assimilation in the pbl
TRANSCRIPT
![Page 1: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/1.jpg)
Assimilation in the PBLJoshua Hacker
National Center for Atmospheric Research,
Research Applications Program
Data Assimilation Initiative review, Sept 2004 – p.1/17
![Page 2: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/2.jpg)
Outline
DAI in my world
Why ensemble approaches?
Data assimilation in the PBL
Research with a column model
DART and assimilation in the PBL
Summary and future DAI interaction
Data Assimilation Initiative review, Sept 2004 – p.2/17
![Page 3: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/3.jpg)
DAI and me
No formal relationship
Collaborative opportunity
Collection of expertise
Data Assimilation Initiative review, Sept 2004 – p.3/17
![Page 4: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/4.jpg)
Predictability Lessons
The forecast problem, particularly at small scales, isinherently probabilistic.
We are obligated to include estimates of uncertainty inobservations, analyses, and forecasts and how theyrelate to the “flow of the day.”
The only way we know how to do this is to combineensemble forecasts with a data assimilation system,using our best dynamic models.
Data Assimilation Initiative review, Sept 2004 – p.4/17
![Page 5: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/5.jpg)
PBL characteristics and assimilation
Transient strong coupling with the Earth’s surface andthe free atmosphere.
Unknown and highly variable (space and time) errorgrowth that is probably not well represented inmesoscale models.
Irreversible processes.
Nonhydrostatic processes and lack of dependable“balances.”
Data Assimilation Initiative review, Sept 2004 – p.5/17
![Page 6: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/6.jpg)
Column model experiments
Hacker/Snyder
Variance-covariance structures in a current mesoscalemodel.
An off-line 1-D PBL modeling framework.
Application of the EnKF to fixed surface (screen-height)observations.
Conclusions and future work.
Data Assimilation Initiative review, Sept 2004 – p.6/17
![Page 7: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/7.jpg)
WRF Climatology
Summer southern great plains variance in the column
Data Assimilation Initiative review, Sept 2004 – p.7/17
![Page 8: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/8.jpg)
WRF Climatology
Summer southern great plains correlation with near-surfacestate.
Data Assimilation Initiative review, Sept 2004 – p.8/17
![Page 9: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/9.jpg)
Assimilation Example: Nighttime
TRUTH and ENSEMBLE MEAN
Data Assimilation Initiative review, Sept 2004 – p.9/17
![Page 10: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/10.jpg)
Assimilation Example: Daytime
TRUTH and ENSEMBLE MEAN
Data Assimilation Initiative review, Sept 2004 – p.10/17
![Page 11: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/11.jpg)
Average Error Reduction for Assimilation
Data Assimilation Initiative review, Sept 2004 – p.11/17
![Page 12: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/12.jpg)
Can the ensemble quantify skill?
Compare spread and error at � �500 m
Data Assimilation Initiative review, Sept 2004 – p.12/17
![Page 13: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/13.jpg)
Add Model Error and Estimate
Augment the state vector with the “moisture availability”and allow the observations to modify the distribution.
Data Assimilation Initiative review, Sept 2004 – p.13/17
![Page 14: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/14.jpg)
Summary of results
The state near the surface is strongly coupled to thePBL through most of the diurnal cycle.
The covariances can be exploited to determine thestructure of the PBL with surface observations.
Model error can be mitigated by augmenting the statevector with model parameters, and estimating theirdistributions.
Data Assimilation Initiative review, Sept 2004 – p.14/17
![Page 15: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/15.jpg)
Future column-model plans
Install more sophisticated column model into DART.
Investigate forward operator error and more parameterestimates.
Attempt real-data experiments for augmenting profilernetworks.
Data Assimilation Initiative review, Sept 2004 – p.15/17
![Page 16: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/16.jpg)
DART and PBL assimilation research
Toward 3DMultiple ways to localize ensemble covariances.
Ease of adding state variables or parameters to thestate vector.
Ease of adding new observation types.
Natural transition to real-data assimilation experiments.
The ability to use both GCMs and mesoscale models.
Data Assimilation Initiative review, Sept 2004 – p.16/17
![Page 17: Assimilation in the PBL](https://reader030.vdocuments.us/reader030/viewer/2022012619/61a05c3f3220ba4ca4547e80/html5/thumbnails/17.jpg)
Primary challenge to DAI
General: data assimilation issues
1. Strongly-forced, highly dissipative
2. Inappropriate closure assumptions (model error)
3. Extremely variable representativeness error
WRF: implementation issues
1. Boundary conditions
2. Cold-start initialization
Data Assimilation Initiative review, Sept 2004 – p.17/17