Applications of regional ocean Ensemble Kalman Filter data assimilation
File(s)
Author(s)
Li, Yi
Type
Thesis or dissertation
Abstract
Data assimilation has been widely used in the forecast of oceanic states and tropical cyclones. In this thesis, the Ensemble Kalman Filter (EnKF) based data assimilation algorithm is applied to two applications, a regional ocean data assimilation system for the South Australian Sea and a coupled ocean-atmosphere tropical cyclone (TC) forecast system.
The regional ocean data assimilation system consists of the data assimilation algorithm provided by the NCAR Data Assimilation Research Testbed (DART) and the Regional Ocean Modelling System (ROMS). We describe the first implementation of a physical balance operator (temperature-salinity, hydrostatic and geostrophic balance) to DART, to reduce the spurious waves which may be introduced during the data assimilation process. The effect of the balance operator is validated in both an idealised shallow water model and the ROMS model real case study. In the shallow water model, the geostrophic balance operator eliminates spurious ageostrophic waves and produces a better sea surface height (SSH) and velocity analysis and forecast. Its impact increases as the sea surface height and wind stress increase. In the real case, satellite-observed sea surface temperature (SST) and SSH are assimilated in the South Australian Sea with 50 ensembles using the Ensemble Adjustment Kalman Filter. Assimilating SSH and SST enhances the estimation of SSH and SST in the entire domain, respectively. Assimilation with the balance operator produces a more realistic simulation of surface currents and subsurface temperature profile. The best improvement is obtained when only SSH is assimilated with the balance operator. A case study with a storm suggests that the benefit of the balance operator is of particular importance under high wind stress conditions. Implementing the balance operator could be a general bene t to ocean data assimilation systems.
The TC forecast system consists of DART and coupled ROMS - WRF (the Weather Research Forecast model). High-frequency (HF) radars can provide high-resolution and frequent ocean surface currents observations during the TC landfall. We describe the first assimilation of such potential observations using idealised Observing System Simulation Experiments. In this system, synthetic HF radar observed coastal currents are assimilated and the forecast performances for weak (Category 2) and strong (Category 4) TCs are examined. Assimilating coastal surface currents improves the 24-hour forecasts of both intensity and track. For the strong case, the errors of the maximum wind speed (Vmax) and the integrated power dissipation (IPD) forecast reduce up to 50%. For the weak case, the improvements in Vmax and IPD forecast are lower (20%), but the track forecast improves 30%. These improvements are similar to the magnitude of the current operational TC forecast error, so that assimilating HF radar observations could be a substantial benefit.
The regional ocean data assimilation system consists of the data assimilation algorithm provided by the NCAR Data Assimilation Research Testbed (DART) and the Regional Ocean Modelling System (ROMS). We describe the first implementation of a physical balance operator (temperature-salinity, hydrostatic and geostrophic balance) to DART, to reduce the spurious waves which may be introduced during the data assimilation process. The effect of the balance operator is validated in both an idealised shallow water model and the ROMS model real case study. In the shallow water model, the geostrophic balance operator eliminates spurious ageostrophic waves and produces a better sea surface height (SSH) and velocity analysis and forecast. Its impact increases as the sea surface height and wind stress increase. In the real case, satellite-observed sea surface temperature (SST) and SSH are assimilated in the South Australian Sea with 50 ensembles using the Ensemble Adjustment Kalman Filter. Assimilating SSH and SST enhances the estimation of SSH and SST in the entire domain, respectively. Assimilation with the balance operator produces a more realistic simulation of surface currents and subsurface temperature profile. The best improvement is obtained when only SSH is assimilated with the balance operator. A case study with a storm suggests that the benefit of the balance operator is of particular importance under high wind stress conditions. Implementing the balance operator could be a general bene t to ocean data assimilation systems.
The TC forecast system consists of DART and coupled ROMS - WRF (the Weather Research Forecast model). High-frequency (HF) radars can provide high-resolution and frequent ocean surface currents observations during the TC landfall. We describe the first assimilation of such potential observations using idealised Observing System Simulation Experiments. In this system, synthetic HF radar observed coastal currents are assimilated and the forecast performances for weak (Category 2) and strong (Category 4) TCs are examined. Assimilating coastal surface currents improves the 24-hour forecasts of both intensity and track. For the strong case, the errors of the maximum wind speed (Vmax) and the integrated power dissipation (IPD) forecast reduce up to 50%. For the weak case, the improvements in Vmax and IPD forecast are lower (20%), but the track forecast improves 30%. These improvements are similar to the magnitude of the current operational TC forecast error, so that assimilating HF radar observations could be a substantial benefit.
Version
Open Access
Date Issued
2018-02
Online Publication Date
2019-03-31T06:00:35Z
2019-04-02T15:35:09Z
Date Awarded
2018-10
Advisor
Toumi, Ralf
Publisher Department
Physics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)