Modelling Soil Organic Matter Dynamics – Global Challenges - Sustainable Management of Soil Organic Matter - page 37

Long-term field experiments have been a mainstay for the development
and evaluation of soil organic matter models (Powlson, 1996). Since soil
organic matter levels change relatively slowly against a large and variable
background level, long-term experiments with well-documented manage-
ment histories are well suited to elicit and detect changes. Many SOM
models originally were developed and tested using such experiments, but
often for a restricted set of ecosystem, soil or climate conditions. However,
with the increased use of models for regional, national and even global
applications, testing of models to determine their generality (or lack
thereof) over the entire domain of their application becomes crucial. This is
particularly true given that these models are used increasingly for predictive
purposes to guide policy-making and management decisions. The recent
compilation of many of the long-term experiment data sets for the US and
Canada (Paul et al ., 1997) and Europe (Powlson et al ., 1998) have helped
to address this need by providing data, in standardized formats, which
can be used to test and validate models for broad geographic regions and
multiple types of land use, management, soil and climate conditions. A
formal network of global metadata, describing both long-term experimental
data and SOM models, SOMNET, has been established to facilitate
collaboration between modellers and data holders (Smith et al ., 1997c).
Further expansion of such networks and compilation of other networked
field data, such as long-term ecosystem C flux estimates, will play an
important role in continued improvements in our models.
Measurement and Modelling Integration
A recurrent theme in discussions about the state-of-the-art of SOM model-
ling is the disparity between modelled pools and analytical SOM fractions
measured in the laboratory. Most models represent the heterogeneity of
SOM by defining several pools, typically 3-5, each with a characteristic
specific decay rate or mean residence time, usually assuming first-order
kinetics (i.e. constant proportional mass loss per unit time) (see reviews
by Paustian, 1994; McGill, 1996). As fresh residues and SOM fractions
decompose, a portion of the residual products is more resistant to further
decay, which is represented by transfers to more slowly decomposing pools,
in a decomposition cascade (see Swift et al ., 1979). Alternatively, specific
decomposition rate can be made a function of a continuous SOM quality
spectrum, which varies according to characteristics of the plant residues
and the duration of decay (e.g. Bosatta and Agren, 1996). In either case,
the representation of the model pools (or quality spectrum) is primarily
conceptual in nature.
The sum of model pools is equal to the total organic carbon, a measur-
able and unambiguous quantity. However, individual pools generally are