Calibration and validation are standardized practices to establish the credibility of biogeochemical models for understanding agroecosystem nutrient dynamics. We evaluated three automatic calibration approaches, including simultaneous, sequential, and separate, for the calibration of model parameters of the biogeochemical DeNitrification DeComposition (DNDC) model through inverse modeling using PEST, open-source parameter estimation, and uncertainty analysis software. While manual calibration by experts performed the best during calibration period, followed by simultaneous calibration, sequential calibration had the best model performance during the validation period. Model sensitivity analyses demonstrated water leaching to be sensitive to curve number and drain spacing, nitrate leaching to be sensitive to porosity and clay content, and corn yield to be sensitive to accumulative temperature and grain C/N ratio. While some level of expertise is required to inform the automated calibration procedure, it represents a more efficient and robust approach toward increasing the performance of biogeochemical models.

校准和验证是建立生物地球化学模型可信度以了解农业生态系统养分动态的标准化做法。我们通过使用 PEST、开源参数估计和不确定性分析软件的逆建模，评估了三种自动校准方法，包括同时、顺序和单独，用于校准生物地球化学反硝化分解 (DNDC) 模型的模型参数。虽然专家手动校准在校准期间表现最好，其次是同步校准，但顺序校准在验证期间具有最佳模型性能。模型敏感性分析表明水浸出对曲线数和排水间距敏感，硝酸盐浸出对孔隙度和粘土含量敏感，玉米产量对积温和籽粒C/N比敏感。虽然需要一定程度的专业知识来为自动校准程序提供信息，但它代表了一种提高生物地球化学模型性能的更有效和更稳健的方法。