In the description above, sampling one pedon of different sizes was considered. These calculations are particularly useful in field investigations, where a large number of pedons are sampled. In this case, the total time that each sampling design would require must be calculated, against the variability of data between pedons, or between samples within a pedon. Considerable time can be saved if an estimate of these parameters is known in advance, to avoid oversampling of the study area, or undersampling that fails to resolve differences over time or between treatments. Two studies with nematodes have shown that 16 X 2 cm cores, removed at random from a forest site, were sufficient to obtain 96% of species diversity (Johnson et al., 1972) and that ten random 2 cm cores were sufficient to determine abundances in a 7 ha agricultural field (Prot and Ferris, 1992). These experimental values are far below those estimated theoretically without field data and those recommended by the Society of Nematology in 1978 (Barker, 1978).
Since several samples are removed from each pedon, and each sample is then subsampled in the laboratory, the correct statistical procedure for error analysis is described by multistage sampling analysis. The number of subsamples should be sufficient to obtain a robust mean, with about 10% coefficient of variance about the mean. An inescapable fact whilst enumerating samples is that different microscopists have different efficiency and skills. It is particularly important for beginners that samples are enumerated consistently. Thus, recounts of the same samples by the same person can be analysed by one-way ANOVA (analysis of variance), followed by a measure of repeatability. It is calculated from R = s2a/s2b + s2a, where R, the measure of repeatability, is obtained from s2a the variance between repeats, and s2b, the variance between counts in subsamples (for further details, see Krebs, 2000). To obtain useful data requires that enumeration be reproducible. These will improve with experience and with the skill of the microscopist.
The defined study area normally will require several pedons to be sampled adequately. There are a number of ways that additional pedons can be positioned on the study area. Since the soil is spatially heterogeneous and patchiness in abundances is expected, two methods are recommended. One is to sample a small number of pedons at random across the area, the other is to sample 3-5 pedons along 2-5 transects of fixed length. The latter resemble long and narrow rectangular quadrats which are usually best for clustered or heterogeneous distributions. They differ in that they are discontinuous and several samples are removed from each pedon within the transect (Fig. 3.2). Although applying equally spaced pedons along a transect is acceptable in most situations, there are
two situations where caution is warranted. One is in agricultural fields where there is a periodicity to the plough line or seeded line, and the other is in managed forests or replanted forests. In these study areas, it is best to avoid a systematic pedon distribution (same spacing between pedons) in favour of unequal (or randomized) pedon position, or to ensure the sampling period does not correspond to the site periodicity. However, in most situations, a systematic pedon distribution, imposed on a heterogeneous landscape, provides a random sampling.
The frequency of individuals counted from samples is then used to obtain the statistical distribution of species. If the frequency data fit a Poisson distribution, then the individuals must be assumed to be randomly distributed in geographical space. In this case, it is best to obtain an estimate of the mean number of individuals per sample (or in each pedon) that is statistically robust. Soil species rarely have a completely random pattern. Instead, there is a tendency to find clusters of higher abundance of one or more species or functional groups. If the spatial clustering of individuals is statistically significant, as it could be in a heterogeneous spatial space such as a forest floor, then a negative binomial distribution is often a better statistical descriptor of the spatial pattern. It is important to distinguish between these two mathematical frequency distributions, as they determine the correct methods to chose for calculating confidence limits for the data. The spatial scale at which the distribution of populations is being described should be tested. Since the investigator samples from within a pedon (<1 m2), to several pedons across a landscape (>100 m2), spatial clustering or non-random distribution of species can occur within the pedons or between pedons on a larger scale. The data can be analysed further with indices of dispersion for quadrat counts.
Normally, the soil samples removed from each pedon can be mixed into a single composite sample, prior to subsampling and extraction. The resolution of this sampling design can be improved using a stratified random sampling procedure. The procedure is particularly useful when sampling soil, because of the variability in habitat over short distances and the diversity of microhabitats. The idea is to group data from similar soil samples (each being one stratum). For instance, when sampling a pedon in a crop field, one could remove, from within the same pedon, samples from the base of the plants (stratum 1, enriched in rhi-zosphere soil), and from between the plants (stratum 2). During data analysis, all the samples from the base of the plant from all the pedons can be compared together, and those from between the plants together. Another situation would be to sample surface litter in forest quadrats, but to subsample coarse woody debris, fresh leaf fall or insect frass separately. Lastly, it may be useful to compare the distribution of species and abundances through the profile. In this case, each sample would be sub-sampled into depths along the profile (0-3 cm, 3-8 cm, 8-15 cm). This procedure provides improved resolution from the sampling design. It is more precise than simple random sampling within the pedon, and should be applied wherever possible.
Was this article helpful?