Modelling Soil Organic Matter Dynamics – Global Challenges - Sustainable Management of Soil Organic Matter - page 38

only loosely associated with measurable quantities obtained through exist-
ing chemical and/or physical fractionations performed in the laboratory.
Consequently, it is not possible to falsify the internal dynamics of SOM
models with conceptual pool definitions via a direct comparison to
measured pool changes. Thus, a closer linkage between theoretical and
analytical representations of SOM heterogeneity can be advanced through
revising model definitions to coincide with measurable quantities or by
devising more functional laboratory fractionation procedures, or both. The
phrase 'modelling the measurable or measuring the modellable' has been
coined as representing the two approaches towards a closer reconciliation
between theoretical and experimental work on SOM (Christensen, 1996;
Elliott et al ., 1996).
Various attempts have been made to correlate laboratory fractions with
model pool definitions and to devise more general guidelines for initializing
the soil organic matter pools as defined in existing models. For example,
Motavalli et al . (1994) compared laboratory measurements of C mineraliza-
tion with simulations by the Century model (Parton et al ., 1994) for several
tropical soils. When the active and slow pools in the model were initialized
using laboratory determinations of microbial + soluble C for the active pool
and light fraction for the slow pool, C mineralization was consistently
underestimated, although all fractions were highly significantly correlated
to C mineralization in a regression analysis. In contrast, using the standard
procedure for internally initializing model pools, by running the model
to equilibrium with estimated climate and primary productivity driving
variables, provided the best fit to measured C mineralization. Similarly,
Magid et al . (1997) tested the DAISY model using field measurements
of litter decomposition, soil microbial biomass and particulate organic
matter (POM), and endeavoured to 'measure the modellable' ' by relating
the analytical fractions and residue quality indices to model pools. They
concluded 'there is no firm relationship between the standard set of
measurable quality parameters of the added plant materials and an
adequate parameterization of the model.' Standard measures of residue
quality such as lignin/N ratios and water-soluble/insoluble fractions were
not able to account for the initial N dynamics during decomposition.
An alternative approach to modelling the measurable has been
proposed by Arah (2000) and others (see Gaunt et al ., Chapter 2.6) using
analytically defined pools and measurement of 13 C and 15 N stable isotope
tracers to derive parameters in the model. The approach considers all
possible transformations between measured C and N pools and devises
a system of equations using observed changes in total C and N and 13 C
and 15 N for each fraction to solve all model unknowns. Necessary
requirements of such an approach are that the analytical fractions are
distinct and together account for the total carbon inventory. It must also
be parsimonious (no more than 4-5 pools each for C and N) such that