Presentation is loading. Please wait.

Presentation is loading. Please wait.

Korea Institute of Atmospheric Prediction Systems (KIAPS) ( 재 ) 한국형수치예보모델개발사업단 Progress and plans for the observational data assimilation module development.

Similar presentations


Presentation on theme: "Korea Institute of Atmospheric Prediction Systems (KIAPS) ( 재 ) 한국형수치예보모델개발사업단 Progress and plans for the observational data assimilation module development."— Presentation transcript:

1 Korea Institute of Atmospheric Prediction Systems (KIAPS) ( 재 ) 한국형수치예보모델개발사업단 Progress and plans for the observational data assimilation module development at KIAPS Hyo-Jong Song, Hyoung-Wook Chun, Su Jin Ha, Youngsoon Jo, Byoung-Joo Jung, Jeon-Ho Kang, Ji-Sun Kang, Hataek Kwon, Jihye Kwun, Sangil Kim, Ju-Hye Kim, Sihye Lee, Jooyeon Lim, Jong-Im Park, and Young-Joon Kim KIAPS International Workshop Apr. 17-19, 2013

2 Introduction  Data assimilation systems (DAS)  DA estimates initial conditions for numerical weather prediction models.  The more accurate and faster the performance of such a system is, the better the quality of the resultant prediction is.  Assimilation of observed data into numerical models is crucial for the short and medium-range forecasts.  Observation processing systems (OPS)  DA technique assumes that observations are unbiased and not correlated to each other.  Observations are often contaminated by various gross errors.  Construction of a system for pre-processing and quality control of observational data is necessary.

3 Diagram for KIAPS OPS (KOPS)  KOPS framework design and proto-type development Kops_util: mpi_utils and physical constants rttov/ropp: open-source package as an observation operator amsua/iasi/gps_prep: main process for each obs_types including bias-correction, quality-control, and so on. bkgnd_interpolation: read bkg_field and matchup model grid to obs_space with vert. or horiz. Interpolation lib/include: mod and lib files from utils and bkg_process script/jobs: setup environments and job-scheduler, etc…

4 Progress % KOPS AMSU-A Process  Gross Quality Control Physical reality check Quality control flag  Check on geolocation, beam position, and strange Tbs  Flagging sea-ice index, scattering index, and cloud liquid water  Flagging surface type  Bias Correction  Scan bias calculated as O-B mean  Calculation monthly coefficient and offset for all scan position, latitude, and surface type Scan bias correction Air-mass bias correction  Residual bias calculated by 2 predictors (Thickness of 850-300 and 200-50 hPa)

5 Progress  AMSU-A  Analysis of innovation [Decoded AMSU-A 1d data using ECMWF BUFRDC]][Observed TBs – Simulated TBs]] [Observed TBs – Simulated TBs from 15 channels]

6 Progress  AMSU-A  Gross Quality Control  Bias Correction latitude band Before Bias Correction After Bias Correction Cloud Liquid Water Scattering Index Sea Ice %For latitudes beyond 50 degrees of the equator

7 Progress Animation of the 6-hourly mean (solid back dots) and the standard deviation (grey vertical bars) of the spectral O-B during Nov. 2012 over globe (top), ocean (middle), and land (bottom). Channel IndexDescription 0-120 Temperature sounding 121-170 Window and ozone 171-314Water vapor  IASI

8 Progress  GPS Radio Occulation  Design framework for GPS-RO data processing system  Development of main components for GPS-RO data processing system  Development of IO module for GPS-RO data by using ECMWF bufr decoder  Build up ROPP(Radio Occultation Processing Package) in KOPS framework  Development of observation operator for bending angle by using ROPP A global mean bending angle innovation as a function of impact parameter from 07Dec2012 at 00 UTC Zonal mean GPS-RO bending angle innovation from 07Dec2012 at 00 UTC

9 Progress  Radiosonde  Analysis of SONDE observation preprocessing and QC modules  Development of framework and main preprocessing modules for radiosonde data - IO modules for obstore file format - Variable transform(wind speed/direction to u, v component), vertical interpolation modules [ Temperature and RH bias correction scheme in UM OPS system ] Obs. stations: TEMP/PILOT/WINDPRO at 500 hPa [ u, v component at 500 hPa from radiosonde observation preprocessing ]

10 Plans  Satellite data  Development of satellite radiance (AMSU-A, IASI) data processing system - Thinning scheme: priority number flag for grid box, analysis time and reliable satellite - statistical innovation quality control check - Quality control module utilizing CLW retrieval and sea-ice (scattering) index  Development of GPS-RO processing system – beta version - Gross quality check - Background check - Error correction caused by ice and liquid water in the atmosphere

11 Plans  Synoptic  Development of radiosonde observation processing modules in KOPS system - Internal consistency check modules: V, T, RH check, hydrostatic check - T and RH bias correction modules caused by solar radiation - I/O modules for observation with ODB format  Development of aircraft observation processing systems - Variable transform (pressure to height) - Duplicate check and thinning - Track check, rejection of calm winds and zero latitude and longitude - Background check  Development of surface observation processing systems - Variable transform (dew point to relative humidity) - Correction of observed surface temperature and pressure from station level to model surface

12 Diagram for KIAPS DAS  4D data assimilation systems Background Error Covariance Model Ensembl e Forecast Initial conditions Global Analysis (optimization) Total Sanity Check Local Analysis Flow- dependency Re-centering Sanity check Debug Tangent linear/ Adjoint Model

13 Progress  Local Ensemble Transform Kalman Filter (LETKF; Hunt et al. 2007) Data Assimilation System  We have started implementing LETKF data assimilation system to NCAR CAM-SE (Community Atmospheric Model with Spectral Element Dynamic Core)  As a test bed, LETKF data assimilation has been implemented to an intermediate-complexity general circulation model SPEEDY (Molteni, 2003)  It is very useful tool to test and to confirm an impact of new data assimilation methodologies and observation datasets on resultant analysis, under Observing System Simulation Experiments (OSSEs).  Many recently developed techniques of ensemble Kalman filter data assimilation have been examined (e.g. Miyoshi 2011; Kang et al. 2011, 2012) within LETKF-SPEEDY system.

14 Progress  ERA Interim  Obs: 500 hPa wind fields(u, v)  Background: HOMME SWE  2DVAR analysis 30 m/s  2DVAR

15 Progress  3DVAR  Implementation of conjugate gradient (CG) algorithm for inverse problems Makes the last residue vector orthogonal to the previous residue Requires storage of the last two residue vectors and the last search direction Parallelizes CG using the data structure of elements in the spectral element method  Background error covariance module (e.g. spectral transform on the cubed- sphere)

16 Progress  Representer method  Components and sanity test a b c d e Symmetric Matrix of Representers Vector of Representer coefficients a. Make a single impulse forcing of given k th observation b. Adjoint model run c. Apply background error covariance, including variable-balance, spatial correlation, and error variance. d. Tangent linear model run e. Sampling of TOTAL observations

17 Progress  Development of tangent linear and adjoint models of a shallow water equation

18 Plans  LETKF Data Assimilation System  We plan to test LETKF implemented to NCAR CAM-SE model under Observing System Simulation Experiments including Rawinsonde, surface station data, and satellite data of AMSU-A and IASI in 2013  Since EnKF analysis system is model-independent and NCAR CAM-SE has the same dynamic core as HOMME/KIAPS, the LETKF-NCAR CAM system can be immediately used for the HOMME/KIAPS when released. Observations rawinsonde, surface obs AMSU-A, and IASI LETKF Data Assimilation Forecast model SPEEDY, CAM-SE, or HOMME/KIAPS Observation operator Ensemble analysis Ensemble forecast (background) Ensemble background in the observation space

19 Plans  3DVAR/LETKF Hybrid  Based on the formulation of Lorenc (2003)  Uses ensemble forecast from LETKF  Re-centering the analysis ensemble of LETKF under consideration Spatial correlation model Gaussian, Second-Order Auto- Regression, Gaspari and Cohn Spectral transform 3DVAR Conjugate gradient parallelized using the data structure of the spectral element method Balance equation model Wind decoupling, Mass/wind balance, hydrostatic balance, Relations in dynamic/moisture variables LETKF Ensemble forecast for a background, error variances, flow- dependent error correlations Schur product Operate on a guess Re-distributed on memory Re-centering (if useful)

20 Plans  Representer method Read obs. data Specify initial guess state NL_model Get the innovation Get the Representer matrix & Sanity test Calculate the Rep. coeff. Construct analysis NL_model Single impulse forcing ADJ_model Covariance model TL_model Sampling of observations

21 Summary  Progress  An observation processing system includes observation operators, which interpolate model state into observation space, in order to provide the operator with a first guess.  We have adopted and evaluated public software for satellite observation processing – RTTOV, ROPP, BUFR decoder.  We have also developed correction modules for bias from scanning and air mass difference based on Harris and Kelly (2001).  We have developed a 2-dimensional variational assimilation scheme (2D-Var). which can assimilate radiosonde observations at 500 hPa and a representer method, which is a prototype of a 4D-Var system (Xu et al., 2005).  In the 2D-Var, a parameter transform and a spectral transform have been implemented.  For the representer method, we used tangent linear and adjoint models corresponding to HOMME shallow water equation model.

22 Summary  Plans  We plan to upgrade the preliminary version and construct the KIAPS Observation Processing System (KOPS) with pre-processing and quality control modules for satellite radiance, GPS radio occultation and synoptic data.  AMSU-A and IASI are selected as representative species for microwave and infrared radiance data, respectively.  LETKF could be the first system to be constructed as an operational system with the analysis cycle.  We are expanding 2D-Var into 3D-Var and then combine it with LETKF.  We intend to code the core components of the linear models of hydrostatic equations such as time integration schemes, dynamic forcing calculation method, and vertical advection schemes, and wrap them up as a direct representer system.

23 Thank You


Download ppt "Korea Institute of Atmospheric Prediction Systems (KIAPS) ( 재 ) 한국형수치예보모델개발사업단 Progress and plans for the observational data assimilation module development."

Similar presentations


Ads by Google