Transcript Weather Radar Data

Weather Radar Data
Doppler Spectral Moments
Reflectivity factor Z
Mean Velocity v
Spectrum width v
Polarimetric Variables
Differential Reflectivity ZDR
Specific Differential Phase
Correlation Coefficient hv
Linear Depolarization Ratio LDR
Contributors to Measurement Errors
*1) Widespread spatial distribution of scatterers
(range ambiguities)
*2) Large velocity distribution (velocity ambiguities)
3) Antenna sidelobes
4) Antenna motion
*5) Ground clutter (regular and anomalous
propagation)
*6) Non meteorological scatterers (birds, etc.)
*7) Finite dwell time
8) Receiver noise
*9) Radar calibration
*--- these can be somewhat mitigated
Mitigation of Range Ambiguities
Uniform PRTs
Alternate batches of long (for Z) and short
(for velocity) PRTS.
Long PRTs (first PPI scan) for reflectivity
ra>460 km;
Short PRTs (second PPI scan) for
velocity, ra <200km; typically 150 km
El = 19.5o
7 Scans
= 5.25
= 4.3
5 Scans
= 2.4
= 1.45
4 Scans
= 0.5
Reflectivity
Field of
Widespread
Showers
(Data
displayed to
460 km)
Velocity Field:
Widespread
Showers (5dB
overlaid
threshold;
data
displayed to
230 km)
Spectrum Width
Field:
Widespread
Showers (20 dB
threshold)
Echoes from Birds leaving a Roost; Spectrum Width
Field
Measurements of Rain
¶ R(Z) relations
¶ Error sources
¶ Procedure on the WSR-88D
Reflectivity Factor
Rainfall Rate Relations
Marshall-Palmer:
Z = 200 R1.6
Z(mm6 m-3); R(mm h-1)
For WSR-88D:
Z = 300 R1.4 - convective rain
Z = 200 R1.2 - tropical rain
Rain Rate Error Sources
*1) Radar calibration
2) Height of measurements
*3) Attenuation
4) Incomplete beam filling
*5) Evaporation
*6) Beam blockage
7) Gradients of rain rate
8) Vertical air motions
*9) Variability in DSD
DSDs, R(Z), and R(disdrometer)
Log(N)
Sep 11, 1999
Dec 3, 1999
Log(N)
DSD’s, R(Z), and R(disrometer)
Locations of Z Data used in the
WSR-88D for Rain Measurement
3.5°
HEIGHT
2.4°
0
1.5°
0.5°
135
20 35
RANGE
230 km
Applications of Polarization
♪
♪
♪
♪
Polarimetric Variables
Measurements of Rain
Measurements of Snow
Classification of Precipitation
Polarimetric Variables
Quantitative - Zh, ZDR, KDP
Qualitative - |hv(0)|, , LDR,  xv,  hv
Are not independent
Are related to precipitation parameters
Relations among hydrometeor
parameters allow retrieval of bulk
precipitation properties and amounts
Rainfall Relation R(KDP, ZDR)
R(KDP, ZDR) = 52 KDP0.96 ZDR-0.447
- is least sensitive to the variation of the
median drop diameter Do
- is valid for a 11 cm wavelength
Scatergrams:
R(Z) and
R(KDP, ZDR)
vs Rain
Gauge
Sensitivity to Hail
R(gauges)
R(gauges)-R(KDPZDR) R(gauges)-R(Z)
Area Mean
Rain Rate
and Bias
R(gauges)-R(radar)
Fundamental Problems in Remote
Sensing of Precipitation
♥Classification - what is where?
♥Quantification - what is the amount?
Weighting Functions
Partitions in the Zh, ZDR Space into
Regions of Hydrometeor Types
Weighting Function for
Moderate Rain WMR(Zh, ZDR)
Scores for hydrometeor classes
M
 AiW j ( Z ,Yi )
S j  i 1
M
 Ai
i 1
Ai = multiplicative factor 1
Wj = weighting function of two variables assigned to the class j
Yi = a variable other than reflectivity (T, ZDR, KDP,  hv, LDR)
j = hydrometeor class, one the following: light rain,
moderate rain, rain with large drops, rain/hail mixture,
small hail, dry snow, wet snow, horizontal crystals,
vertical crystals, other
Class j for which Sj is a maximum is chosen as the correct one
Florida
Florida
Florida
Florida
Florida
Fields of classified Hydrometeors - Florida
Fields of classified Hydrometeor - Florida
Fields of classified Hydrometeors - Florida
Suggestions
Data quality - develop acceptance tests
Anomalous Propagation - consider
“fuzzy logic” scheme
Classify precipitation into type (snow,
hail, graupel, rain, bright band) even if
only Z is available
Calibrate the radar (post operationally,
use data, gauges, ..anything)
Specific Differential Phase at short
wavelengths (3 and 5 cm)
• Overcomes the effects of attenuation
• Is more sensitive to rain rate
• Is influenced by resonant scattering
from large drops
Suggestions for Polarimetric
measurements at =3 and 5 cm
Develop a classification scheme
Develop a R(KDP, ZDR) or other
polarimetric relation to estimate rain
Correct Z for attenuation and ZDR for
differential attenuation (use DP)
Use KDP to calibrate Z
Radar Echo Classifier
•
•
•
•
Uses “fuzzy logic” technique
Base data Z, V, W used
Derived fields (“features”) are calculated
Weighting functions are applied to the
feature fields to create “interest” fields
• Interest fields are weighted and
summed
• Threshold applied, producing final
algorithm output
AP Detection Algorithm
• Features derived from base data are:
– Median radial velocity
– Standard deviation of radial velocity
– Median spectrum width
– “Texture” of the reflectivity
– Reflectivity variables “spin” and “sign”
• Similar to texture
• Computed over a local area
Investigate data “features”
• Feature
distributions
Clutter
mean V
Clutter
texture Z
Weather
mean V
Weather
texture Z
– AP Clutter
– Precipitation
• Best features have
good separation
between echo
types
AP Weighting Functions
1
a) Mean Radial Velocity
(MVE)
Median Radial Velocity
0
-50
-2.3
0
2.3
50
1
b) Mean Spectrum Width
(MSW)
Median Spectrum Width
0
-30
0
3.2
30
1
c) Texture of SNR
(TSNR)
“Texture” of Reflectivity
0
0
1000
45
1
d) Standard Deviation of
Radial Velocity (SDVE)
Standard Deviation of Radial Velocity
0
-30
1
30
0 0.7
1
e) Vertical Difference of
Reflectivity (GDZ)
F)
Spin
“Reflectivity
0
Spin”
0
-1000
-18
0
100
100
50
1
G) Sign
“Reflectivity Sign”
0
-10
-0.6
0
0.6
10
Field of Weights for AP Clutter
Weighting functions are
applied to the feature field
to create an “interest” field
AP Clutter
Clutter
Values scaled between 0-1
For median velocity field,
the weighting function is:
1
0
Interest Field
Radial
Velocity
Radial Velocity
-2.3
0
2.3
+3 m/s
m/s
Example of APDA
using S-Pol data
from STEPS
Polarimetric truth field
given by the
Particle Identification
(PID) output
Reflectivity
Radial Velocity
PID
APDA
APDA is thresholded
at 0.5
Good agreement
between PID clutter
and APDA
Clutter
Rain
20 June 2000, 0234 UTC
0.5 degree elevation
Storm-Scale Prediction
• Sample 4-hour forecast from the Center for Analysis and
Prediction of Storms’ Advanced Regional Prediction
System (ARPS) – a full-physics mesoscale prediction
system
• For the Fort Worth forecast
– 4-hour prediction
– 3 km grid resolution
– Model initial state included assimilation of
• WSR-88D reflectivity and radial velocity data
• Surface and upper-air data
• Satellite and wind profiler data
7 pm
8 pm
Forecast w/Radar
Radar
6 pm
2 hr
3 hr
4 hr
7 pm
8 pm
Fcst w/o Radar
Radar
6 pm
2 hr
3 hr
4 hr
R(Z) for Snow and Ice Water
Content
Snow fall rate:
Z(mm6m-3) =75R2 ; R in mm h-1 of water
Ice Water Content:
IWC(gr m-3)= 0.446 (m)KDP(deg km-1)/(1-Zv/Zh)
Vertical Cross Sections
Z
ZDR
KDP
hv
In Situ and Pol Measurements
T-28 aircraft