International workshop on
Development of Models and Forest Soil Surveys for Monitoring of Soil Carbon
April 5-8, 2006 at Koli, Finland
Uncertainties in the model of soil organic matter pools dynamics ROMUL
A. Komarov1, S. Bykhovets1, O. Chertov2 and A. Mikhailov1
The description of soil processes of mineralization and humification in forest ecosystems deals with follow main sources of uncertainties: litter fall variables, climate variables, splitting of the soil organic matter into different fractions, which are decomposing with different rates; evaluation of the coefficients of the decomposition rates in dependence on splitting conditions; robustness of these coefficients concerning the accuracy of evaluation; initialization of the model in relation to the suggested splitting.
The ROMUL model (Chertov et al., 2001) has been developed using the classic pedological concept of the ‘humus types’. Each litter fraction is represented as: undecomposed litter, partly humified organic material in the organic layer (forest floor and peat) or the same in the mineral topsoil (fraction of ‘labile humus’). Stable humus bonded with the mineral matrix of the top soil is a summator of the humified fractions. Humification is modeled as a consequence of a successive processes regulated by three communities of saprophages. We have evaluated the decomposition coefficients as a function of the biochemical properties of litter, soil temperature and moisture using a set of laboratory experiments in controlled conditions. Verification have been done on a set of field experiments.
A Monte-Carlo procedure in relation to litter fall and climate variations resulted in small sensitivity of leading variables. Litter fall was taken from the ecosystem model EFIMOD (Chertov et al., 2003; Komarov et al., 2003).
We found that sensitivity of small Monte-Carlo changes of coefficients’ values increase with stage of decomposition. The rate of mineralization of stable humus is most sensitive to changes. It looks realistic and could be taken as a calibration parameter.
A procedure of initialization of soil data uses a concept of primary forest succession and forest types. It was found that in this case simulated results fit well with measured experimental data. The dynamics of leading variables converges to any equifinal state due to linear structure of the model. Splitting of the soil organic matter into labile and stable humus also results in converging of leading variables due to the same reason.