Environmental Sampling Design - Fundamentals of Environmental Sampling and Analysis - page 56

3.3.4 Systematic Sampling
Systematic sampling involves selecting sample units according to a specified pattern
in time or space, for example, at equal distance intervals along a line or a grid
pattern. Some slight variations exist for systematic sampling as shown in Figure 3.5
for a two-dimensional space sampling, systematic grid sampling (Fig. 3.5c), and
systematic random sampling (Fig. 3.5d). The systematic grid sampling subdivides
the area of concern by using a square or triangular grids and then collects samples
from the nodes (the intersections of the grid line) or a fixed location (e.g., center) of
each grid. The first sample to be collected from a population is randomly selected
(e.g., through a random number table), but all subsequent samples are taken at a
fixed space or time interval. The systematic random sampling subdivides the area
into grids and then collects a sample from within each grid cell using simple random
sampling.
The systematic grid sampling is easier to implement and more convenient to
field personnel than simple random sampling. An important application of this
approach is groundwater sampling at a fixed well over time. There are several
reasons for this preference: (a) extrapolation from the same period to future periods
is easier with a systematic sample; (b) seasonal cycles can be easily identified and
accounted for in the data analysis; (c) a systematic sample will be easier to
administrate because of the fixed sampling schedule; (d) most groundwater samples
have been traditionally collected using a systematic sample, making comparisons to
background more straightforward (EPA, 2002).
The second advantage of grid sampling is its more uniform distribution over the
space or time domain, which helps to delineate the extent of contamination and
define contaminant concentration gradients. This has been proved to be efficient for
a full characterization of soil contamination, because grid sampling insures that all
areas are represented in the sample and provides confidence that a site has been fully
characterized. The same is true for its application to geostatistical applications to
delineate the temporal or spatial patterns for correlated data in time and space (see
Section 2.2.6).
The calculation of mean and standard deviation for systematic sampling is
straightforward. The simplest way to analyze the data from systematic sampling is to
treat it as though it was collected using simple random sampling (Eqs 2.12 and 3.1).
A critical part of the systematic sampling design is to choose the right grid
spacing (L in Fig. 3.5). The grid spacing should be small enough to detect the
spatial/temporal patterns or to search hot spots. Otherwise, grid sampling will
likely either overestimate or underestimate the population. This also occurs when
the grid spacing L coincides with the spatial or temporal pattern of the variable
of interest. For example, temperature or other temperature-dependent parameters
(e.g., dissolved oxygen in water) will have a 24-h cyclic pattern; if all samples
are taken at the late afternoon of sunny days, it will overestimate the temperature
but will underestimate the dissolved oxygen. These problems associated with the
grid sampling, however, can be minimized by the use of systematic random
sampling.