Assimilation of stream discharge for flood forecasting: The benefits of accounting for routing time lags

Abstract

[1] General filtering approaches in hydrologic data assimilation, such as the ensemble Kalman filter (EnKF), are based on the assumption that uncertainty of the current background prediction can be reduced by correcting errors in the state variables at the same time step. However, this assumption may not be valid when assimilating stream discharge into hydrological models to correct soil moisture storage due to the time lag between the soil moisture and the discharge. In this paper, we explore the utility of an ensemble Kalman smoother (EnKS) for addressing this time-lag issue. The EnKF and the EnKS are compared for two different updating schemes with the probability distributed model (PDM) via synthetic experiments: (i) updating soil moisture only and (ii) updating soil moisture and routing states simultaneously. The results show that the EnKS is superior to the EnKF when only soil moisture is updated, while the EnKS and the EnKF exhibit similar results when both soil moisture and routing storages are updated. This suggests that the EnKS can better improve the streamflow forecasting for models that do not adopt storage-based routing schemes (e.g., unit-hydrograph-based routing). For models with dynamic routing stores, errors in soil moisture are transferred to the routing stores, which can be corrected effectively by real-time filters. The EnKS-based soil moisture updating scheme is also tested with the GR4H model, for which unit-hydrograph-based routing is used. The result confirms that the EnKS is superior to the EnKF in improving both soil moisture and streamflow forecasting.

Multivariate hydrological data assimilation of soil moisture and groundwater head

“…Localization not only provides better results, but also reduces the computational cost, as only a section of the full state is used within the filter. Similar localization approaches have been reported in hydrological models with discharge involved (Li et al, 2013) as well as in other models (e.g. Kang et al, 2011).…”

Operational aspects of asynchronous filtering for flood forecasting

“…The ensemble Kalman smoother (EnKS) is a common example of an asynchronous method (e.g. Evensen and Van Leeuwen, 2000;Dunne and Entekhabi, 2006;Crow and Ryu, 2009;Li et al, 2013). The EnKS extends the EnKF by introducing additional information by propagating the contribution of future measurements backward in time.…”

Data assimilation in integrated hydrological modeling using ensemble Kalman filtering: evaluating the effect of ensemble size and localization on filter performance

“…Adaptive localization is not only easily implemented in the ETKF, it also automatically ensures that the cross-process correlation is localized differently than the intra-process correlation, making it particularly suitable for data assimilation in coupled surfacesubsurface models. Others have encountered the problem with cross-process correlation, notably Zupanski (2013), Li et al (2013) and Wanders et al (2014), although no definitive solution to the problem has been presented. Adaptive localization, such as the method applied in this study, may be one possible solution.…”