Modelling Soil Carbon - Soil Carbon: Science, Management and Policy for Multiple Benefits - page 209

allow 'non-experts' to run SOC models and
advances in web-based technologies for the
management of spatial data. However, there
are limitations, and more investment is still
needed in the development of simple tools
that can be applied in those areas of the globe
most likely to experience dramatic change in
SOC over the coming decades, e.g. the trop-
ics and areas dominated by peatland.

As the scale of simulation increases,
the precision of input data tends to de-
crease, and it becomes more difficult to
quantify accurately the uncertainty in the
simulations. Reducing and quantifying the
uncertainty in large-scale simulations is a
key issue in SOM modelling if we are to
have confidence in the figures they report
and if they are to be of real value in in-
forming policy decisions. However, most
progress in reducing uncertainty is likely
to be made by improving the quality of
data available to drive models, in combin-
ation with developments in the models
themselves.
The accuracy of large-scale simulations
could be improved by developing better
methods to initialize SOM models. A range
of different methods has been used to ini-
tialize models. An approach that is robust at
large scales uses the assumption of steady
state to interpolate national soil C databases
and to initialize the pools in the model in a
way that characterizes the activity of the
SOM (Smith et al ., 2005). This approach
can reduce errors in simulations compared
to the alternative approaches that estimate
the initial pool size using typical values for
a particular soil type.
However, one problem associated with
the approach is the availability of accurate data
on which to base the initialization. Although
soil C is a standard soil parameter, reported in
most soils databases, it is often estimated from
a statistical sample of measurements for the
particular soil type; the soil C value is then
given for the major soil classifications in the
grid cell (e.g. Batjes, 2009). More accurate esti-
mates of soil C for each grid cell would im-
prove the accuracy of simulations greatly, and
it is in this area that a paradigm shift might
generate most advances.
The assumption of steady state does not
hold for some soil types, and development
of mathematical approaches to initialize
soils that are not in steady state would be
a huge step forward. The development of
methods to measure the soil pools rapidly
would provide another way of solving this
problem. While a range of fractionation tech-
niques encompassing physical, biological
and chemical fractionations have shown

Conclusions

Some researchers have recently suggested
that, in order to make any further progress,
a paradigm shift is needed in SOM model-
ling (Schmidt et al ., 2011). They argue that
because the biotic and abiotic environment
of the organic matter are as important in de-
termining the C residence times of organic
matter as the molecular structure of the ma-
terial itself, organic matter should be de-
scribed not by decay rate, pool stability or
level of 'recalcitrance', as in the current
models, but instead by quantifiable envir-
onmental characteristics governing stabil-
ization, such as solubility, molecular size
and 'functionalization'. While it is still un-
clear how such factors would be carried
forward or used in dynamic simulations,
it should also be noted that the character-
ization of pools in the current approaches
already encompass both the biotic and abi-
otic factors impacting decomposability. We
should guard against denying the huge
amount of progress that has already been
made in SOM modelling using the current
approaches.
In the above discussion, we have shown
how models can be used to quantify and re-
port changes in soil C. We have discussed
how they can be used to determine the im-
pacts of land cover, land management and
climate change on ecosystems in a range of
different environments and at scales from
plot to global. There are some environments
to which the models are not yet well suited,
but little by little, we are progressing our de-
scription of SOM dynamics to one that is
able to cover the majority of soil environ-
ments given the appropriate input data to
drive the model.