Evaluation of denitrification and decomposition from three biogeochemical models using laboratory measurements of N<sub>2</sub>, N<sub>2</sub>O and CO<sub>2</sub>

Abstract

Abstract. Biogeochemical models are essential for the prediction and management of nitrogen (N) cycling in agroecosystems, but the accuracy of the denitrification and decomposition sub-modules is critical. Current models were developed before suitable soil N2 flux data were available, which may have led to inaccuracies in how denitrification was described. New measurement techniques, using gas chromatography and isotope-ratio mass spectrometry (IRMS), have enabled the collection of more robust N2, N2O and CO2 data. We incubated two arable soils – a silt-loam and a sand soil – for 34 and 58 d, respectively, with small field-relevant changes made to control factors during this period. For the silt-loam soil, seven treatments varying in moisture, bulk density and NO3- contents were included, with temperature changing during the incubation. The sandy soil was incubated with and without incorporation of litter (ryegrass), with temperature, water content and NO3- content changing during the incubation. The denitrification and decomposition sub-modules of DeNi, Coup and DNDC were tested using the data. No systematic calibration of the model parameters was conducted since our intention was to evaluate the general model structure or “default” model runs. Measured fluxes generally responded as expected to control factors. We assessed the direction of modeled responses to control factors using three categories: no response, a response in the same direction as measurements or a response in the opposite direction to measurements. DNDC responses were 14 %, 52 % and 34 %, respectively. Coup responses were 47 %, 19 % and 34 %, respectively. DeNi responses were 0 %, 67 % and 33 %, respectively. The magnitudes of the modeled fluxes were underestimated by Coup and DNDC and overestimated by DeNi for the sandy soil, while there was no general trend for the silt-loam soil. None of the models was able to determine litter-induced decomposition correctly. To conclude, the currently used sub-modules are not able to consistently simulate the denitrification and decomposition processes. For better model evaluation and development, we need to design better experiments, take more frequent measurements, use new or updated measurement techniques, address model complexity, add missing processes to the models, calibrate denitrifier microbial dynamics, and evaluate the anaerobic soil volume concept.

Modeling Denitrification: Can We Report What We Don't Know?

“…The presented results in Table 1 did not cover the statements and results of the entire study by Grosz et al. (2021).…”

“…Some models (Del Grosso et al., 2000; Parton et al., 1996) have been partly parameterized with N 2 data that are no longer considered reliable (e.g., based on the acetylene inhibition technique (Weier et al., 1993)) and other model calibrations are simply incomplete. Given the lack of empirical data, approaches to describe soil N 2 are mostly process‐oriented, with the sensitivity of both N 2 and N 2 O to controlling factors constrained based solely on N 2 O data (Grosz et al., 2021; Zhang et al., 2022). In these models, having data from frequent measurements (beyond the common weekly or fortnightly intervals) is crucial.…”

“…That level of complexity is challenging to model. Laboratory studies under controlled conditions help isolate specific controlling factors' effects (Grosz et al., 2021; Müller & Clough, 2014; Weier et al., 1993), with both field and laboratory measurements being used to refine biogeochemical models under differing conditions (Deng et al., 2016; Hergoualc'h et al., 2021). However, in many cases, N 2 O is neither the final product nor the main product of denitrification (Scheer et al., 2020).…”

Modeling nitrous oxide emissions from agricultural soil incubation experiments using CoupModel

“…However, single-treatment calibration showed more success to simulate the effects of different soil moisture levels on N 2 O fluxes. In a recent study, Grosz et al (2021) also found it difficult to describe treatments in incubation studies with three N 2 O models (DNDC, CoupModel, and DeNi) that responded to controlling factors in the same direction as measurements with frequencies from only 19 % to 67 %. In CoupModel, soil moisture effects have indirect effects on denitrification through decomposition, nitrification, and gas diffusion processes in which the soil moisture response function and its association with anaerobic conditions were accounted for.…”

“…Having explicit measurements in each layer (i.e., soil-residue mixture and bulk soil layer) of incubated soil cores and having higher measurement frequencies of mineral N and gas emissions within the first 24 h of the experiment are promising tools to improve the reliability of experimental data for modeling. Thirdly, in soil N 2 O modeling, measuring more ancillary variables regarding N cycling (e.g., N 2 , NO) and using state-of-the-art techniques (e.g., 15 N gas flux methods) could help test the validity of model principles (Grosz et al, 2021). Although collecting all data types discussed here is not always possible or practical, we consider that reporting ancillary model outputs regarding N cycle processes even in the absence of observations, particularly the denitrification products, soil oxygen content, and anaerobic fraction, will aid model evaluation efforts, which was not done very often in previous studies.…”