Explicit Treatment of Model Error Simultaneous State and Parameter Estimation with an Ensemble Kalman Filter Altuğ Aksoy*, Fuqing Zhang, and John W. Nielsen-Gammon.

Slides:



Advertisements
Similar presentations
What’s quasi-equilibrium all about?
Advertisements

Meteorologisches Institut der Universität München
The Structural Evolution of African Easterly Waves Matthew A. Janiga and Chris Thorncroft DEPARTMENT OF ATMOSPHERIC AND ENVIRONMENTAL SCIENCES University.
Günther Zängl, DWD1 Improvements for idealized simulations with the COSMO model Günther Zängl Deutscher Wetterdienst, Offenbach, Germany.
HEDAS ANALYSIS STATISTICS ( ) by Altug Aksoy (NOAA/AOML/HRD) HEDAS retrospective/real-time analyses have been performed for the years
Effects of model error on ensemble forecast using the EnKF Hiroshi Koyama 1 and Masahiro Watanabe 2 1 : Center for Climate System Research, University.
Sensitivity of High-Resolution Simulations of Hurricane Bob (1991) to Planetary Boundary Layer Parameterizations SCOTT A. BRAUN AND WEI-KUO TAO PRESENTATION.
What controls the climatological PBL depth? Brian Medeiros Alex Hall Bjorn Stevens UCLA Department of Atmospheric & Oceanic Sciences 16th Symposium on.
Assimilating Sounding, Surface and Profiler Observations with a WRF-based EnKF for An MCV Case during BAMEX Zhiyong Meng & Fuqing Zhang Texas A&M University.
ASSIMILATION of RADAR DATA at CONVECTIVE SCALES with the EnKF: PERFECT-MODEL EXPERIMENTS USING WRF / DART Altuğ Aksoy National Center for Atmospheric Research.
Towards Assimilating Clear-Air Radar Observations with an WRF-Based EnKF Yonghui Weng, Fuqing Zhang, Larry Carey Zhiyong Meng and Veronica McNeal Texas.
Click to edit Master title style Click to edit Master text styles Second level Third level Fourth level Fifth level 1 Click to edit Master title style.
Turbopause and Gravity Waves Han-Li Liu HAO National Center for Atmospheric Research.
Advanced data assimilation methods- EKF and EnKF Hong Li and Eugenia Kalnay University of Maryland July 2006.
A comparison of hybrid ensemble transform Kalman filter(ETKF)-3DVAR and ensemble square root filter (EnSRF) analysis schemes Xuguang Wang NOAA/ESRL/PSD,
“1995 Sunrise Fire – Long Island” Using an Ensemble Kalman Filter to Explore Model Performance on Northeast U.S. Fire Weather Days Michael Erickson and.
Initial Experiments on Simulation of Windshear and Significant Convection Events using Aviation Model (AVM) Wai-Kin Wong 1, C.S. Lau 2 and P.W. Chan 1.
Ensemble-variational sea ice data assimilation Anna Shlyaeva, Mark Buehner, Alain Caya, Data Assimilation and Satellite Meteorology Research Jean-Francois.
ASSIMILATION OF GOES-DERIVED CLOUD PRODUCTS IN MM5.
Chris Birchfield Atmospheric Sciences, Spanish minor.
Mesoscale Modeling Review the tutorial at: –In class.
EnKF Overview and Theory
RCM sensitivity to domain size in summer and winter With the collaboration of: Jean-Philippe Morin (simulations) and Mathieu Moretti (diagnostics) By Martin.
Sea Ice Deformation Studies and Model Development
Observing Strategy and Observation Targeting for Tropical Cyclones Using Ensemble-Based Sensitivity Analysis and Data Assimilation Chen, Deng-Shun 3 Dec,
Diagnosing Climate Change from Satellite Sounding Measurements – From Filter Radiometers to Spectrometers William L. Smith Sr 1,2., Elisabeth Weisz 1,
The fear of the LORD is the beginning of wisdom 陳登舜 ATM NCU Group Meeting REFERENCE : Liu., H., J. Anderson, and Y.-H. Kuo, 2012: Improved analyses.
1.Introduction 2.Description of model 3.Experimental design 4.Ocean ciruculation on an aquaplanet represented in the model depth latitude depth latitude.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss High-resolution data assimilation in COSMO: Status and.
MPO 674 Lecture 20 3/26/15. 3d-Var vs 4d-Var.
In collaboration with: J. S. Allen, G. D. Egbert, R. N. Miller and COAST investigators P. M. Kosro, M. D. Levine, T. Boyd, J. A. Barth, J. Moum, et al.
Development of an EnKF/Hybrid Data Assimilation System for Mesoscale Application with the Rapid Refresh Ming Hu 1,2, Yujie Pan 3, Kefeng Zhu 3, Xuguang.
1/26 APPLICATION OF THE URBAN VERSION OF MM5 FOR HOUSTON University Corporation for Atmospheric Research Sylvain Dupont Collaborators: Steve Burian, Jason.
Assimilating Reflectivity Observations of Convective Storms into Convection-Permitting NWP Models David Dowell 1, Chris Snyder 2, Bill Skamarock 2 1 Cooperative.
Budgets of second order moments for cloudy boundary layers 1 Systematische Untersuchung höherer statistischer Momente und ihrer Bilanzen 1 LES der atmosphärischen.
Richard Rotunno NCAR *Based on:
MODEL ERROR ESTIMATION EMPLOYING DATA ASSIMILATION METHODOLOGIES Dusanka Zupanski Cooperative Institute for Research in the Atmosphere Colorado State University.
Sensitivity Analysis of Mesoscale Forecasts from Large Ensembles of Randomly and Non-Randomly Perturbed Model Runs William Martin November 10, 2005.
Modern Era Retrospective-analysis for Research and Applications: Introduction to NASA’s Modern Era Retrospective-analysis for Research and Applications:
Development of an EnKF to estimate CO 2 fluxes from realistic distributions of X CO2 Liang Feng, Paul Palmer
11 Background Error Daryl T. Kleist* National Monsoon Mission Scoping Workshop IITM, Pune, India April 2011.
Multivariate Data Assimilation of Carbon Cycle Using Local Ensemble Transform Kalman Filter 1 Ji-Sun Kang, 1 Eugenia Kalnay, 2 Junjie Liu, 2 Inez Fung,
DATA ASSIMILATION FOR HURRICANE PREDICTION Experimental system and results of semi-operational implementation during the 2010 Atlantic Hurricane Season.
Multiscale data assimilation on 2D boundary fluxes of biological aerosols Yu Zou 1 Roger Ghanem 2 1 Department of Chemical Engineering and PACM, Princeton.
Deutscher Wetterdienst Vertical localization issues in LETKF Breogan Gomez, Andreas Rhodin, Hendrik Reich.
On the Definition of Precipitation Efficiency Sui, C.-H., X. Li, and M.-J. Yang, 2007: On the definition of precipitation efficiency. J. Atmos. Sci., 64,
Local Predictability of the Performance of an Ensemble Forecast System Liz Satterfield and Istvan Szunyogh Texas A&M University, College Station, TX Third.
Modes of variability and teleconnections: Part II Hai Lin Meteorological Research Division, Environment Canada Advanced School and Workshop on S2S ICTP,
Studying impacts of the Saharan Air Layer on hurricane development using WRF-Chem/EnKF Jianyu(Richard) Liang Yongsheng Chen 6th EnKF Workshop York University.
Determining Key Model Parameters of Rapidly Intensifying Hurricane Guillermo(1997) Using the Ensemble Kalman Filter Chen Deng-Shun 16 Apr, 2013, NCU Godinez,
Module 6 MM5: Overview William J. Gutowski, Jr. Iowa State University.
1 Simulations of Rapid Intensification of Hurricane Guillermo with Data assimilation Using Ensemble Kalman Filter and Radar Data Jim Kao (X-2, LANL) Presentation.
The application of ensemble Kalman filter in adaptive observation and information content estimation studies Junjie Liu and Eugenia Kalnay July 13th, 2007.
Matthew J. Hoffman CEAFM/Burgers Symposium May 8, 2009 Johns Hopkins University Courtesy NOAA/AVHRR Courtesy NASA Earth Observatory.
Mesoscale Assimilation of Rain-Affected Observations Clark Amerault National Research Council Postdoctoral Associate - Naval Research Laboratory, Monterey,
A modeling study of cloud microphysics: Part I: Effects of Hydrometeor Convergence on Precipitation Efficiency. C.-H. Sui and Xiaofan Li.
Meteorological Variables 1. Local right-hand Cartesian coordinate 2. Polar coordinate x y U V W O O East North Up Dynamic variable: Wind.
Matt Vaughan Class Project ATM 621
Numerical Weather Forecast Model (governing equations)
Advisor: Dr. Fuqing Zhang
Simulation of the Arctic Mixed-Phase Clouds
Jeff Anderson, NCAR Data Assimilation Research Section
The Subtropical Sea Breeze
Mark A. Bourassa and Qi Shi
background error covariance matrices Rescaled EnKF Optimization
FSOI adapted for used with 4D-EnVar
RegCM3 Lisa C. Sloan, Mark A. Snyder, Travis O’Brien, and Kathleen Hutchison Climate Change and Impacts Laboratory Dept. of Earth and Planetary Sciences.
Lightning Assimilation Techniques
Project Team: Mark Buehner Cecilien Charette Bin He Peter Houtekamer
Hartmut Bösch and Sarah Dance
Presentation transcript:

Explicit Treatment of Model Error Simultaneous State and Parameter Estimation with an Ensemble Kalman Filter Altuğ Aksoy*, Fuqing Zhang, and John W. Nielsen-Gammon Texas A&M University * Current affiliation: National Center for Atmospheric Research

Two-dimensional, irrotational, incompressible flow with prognostic variables buoyancy (b′, perturbation tempertaure) and vorticity (η′): Explicit heating function: Estimated model parameters: Mean horizontal wind Vertical diffusion coefficients Static stability Heating amplitude Heating depth The Sea Breeze Model: Equations * (Aksoy et al. 2005, JGR) * Similar to Rotunno’s (1983) linear approach

The Sea Breeze Model: Numerics Model domain: Numerical features: Leapfrog time integration Cranck-Nicholson implicit trapezoidal vertical diffusion Rayleigh-damping sponge layers for vorticity Second-order lagged horizontal diffusion for both model variables Asselin-type filtering to control computational mode of the leapfrog scheme Land Sea Forecast Domain Sponge Layer Sponge Layer Sponge Layer 500 km300 km Grid resolution: Horizontal:4 km Vertical:50 m 2 km 3 km

The Sea Breeze Model: Perfect-model behavior LandSeaLandSea Winds and Streamfunction VorticityTemperature LandSea 48H Forecast Noon Maximum Heating 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km

LandSeaLandSeaLandSea 51H Forecast 3:00PM Onset of Sea Breeze TemperatureVorticity Winds and Streamfunction 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 54H Forecast 6:00PM Warmest Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 57H Forecast 9:00PM Strongest Sea Breeze Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 60H Forecast Midnight Maximum Cooling Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 63H Forecast 3:00AM Onset of Land Breeze Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 66H Forecast 6:00AM Coldest Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 69H Forecast 9:00AM Strongest Land Breeze Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

LandSeaLandSea Winds and Streamfunction Vorticity LandSea 72H Forecast Noon Maximum Heating Temperature 3 km Surface 250 km 500 km 250 km 500 km 250 km 500 km The Sea Breeze Model: Perfect-model behavior

Model Error - Enkf Properties (Aksoy et al. 2006, MWR) Observations:Surface buoyancy observations on land Observational error:Standard deviation of ms -2 Observation spacing:40 km (10 grid points) Ensemble size:50 members Ensemble initialization:Perturbations from model climatology Covariance localization:Gaspari and Cohn’s (1999) fifth-order correlation function with 100 grid-point radius of influence Observation processing:Sequential with no correlation between observation errors (Snyder and Zhang 2003) Filter:Square-root after Whitaker and Hamill (2002) with no perturbed observations

Buoyancy Vorticity 3H Prior 3H Posterior Perfect-Model EnKF Results

Buoyancy Vorticity 36H Prior 36H Posterior Perfect-Model EnKF Results

Estimation Performance Buoyancy Vorticity MRE = 60%MRE = 54%

Estimation Performance Mean horizontal windStatic stabilityVorticity diff. coef.Buoyancy diff. coef.Heating amplitudeHeating depth

Mean horizontal wind Parameter Identifiability Static Stability Vorticity Diffusion Coef.Heating Depth :RMS correlation M :Any spatial domain  :Any parameter b :Buoyancy Distinct differences among parameters Static stability and heating depth sensitive to observation location Vort. diff. coef. with smallest correlation, appears to exhibit smallest identifiability

MM5 Experiments: Experimental Setup (Aksoy et al. 2006, GRL, submitted) 36-km resolution with 55×55 grid-point domain 43 vertical sigma layers with 50 hPa model top Initialized: 00Z 28 Aug 2000 Parameterizations: MRF PBL scheme Grell cumulus scheme with shallow cumulus option Simple-ice microphysical parameterization Prognostic variables Winds (u, v, w ) Temperature (T ) Water vapor mixing ratio (q ) Pressure perturbation (p ‘ )

Control Forecast Evidence of the clockwise turning of winds Penetration of the temperature and moisture gradients inland during the sea breeze phase Well-established return flow during the land breeze phase Sea breeze phase (7pm local) Land breeze phase

Ensemble and Filter Properties Ensemble size:40 members Observations:Surface and sounding observations of u, v, and T Observational error:Std. deviations of 2 ms -1 for u and v ; 1 K for T Observation spacing:72 km for surface, 324 km for sounding Covariance localization:Gaspari and Cohn’s (1999) fifth-order correlation function with 30 grid-point radius of influence Observation processing:Sequential (Snyder and Zhang 2003) Filter:Square-root (Whitaker and Hamill 2002)

Parameter Estimation Details Not attempting to identify individual error sources within the PBL scheme associated with different empirical parameters: –Multiplier (m) of the vertical eddy mixing coefficient implanted into the MRF PBL code → for a value of 1.0, the original MRF PBL computation is simply repeated Variance limit applied at 1/4 of initial parameter error Updating is carried out spatially: –Prior parameter value converted to 2-d matrix assumed at surface –Spatial updating is performed with same covariance localization properties as the updating of the state –Updated 2-d distribution is averaged to obtain posterior global parameter value

Correlation Signal – (T,m) and (U,m) Relatively strong overall correlation signal with both temperature and winds (signal strength “comparable” to idealized sea breeze model experiments) Spatially and temporally varying correlation structure Stronger signal near the surface Smaller-scale variability with horizontal winds

Estimation Performance – 3 Cases Initial Mean Error = +0.2Initial Mean Error = +0.65Initial Mean Error = -0.3

Concluding Remarks EnKF demonstrated to be promising for explicit treatment of model error through simultaneous state and parameter estimation Lessons from the idealized sea breeze model experiments: –Sensitivity to observation location, radius of influence, and variance limit is parameter-specific –Counter-acting correlations do lead to identifiability issues with some parameter pairs (do we really need to estimate every single parameter?) A more global approach to the MRF PBL scheme in MM5 appears to be responding well Updating of a global parameter through observations that contain spatial information is an issue and does lead to divergence as number of observations increases: –We have approached this problem through our “spatial updating” technique – ad hoc but effective