by a factor of 2, whereas the model predictions
for fragmented litter and humus were closer to
the actual C mineralization. The simulated N
mineralization rates resembled the observed
rates in litter, but for fragmented litter and
humus the simulated values accounted for only
approximately 40% and 20% of the actual N
mineralization (Berg et al. 2001). However,
when we based our model predictions on the
sampling biomasses, not averaged over time,
and averaged the derived mineralization rates
over 1 year, the model significantly overesti-
mated mineralization rates (data not published).
Sensitivity analyses of this food web model have
shown that small changes in the biomass of par-
ticular functional groups, for example basal or-
ganisms or predators, may have a marked and
disproportionate effect on the mineralization of
nutrients (Huntet al.1987;deRuiteret al. 1993).
The discrepancy between model predictions,
when based on temporal variability in functional
group biomass versus their annual biomass,
might relate to the existence of two distinct en-
ergy channels within detrital food webs (Rooney
et al. 2006). One energy channel is based on
bacteria, the other on fungi, and these two chan-
nels differ in both productivity and turnover
rate. Size differences between fungi and fungi-
vores, on the one side, and bacteria and bacter-
ivores, on the other, result in dissimilar
population dynamics, with faster turnover in
the bacteria-dominated energy channel com-
pared with the fungi-dominated channel. Aver-
aging of time in these compartmentalized food
web models may account for deviation in C and
N flows found between observed values and
model predictions. It may be more informative
to take temporal variability in food web struc-
ture into account when examining C and N dy-
namics, especially when effects of variability are
non-linear. Moreover, identification of which an-
imals are of prime importance for flow of C and
N might change when the organization of the
food web is not fixed but dynamic. For a better
understanding of how food web structure affects
food web stability and nutrient fluxes through
food webs we may need to include more tempo-
ral details.
6.5 Variability across horizontal space in soil communities
Soil organisms are not homogeneously distributed
across space, both horizontally and vertically. A
limitation of studies on spatial variability is that
they have been done on individual organism
groups rather than on whole communities. In our
example, the horizontal spatial variability in com-
munity composition is rather low and does not
differ between horizons (BCL¼0.690.062, BCF
¼0.720.046 and BCH¼0.700.047). Similarly, a
low variability in mass loss of organic matter across
space is found for all three subsequent horizons.
Moreover, a horizontal spatial structure in commu-
nity composition and mass loss of detritus is not
observed, which means that the level of variability
does not depend on the distance between sampling
points. These observations emphasize that organic
matter turnover is an important factor explaining
variability in community composition, and in hori-
zontal space.
The observed low variability in community com-
position is not expected, as the horizontal variabil-
ity measures are based on single sets of litterbags
rather than on averages of six litterbags (see Fig. 6.3
for details), and horizontal variability in plants,
microbes and animals has been shown to be sub-
stantial (Saetre and Ba ̊a ̊th 2000; Ettema and Wardle
2002; Lavermanet al. 2002). A likely explanation is
that the study site lacks ecosystem engineers, such
as earthworms, because of the acid, non-calcareous
soil type. Soil engineers can create small-scale
variability in the environment by forming casts,
redistributing organic matter and altering soil tex-
ture (for an overview, see Lavelle and Spain 2001).
Plants have been shown to increase spatial varia-
bility (Klironomoset al. 1999; Wardle 2002). The
possibility of vegetation inducing variability was
greatly reduced because the litterbags were incu-
bated at a more or less constant distance from equi-
distantly dispersed trees, in a forest with
homogeneous grass–moss undergrowth. The rela-
tively low spatial variability and absence of hori-
zontal spatial structure may thus be indicative
of the absence of factors that create variation,
especially given the short duration of the study –
only 2.5 years. In older, more mature forests with aSPATIO-TEMPORAL STRUCTURE 75