Polly J. Smith

Estimating Forecast Error Covariances for Strongly Coupled Atmosphere–Ocean 4D-Var Data Assimilation

AbstractStrongly coupled data assimilation emulates the real-world pairing of the atmosphere and ocean by solving the assimilation problem in terms of a single combined atmosphere–ocean state. A significant challenge in strongly coupled variational atmosphere–ocean data assimilation is a priori specification of the cross covariances between the errors in the atmosphere and ocean model forecasts. These covariances must capture the correct physical structure of interactions across the air–sea interface as well as the different scales of evolution in the atmosphere and ocean; if prescribed correctly, they will allow observations in one medium to improve the analysis in the other. Here, the nature and structure of atmosphere–ocean forecast error cross correlations are investigated using an idealized strongly coupled single-column atmosphere–ocean 4D-Var assimilation system. Results are presented from a set of identical twin–type experiments that use an ensemble of coupled 4D-Var assimilations to derive estima...

Data assimilation for morphodynamic prediction and predictability

This paper gives an overview of the project Changing coastlines: data assimilation for morphodynamic prediction and predictability. This project is investigating whether data assimilation could be used to improve coastal morphodynamic modeling. The concept of data assimilation is described, and the benefits that data assimilation could bring to coastal morphodynamic modeling are discussed. Application of data assimilation in a simple ID morphodynamic model is presented. This shows that data assimilation can be used to improve the current state of the model bathymetry, and to tune the model parameter. We now intend to implement these ideas in a 2D morphodynamic model, for two study sites. The logistics of this are considered, including model design and implementation, and data requirement issues. We envisage that this work could provide a means for maintaining up-to date information on coastal bathymetry, without the need for costly survey campaigns. This would be useful for a range of coastal management issues, including coastal flood forecasting.

Observation impact, domain length and parameter estimation in data assimilation for flood forecasting

Abstract Accurate inundation forecasting provides vital information about the behaviour of fluvial flood water. Using data assimilation with an Ensemble Transform Kalman Filter we combine forecasts from a numerical hydrodynamic model with synthetic observations of water levels. We show that reinitialising the model with corrected water levels can cause an initialisation shock and demonstrate a simple novel solution. In agreement with others, we find that although assimilation can accurately correct water levels at observation times, the corrected forecast quickly relaxes to the open loop forecast. Our new work shows that the time taken for the forecast to relax to the open loop case depends on domain length; observation impact is longer-lived in a longer domain. We demonstrate that jointly correcting the channel friction parameter as well as water levels greatly improves the forecast. We also show that updating the value of the channel friction parameter can compensate for bias in inflow.

Treating Sample Covariances for Use in Strongly Coupled Atmosphere‐Ocean Data Assimilation

Strongly coupled data assimilation requires cross-domain forecast error covariances; information from ensembles can be used, but limited sampling means that ensemble derived error covariances are routinely rank deficient and/or ill-conditioned and marred by noise. Thus they require modification before they can be incorporated into a standard assimilation framework. Here, we compare methods for improving the rank and conditioning of multivariate sample error covariance matrices for coupled atmosphere-ocean data assimilation. The first method, reconditioning, alters the matrix eigenvalues directly; this preserves the correlation structures but does not remove sampling noise. We show it is better to recondition the correlation matrix rather than the covariance matrix as this prevents small but dynamically important modes from being lost. The second method, model state-space localisation via the Schur product, effectively removes sample noise but can dampen small cross-correlation signals. A combination that exploits the merits of each is found to offer an effective alternative.