 |
| Titel |
Variational fine-grained data assimilation schemes for atmospheric chemistry transport and transformation models |
| VerfasserIn |
Alexey Penenko, Vladimir Penenko, Elena Tsvetova |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250101819
|
| Publikation (Nr.) |
 EGU/EGU2015-1055.pdf |
|
|
|
|
|
| Zusammenfassung |
| The paper concerns data assimilation problem for an atmospheric chemistry transport and
transformation models. Data assimilation is carried out within variation approach on a single
time step of the approximated model. A control function is introduced into the model source
term (emission rate) to provide flexibility to adjust to data. This function is evaluated as the
minimum of the target functional combining control function norm to a misfit between
measured and model-simulated analog of data. This provides a flow-dependent and
physically-plausible structure of the resulting analysis and reduces the need to calculate
model error covariance matrices that are sought within conventional approach to data
assimilation.
Extension of the atmospheric transport model with a chemical transformations module
influences data assimilation algorithms performance. This influence is investigated with
numerical experiments for different meteorological conditions altering convection-diffusion
processes characteristics, namely strong, medium and low wind conditions. To
study the impact of transformation and data assimilation, we compare results for
a convection-diffusion model (without data assimilation), convection-diffusion
with assimilation, convection-diffusion-reaction (without data assimilation) and
convection-diffusion-reaction-assimilation models.
Both high dimensionalities of the atmospheric chemistry models and a real-time mode of
operation demand for computational efficiency of the algorithms. Computational issues with
complicated models can be solved by using a splitting technique. As the result a model is
presented as a set of relatively independent simple models equipped with a kind of coupling
procedure. With regard to data assimilation two approaches can be identified. In a
fine-grained approach data assimilation is carried out on the separate splitting stages [1,2]
independently on shared measurement data. The same situation arises when constructing a
hybrid model out of two models each having its own assimilation scheme. In integrated
schemes data assimilation is carried out with respect to the split model as a whole. First
approach is more efficient from computational point of view, for in some important
cases it can be implemented without iterations [2]. Its shortcoming is that control
functions in different part of the model are adjusted independently thus having less
evident physical sense. With the aid of numerical experiments we compare the two
approaches.
Work has been partially supported by COST Action ES1004 STSM Grants #16817 and #21654,
RFBR 14-01-31482 mol a and 14-01-00125, Programmes # 4 Presidium RAS and # 3 MSD RAS,
integration projects SB RAS #8 and #35.
References:
[1] V. V. Penenko Variational methods of data assimilation and inverse problems for studying
the atmosphere, ocean, and environment Num. Anal. and Appl., 2009 V 2 No 4,
341-351.
[2] A.V. Penenko and V.V. Penenko. Direct data assimilation method for convection-diffusion
models based on splitting scheme. Computational technologies, 19(4):69–83, 2014. |
|
|
|
|
|
|