Agriculture Reference
In-Depth Information
pathology has been devoted to sampling methods appropriate for assessing disease
incidence and severity (Hughes, 1999). Traditional sampling methods involve
diagonal sampling in farmers' fields where at least 50 tillers are sampled at random
along each diagonal; the pattern of disease spread, whether it is scattered or uniform,
may influence the number of samples taken, which in turn is related to the standard
deviation of disease incidence (Church, 1971). In small experimental plots, sampling
may not be customary as replication produces the needed accuracy (Zadoks and
Schein, 1979); however, a minimum of 10 samples is often used for small cereal
plots. Depending on the disease, the usual emphasis in disease measurement is given
to incidence or severity within the sampling unit. Chaube and Singh (1991)
reviewed a number of terms used in sampling, including entity, sample size, sample
point and sampling fraction, all of which need to be considered for the satisfactory
measurement of disease, particularly over large field areas.
Disease incidence, severity and spatial pattern depend on data obtained from
field samples. The accuracy of these data, as well as the time and effort required to
obtain them, are affected by the sampling technique used. In a study of three
naturally occurring epidemics of leek rust (caused by
Puccinia allii
),
de Jong and de
Bree (1995) concluded that in the development of a practical sampling method for
detection of the disease, it was necessary to take into account a clustered distribution
of diseased plants. In their study, the Black-White (BW) join-count statistic was
used to detect non-randomness in the spatial distribution of rust-infected leek plants.
Delp
et al
. (1986a) developed a computer software system called Field Runner to
simplify the task of sampling fields. The system uses the stratified random sampling
design (SRSD) with single-stage cluster sampling; this provides an unbiased sample
and a lower error of disease incidence estimates than conventional diagonal, 'X' or
'W' sampling designs. A hand-held microcomputer is used to direct the operator to
each sample site within a sector, each site being composed of a cluster or transect
(see Fig. 2.1); fields can be assessed for severity of one disease or for incidence of
one to several diseases.
In a further paper, Delp
et
al
. (1986b) evaluated field sampling techniques for
estimation of disease incidence using computer-simulated field tests. Disease
incidence and aggregation were varied to determine their effects on sampling
techniques. The authors concluded that SRSD required the least number of samples
and the lowest sample intensity to estimate disease incidence within a 95%
confidence interval for all field types.
In a three-year study of winter wheat diseases, Parker and Royle (1993)
developed a novel large-scale sampling procedure using randomly positioned
transects based on the theory of autocorrelation. The procedure allowed valid tests
of significance to be made on the autocorrelation coefficients calculated; the sample
data obtained were also suitable for use in mapping analysis and the production of
semivariograms.
A series of papers (Fleischer
et al
.
,
1999; Hughes, 1999; Madden and Hughes,
1999; Morrison, 1999 and Nyrop
et al
.
,
1999) confirmed that data obtained by
sampling are crucial for improved decision making by farmers and growers in crop
loss assessment and disease management. However, decision making will be imperfect
