on the problem of initializing climate prediction · 6th wmo symposium on data assimilation -- 2013...

6th WMO Symposium on Data Assimilation -- 2013

On the Problem of Initializing Climate Prediction

Alberto Carrassi a, Robin Weber a,b, Virginie Guemas a,c, Francisco Doblas-Reyes a,d, Muhammad Asif a and Danila Volpi a

a Institut Català de Ciències del Clima (IC3), Barcelona, Spain, [email protected], b Dept. of Physics, University of Heidelberg, Germany, c Centre National de Recherches Météorologiques/Groupe d’Etude de

l’Atmosphère Météorologique, UMR3589, France, d Institució Catalana de Recerca i Estudis Avançats, Spain

Initialization techniques for seasonal-to-decadal climate predictions fall into two main categories, namely Full Field Initialization (FFI) and Anomaly Initialization (AI). In the FFI case the initial model state is replaced by the best-possible available estimate of the actual state. By doing so the initial error is efficiently reduced but, due to the unavoidable presence of model deficiencies, once the model is let free to run a prediction its trajectory drifts away from the observations no matter how small the initial error is. This problem is partly overcome with the AI where the aim is to forecast future anomalies by assimilating observed climate anomalies on an estimate of the model climate. This way, the initial model state is kept on (or closer to) its own attractor.

The large variety of experimental setups, models and observational networks adopted worldwide make difficult to draw firm conclusions on the respective advantages and drawbacks of the FFI and AI, let alone identifying distinctive lines for improvement. The lack of a unified mathematical framework adds an additional difficulty toward the design of adequate initialization strategies that fit the desired forecast horizon, observational network and model at hand.

In this study we use the notation and concepts of data assimilation theory to propose a unified formalism from which the different initialization approaches can be easily derived and seen as a particular case. Using the idealized coupled dynamics introduced by [1], FFI and AI are compared and studied in a range of different observational and model error scenario, helping clarifying under which conditions one approach outperforms the other. Moreover two advanced formulations of FFI/AI are also proposed to improve the fit to the observations and the bias reduction. Finally, preliminary results using the climate model EC-Earth are used to illustrate the impact of the initialization on climate forecast quality with a state-of-the-art dynamical climate prediction system.

References [1] Pena M., and E. Kalnay, “Separating fast and slow modes in coupled chaotic systems”, Nonlin. Proc. Geophysics, 11, 319–327, 2004.

Evaluating the Fidelity of a Community Coupled Ocean-Atmosphere Data Assimilation System

Abhishek Chatterjeea, Alicia R. Karspeckb, Jeffrey L. Andersonc, Nancy Collinsc, Gokhan Danabasoglub, Timothy J. Hoarc, Kevin D. Raederc, Joseph J. Tribbiab

a CGD/IMAGe, National Center for Atmospheric Research, USA, Email: [email protected], b CGD,

NESL, National Center for Atmospheric Research, USA, c IMAGe, CISL, National Center for Atmospheric Research, USA

Several prototypes of coupled ocean-atmosphere data assimilation (DA) frameworks have been developed or are under development by different groups worldwide [e.g. 1, 2, 3, 4]. Almost all of these systems, however, tend to analyze the atmosphere and the ocean separately, i.e., coupled single-component DA, thereby limiting the impact of observations across the air-sea interface. While such a framework serves as a valuable ‘intermediate’ step for operational centers such as the NCEP [3], several fundamental questions remain unanswered. For example, it is not yet clear what will be the potential benefit of using near surface observations (SST, for example) to correct both the ocean and the atmospheric states simultaneously within a coupled framework. At a more fundamental level, the relative differences in accuracy between different coupled frameworks, which can include assimilation of data either within a single-component, within multiple components or allow cross-component interactions, remain to be established.

Recently, the Community Earth System Model (CESM; previously known as the Community Climate System Model -CCSM) [5] has been interfaced to a community facility for ensemble data assimilation (Data Assimilation Research Testbed – DART) [6], which allows us to answer such questions critically by examining a suite of coupled system configurations. First a coupled multi-component DA system is being run in which data is assimilated into each of the respective ocean/atmosphere model components during the assimilation step, and information is exchanged between the model components during the forecast step. Secondly, two coupled single-component DA systems are being run in which observations are assimilated into only one of the ocean/atmosphere components during the assimilation step. Work is ongoing to expand the coupled multi-component DA framework to allow for cross-component DA, in which the data will be assimilated into each of the ocean and the atmosphere components but the information immediately transferred to the other component through the ensemble filter, and both components updated simultaneously at each assimilation step. Initial runs are being conducted to evaluate the differences between these coupled system configurations after assimilation experiments of length one year. In this presentation, specific analyses (e.g., development of tropical storms, sea-ice formation around the Arctic) as well as correlations of states with observations across model components will be discussed. Details of the air-sea fluxes (heat, momentum and moisture) and corresponding ocean-atmosphere dynamical feedbacks will also be presented. The knowledge gained through this study is expected to improve our understanding of the fidelity and applicability of coupled ocean-atmosphere DA systems, and potentially their long-term climate prediction capabilities. References [1] Zhang, S. and co-authors. "System design and evaluation of coupled ensemble data assimilation for global oceanic studies," Mon. Wea. Rev., 135, 10, pp. 3541-3564, October 2007. (Journal Article) [2] Sugiura, N. and co-authors. "Development of a four-dimensional variational coupled data assimilation system for enhanced analysis and prediction of seasonal to interannual climate variations," J. Geophys. Res., 113, C10, pp. 1-21, October 2008. (Journal Article) [3] Saha, S. and co-authors. "The NCEP Climate Forecast System Reanalysis," Bull. Amer. Meteor. Soc, 91, 8, pp. 1015-1057, August 2010. (Journal Article) [4] Holt, T. R. and co-authors. "Development and testing of a coupled ocean–atmosphere mesoscale ensemble prediction system," Ocean Dynamics, 61, 11, pp. 1937-1954, June 2011. (Journal Article) [5] Gent, P. R. and co-authors. "The Community Climate System Model version 4," J. Climate, 24, 19, pp. 4973-4991, October 2011. (Journal Article) [6] Anderson, J. L. and co-authors. "The Data Assimilation Research Testbed: A Community Facility," Bull. Amer. Meteor. Soc., 90, 9, pp. 1283-1296, September 2009. (Journal Article)


CERA: The ECMWF Coupled Data Assimilation System

Eric de Boisséson1, Patrick Laloyaux1, Magdalena Balmaseda1, Kristian Mogensen1, Peter Janssen1, Dick Dee1,

1European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading, RG2 9AX, United Kingdom.

[email protected]

A coupled data assimilation system for reanalysis called CERA (Coupled ECMWF ReAnalysis) is being developed at ECMWF. The CERA project aims at generating a self-consistent ocean-atmosphere state by assimilating both atmospheric and oceanic observations within a coupled model. CERA uses the ECMWF coupled model where the atmospheric component is based on the IFS software and the oceanic component is based on the NEMO framework. While the computation of the nonlinear trajectories needed in the data assimilation uses the coupled model, the computation of the increments is still performed separately for the atmosphere and ocean components and any covariance between them are ignored. This framework is aimed at being flexible enough to adapt to the initialization of medium range, monthly and seasonal forecasting activities. The building of the CERA system follows several intermediate steps. The first task (called CERA-S) consists in introducing a Sea Surface Temperature (SST) constraint in the coupled model to avoid the model drift while allowing the simulation of coupled processes. The second task, called CERA-A, is based on the existing EMCWF atmospheric data assimilation framework but uses a coupled model in the outer loops. Similarly, the third task, CERA-O, is based on the ocean data assimilation framework used at ECMWF but uses a coupled model in the outer loops. The CERA final product will consist in a merge of the CERA-S, CERA-A and CERA-O systems. This presentation will describe the concept of the different tasks and the first validation and results obtained from each of them.


Dual States Estimation of a Subsurface Flow-Transport Coupled Model Using Ensemble Kalman Filtering

Mohamad El Gharamtia, Ibrahim hoteitb, and Johan Valstarc

a Earth Sciences and Engineering, King Abdullah University of Science and Technology, Saudi Arabia,

[email protected], b Applied Mathematics and Computational Sciences, King Abdullah University of Science and Technology, Saudi Arabia, c Subsurface and Groundwater Systems, Deltares,

The Netherlands. Modeling subsurface contaminant spreading requires coupling a groundwater flow model with a contaminant transport model. This coupling may provide accurate future estimates of the subsurface hydrologic state if assisted with essential flow and contaminant data through data assimilation. Assuming perfect flow, an ensemble Kalman filter (EnKF) can be directly used for data assimilation into the transport model. This is, however, a crude assumption as flow models can be subject to many sources of uncertainties. If the flow is not accurately simulated, contaminant predictions will likely be inaccurate even after successive Kalman updates of the contaminant with the data. The problem is usually better handled when both flow and contaminant states are concurrently estimated using the traditional joint state augmentation approach. In this study, we introduce a dual estimation strategy for data assimilation into this one-way coupled system by treating the flow and the contaminant models separately while intertwining a pair of distinct EnKFs, one on each model. This EnKF-based dual states estimation suggests a number of novel features: (1) it allows for simultaneous estimation of both flow and contaminant states in parallel, (2) it provides a time consistent sequential updating scheme between the two models (first flow, then transport), (3) it simplifies the implementation of the filtering system, and (4) yields more stable and accurate solutions than the standard joint approach. Synthetic numerical experiments are carried based on various time stepping and observation strategies to evaluate the dual approach and compare its performance with the joint state augmentation approach. Experimental results show that under uncertain modeling conditions, the dual strategy could reduce the estimation error of the coupled states, on the average, 15% more than the joint approach. Furthermore, computationally the dual estimation is proven to be very effective, recovering accurate estimates at a reasonable cost.


Development of strongly coupled atmosphere-ocean data assimilation scheme

at the U.S. Naval Research Laboratory Sergey Frolova, Craig Bishopb, Teddy Holtb, James Cummingsb

a UCAR visiting scientists with the Naval Research Laboratory, USA [email protected], b Naval Research Laboratory, USA.

To better utilize sparse observational data and to improve prediction of strongly coupled ocean-atmosphere phenomena (such as hurricanes, deep-water formation, and wind-wave generation), we are developing a coupled data assimilation system for short-range oceanic, atmospheric, and wave forecasts. We are developing our system based on a combination of existing operational models: the COAMPS® atmospheric model [1], NCOM [2] ocean models, and the WAVEWATCH wave model [3]. To characterize the error distribution terms in the coupled model, we use a meta-ensemble drawn that utilizes: an ET flow-dependent ensemble [4,5,6] a method for generating smooth random fields with smoothing scales consistent with those used by existing 3DVAR assimilation schemes, and, possibly, static ensemble drawn from historic model runs. To mitigate for possible spurious correlations in the error covariance, we use both adaptive and static localization. Given this localized ensemble error covariance, we solve for the minimum variance estimator of the coupled state using a 3DVAR solver [7]. We test the developed system using a case of a deep-water formation in the Western Mediterranean associated with Mistral—periods of strong, cold winds over the Gulf of Lions. In this poster, we will present our initial experiments with a coupled system, including studies of the cross-fluid ensemble error covariance functions and convergence properties of a 3DVAR solver.

In a long-run our scientific goals include quantifying the impact of strongly coupled data assimilation, as compared to the COAMPS weakly coupled data assimilation system. We expect that localization of error covariances and the conditioning of the 3DVAR problem to present the main implementation difficulty for this project. References

[1] Hodur, 1997: The Naval Research Laboratory’s coupled ocean/atmosphere mesoscale prediction system (COAMPS). Monthly Weather Review, 125, 1414–1430.

[2] Rhodes, R. C. and Coauthors, 2002: Navy Real-time Global Modeling Systems. Oceanography, 15, 29–43.

[3] Tolman, H. L., 1991: A third-generation model for wind waves on slowly varying, unsteady and inhomogeneous depths and currents. Journal of Physical Oceanography, 21, 782–797.

[4] Bishop, C. H., T. R. Holt, J. Nachamkin, S. Chen, J. G. McLay, J. D. Doyle, and W. T. Thompson, 2009: Regional Ensemble Forecasts Using the Ensemble Transform Technique. Monthly Weather Review, 137, 288–298.

[5] Holt, T. R., J. A. Cummings, C. H. Bishop, J. D. Doyle, X. Hong, S. Chen, and Y. Jin, 2011: Development and testing of a coupled ocean–atmosphere mesoscale ensemble prediction system. Ocean Dynamics, 61, 1937–1954.

[6] McLay, J. G., C. H. Bishop, and C. A. Reynolds, 2008: Evaluation of the Ensemble Transform Analysis Perturbation Scheme at NRL. Monthly Weather Review, 136, 1093–1108.

[7] Daley, R., and E. Barker, 2001: NAVDAS: Formulation and diagnostics. Monthly Weather Review, 129, 869–883.

mailto:[email protected]�


Development of an Ensemble-based Data Assimilation System with a Coupled Atmosphere–Ocean GCM

Nobumasa Komori a, Takeshi Enomoto a,b, Takemasa Miyoshi a,c, Akira Yamazaki a, and Bunmei Taguchi a

a Earth Simulator Center, Japan Agency for Marine-Earth Science and Technology, Japan, [email protected],

b Disaster Prevention Research Institute, Kyoto University, Japan, c RIKEN Advanced Institute for Computational Science, Japan.

To enhance the capability of the local ensemble transform Kalman filter (LETKF) with the Atmospheric general circulation model (GCM) for the Earth Simulator (AFES) [1], a new system has been developed by replacing AFES with the Coupled atmosphere–ocean GCM for the Earth Simulator (CFES). An initial test of the prototype of the CFES–LETKF system has been completed successfully, assimilating atmospheric observational data (NCEP PREPBUFR archived at UCAR) every 6 hours to update the atmospheric variables, whereas the oceanic variables are kept unchanged throughout the assimilation procedure. An experimental retrospective analysis–forecast cycle with the coupled system (CLERA-A) starts on August 1, 2008, and the atmospheric initial conditions (63 members) are taken from the second generation of AFES–LETKF experimental ensemble reanalysis (ALERA2) [2, 3]. The ALERA2 analyses are also used as forcing of stand-alone 63-member ensemble simulations with the Ocean GCM for the Earth Simulator (EnOFES), from which the oceanic initial conditions are taken. In both CLERA-A and EnOFES, the ensemble spread of sea surface temperature (SST) is much larger in the summer hemisphere than in the winter hemisphere in spite that atmospheric disturbances and their spread are much larger in the winter hemisphere. This is likely caused by the difference in the oceanic mixed-layer depth between the summer and winter hemispheres: the mixed layer thins in summer, and SST becomes sensitive to the atmospheric disturbances. The ensemble spread of SST induced by atmospheric disturbances is larger in CLERA-A than in EnOFES, suggesting positive feedback from the ocean to the atmosphere. On the other hand, the spread caused by ocean dynamics such as tropical instability waves is slightly smaller in CLERA-A than in EnOFES. This could be a result of the damping process through the momentum flux [4]. Although the ensemble spread of SST does not grow sufficiently to affect the ensemble spread of air temperature, the ensemble spread of specific humidity in the lower troposphere is larger in CLERA-A than in ALERA2. Thus replacement of AFES with CFES successfully contributes to mitigate an underestimation of the ensemble spread near the surface resulting from the common boundary conditions among all ensemble members and the lack of atmosphere–ocean interaction. References [1] T. Miyoshi and S. Yamane. “Local ensemble transform Kalman filtering with an AGCM at a T159/L48 resolution,” Mon. Wea. Rev., vol. 135, no. 11, pp. 3841–3861, November 2007. [2] T. Miyoshi, S. Yamane, and T. Enomoto. “The AFES-LETKF experimental ensemble reanalysis: ALERA,” SOLA, vol. 3, pp. 45–48, April 2007. [3] T. Enomoto, T. Miyoshi, Q. Moteki, J. Inoue, M. Hattori, A. Kuwano-Yoshida, N. Komori, and S. Yamane. “Observing-system research and ensemble data assimilation at JAMSTEC,” In Data Assimilation for Atmospheric, Oceanic and Hydrological Applications (Vol. II), S. K. Park and L. Xu (Eds.), Springer, in press. [4] H. Seo, M. Jochum, R. Murtugudde, A. J. Miller, and J. O. Roads. “Feedback of tropical instability-wave-induced atmospheric variability onto the ocean,” J. Climate, vol. 20, no. 23, pp. 5842–5855, December 2007.


Assimilation of Surface Observations Using an Atmosphere-ocean Coupled EnKF

Hiroshi KOYAMA a, Masayoshi ISHII b, Hiroaki TATEBE a, Teruyuki NISHIMURA a,

and Masahide KIMOTO c

a Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan, [email protected],

b Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan, c Atmosphere and Ocean Research Institute, University of Tokyo, Kashiwa, Japan.

An ensemble data assimilation system has newly been developed, aiming at improving the skill in seasonal-to-decadal climate predictions and at investigating the feasibility of dynamical reconstruction of a global centennial climate data base. The system is based on an Ensemble Kalman Filter (EnKF) applied to a coupled atmosphere-ocean model, MIROC. The system is uniquely configured, with which oceanic and atmospheric observations are assimilated to reproduce long-term climatic variations by taking account of the covariance between atmospheric and oceanic variables. In preliminary experiments with the system, the reproducibility of global atmospheric and oceanic circulations was investigated using only surface pressure (Ps) and sea-surface temperature (SST) observations for the period from 2004 to 2010; these observations are ISPD ver. 2 (International Surface Pressure Databank version 2; NOAA) and COBE-SST (JMA), respectively. The result indicated that the tropospheric circulation and ocean temperature variation associated with El Nino and Southern Oscillation were reproduced as observed. In particular, assimilating the Ps observation was considerably effective in reproducing atmospheric circulations in middle-to-high latitudes because of geostrophic adjustment process seen dominantly there. In low latitudes, Ps and SST data assimilation succeeded in reproducing short-term atmospheric changes and in reducing errors in the mean state of the troposphere. Moreover, it is found that these low-frequency errors were reduced significantly by introducing covariance between atmospheric and oceanic variables. Further progress of this research will be reported in the meeting.

mailto:[email protected]�

Improved Oceanic Component within the NCEP GFS

Xu Li, IMSG at STAR/NESDIS/NOAA

John Derber, EMC/NCEP/NOAA

Abstract

In the current Numerical Weather Prediction (NWP) systems, the oceanic component is represented by a single thermal variable, the Sea Surface Temperature (SST).

In the atmospheric prediction and analysis, the time dependent SST is a combination of an independent analysis at the initial time, the SST climatology with seasonal variability, and possibly a decay factor of the initial SST anomaly.

The oceanic component has been improved within the NCEP GFS.

In the new scheme, the oceanic component is extended from SST to NSST (Near-Surface Sea Temperature), which is the T-Profile due to diurnal warming layer and sub-layer cooling layer. The new oceanic analysis variable, the foundation temperature, is analyzed together with the atmospheric variables. The satellite radiance of AVHRR, AMSRE and in situ sea temperature observations are newly introduced into GSI (Gridded Statistical Interpolation), the atmospheric assimilation system in NCEP GFS.

NSST model, including a diurnal warming model, which is based on COARE V3.0 (Fairall et al, 1996) but with more physical processes, and a sub-layer cooling model of COARE V3.0, and the radiative transfer model together make it possible to relate all the observations, including wave length dependent satellite radiance and depth dependent in situ sea temperature, to foundation temperature, and therefore to assimilate all the observation directly.

The cycling run has been done for one summer and one winter season. The verification against the own analysis and observation has shown the improvement in the analysis and prediction of SST. The impact on atmospheric prediction is neutral in northern and southern hemisphere, and positive in tropics.

The verification of the momentum flux and heat fluxes at the air-sea interface, the analysis of the function of the SST climatology suggest an oceanic prediction model should reduce the error growth of SST.

A scheme has been proposed to combine the NSST and current NCEP CFS (Climate Forecasting System) towards a coupled data assimilation and prediction system for weather forecasting and climate prediction.


Upper Atmospheric Data Assimilation with an Ensemble Kalman Filter

Tomoko Matsuo ab, I-Te Lee cd, and Jeffrey L. Anderson e

a Cooperative Institute for Research in Environmental Sciences, University of Colorado, USA,

[email protected], b Space Weather Prediction Center, NOAA, USA, c Ionospheric Radio Science Lab, National Central University, Taiwan, d High Altitude Observatory, NCAR, USA, e Institute

for Mathematics Applied to Geosciences, NCAR, USA. Recent availability of global observations of ionospheric parameters, especially from GPS receivers on low Earth orbiting platforms, has motivated a number of attempts at assimilating ionospheric data. However, assimilation of sparse, irregularly distributed thermosphere observations to global models remains a daunting task. In this presentation we demonstrate the utility of ensemble Kalman filtering (EnKF) techniques to effectively assimilate a realistic set of space- and ground-based observations of the thermosphere and ionosphere into a general circulation model. An EnKF assimilation procedure has been constructed using the Data Assimilation Research Testbed (DART) and the Thermoshere-Ionosphere-Electrodynamics General Circulation Model (TIEGCM), two sets of community software offered by NCAR. An important attribute of this procedure is that the thermosphere-ionosphere coupling is self-consistently treated both in a forecast model as well as in assimilation schemes [1]. It effectively facilitates solving the inverse problem of inferring unobserved variables from observed ones, for instance, thermospheric states from better observed ionospheric states. We demonstrate this point using specific observations including (i) neutral mass densities obtained from the accelerometer experiment on board the CHAMP satellite [2], and (ii) electron density profiles obtained from the COSMIC/FORMOSAT-3 mission [3]. Furthermore, we discuss some of the issues specific to upper atmospheric EnKF applications and the roles of auxiliary filtering algorithms, such as adaptive covariance inflation and localization of covariance, to cope with these issues.

References [1] Matsuo, T., and E. A. Araujo-Pradere, "Role of thermosphere-ionosphere coupling in a global ionosphere specification", Radio Science, 46, RS0D23, doi:10.1029/2010RS004576, 2011. [2] Matsuo, T., I.-T. Lee, and J. L. Anderson, "Thermospheric mass density specification using an ensemble Kalman filter", J. Geophys. Res. Space Physics, 118, 1339–1350 doi:10.1002/jara.50162, 2013. [3] Lee, I. T., T, Matsuo, A. D. Richmond, J. Y. Liu, W. Wang, C. H. Lin, J. L. Anderson, and M. Q. Chen "Assimilation of FORMOSAT-3/COSMIC electron density profiles into thermosphere/Ionosphere coupling model by using ensemble Kalman filter", J. Geophys. Res. Space Physics, 117, A10318, doi:10.1029/2012JA017700, 2012.

The Met Office Weakly-Coupled Atmosphere/Land/Ocean/Sea-Ice Data Assimilation System

Isabelle Mirouze, Daniel Lea, Adrian Hines and Matthew Martin

Met Office, FitzRoy Road, Exeter EX1 3PB, UK.

[email protected]

The Met Office has developed a weakly-coupled data assimilation (DA) system around the global coupled model HADGEM3 (Hadley Centre Global Environment Model, version 3). This model combines the atmospheric model UM (Unified Model) at 60 km horizontal resolution on 85 vertical levels, with the ocean model NEMO (Nucleus for European Modeling of the Ocean) at 25 km (at the equator) horizontal resolution on 75 vertical levels, and with the sea-ice model CICE at the same resolution as NEMO. The atmosphere and the ocean/sea-ice are coupled every 1-hour using the OASIS coupler. The initial condition of the coupled model is corrected using two separate 6-hour window data assimilation systems: a 4D-Var for the atmosphere with associated soil moisture content nudging and snow analysis schemes on the one hand, and a 3D-VarFGAT for the ocean and sea-ice on the other hand. The background information in the DA systems comes from a previous 6-hour forecast of the coupled model.

To assess the benefit of the weakly-coupled DA, one-month experiments have been carried out, including 1) a full atmosphere/land/ocean/sea-ice coupled DA run, 2) an atmosphere-only run forced by OSTIA SSTs and sea-ice with atmosphere and land DA, and 3) an ocean-only run forced by atmospheric fields from run 2 with ocean and sea-ice DA. Moreover, 15-day forecast runs, started once per day, have been produced from initial conditions generated by either run 1 or a combination of runs 2 and 3. The different results have been compared to each other and, whenever possible, to other references such as the Met Office atmosphere and ocean operational analyses or the OSTIA data. Evidence of imbalances and initialisation shocks has also been searched for.

The weakly-coupled data assimilation system will be described, and results from these assessments will be presented in order to provide some insight into whether the weakly-coupled DA system offers improvements over starting from separate atmosphere/ocean initial conditions.


mailto:[email protected]

6th

WMO Symposium on Data Assimilation -- 2013

Exploring Coupled 4D-Var Data Assimilation Using an Idealized

Atmosphere-Ocean Model

Polly Smith, Alison Fowler, and Amos Lawless

Department of Mathematics and Statistics, University of Reading, UK, Email: [email protected]

The successful application of data assimilation techniques to operational numerical weather

prediction and ocean forecasting systems has led to an increased interest in their use for the

initialization of coupled atmosphere-ocean models in seasonal to decadal timescale prediction.

Coupled data assimilation presents a significant challenge but offers a long list of potential

benefits including the improved use of near-surface observations, reduction of forecast

initialization shocks, and generation of a consistent system state for the initialization of coupled

forecasts across all timescales.

This presentation will describe work from projects funded by the European Space Agency (ESA)

and UK Natural Environment Research Council (NERC). The ESA funded component is part of

the Data Assimilation Projects - Coupled Model Data Assimilation initiative whose goal is to

advance data assimilation techniques in fully coupled atmosphere-ocean models (see

http://www.esa-da.org/). Our research aims to investigate some of the fundamental questions in

the design of coupled data assimilation systems within the context of an idealized one-

dimensional system. It is being conducted in parallel to the development of the new Coupled

European Centre for Medium-Range Weather Forecasts (ECMWF) Re-analysis system (CERA),

a prototype weakly coupled data assimilation system.

Here, we will describe the development of our simplified single-column coupled atmosphere-

ocean 4D-Var assimilation system which is based on the ECMWF Integrated Forecast System

(IFS) atmosphere model and a K-Profile Parameterization (KKP) mixed layer ocean model

developed by the National Centre for Atmospheric Science (NCAS) climate group at the

University of Reading. The system employs a strong constraint incremental 4D-Var scheme and

is designed to enable the effective exploration of various approaches to performing coupled

model data assimilation whilst avoiding many of the issues associated with more complex

models.

The work within this simple framework will facilitate a greater theoretical understanding of the

coupled atmosphere-ocean data assimilation problem and thus help inform the design and

implementation of the different coupling strategies proposed for the CERA system.

We will present preliminary results from identical twin experiments devised to investigate and

compare the behavior and sensitivities of different coupled data assimilation methodologies.

http://www.esa-da.org/

6th


Exploring Strategies in Coupled Atmosphere-Ocean Data Assimilation with a

Low-Order Climate Model and CMIP5 Data

Robert Tardifa, Chris Snyder

b, and Gregory J. Hakim

c

a Department of Atmospheric Sciences, University of Washington, United States,

[email protected], bNational Center for Atmospheric Research, United States,

cDepartment of

Atmospheric Sciences, University of Washington, United States.

Motivated by the need for initial conditions for decadal climate predictions, we explore the utility

of assimilating time-averaged observations and choice of atmospheric variables in a coupled

atmosphere-ocean framework. Of particular interest is the initialization of the Atlantic Meridional

Overturning Circulation (AMOC), a major driver of the low-frequency variability of North

Atlantic climate and its predictability. AMOC initialization under contrasting scenarios of

observation availability is investigated, including the extreme case of assimilating atmospheric

observations only. Experiments are carried out with an ensemble Kalman filter and the idealized

low-order coupled climate model described in [1], representing interactions between large-scale

atmospheric circulation and an ocean basin driven by the AMOC. Results from this simplified

system are complemented by "no cycling" data assimilation experiments based on comprehensive

coupled model output from CMIP5 simulations. Results from the low-order model indicate that

AMOC analyses of comparable accuracy are obtained whether time-averaging of observations is

applied or not when the ocean is well-observed. Time-averaged data assimilation is however

more effective at initializing the low-frequency component of the AMOC when the ocean is only

partially observed or not observed at all. Results from assimilation experiments using CMIP5 data

further support these findings.

References

[1] Roebber, P. J. “Climate variability in a low-order coupled atmosphere-ocean model.”

Tellus, vol. 47A, pp. 473-494, 1995. (Journal Article)


Ocean-Atmosphere Interactions in CMIP5 Coupled Model Decadal Data

Robin Wedda and Keith Hainesb

a Centre for Australian Weather and Climate Research, Australian Bureau of Meteorology, Australia, [email protected], b Reading e-Science Centre, University of Reading, United Kingdom.

The dynamics of ocean-atmosphere coupled covariances require investigation if they are to be used in strongly coupled data assimilation systems. The design of coupled DA systems will be most likely be based on the covariances of the individual model used. Investigation of the similarities and differences of ocean-atmosphere dynamics in different models and reanalyses can help inform the development of these systems. The fifth round of the Coupled Model Intercomparison Project provides a large data resource with which to investigate ocean-atmosphere covariances in various coupled models. CMIP5 includes data from 27 institutes and 25 different models (plus variants). The core decadal runs in CMIP5 are 10- and 30-year hind-casts initialised with observed ocean states and forced with observed fields. Results of an investigation of ocean-atmosphere coupled cross-covariances in a selection of contributing CMIP5 models will be shown, including SST-precipitation, SST-sea level pressure and SST-surface wind covariances. Ensembles of low- and high-frequency events are investigated for state-dependant coupled interactions.

6th


A Coupled Ensemble Data Assimilation System for Seasonal Prediction in

Australia

Yonghong Yin, Oscar Alves, Debra Hudson, Li Shi, and Harry Hendon

Centre for Australian Weather and Climate Research, and Bureau of Meteorology, Australia,

[email protected]

A new coupled ensemble-based ocean analysis system called the POAMA

(http://poama.bom.gov.au) Ensemble Coupled Data Assimilation System (PECDAS) has been

developed. PECDAS is an approximate form of ensemble Kalman filter system, its

approximations being necessary to reduce its computational cost. It is based on the multivariate

ensemble optimum interpolation of Oke et al (2005) [1]

, but uses covariances estimated from a

time evolving model ensemble. The first version of the system is weakly coupled, only ocean

observations are assimilated into the coupled model and the atmospheric component is nudged

towards pre-existing atmospheric analyses.

A reanalysis from 1980 to present has been completed with this system. Both in situ temperature

and salinity observations are assimilated, and ocean current corrections are generated based on

the ensemble covariances. The performance of PECDAS is evaluated through a series of

comparisons with assimilated and independent observational data. The comparison of PECDAS

reanalysis to a non-coupled reanalysis [2]

and other state-of-the-art international reanalyses are

presented, with a particular focus on the representations of the main modes of climate variability.

The impact of the coupled assimilation on seasonal forecasts will also be presented.

References

[1]Oke, P. R., A. Schiller, D. A. Griffin, and G. B. Brassington. Ensemble data assimilation

for an eddy-resolving ocean model of the Australian region, Quart. J. Roy. Meteor. Soc.,

131, 3301-3311, 2005.

[2] Yin, Y., O. Alves, and P. R. Oke. An ensemble ocean data assimilation system for

seasonal prediction, Mon. Wea. Rev., 139, 786-808, 2011.

6th


The Energy Balance in Climate Estimation by Combining Data with a Biased Coupled

Model

Shaoqing Zhang

a, You-Soon Chang

b, Xiaosong Yang

c and Anthony Rosati

a

a GFDL/NOAA, Princeton USA, [email protected],

b Department of Earth Science Education,

Kongju National University, South Korea, c UCAR/GFDL, Princeton, USA.

When a biased coupled model is combined with the climate observing system,

understanding its behavior in climate estimation is of critical importance for producing

accurate climate analysis and prediction initialization. Due to incomplete representation of

observations to the real climate system, especially for the ocean where most of the

measurements are available only for the upper part, the knowledge of model bias and its

influence on climate estimation and prediction still remain limited. With two CGCMs which

are biased with respect to each other, a ?biased? twin experiment, in which atmospheric

and oceanic ?observations? that sample one model based on the modern climate observing

system are assimilated into the other, is designed to study the impact of coherent

atmosphere-ocean data constraints on climate estimation. Given the assimilation model bias

of warmer atmosphere and deep ocean as well as colder upper ocean, while the

atmospheric-only (oceanic-only) data constraint produces an over-cooling (over-warming)

ocean, the constraints with both atmospheric and oceanic data create a balanced and

coherent ocean estimate as the observational model. Moreover, the consistent atmosphere-

ocean constraint produces the most accurate estimate for North Atlantic Deep Water

(NADW), while the NADW appears too strong (weak) if the system is only constrained by

atmospheric (oceanic) data. These twin experiment results address the importance of

consistent data constraints of multiple components in climate estimation and initialization

of a coupled model for decadal predictions


Coupled assimilation of both atmospheric and oceanic observations for ENSO prediction using an intermediate coupled model

Fei Zhenga, and Jiang Zhua

a Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

The initial surface and subsurface temperature fields are important for El Niño Southern Oscillation (ENSO) prediction due to the ocean’s long memory. As a result, most ENSO predictions are initialized by ocean data assimilation. However another key for the development of an ENSO event is the wind anomalies in the western Pacific associated with air-sea coupled tropical convection systems. Therefore the assimilation of both atmospheric and oceanic observations should be taken into account for ENSO prediction. However one difficulty is that the atmosphere has relative short memory on its initial conditions. This can be even worse if one uses the intermediate coupled models in which the atmospheric components are statistical and slavery to ocean components. In this case, the atmospheric models do not have any memory on its initial conditions and the adjustment of the atmospheric state made by assimilation cannot impact on model forecast. In this study we use the coupled covariance to assimilate both atmospheric and oceanic observations based on ensemble Kalman filter with an intermediate coupled model. The coupled covariance enables the assimilation of atmospheric observations to adjust the ocean states and can overcome the above mentioned difficulty technically. By a series of experiments using both simulated and realistic observations, we found that the assimilation of atmospheric observations (i.e., wind in this study) can help to improve the surface and subsurface currents in ocean because the correlation between the wind and ocean currents is stronger than that between ocean temperature and ocean current in the equatorial Pacific [1]. Finally we demonstrate that initialized by the coupled assimilation the prediction lead time with useful skills can reach up to two years. References [1] Zheng, F., and J. Zhu. "Coupled assimilation for an intermediated coupled ENSO prediction model," Ocean Dyn., 60, 1061−1073, 2010. (Journal Article)


Forecast error covariance in a coupled Land-Atmosphere Ensemble Data Assimilation with NASA-Unified WRF model

Milija Zupanskia, Sara Q. Zhangb, Arthur Y. Houb, and Christa Peters-Lidardb

a Cooperative Institute for Research in the Atmosphere, Colorado State University, USA,

([email protected]), bNASA Goddard Space Flight Center, USA

Complexity of coupled models and high-dimensional nature of the augmented state vector require an advanced data assimilation system. One of the most important components of such data assimilation system is the augmented forecast error covariance that has a capability of representing complex and sometimes unknown correlations and cross-correlations between model variables. Such complex systems may be best represented by an ensemble or hybrid variational-ensemble error covariance. In this work we present preliminary development of an ensemble-based coupled land-atmosphere data assimilation system to that employs NASA-Unified Weather Research and Forecasting (NU-WRF) model and incorporates assimilation of all-sky satellite radiances. At this stage of development we focus on the forecast error covariance structure and examine the error correlations between model variables for land and atmosphere. We will also discuss other issues of land-atmosphere coupled data assimilation including the choice of control variables, high dimensions, computational limitations, and dynamical balances.


Assimilation for Sea Surface Skin Temperature in GEOS-DAS

Santha Akellaa,b, Ricardo Todlinga, Max Suareza,and Michele Rieneckera

aGlobal Modeling and Assimilation Office, Goddard Space Flight Center,

Greenbelt, MD 20771 USA , [email protected], bScience Systems & Applications Inc, 10210 Greenbelt Rd, Lanham, MD 20706 USA

In the context of coupled atmosphere-ocean data assimilation, surface skin temperature (Ts) is an important variable for the calculation of air-sea fluxes and near-surface processes. A bulk retrieved-SST product is an inadequate representation of Ts, since it does not account for the near-surface ocean stratification, which is related to diurnal variations (solar insolation), wind stress and wave (surface currents) induced mixing.

We model Ts: (i) based on a near-surface ocean temperature- which has minimal contribution from diurnal effects, (ii) a diurnal warming model which takes into account the penetration of shortwave heat flux at the surface and turbulent diffusion effects [1], (iii) a cool-skin layer that is dominated by molecular diffusion of net longwave, latent and sensible heat fluxes within a few microns of the air-sea interface [2,3]. The assimilation for Ts is conducted using the Gridpoint Statistical Interpolation (co-developed with NCEP). The analysis includes a state vector representative of the entire atmosphere and Ts. Results and discussion will focus on assimilation of brightness temperatures (from AVHRR Channels 3B, 4 & 5). Emphasis will be on the statistics of observation-minus-background residuals and impact of assimilation on air-sea fluxes. References [1] Zeng and Beljaars, ''A prognostic scheme for sea surface skin temperature for modeling and data assimilation,'' Geophysical Research Letters, vol. 32, L14605, 2005. [2] Saunders PM, ''The temperature at the ocean-air interface,'' Journal of Atmospheric Sciences, vol. 24, 1967. [3] Fairall et al., ''Cool-skin and warm-layer effects on sea surface temperature,'' Journal of Geophysical Research, vol. 101, 1996.


Development of a meso 4D-Var data assimilation system using a coupled atmosphere-ocean model for typhoon intensity prediction

Kosuke Itoa,b, Masaru Kuniib, Takuya Kawabatab, Tohru Kurodaa,b, Yuki Hondac, and Kazuo Saitob

a Japan Agency for Marine-Earth Science and Technology, Japan, [email protected],

b Meteorological Research Institute, Japan, c Japan Meteorological Agency.

Since typhoons are often highly destructive, a better prediction of their intensity is important for disaster prevention. One of the promising approaches is to use a sophisticated high-resolution data assimilation system with a treatment of inner core dynamics and oceanic features. The Japan Meteorological Agency has developed the JMA-Nonhydrostatic model based Variational data Assimilation system (hereafter, JNoVA; [1]), which has been used operationally to construct a mesoscale objective analysis dataset and for regional forecasts since April 2009. JNoVA employs an adjoint-based 4D-Var technique using an incremental approach [2]. The analysis field is obtained by performing a high-resolution model run with the horizontal grid spacing of 5km, which is sufficient to represent the typhoon inner core structure, whereas the grid spacing is 15 km in the adjoint calculations. The purpose of this presentation is to summarize the recent progress toward improving the typhoon intensity prediction, including the coupling to the ocean mixed layer model and the use of ensemble-based background-error covariances. To represent the impacts of sea surface temperature (SST) decrease along the typhoon passage, a simple one-dimensional ocean mixed layer model [3] is coupled to the high-resolution model. The initial condition of SST at the first assimilation cycle is set to the same as in the original system. Ocean temperature anomaly relative to the SST and salinity are taken from the World Ocean Atlas 2009. The SST is restored toward the values originally employed in JNoVA in a time-scale of one day because the ocean model is too simple to represent the long-term tendency. The assimilation experiment is conducted for Typhoon Talas (2011). With the coupled configuration, the SST decreases particularly in the right of the typhoon pathway with the maximum value of 1-2 K, which is consistent with the observation. The forecast experiment from 0000UTC Sep. 2 exhibits that the typhoon intensity is overestimated in the original system (minimum sea level pressure = 932 hPa) at FT=36 h, while the coupled configuration brings about the closer estimate (971 hPa) at FT=36 h to the JMA best track (970 hPa). Currently, we are testing the impact of introducing the background-error covariances that are obtained from a local ensemble transform Kalman filter with 51 ensemble members. As a result, an observational data near an inner core yields the analysis increment according to nearly a gradient wind balance, contrary to a geostrophic balance which was obtained from the original system. This difference becomes more obvious when the observation is added in the beginning of an assimilation window. References [1] Honda, Y., and K. Sawada, 2009: Upgrade of the operational mesoscale 4D-Var at the Japan Meteorological Agency. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 39, 01.11–01.12. [2] Courtier, P., Thépaut, J.-N. and Hollingsworth, A. (1994), A strategy for operational implementation of 4D-Var, using an incremental approach. Q.J.R. Meteorol. Soc., 120: 1367–1387. doi: 10.1002/qj.49712051912 [3] Price, J.-F., Robert A. Weller and Robert Pinkel. "Diurnal cycling: Observations and models of the upper ocean response to diurnal heating, cooling, and wind mixing," Journal of Geophysical Research, vol. 91, no. C7, pp. 8411-8427, July1986.

6th


Implementation of GLDAS in the NCEP Operational

Global Climate and Weather Forecast Systems

Jesse Mengab

, Helin Weiab

, Rongqian Yangab

, and Michael Eka

a Environmental Modeling Center, NOAA NCEP, USA, [email protected],

b I.M. Systems Group, Inc., USA.

This presentation introduces the preliminary results of the Global Land Data Assimilation System

(GLDAS) implementation in the NCEP operational Climate Forecast System (CFS) for seasonal

climate prediction [1] and the Global Forecast System (GFS) for mid-range weather prediction.

Accurate initialization of land surface states is critical in global climate and weather prediction

systems because of their regulation of water and energy fluxes between the land surface and

atmosphere over a variety of spatial and temporal scales. Since measurements of many land

surface states are generally not available on global scales, traditional coupled land-atmosphere

prediction systems rely on their land surface models to predict the land surface states and fluxes.

It is widely acknowledged that bias in the land surface forcing predicted by the atmospheric

model, particularly precipitation, may lead to nontrivial bias in predicted land surface states and

fluxes.

In order to provide enhanced land surface states for operational prediction systems, the NCEP

GLDAS is implemented using the NASA Land Information System (LIS). Global observed

precipitation is used as direct forcing to drive GLDAS/LIS. Global observed snow cover

and depth are used to constrain the predicted snow field. The NCEP GLDAS/LIS has been

used in the CFS Reanalysis and Reforecast project (CFSRR) to provide land surface initial

conditions to the reforecast experiments. The coupled CFSRR prediction and assimilation system

was transitioned into NCEP operations for seasonal climate prediction (CFSv2). Meanwhile,

GLDAS/LIS is also being tested as part of the development of the coupled NCEP GFS Data

Assimilation System (GDAS), anticipating improving the GFS mid-range weather prediction.

References

[1] Meng, J., R. Yang, H. Wei, M. Ek, G. Gayno, P. Xie, K. Mitchell. “The Land Surface

Analysis in the NCEP Climate Forecast System Reanalysis,” J. Hydrometeor, vol. 13, pp.

1621–1630, October 2012.


An Ensemble Kalman Smoother for the Coupled Greenhouse Gas and Flux estimation problem: Design of the Environment Canada Carbon Assimilation

System

Saroja Polavarapua, Michael Neishb, Shuzhan Renb, Wei Lub, Douglas Chana and Dylan Jonesb

a Climate Research Division, Environment Canada, Canada, [email protected] b Department of Physics, University of Toronto, Canada.

Climate change is one of the pressing issues of our time. In an effort to understand and quantify the biospheric and anthropogenic contributions to the atmospheric rise of CO2, climate scientists are building comprehensive carbon monitoring systems of the global carbon cycle. The core of such systems is an atmospheric transport model constrained by atmospheric measurements. Eventually, this will be coupled to terrestrial biospheric models and ocean carbon cycle models each constrained by observations. The primary goal of the system is to estimate the atmospheric exchange of carbon with the Earth’s surface. There are many challenges particular to the carbon flux estimation problem. Unlike the weather forecasting case, the background state derives not from a geophysical model, but rather, from a combination of industry and government reported anthropogenic emissions along with ecosystem model outputs of biospheric fluxes and an ocean model of carbon fluxes. The propagation of the background state does not follow a dynamical model yet has clear biases, and temporal (diurnal to annual) scales of variability. This renders covariance estimation of the prior flux a daunting prospect. Yet, the prior flux is critical to the estimation problem, because the surface measurement network is sparse and satellite measurements are very new—the first dedicated greenhouse gas instrument (GOSAT) was launched in 2009 and OCO-2 will be launched in 2014. The promise of the new satellite measurements is improved spatial scales of retrieved fluxes. The promise of advanced assimilation methods such as ensemble Kalman smoothers is improved uncertainty estimates since conventional methods of flux estimation typically assume that transport is perfect (i.e. no initial CO2 concentration, model or analysed wind errors). In this work, the formulation of the data assimilation problem through the design of the Environment Canada Carbon Assimilation System is presented. The core of the system is the operational weather and environmental prediction model. Key model adaptations needed for greenhouse gas simulation included mass conservation, improvement of boundary layer mixing, and the addition of convective transport of tracers. The assimilation system is an extension of the operational Ensemble Kalman filter (to be adapted for constituents and for the smoothing problem). In addition to covariance estimation of prior fluxes, the assimilation challenge is one of multiple time scales. The large diurnal cycle of the terrestrial biosphere in summer provides gradients which, if detected by measurements, can be used to infer upstream fluxes. However, the sparse measurement network dictates update cycles of days to weeks, and the signal of an instantaneous pulse disperses on the order of months. At surface measurement sites, synoptic to inter-hemispheric transport operating on daily to inter-annual time scales determines concentrations. The issues in designing the assimilation system (i.e. choosing the update frequency, prior covariance modeling for additive inflation, ensemble generation and localization, etc.) will be presented along with preliminary results from the system development.

on the problem of initializing climate prediction · 6th wmo symposium on data assimilation -- 2013...

Documents