The complex conditions in paddy fields introduce significant challenges, increasing model parameters and simulation uncertainties. To address this, a predictive model for soil nitrogen transport into surface runoff in flat bare paddy fields was developed, by modifying the HYDRUS-1D source code based on the soil mixing layer (SML) theory. The SML is the unique solute source for surface runoff in the vertical one-dimensional soil profile. The model was integrated with a data assimilation (DA) method via the ensemble Kalman filter (EnKF). The least time-averaged RMSE (called dif) of DA concentrations were obtained with optimal parameters for ensemble size of 200, error variances of 10% in initial guessed parameter and 3.34% in observation, inflation factor of 2.5 and largest measurement time number of 22. Two assimilation scenarios for simultaneous assimilation of both NH4+−N and NO3−−N concentrations and assimilation of only one nitrogen species were investigated. The former scenario predictions were called the coupled data. And the later scenario predictions were named assimilated and updated data for the same assimilation and the other species. Predictions without DA were labeled baseline outputs, and measured nitrogen concentrations in surface runoff were denoted as measurements. NH4+−N results indicated dif = 0.10, 0.20, 0.42 and 0.55 mg L− 1 for the coupled data, assimilated data, baseline outputs and updated data, respectively. The corresponding values for NO3+−N were 0.90, 2.41, 3.42 and 9.02 mg L− 1. So the prediction accuracy improved in order as coupled data > assimilated data > baseline outputs > updated data. These results indicated that assimilating measurements of a single nitrogen species sufficiently improved predictions for that species, whereas updating the other species introduced noise and reduced accuracy. Therefore, simultaneous assimilation of both NH4+−N or NO3−−N concentrations proved to be the best strategy for improving prediction accuracy in flat bare paddy fields. Temporal analysis revealed that the inversed water transfer rate and SML depth exhibited a fluctuation followed by an overall increasing trend. In contrast, inversed nitrogen reaction parameters increased with time overall and converged to similar values in deeper soil layers. The influence of deeper soil processes on surface runoff solute concentrations diminished with depth. These findings provide valuable insights for reducing simulation errors in nitrogen transport modeling for flat paddy fields.

Predicting NH₄⁺–N and NO₃⁻–N transport from soil to runoff in flat bare paddy fields using dual-species data assimilation

期刊论文
发表日期：

2026年1月4日

查看全文

Juxiu Tong,
Yang Gu

摘要

The complex conditions in paddy fields introduce significant challenges, increasing model parameters and simulation uncertainties. To address this, a predictive model for soil nitrogen transport into surface runoff in flat bare paddy fields was developed, by modifying the HYDRUS-1D source code based on the soil mixing layer (SML) theory. The SML is the unique solute source for surface runoff in the vertical one-dimensional soil profile. The model was integrated with a data assimilation (DA) method via the ensemble Kalman filter (EnKF). The least time-averaged RMSE (called dif) of DA concentrations were obtained with optimal parameters for ensemble size of 200, error variances of 10% in initial guessed parameter and 3.34% in observation, inflation factor of 2.5 and largest measurement time number of 22. Two assimilation scenarios for simultaneous assimilation of both NH₄⁺−N and NO₃⁻−N concentrations and assimilation of only one nitrogen species were investigated. The former scenario predictions were called the coupled data. And the later scenario predictions were named assimilated and updated data for the same assimilation and the other species. Predictions without DA were labeled baseline outputs, and measured nitrogen concentrations in surface runoff were denoted as measurements. NH₄⁺−N results indicated dif = 0.10, 0.20, 0.42 and 0.55 mg L^− 1 for the coupled data, assimilated data, baseline outputs and updated data, respectively. The corresponding values for NO₃⁺−N were 0.90, 2.41, 3.42 and 9.02 mg L^− 1. So the prediction accuracy improved in order as coupled data > assimilated data > baseline outputs > updated data. These results indicated that assimilating measurements of a single nitrogen species sufficiently improved predictions for that species, whereas updating the other species introduced noise and reduced accuracy. Therefore, simultaneous assimilation of both NH₄⁺−N or NO₃⁻−N concentrations proved to be the best strategy for improving prediction accuracy in flat bare paddy fields. Temporal analysis revealed that the inversed water transfer rate and SML depth exhibited a fluctuation followed by an overall increasing trend. In contrast, inversed nitrogen reaction parameters increased with time overall and converged to similar values in deeper soil layers. The influence of deeper soil processes on surface runoff solute concentrations diminished with depth. These findings provide valuable insights for reducing simulation errors in nitrogen transport modeling for flat paddy fields.

Predicting NH4+–N and NO3−–N transport from soil to runoff in flat bare paddy fields using dual-species data assimilation

摘要

Predicting NH₄⁺–N and NO₃⁻–N transport from soil to runoff in flat bare paddy fields using dual-species data assimilation