Predictive modelling of C dynamics in the long-term fertilization experiment at Bad Lauchstadt with the Rothamsted Carbon Model

Models are important for predicting how soil organic carbon alters with changing environment and management. We tested three different ways of parameterizing the Rothamsted Carbon Model in four treatments from the long-term static fertilization experiment at Bad Lauchstadt. Four bare fallow plots were also used to test different ways of parameterizing the model. Model version 1 used Delta C-14 to estimate the amount of inert organic matter whereas in versions 2 and 3 a long-term bare fallow treatment was used to estimate this amount. In version 1, C inputs were optimized from the long-term data. In version 2, crop inputs were estimated from published functions that relate C inputs to crop yields. In version 3, C inputs (average or minimum data) were taken from actual measurements of crop and root residues. In both versions 2 and 3, rhizodeposition was included as additional input of 50% (winter wheat, spring barley) or 35% (potatoes, sugar beet) of the C input by crop and root residues. The performance of the three versions decreased in the order 3 > 2 > 1. Model efficiency and root mean square error were 0.86 and 6.1 for version 3 and 0.81 and 7.0 for version 2. Overall, our results indicate the need for a long-term treatment for calibration. Setting total C inputs as a function of crop yield performed satisfactorily. Measurements of crop and root residues gave a good representation of total C inputs when carbon from rhizodeposition was included as additional input.