INTERNATIONAL WORKSHOP ON COUPLED DATA ASSIMILATION

What: Representatives from major operational centers and research institutions around the world gathered to assess the current state and future of coupled data assimilation (CDA). Starting with a few pioneering efforts over a decade ago, the field of CDA has grown to the point at which today many operational centers have indicated a goal of using coupled modeling and CDA as a primary prediction tool. CDA is at a stage where rapid progress is possible and can benefit a wide range of applications, including weather, subseasonal to seasonal (S2S), seasonal, and interannual climate prediction and climate reanalysis, to name a few.

When: 18–21 October 2016

Where: Météo-France, Toulouse, France

An international workshop on coupled data assimilation (CDA) hosted at Météo-France was conducted to mark progress in CDA made at operational centers and highlight developments in the research community that are helping to advance the field. The workshop consisted of daily presentations, breakout sessions to discuss aspects of DA specifically in relation to coupled DA, and plenary discussions reporting the outcomes of the breakout sessions to the full group of attendees. (Presentations are available online at www.meteo.fr/cic/meetings/2016/CDAW2016/.)

Coupled data assimilation addresses two overarching goals of the World Meteorological Organization (WMO) World Weather Research Programme (WWRP) strategic plan: 1) to improve environmental prediction and 2) develop a seamless predictive capability (Brunet et al. 2015). The WWRP has objectives to improve forecast skill of subseasonal-to-seasonal (S2S) time scales with a focus on high-impact events and promote the adoption of new initiatives by operational centers.

DAY 1 THEME: PLANS OF THE OPERATIONAL CENTERS.

Pioneering CDA efforts, such as the weakly coupled three-dimensional variational data assimilation (3DVar) used by the National Centers for Environmental Prediction (NCEP) Climate Forecast System (CFS) used to generate the CFS reanalysis (Saha et al. 2006, 2010) and the coupled ensemble Kalman filter (EnKF) developed by Zhang et al. (2007) at the Geophysical Fluid Dynamics Laboratory (GFDL), have led to many agencies now attempting to build coupled Earth system models (e.g., including atmosphere, ocean, sea ice, wave, land, aerosols, ionosphere, and biogeochemistry) to help extend forecast skill and achieve a seamless prediction capability.

As noted in the European Centre for Medium-Range Weather Forecasts (ECMWF) Strategy 2016–2025 (ECMWF 2016, p. 13), CDA methods are required to initialize these coupled Earth system modeling systems: “This [DA] framework must cater for coupled initialization requirements, in particular for ocean, sea-ice and land.” As a result, a spectrum of CDA methods is in development at each of these agencies. Examples include 1) an iterated weakly coupled four-dimensional variational data assimilation (4DVar) atmosphere/3DVar ocean DA system applied to reanalysis applications (Laloyaux et al. 2016) with a plan to extend to a coupled Earth system model (P. Laloyaux, ECMWF), 2) a weakly coupled EnKF applied to a coupled Earth system model with a path toward the implementation of hybrid methods (Hamill and Snyder 2000) and strongly coupled DA (X. Wu, Unified Global Coupled System at NCEP), 3) a full-scale 4DVar including a coupled tangent linear model [TLM; H. Ngodok, Naval Research Laboratory (NRL)], 4) an interface technology providing interaction across domains by facilitating strongly coupled DA with existing DA systems (S. Frolov, NRL) (Frolov et al. 2016), 5) weakly coupled DA with sophisticated assimilation of skin-layer sea surface temperature (SST) observations to constrain the air–sea interface and resolve the diurnal cycle [S. Akella, National Aeronautics and Space Administration (NASA)], 6) weakly coupled 4DVar atmosphere/3DVar ocean with weather-scale variations damped to improve seasonal prediction [Y. Fujii, Japan Meteorological Agency (JMA)], and 7) improved ensemble initialization using a coupled model breeding method [Y. Yin, Bureau of Meteorology (BOM)].

DAY 2 THEME: CDA METHODS.

Exciting advances in strongly coupled DA were presented by T. Sluka, who described one of the first successful demonstrations of strongly coupled DA implemented on an operational system [Climate Forecast System, version 2 (CFSv2)], expanding on earlier results that had shown 40%–60% improvement using strongly coupled versus weakly coupled DA in an intermediate-complexity coupled model (Sluka et al. 2016). While this method implemented a uniform data assimilation system for both the atmosphere and ocean, an alternative approach was proposed by S. Frolov who gave details of the interface coupler being designed at NRL (Frolov et al. 2016), identifying a primary benefit in that it allows inhomogeneous legacy systems to interact like strongly coupled DA during the analysis. Strongly coupled DA was also demonstrated in a slightly simplified scenario by A. Storto using an ocean general circulation model (OGCM) and atmospheric boundary layer model. In this application an air–sea balance operator was used to mimic a thermodynamic TLM, and positive impacts were found in the atmospheric boundary layer in the tropics.

While some efforts have already been mentioned in which coupled atmosphere–ocean TLMs were derived in order to implement a strongly coupled 4DVar, the construction and maintenance of a TLM for a full coupled Earth system model presents a significant challenge. As an alternative, the TLM can be approximated locally using ensemble estimates (Bishop et al. 2017). Such an approach provides a tool that enables coupled 4DVar for a coupled Earth system model without the need to explicitly build and maintain software for the TLM and adjoint.

The character of coupled model error covariances across model domains is an important aspect of coupled models that requires deep investigation. A thorough investigation of temporally and spatially dependent error covariances between the atmosphere and ocean was given by A. Karspeck using reanalysis experiments with the Community Earth System Model (CESM), while X. Feng investigated such covariances in the CERA-20C. CERA-20C is a 10-member ensemble coupled climate reanalysis of the 20th century, from 1901-2010 based. It is based on ECMWF’s CERA data assimilation system (Laloyaux et al., 2016) which assimilates surface pressure, marine wind observations, and profiles of ocean temperature and salinity.

A methodical investigation of more intrinsic model errors in the CESM was also conducted by A. Subramanian. These studies must be expanded upon to investigate cross covariances between all component models of the coupled Earth system.

As an alternative to testing on operational-scale coupled models, simple coupled models with similar characteristics provide great value in exploring fundamental questions and new ideas for CDA. One example of such a simple coupled model was provided by A. Lawless (Smith et al. 2015), who examined strongly coupled incremental 4DVar, weakly coupled incremental 4DVar, and uncoupled incremental 4DVar in the presence of model error using a simplified version of the ECMWF single-column model (Fowler and Lawless 2016). Results indicated that strongly coupled DA has advantages over weakly coupled DA as observations become sparse and that CDA in general provides better forecasts even if initial conditions are not always more accurate than uncoupled DA.

DAY 3 THEME: CLIMATE AND REANALYSIS APPLICATIONS.

Day 3 focused on longer-time-scale processes, including climate and reanalysis applications ranging from interannual to centennial. The coupled Earth system is noted for the presence of numerous processes occurring on disparate time scales. A number of the efforts (R. Tardif and S. Masuda) used methods to smooth the data to eliminate faster scales in the atmosphere that might be interpreted as noise and increase the correlations with the slow ocean component (Tardif et al. 2014, 2015).

Treatment of SSTs in CDA systems were a primary focal point in a number of studies. F. Counillon discussed results with a coupled model assimilating only ocean observations (SST anomalies), finding that SST data could be useful for corrections to the interior ocean due to the ensemble-derived cross covariance (Counillon et al. 2016). The Met Office has begun to supplement the diffusion operator in their 3DVar system with empirical orthogonal functions (EOFs) in the hopes of extending the influence of SST and in situ profile observations in time periods with very sparse data. NASA and NCEP both have efforts to constrain the skin-layer SST using near-surface modeling and direct assimilation of brightness temperatures.

DAY 4 THEME: NON-ATMOSPHERE/OCEAN DA.

To give due attention to coupling outside of the atmosphere–ocean paradigm, the final day of the workshop focused on coupled DA between other components, including sea ice, land, and aerosols. M. Buehner provided a review of Environment and Climate Change Canada’s developments in sea ice DA using numerous sources of sea ice concentration data at varying resolutions. He emphasized that coupled DA is important near sea ice because 1) atmospheric and oceanic data are sparse here, and 2) forecast models can be very sensitive to inconsistencies between the ocean–sea ice–atmosphere states. C. Draper followed with a description of NASA developments in land DA in the Modern-Era Retrospective Analysis for Research and Applications (MERRA2), highlighting that precipitation is the main driver of land surface hydrology. Thus, in MERRA2, observed precipitation is directly inserted at the land surface with the intent to improve land surface moisture storage, preventing model precipitation errors from feeding back to the atmosphere. A land-only EnKF (with perturbations currently generated statistically) will be adapted to couple weakly with the Global Modeling and Assimilation Office (GMAO) atmospheric ensemble EnKF/3DVar hybrid system. C. Keller detailed NASA’s efforts in developing a multispecies DA approach for atmospheric chemistry in the Goddard Earth Observing System 3D chemical transport model (GEOS-Chem; NASA’s global 3D model of atmospheric composition), emphasizing the importance of boundary conditions (e.g., over ocean and land) to the atmospheric chemistry DA problem.

BREAKOUT SESSIONS AND PLENARIES.

A series of breakout sessions provided a valuable opportunity for workshop attendees to self-select into smaller groups to discuss challenges in CDA. These discussions were documented and discussed in plenary sessions each day. Breakout sessions addressed the following topics: gaps in the overall science for CDA, methods and algorithms for CDA, estimation of coupled forecast error covariances, error and bias in coupled models, simple models for studying CDA, coupled initialization and prediction, the future of the observing system supporting CDA, coupled observation operators, software and hardware challenges for CDA, metrics and diagnostics specific to coupled problems, and CDA for reanalysis.

We briefly summarize some key points. First, regarding observations, many existing observing systems that would help coupled DA (e.g., snow observations over the United States and China) are not getting into the Global Telecommunication System (GTS). A more comprehensive and standardized data collection approach for the Earth system is desired. There is a need for a wider network of flux observations to support model validation and coupling between all model domains. Similarly, collocated observations (e.g., ocean–atmospheric observations at the same location and time) would provide significant value for model validation, bias correction, error estimation, and other DA applications.

For the coupled atmosphere–sea ice–ocean, there are significant methodological challenges, as the linear and Gaussian assumptions underlying most DA algorithms are violated more strongly than in other scenarios. In general, the specific methodologies that will become the “best practices” of the future for CDA are still a subject of research, but there is a solid foundation of research to advance the current state of the art.

FUTURE GOALS.

A white paper detailing the state of the science and listing a comprehensive set of workshop recommendations is forthcoming. For future workshops, it is desirable to expand attendance to include members of the coupled modeling community who can speak in depth on coupled model errors and biases as well as those researchers focusing on collecting observations for both long-term efforts and short-term field campaigns.

ACKNOWLEDGMENTS

We acknowledge the WMO, Météo-France, the European Union and ERA-CLIM2, and the NOAA Climate Program Office (CPO) for providing support for this workshop. We give thanks to Météo-France for hosting the workshop and providing technical support. Penny also acknowledges support from the National Weather Service (NWS) Next Generation Global Prediction System (NGGPS) program (NA15NWS4680016).

REFERENCES

REFERENCES
Bishop
,
C. H.
,
S.
Frolov
,
D. R.
Allen
,
D. D.
Kuhl
, and
K.
Hoppel
,
2017
:
The local ensemble tangent linear model: An enabler for coupled model 4DVAR
.
Quart. J. Roy. Meteor. Soc.
,
143
,
1009
1020
, doi:.
Brunet
,
G.
,
S.
Jones
, and
P. M.
Ruti
, Eds.,
2015
:
Seamless prediction of the Earth system: From minutes to months. WMO Rep. 1156, 483 pp. [Available online at http://library.wmo.int/pmb_ged/wmo_1156_en.pdf.]
Counillon
,
F.
,
N.
Keenlyside
,
I.
Bethke
,
Y.
Wang
,
S.
Billeau
,
M.-L.
Shen
, and
M.
Bentsen
,
2016
:
Flow-dependent assimilation of sea surface temperature in isopycnal coordinates with the Norwegian Climate Prediction Model
.
Tellus
,
68A
,
32437
, doi:.
ECMWF
,
2016
:
ECMWF’s strategy 2016–2025: The strength of a common goal. ECMWF Rep., 32 pp. [Available online at www.ecmwf.int/sites/default/files/ECMWF_Strategy_2016-2025.pdf.]
Fowler
,
A. M.
, and
A. S.
Lawless
,
2016
:
An idealized study of coupled atmosphere–ocean 4D-Var in the presence of model error
.
Mon. Wea. Rev.
,
144
,
4007
4030
, doi:.
Frolov
,
S.
,
C. H.
Bishop
,
T.
Holt
,
J.
Cummings
, and
D.
Kuhl
,
2016
:
Facilitating strongly coupled ocean–atmosphere data assimilation with an interface solver
.
Mon. Wea. Rev.
,
144
,
3
20
, doi:.
Hamill
,
T. M.
, and
C.
Snyder
,
2000
:
A hybrid ensemble Kalman filter–3D variational analysis scheme
.
Mon. Wea. Rev.
,
128
,
2905
2919
, doi:.
Laloyaux
,
P.
,
M.
Balmaseda
,
D.
Dee
,
K.
Mogensen
, and
P.
Janssen
,
2016
:
A coupled data assimilation system for climate reanalysis
.
Quart. J. Roy. Meteor. Soc.
,
142
,
65
78
, doi:.
Saha
,
S.
, and Coauthors
,
2006
:
The NCEP Climate Forecast System
.
J. Climate
,
19
,
3483
3517
, doi:.
Saha
,
S.
, and Coauthors
,
2010
:
The NCEP Climate Forecast System Reanalysis
.
Bull. Amer. Meteor. Soc.
,
91
,
1015
1057
, doi:.
Sluka
,
T.
,
S.
Penny
,
E.
Kalnay
, and
T.
Miyoshi
,
2016
:
Assimilating atmospheric observations into the ocean using strongly coupled ensemble data assimilation
.
Geophys. Res. Lett.
,
43
,
752
759
, doi:.
Smith
,
P. J.
,
A. M.
Fowler
, and
A. S.
Lawless
,
2015
:
Exploring strategies for coupled 4D-Var data assimilation using an idealised atmosphere–ocean model
.
Tellus
,
67A
,
27025
, doi:.
Tardif
,
R.
,
G. J.
Hakim
, and
C.
Snyder
,
2014
:
Coupled atmosphere–ocean data assimilation experiments with a low-order climate model
.
Climate Dyn.
,
43
,
1631
1643
, doi:.
Tardif
,
R.
,
G. J.
Hakim
, and
C.
Snyder
,
2015
:
Coupled atmosphere–ocean data assimilation experiments with a low-order model and CMIP5 model data
.
Climate Dyn.
,
45
,
1415
, doi:.
Zhang
,
S.
,
M. J.
Harrison
,
A.
Rosati
, and
A. T.
Wittenberg
,
2007
:
System design and evaluation of coupled ensemble data assimilation for global oceanic climate studies
.
Mon. Wea. Rev.
,
135
,
3541
3564
, doi:.

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).