All-Sky Satellite Radiance Data Assimilation: Methodology and Challenges - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 459

The main assumption of variational methods is related to the forecast error
covariance, which is modeled, and thus generally static without time depen-
dence. Another characteristic of variational methods is that they are developed
around an iterative minimization algorithm, typically of unconstrained type (e.g.,
conjugate-gradient, quasi-Newton), making them suitable for nonlinear processes
and operators.
Ensemble DA methods use ensemble of forecast models to define a time-
dependent forecast error covariance, however for the price of being a reduced
rank approximation. They are algorithmically simpler than variational methods,
thus easier to develop and maintain. Although ensemble DA can handle very well
nonlinearities of the forecast model, their straightforward application is not very
good for addressing nonlinearities of observation operator because the analysis is
based on using the Kalman filter linear solution.

19.2.2 Assimilation of All-sky Radiances

Most operational weather centers are currently using variational DA methods,
although they actively investigate ensemble and hybrid variation-ensemble methods.
On the other side, ensemble and hybrid variational-ensemble methodologies are typ-
ically developed at research laboratories and universities. However, this distinction
is not that clear and there are several research data assimilation algorithms being
tested for operational use. Even though there is a wealth of information that all-
sky radiances could bring to the prediction system, only a limited research effort
to assimilate such observations exists. Most efforts include the use of variational
methods (e.g., Vukicevic et al. 2004 ; Bauer et al. 2006 , 2010 ; Geer et al. 2010 ; Geer
and Bauer 2010 ; Polkinghorne and Vukicevic 2011 ) with some applications within
ensemble and hybrid variational-ensemble methods (e.g., Zupanski et al. 2011a ,b;
Zhang et al. 2012 ).
Especially relevant is the pioneering work by satellite research group at the
European Centre for Medium Range Weather Forecast (ECMWF) (e.g., Bauer
et al. 2010 ; Geer et al. 2010 ) leading to the first operational assimilation of all-
sky radiances, since 2009. There are similar efforts to assimilate all-sky satellite
radiances in the United States at the National Centers for Environmental Prediction
(NCEP), and other centers will likely follow.

19.3 Challenges

In general, challenges of all-sky satellite radiance assimilation originate due to
their relation to clouds. Observing and simulating clouds is challenging in it
own right. This is magnified in data assimilation, being a method that combines
information from observations and from prediction models. Although problems