Parameter and input data uncertainty estimation for the assessment of long-term soil organic carbon dynamics

Abstract

The use of integrated soil organic matter (SOM) models to assess SOM dynamics under climate change, land use change and different land management practices require a quantification of uncertainties and key sensitive factors related to the respective modelling framework. Most uncertainty studies hereby focus on model parameter uncertainty, neglecting other sources like input data derived uncertainties, and spatial and temporal properties of uncertainty. Sources of uncertainties assessed in this study stem from uncertainties in model parameterisation and from uncertainties in model input data (climate, soil data, and land management assumptions). Thereby, Monte Carlo based global sensitivity and uncertainty analysis using a latin hypercube stratified sampling technique was applied to derive plot scale (focusing on temporal propagation) and river basin scale propagation of uncertainty for long-term soil organic carbon (SOC) dynamics. The model used is the eco-hydrological river basin model SWIM (Soil and Water Integrated Model), which has been extended by a process-based multi-compartment model for SOM turnover. Results obtained by this study can be transferred and used in other simulation models of this kind. Uncertainties resulting from all input factors used (model parameters + model input data) show a coefficient of variation between 5.1 and 6.7% and accounted for ± 0.065 to ± 0.3% soil carbon content (0.06–0.15 t C ha −1 yr −1). Parameter derived uncertainty contributed most to overall uncertainty. Concerning input data contributions, uncertainties stemming from soil and climate input data variations are striking. At the river basin scale, cropland and forest ecosystems, loess and gleyic soils possess the highest degree of uncertainty. Quantified magnitudes of uncertainty stemming from the examined sources vary temporally and spatially due to specific natural settings (e.g. climate, land use and soil properties) and deliver useful information for interpreting simulation results on long-term soil organic carbon dynamics under environmental change. Derived from this analysis, key sensitive model parameters and interactions between them were identified: the mineralization rate coefficient, the carbon use efficiency parameter (synthesis coefficient) along with parameters determining the soil temperature influence on SOM turnover (mainly Q10 value) and the soil input related data ( soil bulk density and initial soil C content) introduced the highest degree of model uncertainty. The here gained information can be transferred to other process-based SOM turnover models to consider stronger most crucial parameters introducing highest uncertainty contribution to soil C storage assessment under changing environmental conditions.