![Page 1: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/1.jpg)
Atmospheric Data Assimilation
Tom Auligné (JCSDA),
John Derber (Lead, EMC),
Jeff Whitaker (ESRL),
Ricardo Todling (NASA GMAO),
Daryl Kleist (U. of Maryland),
Andrew Collard (EMC),
Nancy Baker (NRL Monterey),
Jeff Anderson (NCAR)
![Page 2: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/2.jpg)
Motivation
• Initial conditions for Numerical Weather Prediction (NWP)
• Calibration and validation
• Observing system design, monitoring and assessment
• Reanalysis
• Better understanding (Model errors, Data errors, Physical process interactions, etc.)
![Page 3: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/3.jpg)
α ~ 0.4/day β ~ 2.4m/day E∞ ~ 115m
dE/dt=(αE+β)(1-E/E∞)
1990 2000 2010
Magnusson and Källen 2013
Chaotic ~ Model error contribution
500 hPa RMSE
Motivation
![Page 4: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/4.jpg)
500 hPa Anomaly
Correlations
(SH, 15 Aug – 30 Sep 2010)
Motivation
Source: Jung (2012)
![Page 5: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/5.jpg)
xb yo
Modern data assimilation systems combine together information from a short term forecast, a set of observations and possibly other information to estimate the most probable state of atmosphere.
Introduction
Courtesy: Ménétrier
![Page 6: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/6.jpg)
Introduction
• Compute observation operator H: xy d = yo-Hxb
Source: ECMWF
![Page 7: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/7.jpg)
• Compute observation operator H: xy d = yo-Hxb
• (Extended) Kalman Filter analysis: xa = xb + Kd A = (I-KH)B
• Model forecast: xb M(xa) B MAMT + Q
Hypotheses: Background and observation errors are uncorrelated, unbiased, normally distributed, with known covariances B and R
Introduction
K = BHT(HBHT+R)-1
![Page 8: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/8.jpg)
State of the Art
• Good estimates of the observational AND forecast error structure are necessary. Much of our effort is directed towards improving the specification of these error structures.
• In addition, determining the set of observations to use in an analysis is very important. – Quality control – Forward models (go from analysis variables to observed variables)
• Also, assimilation system must be efficient enough to
complete in operational time window (~20 minutes for current global system). – Approximations necessary.
![Page 9: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/9.jpg)
Horizontal autocorrelations
RH (mid-troposphere)
Vertical auto-correlation
State of the Art
![Page 10: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/10.jpg)
Wang, Sun, Zhang, Huang and Auligné (MWR 2013) Mini-4DVar (10min)
![Page 11: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/11.jpg)
State of the Art
• Good estimates of the observational AND forecast error structure are necessary. Much of our effort is directed towards improving the specification of these error structures.
• In addition, determining the set of observations to use in an analysis is very important. – Quality control – Forward models (go from analysis variables to observed variables)
• Also, assimilation system must be efficient enough to
complete in operational time window (~20 minutes for current global system). – Approximations necessary.
![Page 12: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/12.jpg)
State of the Art
• Technique – 3D/4D-Var – Ensemble – Hybrid – Ensemble used to specify part of model error covariance
• Current NCEP operational system is 3D Hybrid EnVar • Upgrade to 4D Hybrid EnVar under final testing
• Observations – Conventional – Satellite instruments – Bias correction – currently aircraft satellite radiances, radiosondes
(radiation correction).
• Many other details (the devil is in the details)
![Page 13: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/13.jpg)
Application of NWP
Bias Correction for SSMIS F18
Ascending Node Descending Node
Latitude
Unbias & Bias Corrected O-B
O-B Before Bias Correction
Global
Dsc
Asc
O-B After Bias Correction
Global
Dsc
Asc
O-B Before Bias Correction
O-B After Bias Correction
Using Met Office SSMIS
Bias Correction Predictors
T (K)
T (K)
T (
K)
Courtesy: Andrew Collard
![Page 14: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/14.jpg)
Priorities
• Upgrade and maintain use of current observations
• Continue to develop GSI Hybrid
• Enhance use of cloud-impacted (“all sky”) radiances
• Improve specification of observation errors
– Error variances
• Increased granularity (station/instrument type statistics)
– Correlated errors
![Page 15: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/15.jpg)
Priorities
• Improve hydrological and dynamical balance in analysis increments
• Improve stochastic physics in ensemble members
• Add new satellite instruments/channels
• Improve scalability and efficiency of GSI code
![Page 16: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/16.jpg)
Priorities
• Improve quality control and bias correction of conventional data
• Enhanced variational quality control for all observations
• Bias correction of background field
• Assimilation of aerosol and trace gas information
![Page 17: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/17.jpg)
Priorities
• Improve static background error in GSI hybrid
• Increase resolution for data assimilation ensemble members
![Page 18: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/18.jpg)
Externally Funded Projects
1. Improved tropical cyclone initialization for NCEP operations through direct assimilation of storm information. PI Daryl Kleist
1. Testing and implementation of a cycling ensemble data
assimilation system for operational hurricane prediction. PI Jeff Whitaker
2. Development of Advanced Data Assimilation Techniques for Improved use of Satellite-Derived Atmospheric Motion Vectors. PI James Jung
3. Improving Global and Hurricane Prediction by Using Minimum-Cost Large Ensemble in GFS 4DEnVar Hybrid Data Assimilation System. PI Xuguang Wang
![Page 19: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/19.jpg)
Interaction with other projects
• Some people would argue that greatest improvement in forecast skill has been due to improvement in assimilation systems
• To properly evaluate a forecast system, must include data assimilation – Corollary: Details of assimilation system may be different for
different forecast models. Significant effort necessary to ensure appropriate assimilation system for individual model.
Forecast model skill dependent on Assimilation system
Assimilation system dependent on Forecast model
![Page 20: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/20.jpg)
Source: Craig Schwartz
WRF (4km) initialized over CONUS
Examples of DA details important for evaluation of Forecast system
![Page 21: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/21.jpg)
Examples of DA details important for evaluation of Forecast system
– Model error. If two models have different forecast error characteristics, different background error covariances will be necessary to project information on proper scales
– Vertical coordinate. Background error covariances can be very sensitive to the vertical coordinate used. Should be defined in same vertical coordinate as forecast model.
– Grid. Ideally analysis is performed on same grid as forecast model. For non-uniform grids not straightforward to do analysis. Possibilities:
– Calculate O-B directly from model solution, then calculate increment and interpolate to non-uniform grid. (More accurate)
– Interpolate forecast from non-uniform grid then calculate O-B, then calculate increment and interpolate to non-uniform grid. (Easier – more common)
![Page 22: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/22.jpg)
Towards Cloudy Radiance Assimilation
Simulated mismatch in resolution: (Daley 1993, Liu and Rabier 2002, Waller et al. 2014) Perfect observations (high resolution), perfect Background (lower resolution)
Departures Background
Examples of DA details important for evaluation of Forecast system
![Page 23: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/23.jpg)
Examples of DA details important for evaluation of Forecast system
![Page 24: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/24.jpg)
GSI (Hybrid 3DVar 4DEnVar)
(G)LDAS (2DVar)
GSI GODAS (CFS)
NCODA (3DVar)
N/A RTMA (GSI)
?
NCODA GSI
GSI
![Page 25: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/25.jpg)
Coupled Model
Ensemble Forecast
NEMS
OC
EAN
SEA
-IC
E
WA
VE
LAN
D
AER
O
ATM
OS
Ensemble Analysis (N Members)
OUTPUT
Coupled Ensemble Forecast (N members)
INPUT
Coupled Model
Ensemble Forecast
NEMS
OC
EAN
SEA-IC
E
WA
VE
LAN
D
AER
O
ATM
OS
NCEP Coupled Hybrid Data Assimilation and Forecast System
Courtesy: Suru
![Page 26: Atmospheric Data Assimilation - National Weather Service Atmospheric...Atmospheric Data Assimilation Tom Auligné (JCSDA), John Derber (Lead, EMC), Jeff Whitaker (ESRL), Ricardo Todling](https://reader033.vdocument.in/reader033/viewer/2022042613/5faee74f06df5d42901e54c1/html5/thumbnails/26.jpg)
Joint Effort for Data assimilation Integration (JEDI)
• Modular, flexible, object-oriented code
• Improved readability, maintenance and testing
• Collaborative operation & research applications
• Model-agnostic DA components strongly coupled Data Assimilation