Using an ensemble Kalman filter method for a soil nitrogen transport model in the real rice field

Abstract

The overuse of nitrogen fertilizer in rice field of China leads to nitrogen loss and serious water pollution, so it is vital to accurately predict soil nitrogen transport in rice field. But the prediction errors of soil nitrogen transport are great due to complex chemical and reactive conditions and uncertain parameters in real rice fields. In this study, a prediction model of soil nitrogen transport in a rice field was established via modifying the HYDRUS-1D source code, and a data assimilation method called the ensemble Kalman filtering (EnKF) was coupled, based on the observed NH₄⁺-N and NO₃^–-N concentrations at different depths in a real rice field. Study results for two different protocols of assimilating observed NH₄⁺-N and NO₃^–-N concentrations simultaneously and separately were compared. It indicated the predictions accuracy of NH₄⁺-N and NO₃^–-N concentrations was improved significantly via the EnKF method, and the former protocol is better than the latter. Moreover, for the latter protocol, observations of NO₃^–-N concentrations were more efficient than NH₄⁺-N to improve the predictions accuracy of NH₄⁺-N and NO₃^–-N concentrations at different depths. Inversed parameters of urea hydrolysis, NH₄⁺-N volatilization, soil adsorption of NH₄⁺-N, nitrification and denitrification increased over time. On the whole, the inversed model parameters were more stable at deep soil than shallow soil, which were different at different depths. With soil depths increase, parameters of the NH₄⁺-N adsorption and NO₃^–-N denitrification increased, while parameters of urea hydrolysis, NH₄⁺-N volatilization and nitrification decreased. This study improved the model predictions accuracy and inversed the model parameters, revealing the mechanism of nitrogen loss in real rice fields, which can provide scientific basis to reduce serious environmental problems caused by the overuse of nitrogen fertilizer.