Reproducing CO2 exchange rates of a crop rotation at contrasting terrain positions using two different modelling approaches

• CO2 fluxes were measured periodically in a hummocky landscape using chambers.
• The investigation included maize, fodder rye and sorghum.
• CO2 fluxes were simulated using an empirical and a process-based simulation model.
• GPP, Reco, and NEE simulations were similar, only maize GPP did not match well.
• Applying both models together, individual weaknesses may be compensated.

In undulating landscapes erosion is largely responsible for the spatial distribution of C stocks in agricultural soils. Whether these stocks contribute to global atmospheric CO2 concentrations as source or sink of CO2 is under constant debate. Periodic CO2 measurements were carried out at a hummocky ground moraine site grown with maize, fodder rye and sorghum using dynamic non-steady-state transparent and opaque chambers. Flux calculation for CO2 was conducted using the empirical gap-filling model of Hoffmann et al. (2015b), which uses temperature and radiation to simulate ecosystem respiration (Reco) and gross primary production (GPP) and to calculate net ecosystem CO2 exchange (NEE). This model was compared with a process-based agro-ecosystem simulation model, MONICA, which was tested for its ability to simulate Reco, GPP and NEE, using the empirical model as benchmark. Both models simulated GPP and Reco in the same order of magnitude, with MONICA simulating a considerably higher amount of CO2 produced by photosynthesis for maize and less deviating CO2 produced by photosynthesis for the other crops and CO2 consumed by respiration for all crops as compared to the empirical model. Both models largely agree in CO2 flux patterns, but show considerable differences directly after harvest and during bare soil periods. Strengths and weaknesses of both approaches were discussed and synergies of applying both approaches in conjunction were identified in a way that (i) MONICA may act as an independent method to identify significant deviations from the optimum crop growth pattern and thus point at times during which assumptions of the empirical model for simulating NEE may be violated and that (ii) the empirical model may act as a calibration benchmark for MONICA flux simulations.