Transcript IMPROVED SUPER-OB RADAR RADIAL WIND PRECISION FOR THE NCEP DATA ASSIMILATION

IMPROVED SUPER-OB RADAR
RADIAL WIND PRECISION FOR
THE NCEP DATA ASSIMILATION
AND FORECAST SYSTEM
Jordan C. Alpert, NOAA/NWS/EMC
Krishna V. Kumar, NOAA/NWS/NCO
Yukuan Song, NWS/OST/SEC
[email protected]
16th Conf. NWP J9.2 12/09//03
where the nation’s climate and weather services begin
NEXRAD WSR-88D RADARS
WSR-88D raw radar radial wind represent a
rich source of high resolution observations.
First task is obtaining quality returns from the radars at a
central site for model ingest (Alpert et al., 2003).
NEXRAD WSR-88D RADARS (continued)




158 operational NEXRAD Doppler radar systems
deployed throughout the United States.
These observations warn people about dangerous weather
and its location.
The doppler radars provide high resolution estimated radial
velocity data in addition to reflectivity.
The resolution of the radar radial wind is 1/4 km by 1
degree of azimuth with vertical structure information from
16 tilt angles.
Seven hundred million observations: 4 X 200 km X 150 radars X 360 deg
X 16 tilts per radar scan, every ~8 minutes or >0.1 T reports/day.
What is a Superob and why do we need it.
To improve our assimilation of observations for weather
forecast models, meso-scale (small scale) phenomena need
to be defined in the initial condition of the model.


Assimilation of Radar observations for initialization of
weather forecast models (Purser, et al., 2000) involve
processing each datum with repetitive interpolations from
the analysis grid to its location and back again.
The NCEP Operational Meso-Eta at 12 km obtains initial
conditions form the Eta Data Assimilation System, a 3dimensional variational analysis of conventional and nonconventional observations.
There is a significant degree of redundant information in
the radar radial wind observations.
What is a Superob and why do we need it. (continued)
Redundant observations impose a computation burden:


Computer resources and storage expended on mutually
redundant data could be better spent on improving other
aspects of an assimilation system (Purser, et al., 2000)
It is desirable to compress the data and use the highest
precision to minimize degradation of the information
content and minimize the band width to transmit the data
to a central site.
The term for a surrogate datum which replaces several
partially redundant actual data is a “super observation”
or “SUPEROB”.
Level 3 vs Level 2.5 super-ob radial wind data



Super-obs have been applied to operational analysis at
NCEP from the NEXRAD Information Dissemination
Service (NIDS), which we call Level 3 data having 4 bits
of precision for each radar radial wind return.
We designate the super-ob created by utilizing all the
precision at each radar as Level 2.5 and transmit the report
with 16 bits of precision according to the following table.
We compare these 2 data sets and a “No radar” case using
the Eta Data Assimilation System (EDAS) analysis and
Meso-Eta forecast model in operational configuration with
resolution of 12 km.
Adaptable Parameters for the Level 2.5
Superob Product
Parameter
Default
Range
Time Window
60 minutes
[5-90 min]
Cell Range Size
5 km
[1-10 km ]
Cell Azimuth Size
6 degrees
[2-12 deg]
Maximum Range 100 km
[60-230 km]
Minimum Number
of points required
50 unitless [20-200]
Experimental results of the Level 2.5 Super-ob




The EDAS is started 12-h before initial condition time on
28OCT2003, assimilating all conventional and nonconventional observations.
A 3-h window of radar radial wind observations was used
(1 ½ hours on either side of 12Z) to gather the radar winds.
A system for checking duplicate reports and quality control
in terms of a gross check based on the standard deviation
within the super-ob is in place.
The location of precipitation is shown from a 12-h forecast
past the initial time since radial wind returns are increased
in areas which have precipitation along the northern US.
Note the region of convection along the coast.
Eta 12-h forecast MSLP (hPa) and
Accumulated precipitation rate verifying
29OCT 2003 00Z.
Fig 2. Difference in the analysis (00) of wind
magnitude between the level 2.5 and level 3
super-ob are shown in the color fill contours
and the arrows show the vector difference of
the wind fields at 700 hPa
Fig 3. Difference in the analysis (00) of wind
magnitude between the level 3 and No radial
wind are shown in the color fill contours and
the arrows show the vector difference of the
wind fields at 700 hPa.
Fig 4. Difference in the Eta forecasts at 36-h of wind
magnitude between the level 2.5 and level 3 super-ob
are shown in the color fill contours and the arrows
show the vector difference of the wind fields at 700
hPa
Fig 5. Difference in the Eta foreasts at 36-h of
wind magnitude between the level 3 and No
radial wind are shown in the color fill contours
and the arrows show the vector difference of
the wind fields at 700 hPa.
There is only one direction where the regular wind will be identical to the
radial wind otherwise, we can not uniquely compose the radial winds into
regular wind vectors. However, the EDAS uses the information of how the
regular wind increments are related to the radial winds to make an optimal
decision of how the observed radial wind projects onto the regular winds.
For the analysis generated winds from the Radar Radial Wind reports, the
number of analysis wind increments greater (magnitude) than the conventional
analyzed wind for all locations.
Conventional
Wind Magnitude
2 m/s
3 m/s
4 m/s
5 m/s
6 m/s
7 m/s
8 m/s
9 m/s
10 m/s
11 m/s
Number of Level
2.5 Super-ob
1055320
230193
158347
116907
81912
54285
33957
21209
13307
8724
Number of Level
3.0 Super-ob
40352
157944
49268
22451
15160
9809
6141
3533
1953
985
Summary
The level 2.5 radar radial wind super-ob product, composed at each WSR88D
radar site, are received at NCEP and used in the EDAS analysis with other
remotely sensed and conventional observations, providing a high
resolution wind component.
In addition to experiments run with and without the presence of the radial
winds, we compare with the level 3 radar feed and no radar data present
for one period and a 36-h meso-Eta forecast.
The results show that the level 2.5 super-ob provides a larger number of
suitable super-obs compared with the level 3 data and provides high
resolution wind differences that are “drawn” for in the variational analysis.
Experiments to test accuracy in the high resolution forecasts including
verification to rain gauge observations are underway.
 Location of plots and pre-print.
http://wwwt.emc.ncep.noaa.gov/gmb/wd23ja/research/radar/
 http://wwwt.emc.ncep.noaa.gov/gmb/wd23ja/presentations