Warn-on-Forecast and High Impact Weather Workshop, Norman, OK, 8-9 Feb 2012 Challenges of assimilating all-sky satellite radiances Milija Zupanski, Man Zhang, and Karina Apodaca Cooperative Institute for Research in the Atmosphere Colorado State University Fort Collins, Colorado, U. S. A. [ http://www.cira.colostate.edu/projects/ensemble/ ] Acknowledgements: - NOAA NESDIS/GOES-R - NOAA NESDIS/JCSDA - NSF Collaboration in Mathematical Geosciences
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Warn-on-Forecast and High Impact Weather Workshop,Norman, OK, 8-9 Feb 2012
Challenges of assimilating all-sky satellite radiances
Milija Zupanski, Man Zhang, and Karina Apodaca
Cooperative Institute for Research in the AtmosphereColorado State University
Fort Collins, Colorado, U. S. A.[ http://www.cira.colostate.edu/projects/ensemble/ ]
Acknowledgements:
- NOAA NESDIS/GOES-R
- NOAA NESDIS/JCSDA
- NSF Collaboration in Mathematical Geosciences
Relevance of all-sky radiance assimilation
Current status
Challenges
Future
Overview
Satellite observations
Remote sensing is the major source of observations - radars- satellites
AMSU-A GOES-11 SNDR
Need to maximize the utility of cloud observations- hurricanes- severe weather
Satellite data are available everywhere- open ocean- polar regions- other isolated areas on the globe
Current status of all-sky radiance data assimilation
Most operational centers assimilate only clear-sky radiances
- The wealth of cloud-related measurements is discarded
- However, most high impact weather events are characterized by the presence of clouds and precipitation
- The consequence is a sub-optimal use of satellite observations
Limited research and operational efforts
- Mostly related to the use of variational methods, only recent use of ensemble and hybrid variational-ensemble methods
- All-sky microwave data assimilation operational at ECMWF since 2009 (Bauer et al. 2010; Geer et al. 2010 - SSM/I, TMI. AMSR-E)
- Pre-operational testing at NCEP, also pursued at other operational weather centers
Potential benefit of all-sky radiance assimilation is generally accepted, but it is difficult to extract the maximum information from these observations
- modeling of clouds (e.g., microphysics)
- data assimilation methodology
- computing resources, high resolution
Re-development of the TS Erin (2007): Distribution of AMSU-B radiance data in the NCEP operational data stream: (a) all observations, (b) accepted observations after cloud clearing. Data are collected during the period 15-18Z, August 18, 2007. Note that almost all observations in the area of the storm got rejected by cloud clearing. (from Zupanski et al. 2011, J. Hydrometeorology)
Impact of cloud clearing (radiance assimilation)
Need assimilation of all-sky radiances to improve the observation information value
Motivation - CIRA DA research
Develop a robust and efficient data assimilation for high impact weather events- tropical cyclones
- severe weather
Focus on assimilation of cloud and precipitation affected satellite measurements, such as all-sky radiance assimilation
Utilize operational codes as much as possible, focus on realistic issues- WRF NMM, HWRF
- GSI
- CRTM
Assimilate cloudy radiance from various sources: - microwave, infrared, lightning
- combine information from different sources to find most beneficial combinations
List of instruments- New: GOES-R (ABI, GLM), JPSS (ATMS, CrIS),
Data assimilation: Methodological and computational issues Microphysical control variables
- allow cloud observations to impact hydrometeors
Forecast error covariance- Forecast error covariance needs to be state-dependent, and also to represent dynamical and microphysical correlations
Nonlinearity and non-differentiability of Radiative Transfer (RT) operator Correlated observation errors Non-Gaussian errors Quantifying all-sky radiance information:
- How to provide a maximum utility of these data, and how to measure success?
Other relevant issues: verification, code maintenance, bias correction, …
Everything is connected, need to take into account all components
Relevance of microphysical control variables
Adjustment of microphysical control variables:- provides a more complete control of initial conditions- allows most direct impact of cloud observations on the analysis- critical for high impact weather (e.g., TC and severe weather)
Microphysics control variables: impact on DA
Physically unrealistic analysis adjustment without hydrometeor control variable (cloud ice in this example)
Temperature analysis increment at 850 hPa
No cloud ice adjustment
> 25 K
With cloud ice adjustment5-10 K
Forecast error covariance
Analysis correction from variational and ensemble DA can be represented as a linear combination of the forecast error covariance singular vectors ui
Singular value decomposition (SVD) of the forecast error covariance
Fundamentally important to have adequate forecast error covariance. For clouds and precipitation, this implies flow-dependent and dynamically meaningful representation of model uncertainties.
Structure of forecast error covariance defines the analysis correction!
The quality of data assimilation can be assessed by examining the structure of forecast error covariance!
Use single observation experiment to assess the structure:
Forecast error covariance: Algebra
Only Pdd is well known:
- Correlations among microphysical variables not well known- Even less known correlations between dynamical and microphysical variables
Complex inter-variable correlations (e.g., standard dynamical variables and microphysical variables)
Correlations between dynamical variables Correlations between microphysical variables
Cross-correlations between dynamical and microphysical variables
Forecast error covariance: Algebra
Both methods have limitations in representing cloud-related correlations
Variational: modeling of cross-covariances, time-dependence
IR: Assimilation of synthetic GOES-R ABI (10.35 mm) all-sky radiances (Kyrill, 2007) (from Zupanski et al. 2011, Int. J. Remote Sensing)
Cloud ice analysis uncertainy Degrees of Freedom for Signal (DFS)
METEOSAT Imagery valid at19:12 UTC 18 Jan 2007
Analysis uncertainty and DFS are flow-dependent, largest DFS in cloudy areas of the storm.
Quantification of Shannon information for all-sky radiances
Use mutual information (I) and entropy (H)
Joint entropy H(Y1,Y2) can be used to quantify the loss of information due to correlations between all-sky radiance observations Y1 andY2
Mutual information of dependent variables is smaller than mutual information of independent variables
By definition
Use
To obtain
• All-sky radiance observations are correlated• How to measure information from correlated observations?
Future
Important to develop capability to extract maximum information from cloudy and precipitation-affected radiances
Critical for improved analysis and prediction of TC and severe weather outbreaks
Take into account all relevant components (e.g., current challenges)
- partial solutions will not bring true progress
- if Gaussian error statistics is not correct, but used, the cost function is inadequate, implying incorrect minimizing solution
- unbalanced Hessian preconditioning will adversely change the adjustment of variables by creating dynamical imbalances of the analysis
Computation- RT computation increases 2-3 times with scattering
- number of observations increases by an order of magnitude due to cloudy information
- 20-30 times more expensive to compute
Improved information measures for all-sky radiances (e.g., assessment) Combine information from various sources: GOES-R, JPSS (MW, IR, Lightning)
References:
Non-differentiable minimization
Steward, J. L., I. M. Navon, M. Zupanski, and N. Karmitsa, 2011: Impact of Non-Smooth Observation Operators on Variational and Sequential Data Assimilation for a Limited-Area Shallow-Water Equation Model. Quart. J. Roy. Meteorol. Soc., DOI: 10.1002/qj.935.
All-sky IR
Zupanski D., M. Zupanski, L. D. Grasso, R. Brummer, I. Jankov, D. Lindsey and M. Sengupta, and M. DeMaria, 2011: Assimilating synthetic GOES-R radiances in cloudy conditions using an ensemble-based method. Int. J. Remote Sensing, 32, 9637-9659.
All-sky MW
Zupanski, D., S. Q. Zhang, M. Zupanski, A. Y. Hou, and S. H. Cheung, 2011: A prototype WRF-based ensemble data assimilation system for downscaling satellite precipitation observations. J. Hydromet., 12, 118-134.