is an essential part of thematic accuracy. This is why it is
strongly recommended that surveyors enter the GPS geographical coordinates and
the observation distance when they are at the surveyed point. Thematic accuracy
refers to the accuracy of land cover types depicted in the survey when compared
with the land cover types in the reference data.
Two main types of thematic errors can be identified: omission and commission
errors. An omission error occurs when a case belonging to a class is not allocated to
that class. A commission error occurs when a case belonging to a class is erroneously
allocated to another class. Thematic accuracy is typically assessed using a confusion
or error matrix. The confusion matrix summarizes the correct classifications and
misclassifications in a contingency table format. The rows of the confusion matrix
typically represent the survey labels, and its columns identify the reference labels.
The entries of the confusion matrix must be estimated for the sampled units to
obtain estimates of the accuracy parameters. Several accuracymeasures can be derived
from the confusion matrix. There is no standard assessment approach for land cover/
land use survey accuracy. Each accuracy measure incorporates specific information
about the confusion matrix and is suitable for a particular purpose (Liu et al. 2007 ).
One of the most popular accuracy measures is the overall accuracy that repre-
sents the overall proportion that has been correctly classified. The overall accuracy
expresses the probability that a randomly selected unit is correctly classified by the
map, and provides a measure of the quality of the map as a whole.