Implications of input estimation, residue quality and carbon saturation on the predictive power of the Rothamsted Carbon Model

Abstract

Effects of fertilisation and cropland management on soil organic carbon (SOC) dynamics can be assessed best in long-term experiments. Using data from the long-term fertilisation experiment in Puch, Germany (part of the series “Internationale Organische Stickstoff Dauerversuche”, IOSDV), we tested the performance of the Rothamsted Carbon Model 26.3 (RothC). The objectives of this work were: (i) quantify the C-input and the efficiency of SOC stabilisation, (ii) test the performance of different input estimates on predictive power of the RothC and (iii) test implementations of residue quality and C-saturation on model predictions. The experiment is a full-factorial strip design, the factors being “organic amendment” and “level of N-fertiliser”. Each treatment was replicated three times. The crop rotation is silage maize–winter wheat–winter barley. Five levels of the factor “organic amendment” were considered: (i) CON: no organic amendment; (ii) SLU: slurry application (on average 0.8MgCha−1year−1); (iii) FYM: application of farmyard manure (30 to 40Mgha−1 fresh mass every third year to maize, on average 1.0MgCha−1year−1); (iv) STR: straw incorporation after harvest of wheat and barley (depending on straw yield on average 0.7 to 2.2MgCha−1year−1); (v) STSL: slurry application plus straw incorporation (on average 1.1 to 2.4MgCha−1year−1). All treatments (including CON) were combined with five different levels of N-fertilisation (N0 to N4), whereas N0 was nil N application and N4 averaged 177kgNha−1year−1. N-rates increased gradually and differed depending on the crop. Starting values for SOC stocks (Mgha−1) were measured in 1983 as a mean among N-rates for organic amendment treatments (CON: 42; SLU: 39.8; FYM: 40.5; STR 39.8; STSL: 40.5). SOC stocks (0–25cm) in 2004 (35.5 to 46.6MgCha−1) were in the order STSL>FYM=SLU>STR=CON (p≤0.001). However, slightly different starting values indicated a higher loss of SOC after 21years in the CON (11–14%) compared to the STR treatments (1–10%). Effect of N-rate was not significant. The observed relation between change of SOC and C-input was quadratic (YO=−13.4+7.5x−0.9x2; R2=0.74, p≤0.001), which contrasted the linear relationship predicted by RothC (YP=−12.9+5.5x; R2=0.97, p≤0.0001). Serious deviation between observed and predicted relationship occurred above C-inputs of 2.5MgCha−1year−1. Mechanistic explanation (e.g. C-saturation or increased mineralisation by N-fertilisation) for the observation needs further exploration, but implication on regional estimates for C-accumulation for different cropland management scenarios is obvious: potential gain in SOC storage by increasing C-inputs may be overestimated, at least under conditions of the Puch site. Independent model predictions (i.e. no parameter adjustment and independent estimation and measurement of C-input) were successful for treatments without straw incorporation (CON, SLU, FYM). Using a regression between crop yields and crop residue input yielded better results than using a constant belowground-to-aboveground biomass ratio. SOC stocks of treatments STR and STSL were seriously overestimated by the model. Using a higher decomposability of crop residue improved result only marginally and required the change of a standard parameter. Using a simple implementation of C-saturation improved predictions for STR and STSL but failed to simulate dynamics in all other treatments. Overall, our results showed that it is important to recognise that relation between SOC change and C-input is not necessarily linear. However, the RothC model predicted SOC dynamics well at lower input levels. Observation that a regression equation for input estimation is superior to a constant biomass ratio for modelling purposes has to be tested further. An implementation of residue quality or saturation capacity in the RothC model may be promising for a better mechanistic understanding of SOC dynamics. However, this requires careful calibration and will increase the number of parameters to be fitted.