24
From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems Division

From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Embed Size (px)

Citation preview

Page 1: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

From LAPS to VLAPSmultiscale hot-start analysisFrom LAPS to VLAPS

multiscale hot-start analysis

NOAA ESRL/GSD/FAB

Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth

Global Systems Division

Page 2: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

OutlineReview of LAPS features;Multigrid variational analysis (Space and

Time Multiscale Analysis System, STMAS);Modernizing LAPS using STMAS:

Multigrid variational analysis;Variational cloud analysis;Balance and constraints;Use of remote sensing data;

Future plan and collaboration with KMA

Page 3: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

LAPS reviewAn objective analysis (modified Barnes) scheme;Meteorological states are analyzed sequentially

and dynamic balance is applied afterward;Hot-start:

Analysis of microphysics;Temperature adjustment;Vertical velocity;Analysis of water vapor;

Efficiency;Ease of use, particularly with local data.

Page 4: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Transition from Traditional to Fully Variational LAPS

state vars, wind (u,v) clouds / precip

balance and constraintsin multi-scale variational

analysis

Windanalysis

Temp/Ht analysis

Humidity analysis

Cloud analysis

balance

Traditional LAPS analysis: Wind, Temp, Humidity, Cloud, Balance

Ultimately

Temporary hybrid system: Traditional LAPS cloud analysis

and balance

NumericalForecast

model

Large Scale Model First Guess

Cycling Option

Var.LAPS

Page 5: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

LAPS assimilates a wide range of datasets and local data

Page 6: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

6

LAPS USER BASE

• NOAA– ~120 WFOs (via AWIPS), ARL, NESDIS

• Other US Agencies– DHS, DoD, FAA, CA DWR, GA Air Qual.

• Academia– Univ of HI, Athens, Arizona, CIRA, UND,

McGill

• Private Sector– Weather Decision Tech., Hydro Meteo,– Vaisala, Greenpower Labs

• International agencies (10+ countries)– KMA, CMA, CWB, Finland (FMI), Italy, Spain, – BoM (Australia), Canary Islands, HKO, – Greece, Serbia

Page 7: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Cloud analysis vs. all sky camera

• Demonstrates high resolution analysis of hydrometeors, aerosols, land surface• Check 3-D cloud placement and microphysical properties• Forecasts can also be visualized• Data assimilation a future possibility

Page 8: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Multigrid variational analysisSTMAS

Inherit traditional LAPS multiscale (Barnes) analysis by a multigrid technique (wavelet and recursive filter were also tested and yielded similar results);Improvement of standard 3dvar;

Enhance the analysis by a fully variational analysis with simultaneous balance and constraints;Improvement of traditional LAPS;

Better assimilate remotely sensed observation data, such as satellite IR/VIS, cloud optical depth and radar;Improvement of traditional LAPS.

Page 9: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

OAR/ESRL/GSD/Forecast Applications Branch

Sequence of 3-4DVARs with proper balances– need for covariance information reduced

Similar to traditional LAPS

Standard 3-4DVARWith banded covariance

Possible ensembleFilter application

Long waves Short waves

Xie et al. “A Space–Time Multiscale Analysis System: A Sequential Variational Analysis Approach”, MWR 2011

Analysis and model initialization may endat different multigrid levels

MULTISCALE VARIATIONAL ANALYSIS

Page 10: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Humidity Analysis resolving discontinuity

LAPS

STMAS

Page 11: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

VLAPS (STMAS) bound constraintsVLAPS uses the L-BFGSB in its variational analysis

and this quasi-Newton software allows users to use bound constraints;

VLAPS can use cloud and/or reflectivity information to constrain its humidity analysis:Currently, if an area is covered with cloud and

reflectivity, VLAPS constrains its humidity to 100% RH.

An on-going evaluation is to make it as weak one for accommodating other obs (e.g., GPS);

Such bound constraints are considered for variational cloud analysis.

Page 12: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

A real time example

Possible collaboration: improving covariance

Page 13: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

VLAPS assimilation of remote sensing observations with collaborators

A long list of datasets:AMSU-A and B for Taiwan now;GPS (TPW now and slant delay next);GOES sounder IPW;Cloud mask and optical depth (testing now);Dual Pol radar (Serbia Meteorological Agency);GOES IR and visible imagery (with CRTM);GOES-R cloud cooling and over-shooting;……

Page 14: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

AMSU-B Up Air Impact

No AMSU-B AMSU-B all channels

Page 15: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

GPS TPW data impact

Page 16: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

General methodology of VLAPS analysis of remote sensing data

A forward operator mapping analysis variables to observations: F(X)≈Y;

An adjoint of this operator, F’(X);

An additional term in the minimization cost function: (F(X)-Y)T O-1 (F(X)-Y);

Minimization is done with added gradient term from the remote sensing data.

Page 17: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

GPS TPW forward operator

vertically

Surface grid box

Domain top

Page 18: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

GPS Examples:A forward operator for GPS TPW(specific humidity):

An integration of vertical specific humidity along a given GPS zenith path;

A forward operator for GPS slant delay:An integration of refractivity along the GPS slant

path;

Both are differentiable in terms of the control variables, sh for the former and sh, T and p for the later.

Page 19: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Variational cloud analysisCurrently, use the traditional LAPS cloud analysis as an

initial guess;Cloud mask as constraints of cloud ice, liquid, rain and

snow (possible graupel), including ;Cloud phase products are also used;Cloud optical depth is being tested;IR and visible data will be assimilated;Temperature is used to constrain cloud ice and liquid;Variationalization of LAPS cloud components, e.g.,

estimated cloud from RH;Sophisticated covariance is needed for filling the data

void regions.

Page 20: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Cloud optical depth vs. cloud ice analyses

VLAPS

LAPS

Cloud Optical Depth OBS

Page 21: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

VLAPS (1km) without GPS and COD

Page 22: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems
Page 23: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Future PlanIdentify and assimilate important observation

data sources, dual pol radar, IR and visible;Improve balance and constraints, particularly

on the hydrometer state variables;Continuity (already in), hydrostatic, etc;WRF FDDA collaborating with US Army;WRF adjoint for short 4DVAR assimilation

window.Improve forecast model parameters, e.g.,

snow cover, land types etc;Parallelization of VLAPS;Object-oriented design of VLAPS.

Page 24: From LAPS to VLAPS multiscale hot-start analysis NOAA ESRL/GSD/FAB Y. Xie, S. Albers, H. Jiang, D. Birkenheuer, J. Peng, H. Wang, and Z. toth Global Systems

Collaboration with KMAForecast model for 200-m resolution run with

tuned model parameters and topography;Local observation datasets, in-situ and

remotely sensed data, including all-sky images;Observation forward operators and their

adjoint;Variational cloud analysis;Hydrometeor constraints;Terrain following VLAPS code development;GIT LAPS software sharing;Object-oriented VLAPS development.