Hybrid ensemble 4DVar assimilation of stratospheric ozone using a global shallow water model

Abstract

Abstract. Wind extraction from stratospheric ozone (O 3 ) assimilation is examined using a hybrid ensemble 4-D variational assimilation (4DVar) shallow water model (SWM) system coupled to the tracer advection equation. Stratospheric radiance observations are simulated using global observations of the SWM fluid height (Z), while O 3 observations represent sampling by a typical polar-orbiting satellite. Four ensemble sizes were examined (25, 50, 100, and 1518 members), with the largest ensemble equal to the number of dynamical state variables. The optimal length scale for ensemble localization was found by tuning an ensemble Kalman filter (EnKF). This scale was then used for localizing the ensemble covariances that were blended with conventional covariances in the hybrid 4DVar experiments. Both optimal length scale and optimal blending coefficient increase with ensemble size, with optimal blending coefficients varying from 0.2-0.5 for small ensembles to 0.5-1.0 for large ensembles. The hybrid system outperforms conventional 4DVar for all ensemble sizes, while for large ensembles the hybrid produces similar results to the offline EnKF. Assimilating O 3 in addition to Z benefits the winds in the hybrid system, with the fractional improvement in global vector wind increasing from ∼ 35 % with 25 and 50 members to ∼ 50 % with 1518 members. For the smallest ensembles (25 and 50 members), the hybrid 4DVar assimilation improves the zonal wind analysis over conventional 4DVar in the Northern Hemisphere (winter-like) region and also at the Equator, where Z observations alone have difficulty constraining winds due to lack of geostrophy. For larger ensembles (100 and 1518 members), the hybrid system results in both zonal and meridional wind error reductions, relative to 4DVar, across the globe.

Extraction of wind and temperature information from hybrid 4D-Var assimilation of stratospheric ozone using NAVGEM

“…In addition, the impact of assimilation of other tracers could be tested in a similar framework. For example, Andersson et al (1994Andersson et al ( , 2007 have shown dynamical impacts from assimilation of radiance channels sensitive to water vapor, and Allen et al (2014), in a shallow water model study, showed that nitrous oxide could additionally improve winds in 4D-Var DA. Given the potential benefit of tracer assimilation on the dynamics in NWP (referred to by Daley, 1995, as a "tantalizing possibility"), it is our hope that this study will motivate future work in this area and eventually result in improved operational analyses and forecast skill.…”

“…The spatial-temporal variability of long-lived tracers such as stratospheric ozone contains dynamical information that can potentially be exploited to improve analyses of wind and temperature in the stratosphere and mesosphere, where direct wind observations are largely absent. Various studies have examined tracer-wind interactions within a variety of data assimilation (DA) systems including extended Kalman filter (EKF) (Daley, 1995(Daley, , 1996, 4-D variational assimilation (4D-Var) (Andersson et al, 1994;Riishøjgaard, 1996;Peuch et al, 2000;Andersson et al, 2007;Peubey and McNally, Published by Copernicus Publications on behalf of the European Geosciences Union. 3000 D. R. Allen et al: Extraction of wind and temperature information from ozone assimilation 2009; Semane et al, 2009;Han and McNally, 2010;Dragani and McNally, 2013;Allen et al, 2013Allen et al, , 2014, ensemble Kalman filter (EnKF) (Milewski and Bourqui, 2011;Allen et al, 2015), and hybrid 4D-Var (Allen et al, 2016).…”

Extraction of wind and temperature information from hybrid 4D-Var assimilation of stratospheric ozone using NAVGEM

“…This is likely due the combined positive effects of the ensemble flow-of-the-day information with the negative aspects of spurious unbalanced modes spawned by the localized ensemble covariance. These aspects were discussed in the shallow water model context in Allen et al (2016), where it was shown that the optimal blending coefficient also depends both on the data being assimilated and on the ensemble size. 15…”

“…Theoretical studies have examined tracer-wind interactions within a variety of data assimilation (DA) systems including extended Kalman Filter (EKF) (Daley, 1995(Daley, , 1996, 4D Variational assimilation (4D-Var) (Riishøjgaard, 1996;Peuch et al, 2000;Andersson et al, 2007;Peubey and McNally, 2009;Semane et al, 2009;Han and McNally, 2010;Dragani and McNally, 2013;Allen et al, 2013Allen et al, , 2014, Ensemble Kalman Filter (EnKF) (Milewski and30 Bourqui, 2011, Allen et al, 2015), and hybrid 4D-Var (Allen et al, 2016). While these studies show that strong potential exists for wind extraction from tracer assimilation, the operational benefit has yet to be obtained.…”

Inferring atmospheric dynamics from aerosol observations in 4D‐Var

“…A number of studies focused on wind extraction from stratospheric ozone, which can be considered a passive tracer, since physical processes affect ozone fields much less than advection. A varying level of realism was applied, from idealized models of different complexities and simulated observations (Riishøjgaard, ; Allen et al ., ; ) to NWP forecast models and idealized observations (Peuch et al ., ; Allen et al ., ). A positive impact of ozone observations on wind analyses was reported not only for perfect observations but also for real observations, although the impact was less significant.…”