摘要：Core Ideas Ensemble Kalman filter data assimilation was used to predict soil water content. Analyzed data assimilation frequencies were 1, 2, 3, 5, 7, 9, 11, and 14 d. Assimilation of observed data every 7 d or more yielded better results. Data assimilation (DA) is a promising alternative to infer soil hydraulic parameters from soil water dynamics data. Frequency of measurements and updates are important controls of DA efficiency; however, no strict guidance exists on determining the optimal frequency. In this study, DA was performed with the ensemble Kalman filter (EnKF) with a state augmentation approach to update both model states and parameters. We analyzed updates every 1, 2, 3, 5, 7, 9, 11, and 14 d. Two soil types (sandy loam and loam) and four climates (hot semiarid [Bwh], cold semiarid [Bsk], humid continental [Dfa], and humid subtropical [Cfa]) were considered. Results demonstrate that DA with high update frequencies does not provide better results than results obtained when using low frequencies. For sandy loam soil, assimilation of data every seven or more days yields better results for whatever climate considered. For loam soil, the same is true after 9 mo of assimilation. The chosen performance metric may affect the results, but the general trend of better results with low assimilation frequencies does not change.
关键词：Bsk; cold semiarid; Bwh; hot semiarid; Cfa; humid subtropical; DA; data assimilation; Dfa; humid continental; EnKF; ensemble Kalman filter; nn; numerical nodes.
