08 DataAssimilation Atlas

Slide1: 

The National Polar-orbiting Operational Environmental Satellite System Assimilation of Polar-Orbiting Space-Based Data and its Impact on Weather Prediction Dr. Robert Atlas Chief Meteorologist NASA GSFC Laboratory for Atmospheres January 13, 2004

Assimilation of Polar-Orbiting Space-Based Data and its Impact on Weather prediction : 

Assimilation of Polar-Orbiting Space-Based Data and its Impact on Weather prediction Has conducted research to assess and improve the impact of satellite data on weather prediction since 1975 Head of NASA Data Assimilation Office from 1996-2002 First to demonstrate the beneficial impact of quantitative satellite data on numerical weather prediction Developed the assimilation methodology to obtain the first beneficial impact of satellite surface wind measurements. Developed methodology to enhance the realism of Observing System Simulation Experiments Ph.D. Meteorology and Oceanography, NYU Dr. Robert Atlas Chief Meteorologist NASA GSFC Laboratory for Atmospheres

Outline : 

Outline The need for space-based data Observing System Experiments (OSE and OSSE) Impact of early and current Polar orbiting satellite data on weather prediction Impact of future Polar-orbiting satellite data Summary

Need for Space-Based Data: 

Need for Space-Based Data Coverage of Ship and Buoy Reports 1999 Sep 13 00Z Conventional observations of temperature, wind, and moisture profiles (by rawinsondes) are concentrated over land areas and primarily in the Northern Hemisphere. Coverage of Rawinsonde Reports Coverage of Ship and Buoy Reports 1999 Sep 13 00Z Over oceans, conventional observations are primarily limited to single level data provided by aircraft, ships, and buoys. The coverage of these and other ground based observing systems is not sufficient for global atmospheric and oceanographic research or weather prediction.

Slide5: 

Satellites have offered, and continue to offer an effective way to provide needed observations in data sparse regions and to provide data at higher resolution than could be afforded by the conventional observing network. AIRS TOVS QuikScat SSM/I

Slide6: 

Satellite Data Indirect measurements Until recently, low vertical resolution in troposphere Area or volume averages Variable temporal resolution – asynoptic Good horizontal coverage High horizontal resolution Automatic data processing High data volumes Conventional Data Direct measurements High vertical resolution Point measurements Generally low temporal resolution Poor horizontal coverage Low horizontal resolution Human link in data chain Low data volumes

Key aspects of Assimilation Development that have been required to use satellite data effectively: 

Key aspects of Assimilation Development that have been required to use satellite data effectively Increased horizontal, vertical and temporal resolution Data selection / data usage Covariance modeling Bias estimation Algorithms Capable of directly assimilating non-state variables (radiance, backscatter) Satellite data >90% of the data volume analyzed Account for roughly half of the forecast skill Reduces major forecast busts This is still below the potential of the data

Observing System Experiments &Observing System Simulation Experiments: 

Observing System Experiments & Observing System Simulation Experiments Observing System Experiments (OSE) are conducted to evaluate the impact of specific observations or classes of observations on analyses and forecasts. Observing System Simulation Experiments (OSSE) are conducted: To provide a QUANTITATIVE assessment of the potential impact of proposed observing systems on data assimilation. To evaluate new methodology for the processing and assimilation of space-based data. To evaluate tradeoffs in the design and configuration of proposed observing systems (e.g. coverage, resolution, accuracy and data redundancy).

Early Simulation Studies: 

Early Simulation Studies PROVIDED AN ANALYSIS OF GARP DATA REQUIREMENTS “USEFUL RANGE” OF PREDICTABILITY NEED FOR REFERENCE LEVEL DATA RELATIVE USEFULNESS OF ASYNOPTIC vs. SYNOPTIC DATA ASSIMILATION INDICATED THAT ALL THREE OF THE PRIMARY VARIABLES (TEMPERATURE, MOISTURE, WIND) COULD BE DETERMINED IF A CONTINUOUS TIME HISTORY OF ANY ONE OF THESE VARIABLES WERE INSERTED INTO A GENERAL CIRCULATION MODEL. (“CHARNEY CONJECTURE”)

Limitations of Previous Studies: 

Limitations of Previous Studies Most important is the use of the same model for nature and assimilation / forecasting (“Identical Twin Experiments”). Model Dependence of results. Treating observational errors as random and uncorrelated.

Slide11: 

OSSE Methodology

Summary of OSSE Results: 

Summary of OSSE Results EVALUATED THE RELATIVE IMPACT OF TEMPERATURE, WIND AND MOISTURE DATA These experiments showed wind data to be more effective than mass data in correcting analysis errors and indicated significant potential for space-based wind profile data to improve weather prediction.The impact on average statistical scores for the northern hemisphere was modest, but in approximately 10% of the cases a significant improvement in the prediction of weather systems over the United States was observed. EVALUATED THE RELATIVE IMPORTANCE OF UPPER & LOWER LEVEL WIND DATA These experiments showed that the wind profile data from 500hpa and higher provided most of the impact on numerical forecasting. EVALUATED DIFFERENT ORBITAL CONFIGURATIONS AND THE EFFECT OF REDUCED POWER FOR A SPACE-BASED LASER WIND SOUNDER (LAWS) These experiments showed the quantitative reduction in impact that would result from proposed degradation of the LAWS instrument. DETERMINED DRAFT DATA REQUIREMENTS OF SPACE-BASED LIDAR WINDS These experiments evaluated different coverages, resolutions, and accuracies for lidar wind measurements to estimate both research and operational requirements for the Global Tropospheric Wind Sounder (GTWS) Mission.

Summary of OSSE Results: 

Summary of OSSE Results DEVELOPED AND TESTED IMPROVED METHODOLOGY FOR ASSIMILATING SATELLITE SCATTEROMETER DATA. Applying this methodology resulted in the demonstration of the first significant positive impact of real scatterometer data in 1983. DEVELOPED AND TESTED DIFFERENT METHODS FOR ASSIMILATING SATELLITE SURFACE WIND SPEED DATA. This led to assimilation of SSM/I wind speed data to improve ocean surface wind analyses. EVALUATED THE QUANTITATIVE AND RELATIVE IMPACT OF ERS AND NSCAT YEARS PRIOR TO THEIR LAUNCH. These results were confirmed after the launch of both instruments. EXPERIMENTS ARE BEING CONDUCTED NOW TO DETERMINE THE POTENTIAL IMPACT OF LIDAR WINDS IN CURRENT DATA ASSIMILATION SYSTEMS. The results continue to show significant potential.

Slide14: 

OSSE Example Error and Forecast Improvement Scatterometer Ocean Surface Wind Field

Slide15: 

OSE Example: SeaWinds Experiments GLOBAL DATA ASSIMILATION SYSTEM USED: GEOS-4: fvDAS Operational System Resolution: 1o x 1.25o PERIOD OF ASSIMILATION: April 10 - May 15, 2003 EXPERIMENTS: CONTROL: All Conventional Data + TOVS + CTW + SSM/I TPW CONTROL + QuikScat CONTROL + SeaWinds CONTROL + QuikScat + SeaWinds FORECASTS: 26 forecasts starting from April 15, 2003

Slide16: 

OSE Example Forecast Improvement with Additional Satellite Coverage/Observations

Slide19: 

AIRS Advanced Sounder Experiments At GSFC we are evaluating the impact of AIRS data in several different forms, NESDIS statistical retrievals AIRS Team physical retrievals 1D VAR interactive retrievals AIRS radiances The impact of clear retrievals or radiances vs. the addition of partially cloudy data will be evaluated. The impact of data over water vs. data over both water and land will be evaluated. The impact of AIRS will be evaluated using several different DAS: FVSSI, FVDAS, EDAS

Slide20: 

AIRS Experiments With FVSSI GLOBAL DATA ASSIMILATION SYSTEM USED: fvSSI: fvGCM - Resolution: 1 o x1.25 o SSI (NCEP) analysis-T62 PERIOD OF ASSIMILATION: 1 January - 31 January, 2003 EXPERIMENTS: CONTROL: All Conventional Data + ATOVS + Radiance (NOAA-14, 15, 16) + CTW + SSM/I TPW+ SSM/I Wind Speed + QuikScat + Ozone CONTROL + AIRS (Clear/Ocean / -40 deg to + 40 deg) CONTROL + AIRS (Clear/Ocean/Global) CONTROL + AIRS (Clear +Partly Cloudy/Ocean/Global) CONTROL + AIRS (w/o sea - ice data with land data, with moisture profiles, updated algorithms) FORECASTS: 13 forecasts run every two days beginning on 6 January, 2003

Slide22: 

Impact of AIRS on Forecast Accuracy

Slide31: 

Impact of Future Polar-Orbiting Satellite Data Objectives of the current OSSE: TO EVALUATE THE POTENTIAL FOR SPACE-BASED WIND PROFILE DATA TO IMPROVE ATMOSPHERIC ANALYSES AND NUMERICAL WEATHER FORECASTS USING THE CURRENT NASA/GLA DATA ASSIMILATION SYSTEMS. TO EVALUATE THE IMPACT OF SPACE-BASED WIND PROFILES ON THE ANALYSIS AND FORECASTING OF PHENOMENA NOT PREVIOUSLY STUDIED IN OSSEs. TO DETERMINE IF THE IMPACT OF SPACE-BASED WIND PROFILES WOULD BE SUBSTANTIALLY REDUCED IN THE PRESENCE OF ADVANCED SOUNDER OR WITH A MORE ACCURATE FORECAST MODEL.

Slide32: 

Description of Experiment A new nature run was generated using the NASA NCAR FVGCM at 0.5 deg resolution for the period September 11 to December 31, 1999. A detailed evaluation of the nature run was performed. All conventional and satellite observations (that are currently assimilated at NASA) were simulated with existing coverages and expected accuracies.

Slide33: 

Description of Experiment Space-based wind profiles were simulated first in a very idealized way, with the same coverage as TOVS, 1m/s accuracy at all levels and no degradation due to cloud effects; in later experiments somewhat more realistic lidar winds with attenuation due to clouds, but still 1m/s accuracy were assimilated. The OSSE system was calibrated through comparison with real data impact experiments. The results of the OSSE were evaluated in terms of standard metrics (rms, anomaly correlation) and new metrics ( cyclones, jet streaks, hydrologic cycle).

Summary of OSSEsUsing FVCCM Nature Run: 

Summary of OSSEs Using FVCCM Nature Run GLOBAL DATA ASSIMILATION SYSTEM USED: GEOS-3 1 X 1 deg horizontal resolution FVDAS 1 x 1.25 deg horizontal resolution SPINUP: 35 days PERIOD OF ASSIMILATION: Sept. 11 - Oct. 31, 1999 CALIBRATION EXPERIMENTS (REAL and SIMULATED): CTRL (Conventional Data + TOVS + CTW + QSCAT) CTRL-ALL SAT (Conventional Data only) CTRL-SAT TEMP (Conventional Data + CTW + QSCAT) CTRL-QSCAT (Conventional Data + TOVS + CTW) SIMULATED DATA EXPERIMENTS: CTRL + Lidar Winds (with varying coverage) CTRL + Advanced Sounder + Lidar winds

Slide42: 

Based on OSSEs at NASA Data Assimilation Office (R. Atlas) Tracks Green: actual track Red: forecast beginning 63 hours before landfall with current data Blue: improved forecast for same time period with simulated wind lidar Save ~ $1M/mile per hurricane for improved landfall forecast Lidar in this one case Reduces landfall prediction error by 66% Potentially save > $165M Potential Impact of New Space-based Observations (Wind Lidar) on Hurricane Track Prediction

Slide44: 

Impact of LIDAR Winds on Cyclone Prediction [NASA/GSFC/DAO] 5-day Average Reduction in Position Error Global: 35 km (10% improvement) N. America: 48 km (11% improvement) 10-day Average Reduction in Position Error Global; 66 km (17% improvement) N.H.X.T: 17 km (5% improvement) S.H.X.T: 48 km (24% improvement) Reduction in Hurricane Landfall Position Error For United States: 239 km (66% improvement) at 63 hr

Summary: 

Summary Observing System Simulation Experiments (OSSEs) provide an effective means to: Evaluate potential impact of proposed observing systems Determine tradeoffs in their design Evaluate new data assimilation methodology OSSEs conducted with 4 different data assimilation systems (from 1985-2003) all showed significant potential for space-based lidar wind profiles to improve atmospheric analyses & weather predictions. Polar-orbiting satellite data are a critical component of the Global Observing System and contribute significantly to atmospheric and oceanographic prediction. New space-based observations to be provided by NPOESS should contribute very substantially to further improvements to both research and forecasting.