the use of wsr-88d radar data at ncep
DESCRIPTION
The use of WSR-88D radar data at NCEP. Shun Liu 1 David Parrish 2 , John Derber 2 , Geoff DiMego 2 , Wan-shu Wu 2 Matthew Pyle 2 , Brad Ferrier 1 1 IMSG/ National Centers of Environmental Prediction, Camp Springs, Maryland - PowerPoint PPT PresentationTRANSCRIPT
The use of WSR-88D radar data at NCEP
Shun Liu1
David Parrish2, John Derber2 , Geoff DiMego2 , Wan-shu Wu2
Matthew Pyle2 , Brad Ferrier1
1IMSG/ National Centers of Environmental Prediction, Camp Springs, Maryland2NOAA/National Centers of Environmental Prediction, Camp Springs, Maryland
OUTLINE
• Radar data processing at NCEP
• High resolution forecast initialization with radar data
WSR88D-Radar Data Processing at NCEP
Problems of radar data processing in operations:
1. the relatively large volume of radar data restricting the data to be transmitted to the operational center in real time
2. the radar data decoding software and storage taking excessive computational resources
3. the quality control (QC) problems of radar data further limiting the applications of radar data for operational use
WSR88D-Radar Data Processing at NCEP
Radar data received at NCEP:
1.Digital Precipitation Arrays (DPA)2.VAD wind (velocity azimuth display)3.WSR88D Level-III (NIDS) data4.WSR88D Level 2.5 data5.WSR88D Level-II data
Flow chart of level-II radar data processing at NCEP
Radar data QC at NCEP
QC Parameters
Mean reflectivity (MRF)
refNnrefMRF /)(
max/ NNVDC vr
bmvrpsc JjIjIVSC /])(/)([
Velocity data coverage (VDC)
Along-beam perturbation velocity sign changes (VSC)
Along-beam velocity sign changes(SN)Standard deviation of radial wind (STD)
Recorded QC parameters
0 400 800 1200 1600 2000 2400
02468
10
time
MR
F (d
BZ)
203040506070
VD
C (%
)
273033363942
VS
C (%
)
0 5 10 15 20 25 300
2000
4000
6000
8000
10000
12000
14000
16000
SN(%)
SN
KFWS 200909110605
KBUF 2009090513
23%
6%
Along beam velocity sign change (SN)
Threshold to reject data
Performance of radar data QC
Observation (m/s)
anal
ysis
(m/s
)
anal
ysis
(m/s
)
Observation (m/s)
before QC after QC
With QC
Goes image
Zoom-in area
HiRes Initialization with radar data• The radial wind is directly analyzed by GSI. • The cloud analysis package developed by GSD is
modified and used to analyze reflectivity with NCEP’s forecast model background.
• Hourly cycle is used
HiRes Initialization with radar data
• After complex radial wind quality control, level II radial winds are used in GSI analysis.
• 3D reflectivity mosaic from level II data are used in cloud analysis.
• Metar and Satellite observations are used in cloud analysis to detect cloud.
• Latent heat estimated from reflectivity is used to adjust background temperature.
• After radial wind and reflectivity assimilation, wind, temperature, rain water mixing ratio, cloud water and cloud ice mixing ratio and specific humidity are upgraded.
Test case on 2011091800
03
04
05
24
Forecast start at 00z
CTL: forecast withoutRadar data assimilation
EXP: forecast with radar data assimilation
Low level
divergence
High level
CREF obs
CREF ctl
EXP-CTLat the end of dataassimilation cyclewind divergence
conv
div
div
conv
psf (mb) uv (m/s) T (k) Rh (%)
bias rms bias rms bias rms bias rms
4h fcstctl 0.06 0.96 0.07 4.69 -0.28 1.92 2.43 10.85
exp 0.32 1.25 -0.06 4.55 -0.29 1.87 3.33 10.77
10h fcstctl 0.56 1.07 0.18 4.83 -0.06 1.55 2.39 15.62
exp 0.68 1.40 -0.05 4.91 -0.07 1.53 2.47 15.86
22h fcstctl 0.25 1.21 0.40 5.22 -0.35 1.56 0.63 16.19
exp 0.33 1.25 0.39 5.09 -0.29 1.55 0.47 15.06
Conventional data verification
Forecast hour
Domain average of absolute pressure change per 3h
03 06 09
12 15 18
CREF ETS score
10 day parallel run from 20110918 --- 20111013
ctlexp
33 3630
21 24 27
Vector windAnalysisIncrement
GSI analyzed wind increment
Most of analyzed wind increments are along the radial direction.
Challenges of radar data assimilation
do we get cross-beam wind information with current 3DVAR system?
Challenges of radar data assimilation
How to get balance between wind and other model variables?
1. Examine if DFI can distribute wind increment from Vr assimilation to other model variables. T increment through DFI is small. Q increment is relative large through DFI
2. Add new constrains in GSI to get balance between wind and other variables.
Future planContinue to test current hourly data assimilation
system for HiRes initialization. Will try to extend current 2 hour assimilation window to 6 hour or change assimilation interval to half hour.
Consider including diabatic digital filter treatment.
May consider to use radar data in hybrid ensemble data assimilation system.
Improvement of radar data quality control package at NCEP is constantly needed. We will need to process TDWR, Dual-pol and Canadian radar data in near future.