Modeling cultivar-specific crop growth and CH4 and N2O emissions using the DeNitrification DeComposition (DNDC) model: Calibration and validation in California rice systems

New understanding and predictability of methane (CH4) and nitrous oxide (N2O) emissions from agricultural ecosystems have become increasingly important due to growing atmospheric concentrations of these greenhouse gases (GHG) and a need for accurate estimates of mitigation potentials of various management practices. The overall objective of this study was to calibrate and validate the DNDC model in California rice systems based on two crop sub-model calibrations with distinct growth and development characteristics across a wide range of management practices and environmental conditions. After model revisions to the soil climate and biogeochemistry sub-models in DNDC, the crop growth sub-model was calibrated for a high-yielding, medium-grain semi-dwarf rice cultivar (M206) and a long-grain, traditional variety, Koshihikari (KOSH), using data collected from rice fields located in California. The ability of the model to predict the observed temporal patterns and magnitude of both CH4 and N2O emissions was evaluated using field data from 9 (n = 60) and 5 (n = 50) site-year combinations, respectively. Correlations between measured and simulated carbon (C) and nitrogen (N) content in rice grain, straw and roots, root to shoot ratios, grain yield, soil temperature and soil N dynamics were determined at a subset of the experimental sites as well. The agreement between measured and simulated cumulative emissions depended on the gas (CH4 or N2O) and period of interpolation (growing season or fallow period). The model was best able to predict seasonal CH4 emissions (R2 = 0.75), with measured seasonal emissions ranging from 6 to 767 kg CH4-C ha-1. However it underestimated CH4 on average by 58% of the observed value. Low predictability of seasonal N2O emissions (R2 = 0.12) is partly attributed to the low seasonal N2O flux at all the sites (i.e., mean = 0.58 kg N-N2O ha-1, maximum =2.2 kg N2O-N ha-1). The model was unable to predict cumulative CH4 or N2O emissions during the fallow period (R2 = 0.07, 0.005, respectively). The aforementioned results are preliminary as additional datasets will be added to the validation. An analysis of the empirical errors of the model predictions using Classification and Regression Tree (CART) will identify which field conditions lead to the breakdown of the model (i.e., excessively high or low predictions). A sensitivity analysis of the model predictions to key input parameters and various GHG mitigation practices, as well as an uncertainty analysis of model predictions due to estimation of the input parameters, will be presented.