10. Herbivory in Terrestrial Ecosystems
with the animal-based approach. There are two major assumptions and, as usual, controlling for the
risk of violating one increases the risk of violating
the other. The first assumption is that initial biomass inside and outside cages is the same. Thus, a
reasonably high number of cages must be installed
if relatively low values of consumption are to be
detected. The second assumption is that growth rate
inside and outside cages will be the same. The risk
of violating this assumption increases with the
length of time the cages are kept in one place (McNaughton et al. 1996). Usually, researchers have
limited time or money to allocate to setting cages
and clipping biomass. That total budget will be the
product of the number of replicated cages, which
controls for the first assumption, and the number of
times the cages are moved, which controls for the
second assumption. That tradeoff must be solved
through individual judgment in each case. Low herbivore load and large spatial heterogeneity in plant
biomass and animal distribution will need more
replicates and less frequent movement, whereas
high herbivore loads and high vegetation and animal homogeneity will call for fewer replicates and
more frequent cage movement.
What is the magnitude of the bias associated to
these two sources of error? The first error, initial
biomass differences between caged and uncaged
plots, is a simple sampling error (see Chapter 2).
Coefficient of variation of biomass for reasonable
sample sizes becomes higher as data on particular
biomass compartments, such as species or functional groups, are required. Since consumption is a
difference between two variables, its variance is actually the sum of the variances of the two variables.
The second error, the differential growth rate between caged and uncaged plots will depend on the
effect of grazing on production. Recent reviews for
grasslands and savannas indicate that grazing may
increase or decrease production by as much as 50%
(Milchunas and Lauenroth 1993; Oesterheld et al.
1999).
For systems dominated by arthropod herbivores,
there is a different method also based on the plants.
Although it is practical to exclude arthropod herbivores by means of poisoning, this is generally
done to study their effects on plant community dynamics or ecosystem function, but not to assess
consumption (e.g., Brown and Gange 1992). This
likely results from the low levels of consumption
153
in systems where the herbivore trophic level is
dominated by arthropods (McNaughton et al. 1989,
1991; Oesterheld et al. 1999). Detecting 0 to 10%
levels of consumption by differences between poisoned and control plots may be unfeasible. Fortunately for researchers, herbivore arthropods, particularly chewing insects in forests, leave marks of
their consumption on the leaves, and these marks
are the core of the plant-based approach to determine consumption levels.
This approach is based on the quantification of
the proportion of leaf area that is accounted for by
various kinds of herbivore damage: holes, mines,
galls, and scraped surfaces (Bray 1964). There are
some problems associated with this determination
of damaged leaf area, particularly for the case of
chewing insects. If damage occurs while the leaf is
growing and the quantification of the damage is
done at the end of the growing season, several distortions caused by the continuous growth of the leaf
and the holes arise (Reichle et al. 1973; Coleman
and Leonard 1995). By measuring the relative increase of hole and leaf area after punching holes in
expanding leaves, Reichle et al. (1973) showed that
consumption is overestimated when damage is
evaluated as percentage of total leaf area at the end
of the growing season. This error may be slight for
tree species with quick and simultaneous leaf
growth, and severe for tree species with continuous
growth throughout the season (Reichle et al. 1973).
The error can be reduced by performing sequential
estimations of leaf area and damaged area throughout the growing season, arriving at a daily percentage of removal (Reichle et al. 1973).
Another potential error is the integration process
of individual leaf measurements into an average
ecosystem level. First, the integration into an average proportion of leaf area for the whole stand
must be done as the ratio of total defoliated area to
total leaf area, instead of the average of the ratios
for individual leaves (Williams and Abbott 1991).
The results from either calculation method may be
considerably different if individual leaves are very
different in size (Williams and Abbott 1991). Second, a determination of percentage of leaf area
damaged, even if it is error-free, will not return directly a value of biomass or energy consumption at
the ecosystem level. In order to have consumption
values in these terms, total leaf area must be deter-
with the animal-based approach. There are two major assumptions and, as usual, controlling for the
risk of violating one increases the risk of violating
the other. The first assumption is that initial biomass inside and outside cages is the same. Thus, a
reasonably high number of cages must be installed
if relatively low values of consumption are to be
detected. The second assumption is that growth rate
inside and outside cages will be the same. The risk
of violating this assumption increases with the
length of time the cages are kept in one place (McNaughton et al. 1996). Usually, researchers have
limited time or money to allocate to setting cages
and clipping biomass. That total budget will be the
product of the number of replicated cages, which
controls for the first assumption, and the number of
times the cages are moved, which controls for the
second assumption. That tradeoff must be solved
through individual judgment in each case. Low herbivore load and large spatial heterogeneity in plant
biomass and animal distribution will need more
replicates and less frequent movement, whereas
high herbivore loads and high vegetation and animal homogeneity will call for fewer replicates and
more frequent cage movement.
What is the magnitude of the bias associated to
these two sources of error? The first error, initial
biomass differences between caged and uncaged
plots, is a simple sampling error (see Chapter 2).
Coefficient of variation of biomass for reasonable
sample sizes becomes higher as data on particular
biomass compartments, such as species or functional groups, are required. Since consumption is a
difference between two variables, its variance is actually the sum of the variances of the two variables.
The second error, the differential growth rate between caged and uncaged plots will depend on the
effect of grazing on production. Recent reviews for
grasslands and savannas indicate that grazing may
increase or decrease production by as much as 50%
(Milchunas and Lauenroth 1993; Oesterheld et al.
1999).
For systems dominated by arthropod herbivores,
there is a different method also based on the plants.
Although it is practical to exclude arthropod herbivores by means of poisoning, this is generally
done to study their effects on plant community dynamics or ecosystem function, but not to assess
consumption (e.g., Brown and Gange 1992). This
likely results from the low levels of consumption
153
in systems where the herbivore trophic level is
dominated by arthropods (McNaughton et al. 1989,
1991; Oesterheld et al. 1999). Detecting 0 to 10%
levels of consumption by differences between poisoned and control plots may be unfeasible. Fortunately for researchers, herbivore arthropods, particularly chewing insects in forests, leave marks of
their consumption on the leaves, and these marks
are the core of the plant-based approach to determine consumption levels.
This approach is based on the quantification of
the proportion of leaf area that is accounted for by
various kinds of herbivore damage: holes, mines,
galls, and scraped surfaces (Bray 1964). There are
some problems associated with this determination
of damaged leaf area, particularly for the case of
chewing insects. If damage occurs while the leaf is
growing and the quantification of the damage is
done at the end of the growing season, several distortions caused by the continuous growth of the leaf
and the holes arise (Reichle et al. 1973; Coleman
and Leonard 1995). By measuring the relative increase of hole and leaf area after punching holes in
expanding leaves, Reichle et al. (1973) showed that
consumption is overestimated when damage is
evaluated as percentage of total leaf area at the end
of the growing season. This error may be slight for
tree species with quick and simultaneous leaf
growth, and severe for tree species with continuous
growth throughout the season (Reichle et al. 1973).
The error can be reduced by performing sequential
estimations of leaf area and damaged area throughout the growing season, arriving at a daily percentage of removal (Reichle et al. 1973).
Another potential error is the integration process
of individual leaf measurements into an average
ecosystem level. First, the integration into an average proportion of leaf area for the whole stand
must be done as the ratio of total defoliated area to
total leaf area, instead of the average of the ratios
for individual leaves (Williams and Abbott 1991).
The results from either calculation method may be
considerably different if individual leaves are very
different in size (Williams and Abbott 1991). Second, a determination of percentage of leaf area
damaged, even if it is error-free, will not return directly a value of biomass or energy consumption at
the ecosystem level. In order to have consumption
values in these terms, total leaf area must be deter-