Download presentation
Presentation is loading. Please wait.
Published byAsher Harper Modified over 9 years ago
1
GSI developments and plans at NCAR/MMM Tom Auligné Aimé Fournier, Hans Huang, Andy Jones, Hui-Chuan Lin, Zhiquan Liu, Yann Michel, Arthur Mizzi, Thomas Nehrkorn, Syed Rizvi, Hongli Wang, Xin Zhang National Center for Atmospheric Research NCAR is supported by the National Science Foundation GSI Data Assimilation Workshop - June 28, 2011
2
Focus at NCAR/MMM –Regional GSI –WRF-ARW model (NetCDF files) Projects funded by AFWA –AFWA Coupled Analysis and Prediction System (ACAPS) –AFWA Data Assimilation –AFWA Aerosols Collaboration with –GSI developers (EMC, GMAO, GSD, DTC) –JCSDA Introduction
3
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
4
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
5
Background and Obs Errors: Community tools “Community GEN_BE” utility: https://svn-wrf-var.cgd.ucar.edu/branches/gen_be https://svn-wrf-var.cgd.ucar.edu/branches/gen_be –Includes all the features of WRFDA V3.2.2 –Multi-variate humidity –Generation of WRF-ARW background errors for GSI Extension of GEN_BE to include –Aerosol concentrations (univariate) –Cloud parameters (Qcloud, Qrain, Qice, Qsnow) Expansion of GSI control variable Observation error tuning with the diagnostic equations (Desroziers 2005)
6
Background Error Covariances: Masked Statistics
7
Michel et al. (MWR, 2011) Background Error Covariances: Masked Statistics
8
Background Error Covariances: Wavelets
10
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
11
Variational/Ensemble Hybrid WRF/GSI Regional Hybrid Testing package: https://svn-mmm-hybrid-testbed.cgd.ucar.edu/HYBRID_TRUNK https://svn-mmm-hybrid-testbed.cgd.ucar.edu/HYBRID_TRUNK Cf. presentation by Arthur Mizzi
12
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
13
Conceptual view of using displacements to characterize errors background error displacements of coherent features additive (residual) error =>+ Displacement Pre-Processing
14
Initial time: 08-28-05 06:00:00z Vortex displaced forward along track
15
18 Hour forecast time: 08-29-05 00:00:00z 18 hours later vortex maintains forward position
16
Collaboration between AER, MIT and NCAR Integration of displacements –Build on the existing API, with enhancements to add: –Support for multiple displacement algorithms Algorithmic developments –Constraints formulated and evaluated specifically for cloud-related fields Candidates: smoothness, non-divergence of displacements Application in: grid point, spectral, or wavelet space –Time evolution of displacements Characterize and model the time evolution of displacements Prepare for integration with 4D-Var –Figures of Merit for cloud-related fields Displacement Pre-Processing: Status and Plans
17
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
18
WRF Adjoint: WRF/GSI 4DVar New TL/AD code: WRFPLUS –Consistent with latest WRF-ARW (v3.3) –Includes simplified physics (surface drag, large-scale condensation, cumulus scheme, Kessler microphysics) WRF/GSI 4DVar –Based on GMAO 4DVar framework –New coupling between GSI and WRF/WRFPLUS Initial testing looks good. Cf. presentation by Xin Zhang
19
WRF Adjoint: Observation Impact Observation (y) WRFDA/GSI Data Assimilation WRF-ARW Forecast Model Forecast (x f ) Derive Forecast Accuracy Background (x b ) Analysis (x a ) Adjoint of WRF-ARW Forecast TL Model (WRF+) Observation Sensitivity ( F/ y) Background Sensitivity ( F/ x b ) Analysis Sensitivity ( F/ x a ) Observation Impact ( F/ y) Adjoint of WRFDA/GSI Data Assimilation Obs Error Sensitivity ( F/ ob ) Gradient of F ( F/ x f ) Define Forecast Accuracy Forecast Accuracy (F) Bias Correction Sensitivity ( F/ k ) Figure adapted from Liang Xu (NRL)
20
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
21
Aerosol Satellite Observations Assimilation of MODIS Aerosol Optical Depth in GSI –Process MODIS AOD data (HDF to BUFR converter) –Use CRTM-AOD (Quanhua Liu) –Couple with WRF-Chem GOCART (14 aerosol species) Status and plans –Assimilate surface PM2.5 (ongoing) –Assimilate MODIS Visible/NIR radiances (planned, pending) Cf. presentation by Zhiquan Liu
22
Cloud Satellite Observations: Retrievals MODIS cloud retrieval products –Cloud liquid/ice water path, cloud optical depth, particle effective radius (1km resolution observations) –Cloud top properties: pressure, temperature, fraction/emissivity (5km resolution observations) Assimilation of MODIS Cloud Water Path –Process MODIS CWP data (HDF to BUFR converter) –Observation Operator (+ TL & AD) Status and plans –Assimilate MODIS CWP at convective scale (ongoing) –Assimilate MODIS Cloud Optical Depth (planned)
23
Very first shot at cloudy radiances, still needs a lot more work… Cloud parameters from WRF-ARW first-guess CRTM forward model and Jacobian Inclusion of cloud (microphysical) parameters in control variable (implemented in both WRFDA and GSI) Cloud Satellite Observations: Radiance Assimilation
24
Observation AIRS (12micron) Background (WRF-DART) Observation – Background
25
Simple B Matrix for cloud parameters copied from humidity Ensemble assimilation using the alpha control variable (no tuning) Cloud Satellite Observations: Radiance Assimilation
26
Clear observations only Cloud Satellite Observations: Radiance Assimilation Simple B Matrix for cloud parameters copied from humidity
27
Remaining issues include: - Bias Correction - Quality Control - Non-linearities in the observation operator - Representativeness Error Cloud Satellite Observations: Radiances
28
Pixel N k1 N k2 N k3 NoNo Cloud Top Pressure (hPa) MODIS Level2 AIRS MMR with Cloud fractions N k are ajusted variationally to fit observations: Cloud Satellite Observations: Radiances
29
CloudSat Reflectivity AIRS MMR Effective Cloud Fraction Cloud Satellite Observations: Radiances
30
Towards Cloudy Radiance Assimilation Pixel N k1 N k2 N k3 NoNo Cloud Satellite Observations: Radiances
31
31 Towards Cloudy Radiance Assimilation Simulated mismatch in resolution: - Perfect observations (high resolution) - Perfect Background (lower resolution) Innovations Background Cloud Satellite Observations: Representativeness
32
32 Towards Cloudy Radiance Assimilation New interpolation scheme: 1. Automatic detection of sharp gradients 2. New “proximity” for interpolation Innovations Background New Innovations Cloud Satellite Observations: Representativeness
34
The raw y o − y b (left) includes errors due to y o and y b coming from completely different representations, that (hypothetically) have been reconciled by the foregoing wavelet- coefficient selection procedure. Cloud Satellite Observations: Representativeness
35
Background and Observation Errors Variational/Ensemble Hybrid Displacement Pre-processing WRF Adjoint: 4DVar and Observation Impact Aerosol and Cloud Satellite Observations Verification Outline
36
Period: 4-17 June 2009 Analyses and 6 hr forecasts from 50-member ensembles using Data Assimilation Research Testbed (DART) system Verification: Test Case (courtesy Glen Romine) 15 km mesoscale, 3 km storm-scale
37
Verification: Validation Data World-Wide Merged Cloud Analysis (WWMCA) Main Archive: Quality-controlled, GOES East and GOES West over CONUS Covers January 1998 – December 2009 Resolution – 4x4 km for all channels except #3 which is 4x6 km Monthly/hourly cloud cleared background for all visible hours Monthly/hourly Cloud % using visible threshold Monthly/every other hour Cloud % using IR threshold since 2003 Addition hours of QC’d GOES West for May-Sept 1999-2009 Example of GOES 8 background image
38
New WRF Adjoint for GSI 4DVar and Observation Impact Community tool for Background Error calculation (GEN_BE) Specific developments for Cloud and Aerosol assimilation Opportunity for inter-comparison Conclusion
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.