Sensitivity of the RothC to carbon inputs and evapotranspiration estimates across a gradient of data availability

The RothC model is widely used to simulate soil organic carbon (SOC) dynamics but requires key inputs, including evapotranspiration (ET) and carbon (C), to be prescribed. However, the sensitivity of RothC to different methods of estimating these prescribed inputs remains uncertain. To test this, we evaluated combinations of ET and C input estimation methods across a gradient of data availability. Initially, we established a benchmark simulation based on ‘optimum’ ET and C inputs and compared the resulting SOC estimates with available measurements from managed and unmanaged plots at the long-term Park Grass experiment in the United Kingdom. We subsequently evaluated eleven ET and C input configurations differing in their data requirements relative to the benchmark simulation. For the unmanaged plot, all configurations produced root-mean-square error (RMSE) values of <10 Mg C ha -1 relative to the benchmark, compared with only five configurations for the managed plot. Varying the ET estimation method produced RMSE values up to 2.02 and 1.84 Mg C ha -1 for the unmanaged and managed plot, respectively, when controlling for the C input method. Similarly, varying the C input methods while controlling for ET method produced RMSE values of up to 4.85 and 17.80 Mg C ha -1 , respectively. RothC was more sensitive to C input than ET estimation, although the magnitude of C input sensitivity varied among estimation methods and between the two plots. Several lower-data requirement configurations produced SOC trajectories comparable to the benchmark, indicating that greater methodological complexity did not consistently result in closer agreement. However, these findings are specific to the site and management conditions evaluated and require testing across broader soil, climate, environment and management conditions. ET and C input estimation methods should therefore be selected according to data availability, site and management conditions, and compatibility with the model structure.