NCARAviation Weather ResearchProgram Management Review
Advanced Weather Radar Techniques Advanced Weather Radar Techniques Product Development TeamProduct Development Team
Lead: Kim ElmoreDistributed Leads:
Cathy Kessinger – NCAR/RAPTim Schneider – ETL
David Smalley - MIT/LL
The AWRT Mission:The AWRT Mission:
Develop and apply new and advanced techniques to data from various radar platforms for the
benefit of the aviation community.
AWRT PDT OrganizationAWRT PDT Organization
NOAA NSSLLead: Kim Elmore
Polarization/Data QualityMulti-Sensor
CONUS 4-D Mosaic
MIT LLAlt: David Smalley
Data qualityFAA Wx Systems Integration
ORPG Implementation
NCAR RAPAlt: Cathy Kessinger
PolarizationData quality
Remote RetrievalsMulti-Sensor
NOAA ETLAlt: Tim SchneiderGRIDS Radar System
AWRP PDTs
WARP SupportWARP Support• WSR-88D data for controllers, not meteorologists.
• Different approaches to data quality needed; techniques long-range L-band radars no longer adequate.
• New approaches to data quality control need to be developed so users have confidence in/don’t feel compelled to second-guess the weather data products displayed to them.
WARP: Reflectivity Reduction WARP: Reflectivity Reduction ProblemProblem
Product 36Product 96
Product 36
Warp: Reflectivity Reduction Warp: Reflectivity Reduction ProblemProblem
Product 96Product 96i
Circulation Detection Circulation Detection DevelopmentDevelopment
•• Violent or longViolent or long--lived storms tend to possess lived storms tend to possess circulations/circulations/mesocyclonesmesocyclones. .
•• Current WSRCurrent WSR--88D algorithms have a very high false 88D algorithms have a very high false alarm rate; unacceptable by controllers.alarm rate; unacceptable by controllers.
•• New robust and reliable circulation detection New robust and reliable circulation detection algorithms needed. algorithms needed.
•• Algorithms that use circulations to diagnose storm Algorithms that use circulations to diagnose storm severity or estimate storm longevity will be severity or estimate storm longevity will be considerably improved by this work.considerably improved by this work.
LLSD: More Accurate Shear EstimatesLLSD: More Accurate Shear Estimates
LLSD Location AccuracyLLSD Location Accuracy
LLSD: 6 h Accumulated LLSD: 6 h Accumulated AzAz ShearShear
Technical FacilitationTechnical Facilitation•Technical facilitation supports the AWRT PDT algorithm
development.
•The interface being developed at the NSSL, the WDSS-II, provides a way to develop, validate, verify anddemonstrate algorithms developed within this PDT.
•WDSS-II provides a route into the Open Radar Product Generation (ORPG) system. WDSS-II supports and incorporates the MITRE Common Operations Development Environment (CODE).
•Transfer of algorithms to the MIT/LL ORPG development arm will be straightforward, as anything within WDSS-II must also conform to CODE standards.
•• A system for developing new applicationsA system for developing new applications•• Common set of interfaces.Common set of interfaces.•• Applications differ only in scientific aspects.Applications differ only in scientific aspects.•• Algorithms can Algorithms can colloboratecolloborate with each other.with each other.
•• Integration of multiple sourcesIntegration of multiple sources•• Treat radars as a network Treat radars as a network –– not limited to singlenot limited to single--
radar applications.radar applications.•• Integrate lightning, surface observations, satellite, Integrate lightning, surface observations, satellite,
GIS, etcGIS, etc..
•• Support research and validationSupport research and validation•• Many tools and easy automation.Many tools and easy automation.
•• Open system: easy to add new data sources and Open system: easy to add new data sources and new applications.new applications.
WDSSWDSS--II Goals: ResearchII Goals: Research
WDSSWDSS--II CapabilitiesII Capabilities
•• RealReal--time Data Integrationtime Data IntegrationIntegrate radar data with other observing systems for total viewIntegrate radar data with other observing systems for total viewof the meteorologyof the meteorology
•• Multiple radar data streams (WSRMultiple radar data streams (WSR--88D, 88D, SpolSpol, SMART, SMART--R, R, •• Surface observations (Surface observations (METARsMETARs, , mesonetsmesonets))
•• Lightning (NLDN, LMA)Lightning (NLDN, LMA)•• Satellite (GOES)Satellite (GOES)•• ModelModel--generated output (RUC20)generated output (RUC20)•• Graphical weather products (watches, warnings, SIGMETS, Graphical weather products (watches, warnings, SIGMETS,
AIRMETS)AIRMETS)All data in a common coordinate systemAll data in a common coordinate system
•• EarthEarth-- and timeand time--centric, 4centric, 4--Dimensional coordinatesDimensional coordinates•• All data sources timeAll data sources time--synchronizedsynchronized
• Data from adjacent radar filled the cone-of-silence
• Complete multi-radar data used to compute VIL, for example.
Multiple RadarMultiple Radar
Radars other than WSR-88DRadars other than WSRRadars other than WSR--88D88D
KTLXKTLXTDWRTDWR
OKC TDWR dBZ overlaid on KTLX WSR-88D dBZOKC TDWR dBZ overlaid on KTLX WSR-88D dBZ• Terminal Doppler Weather Radar (TDWR)
• Research • Dopplers-On-
• Other Radars (GRIDS, commercial, ASR, PAR, etc).
• Terminal Doppler Weather Radar (TDWR)
• Research • Dopplers-On-
• Other Radars (GRIDS, commercial, ASR, PAR, etc).
• Terminal Doppler Weather Radar (TDWR)
• Research • Dopplers-On-
• Other Radars (commercial, ASR, PAR, etc).
• Terminal Doppler Weather Radar (TDWR)
• Research • Dopplers-On-
• Other Radars (commercial, ASR, PAR, etc).
KOUNKOUN
Radars other than WSR-88DRadars other than WSRRadars other than WSR--88D88D
Quality Control Neural Network Quality Control Neural Network (QCNN)(QCNN)
Original dBZeOriginal dBZe
Quality Control Neural Network Quality Control Neural Network (QCNN)(QCNN)
Radar Echo Classifier (REC)Radar Echo Classifier (REC)
Quality Control Neural Network Quality Control Neural Network (QCNN)(QCNN)
Quality Control Neural Network (QCNN)Quality Control Neural Network (QCNN)
Quality Control Neural Network Quality Control Neural Network (QCNN)(QCNN)
•• Two stages:Two stages:RadarRadar--only stage (uses texture features and a only stage (uses texture features and a neural network)neural network)Satellite and surf. Temp (optional)Satellite and surf. Temp (optional)
•• The radarThe radar--only stage alone outperforms existing only stage alone outperforms existing techniques.techniques.
.55.55.41.41.90.90QCNNQCNN
.35.35.37.37.44.44REC (88D)REC (88D)
CSICSIFARFARPODPODTechniqueTechnique
Motion EstimationMotion Estimation
Actual dBZeActual dBZe
• Sophisticated technique using statistical segmentation and error analysis.
• Can be used on dBZ, IR
• Produces high-resolution motion field that can be used to predict hail, precipitation, rotation, lightning, etc.
• Sophisticated technique using statistical segmentation and error analysis.
• Can be used on dBZ, IR
• Produces high-resolution motion field that can be used to predict hail, precipitation, rotation, lightning, etc.
Forecast dBZeForecast dBZe
Better Velocity Better Velocity DealiasingDealiasing
Hurricane Isabel: 9/18/2003
BeforeAfter ORPG
After NEW DA
Hurricane Isabel: 9/18/2003
PolarimetryPolarimetry
•• Volumetric extent of hail, freezing rain, Volumetric extent of hail, freezing rain, snow, and icing conditions, as well as nonsnow, and icing conditions, as well as non--hydrometeor hydrometeor scatterersscatterers
•• Enhanced data quality Enhanced data quality •• Problems associated with seaProblems associated with sea--clutter, ground clutter, ground
Conventional
Polarimetric
• 2003 Storm Intercept VerificationProvided more accurate verification data (hail/no hail) for 60 storms
• Hail Detection StatisticsConventional POD=88% FAR=39% CSI=0.56Polarimetric POD=94% FAR=8% CSI=0.86• Method of Detection
Conventional: Hail probability founded Polarimetric: Classification of hail
• Location of HailConventional: Probability of hail applies Polarimetric: Specific location of hail
PolarimetricPolarimetric Hail DetectionHail Detection
Freezing rainRain
Ice
• Detection of bright band and delineation of rain and snow• Freezing rain can be identified if polarimetric data are used in
Benefits of Dual Polarization Technology Benefits of Dual Polarization Technology Hydrometeor ClassificationHydrometeor Classification
Heavy Snow Event of 24Heavy Snow Event of 24--25 February 200325 February 2003URGENT - WINTER WEATHER MESSAGENATIONAL WEATHER SERVICE NORMAN OK555 PM CST MON FEB 24 2003
OKZ041-043-046>048-050>052-250550-ATOKA OK-BRYAN OK-CARTER OK-COAL OK-JOHNSTON OK-LOVE OK-MARSHALL OK- MURRAY OK-INCLUDING THE CITIES OF...ARDMORE AND DURANT
...HEAVY SNOW WARNING THIS EVENING...SNOWFALL...SOMETIMES HEAVY...IS EXPECTED TO ACCUMULATE 4 TO 6 INCHES BEFORE MIDNIGHT.PERIODS OF HEAVY SNOW WILL REDUCE VISIBILITY SIGNIFICANTLY. ACCUMULATIONS OF HEAVY SNOW IN THE WARNING AREA WILL CAUSE HAZARDOUS DRIVING CONDITIONS. USE EXTREME CAUTION IF TRAVEL PLANS CANNOT BE POSTPONED.
Successful Heavy Snow Warning –“Polarimetric Radar Data significantly increased forecast confidence and likely contributed to several hours additional lead time.”– Dan Miller, NWS Norman Forecaster
Benefits of Dual Polarization Benefits of Dual Polarization Technology Technology –– Data QualityData Quality
INSECTS
BIRDS
Light Rain 08/24/02
(0734 UTC)• Polarimetric classification
• Doppler wind measurement
MCS16 June 2002
• Radar reflectivity factor (Z) can be biased due to radar
• Biases can be addressed if
• Result: Improved echo classification
More Dual Polarization More Dual Polarization Technology Data Quality BenefitsTechnology Data Quality Benefits
ReflectivityReflectivity Correlation Coefficient (Correlation Coefficient (ρρHVHV))
CHAFF~0.2-0.5
CLUTTER~0.5-0.85
SNOW~0.85-1.00
Dual Polarization Data Quality BenefitsDual Polarization Data Quality Benefits
Hydrometeor classification algorithmHydrometeor classification algorithmReflectivity before quality controlReflectivity before quality control Reflectivity after quality controlReflectivity after quality control
Approximately 99% of echoes with SNR > 10 dB are correctlyIdentified as meteorological/non-meteorological by the
Hydrometeor Classification Algorithm.
Dual Polarization Data Quality BenefitsDual Polarization Data Quality Benefits
Potential Benefits to Numerical ModelingPotential Benefits to Numerical Modeling
DSD Retrievals
Model initialization
Phenomena classification
Dual Polarization: Current Dual Polarization: Current Results and Future PlansResults and Future Plans
•• Operational demo shows Operational demo shows polarimetricpolarimetric radar data can be a great benefit radar data can be a great benefit •• With the With the polarimetricpolarimetric hydrometeor classification algorithm, “Cleaned hydrometeor classification algorithm, “Cleaned •• Gaining better understanding of Gaining better understanding of polarimetricpolarimetric radar signatures in radar signatures in •• Light rain and light snow Light rain and light snow polarimetricpolarimetric characteristics heavily overlap. characteristics heavily overlap. •• Numerous Numerous polarimetricpolarimetric signatures of birds have been collected and are signatures of birds have been collected and are •• Still looking into microburst detection. Some promising early rStill looking into microburst detection. Some promising early results.esults.•• A LOT of data left to examine!A LOT of data left to examine!
Multi Radar CompositesMulti Radar Composites• The area for which any arbitrary ARTCC has responsibility
likely encompasses the coverage area of several WSR-88D installations.
• Neither the ROC nor the NWS has in place plans to treat the various WSR-88D installations as a single network, so there are no existing algorithms that use data from more than one radar.
• Serious limitation: treating each radar separately leads to ambiguities when the radar data overlap. Currently, the users must independently mitigate these ambiguities, which requires significant knowledge about meteorological radar data and the nature of the algorithms that are run on these data.
• Algorithms and techniques aimed specifically at multiple radar composites must be developed so that WSR-88Ds may be treated as a network.
Intensity Bias:Intensity Bias:Where Does the Damping Occur?Where Does the Damping Occur?
CREF-Raw CREF - Mosaic CREF - diff
Most damping occurs within 15km range of radar site
Storm Intensity ReductionStorm Intensity Reduction• Results:
The 3D mosaic produces smooth and physically realistic 3D reflectivity analysisNo significant reduction (3 dB or more) to storm intensity beyond 15km range>=5 dB damping only occurs within 5-10km of radar range
• A new composite reflectivity product derived from raw radar reflectivity (after QC) is generated to show full intensity of storms
• Difference field between raw composite refl and mosaic composite refl is
4D Dynamic Grid: No time weighting vs. 4D Dynamic Grid: No time weighting vs. exponential time weightingexponential time weighting
Note smoother reflectivity field with timeNote smoother reflectivity field with time--weightingweighting
No weighting Exponential weighting
3D National Mosaic Products3D National Mosaic Products•• Available in Available in NetCDFNetCDF formatformat
currently to CWPDTcurrently to CWPDTavailable to other available to other PDTsPDTs by requestby request
•• 3D reflectivity mosaic3D reflectivity mosaic
•• Composite ReflectivityComposite Reflectivity•• 2D reflectivity on 21 constant height levels2D reflectivity on 21 constant height levels
1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 6, 7, 8, 9, 10, 11, 12, 13, 11, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 4, 15, 16, and 17km MSLand 17km MSL
•• 2D reflectivity on 9 constant temperature levels2D reflectivity on 9 constant temperature levels20, 10, 0, 20, 10, 0, --10, 10, --20, 20, --30, 30, --40, 40, --50, and 50, and --60°C60°C
But Wait! There’s More…But Wait! There’s More…• CONUS domain
1km × 1km × 500m spatial resolution5 minute update rate (run time CPU < 150s)
• High Resolution National QPE ProductsPrecipitation Rate and TypeRain/Snow line delineation1, 3, 6, 12, 24, 72-h accumulations
• National Radar Calibration Monitoring
National Mosaic Domain National Mosaic Domain -- 19 Tiles19 Tiles
National Mosaic National Mosaic -- DomainDomain SpecificationsSpecifications
TileID
ctrlat(ºN)
ctrlon(ºW)
dx(ºlon)
dy(ºlat)
nx ny nz nradars
P1 45 127.5 0.01 0.01 501 1001 21 9N1 45 120 0.01 0.01 1001 1001 21 18N2 45 110 0.01 0.01 1001 1001 21 24N3 45 100 0.01 0.01 1001 1001 21 31N4 45 90 0.01 0.01 1001 1001 21 41N5 45 80 0.01 0.01 1001 1001 21 36N6 45 70 0.01 0.01 1001 1001 21 16A1 45 62.5 0.01 0.01 501 1001 21 2P2 35 127.5 0.01 0.01 501 1001 21 6S1 35 120 0.01 0.01 1001 1001 21 23S2 35 110 0.01 0.01 1001 1001 21 35S3 35 100 0.01 0.01 1001 1001 21 48S4 35 90 0.01 0.01 1001 1001 21 70S5 35 80 0.01 0.01 1001 1001 21 54A2 35 70 0.01 0.01 1001 1001 21 15G1 27.5 110 0.01 0.01 1001 501 21 11G2 25 100 0.01 0.01 1001 1001 21 16G3 25 90 0.01 0.01 1001 1001 21 30G3 25 80 0.01 0.01 1001 1001 21 21
dx≈ 0.715km-- 1.045km
dy ≈1.112km
Benchmark Test for National Benchmark Test for National ImplementationImplementation
• ComputerLINUX cluster of 10 nodesEach node: 2.8 GHZ CPU, 3GB RAM, 512MB Cache
• Grid -Tile S31001x1001x2147 radars
• Case“Worst Scenario” Simulation
• Mosaic Performance135.9s CPU694MB RAM
A fake case: 47 radars, wide spread precip
3D Mosaic Grid3D Mosaic Grid
Precipitation Rate and TypePrecipitation Rate and Type
Precipitation Rate Precipitation Type
Radar Calibration DiagnosticRadar Calibration Diagnostic
FY04 WorkFY04 Work
•• TDWR data study and integration into TDWR data study and integration into the 3D mosaicthe 3D mosaic
•• Radar data QC improvementsRadar data QC improvements
•• 3D mosaic grid support for other 3D mosaic grid support for other PDTsPDTs
•• Continued development of the 4D Continued development of the 4D dynamic griddynamic grid
The AWRT Takes Off!The AWRT Takes Off!
• The NEPDT has a broadened venue and is now the AWRT.
• New, distributed leadership across four major institutions.
• Bonds forged across several PDTs.• Plenty of work to do; lots of improvements to make.• Exciting times ahead!