relationships (Patra et al. 2007), a central question in
microbial ecology. Populations highly correlated with
measured parameters will also be spotlighted as a means to
identify potential indicator species. The importance of these
species must be tested to confirm their “indicator” status. For
example, the processing of data by CCA has highlighted the
interactions between various abiotic parameters and dynamics of bacterioplankton in the North Sea and the identification of bacterial phylotypes responding specifically to
certain environmental factors (Sapp et al. 2007). Other
examples of the use of CCA to assess the impact of abiotic
factors or pollutants on the structure of microbial mats are
given in this chapter.
8.4
Variables and Methods for Studying
Microbial Diversity
Different levels of diversity are studied in microbial ecology
to answer different questions. A first level of questioning is
that of bacterial species, the only taxonomic level for which
an operational definition exists, based on DNA hybridization
or 16S sequences comparison (cf. Chap. 6). At a finer level,
approaches targeting DNA can also reach subspecies levels.
There are two types of biodiversity used in the literature
(Whittaker 1972). Alpha diversity is defined as the smallscale diversity observed within a community. Beta diversity
is defined as the large-scale diversity observed between
communities. This distinction is rarely used in microbial
ecology, because the definition of “local” microbial
communities is not easy in many environments like soil
(Box 8.1) that are highly heterogeneous even at the microscale so that environmental samples (e.g., a soil core) are
often already a mixture of local communities. In addition,
numerous horizontal gene transfers can jeopardize the definition of species identity and associated functional traits as
classically accepted in ecology of higher organisms. The
Box 8.1: Example of a Study of Biodiversity and Soil
Management
Xavier Le Roux
The following example illustrates the interaction
between statistical tests, indices, and assumptions.
Management strategies are important for soil fertility
and for the abundance of plant pathogens and of
microorganisms that contribute to the degradation and
recycling of compounds. It may therefore be interesting
to evaluate the impact of a fallow period, for example,
(continued)
Box 8.1 (continued)
on the soil microbial community. For a period of
fallow, a field is colonized by a wide variety of weeds
rather than a single species as in classic monoculture.
Therefore, several hypotheses can be posed such as:
1. A fallow season will increase the overall diversity
of microorganisms in the soil.
2. A fallow season will reduce the frequency or abundance of aggressive strains of a plant pathogen specific to a particular host (e.g., Fusarium oxysporum).
In the case of the first hypothesis, it is necessary to
express the overall microbial diversity as an index
incorporating the relative abundance of different taxonomic groups present in the soil. The Shannon index,
for example, could be used for this purpose. This
would involve sampling and characterizing by the
most appropriate method (see below) under field
conditions and comparing microbial communities
before and after the fallow. To calculate the index,
many strains or clones must be characterized before
and after the fallow period. These observations must
then be integrated into a single index value for each
situation. However, parametric tests require replicated
observations for each treatment (e.g., t-test or analysis
of variance) in order to assess the variance. This will
clearly lead to a significant amount of work. Some
researchers have used numerical simulations to estimate the variance of diversity indices in order to
compare the indices without replicate measurements.
It should be kept in mind that the simulation can only
be based on the observed variability in the sample and
cannot be based on “real” variability in situ, especially
if the real variability is greater than that observed in
the sample. An alternative approach might be to compare populations using nonparametric techniques
based on gross multinomial values. For populations
with bi- or trinomial distributions, χ
2 tests could be
used to compare populations. However, if the aim is to
compare global biodiversity, it is likely that the
samples will reveal many taxonomic groups and
require the comparison of multinomial distributions.
At present, few statistical techniques are available to
compare multinomial complexes.
The second hypothesis mentioned above is more
tractable than the first because the target of the study
of biodiversity is more precisely defined than the first
hypothesis. For this study, the abundance of Fusarium
oxysporum in soil will be evaluated and strains tested for
their degree of pathogenicity on different hosts. To
statistically compare the abundance of the Fusarium
(continued)
266
P. Normand et al.
microbial ecology. Populations highly correlated with
measured parameters will also be spotlighted as a means to
identify potential indicator species. The importance of these
species must be tested to confirm their “indicator” status. For
example, the processing of data by CCA has highlighted the
interactions between various abiotic parameters and dynamics of bacterioplankton in the North Sea and the identification of bacterial phylotypes responding specifically to
certain environmental factors (Sapp et al. 2007). Other
examples of the use of CCA to assess the impact of abiotic
factors or pollutants on the structure of microbial mats are
given in this chapter.
8.4
Variables and Methods for Studying
Microbial Diversity
Different levels of diversity are studied in microbial ecology
to answer different questions. A first level of questioning is
that of bacterial species, the only taxonomic level for which
an operational definition exists, based on DNA hybridization
or 16S sequences comparison (cf. Chap. 6). At a finer level,
approaches targeting DNA can also reach subspecies levels.
There are two types of biodiversity used in the literature
(Whittaker 1972). Alpha diversity is defined as the smallscale diversity observed within a community. Beta diversity
is defined as the large-scale diversity observed between
communities. This distinction is rarely used in microbial
ecology, because the definition of “local” microbial
communities is not easy in many environments like soil
(Box 8.1) that are highly heterogeneous even at the microscale so that environmental samples (e.g., a soil core) are
often already a mixture of local communities. In addition,
numerous horizontal gene transfers can jeopardize the definition of species identity and associated functional traits as
classically accepted in ecology of higher organisms. The
Box 8.1: Example of a Study of Biodiversity and Soil
Management
Xavier Le Roux
The following example illustrates the interaction
between statistical tests, indices, and assumptions.
Management strategies are important for soil fertility
and for the abundance of plant pathogens and of
microorganisms that contribute to the degradation and
recycling of compounds. It may therefore be interesting
to evaluate the impact of a fallow period, for example,
(continued)
Box 8.1 (continued)
on the soil microbial community. For a period of
fallow, a field is colonized by a wide variety of weeds
rather than a single species as in classic monoculture.
Therefore, several hypotheses can be posed such as:
1. A fallow season will increase the overall diversity
of microorganisms in the soil.
2. A fallow season will reduce the frequency or abundance of aggressive strains of a plant pathogen specific to a particular host (e.g., Fusarium oxysporum).
In the case of the first hypothesis, it is necessary to
express the overall microbial diversity as an index
incorporating the relative abundance of different taxonomic groups present in the soil. The Shannon index,
for example, could be used for this purpose. This
would involve sampling and characterizing by the
most appropriate method (see below) under field
conditions and comparing microbial communities
before and after the fallow. To calculate the index,
many strains or clones must be characterized before
and after the fallow period. These observations must
then be integrated into a single index value for each
situation. However, parametric tests require replicated
observations for each treatment (e.g., t-test or analysis
of variance) in order to assess the variance. This will
clearly lead to a significant amount of work. Some
researchers have used numerical simulations to estimate the variance of diversity indices in order to
compare the indices without replicate measurements.
It should be kept in mind that the simulation can only
be based on the observed variability in the sample and
cannot be based on “real” variability in situ, especially
if the real variability is greater than that observed in
the sample. An alternative approach might be to compare populations using nonparametric techniques
based on gross multinomial values. For populations
with bi- or trinomial distributions, χ
2 tests could be
used to compare populations. However, if the aim is to
compare global biodiversity, it is likely that the
samples will reveal many taxonomic groups and
require the comparison of multinomial distributions.
At present, few statistical techniques are available to
compare multinomial complexes.
The second hypothesis mentioned above is more
tractable than the first because the target of the study
of biodiversity is more precisely defined than the first
hypothesis. For this study, the abundance of Fusarium
oxysporum in soil will be evaluated and strains tested for
their degree of pathogenicity on different hosts. To
statistically compare the abundance of the Fusarium
(continued)
266
P. Normand et al.