Long-Term Refractivity Comparisons:
Radar vs. Surface Station Measurements
1 Radar Refractivity
2 N Intercomparisons
3 Sources of Bias
4 Discussions
ShinJu Park
Toulouse, June. 16. 2004
5 Conclusions
1
Radar Refractivity
Fixed range: r
S-Pol radar
n
t  2r
c
Ground Targets
,
4fr
  2ft 
n (0  r )
c
2
Radar Refractivity
Refractivity (N) ≡
(n  1)  10
6
refractive index of medium
 Bean and Dutton (1968):
P
e
N  77.6  373000 2
T
T
P : Pressure in hPa
T : Temperature in Kelvin
e : Vapor pressure in hPa
3
Radar Refractivity
 Examples: IHOP_2002 (The Southern Great Plains, Oklahoma May 13~June 25)
Kansas
cold/ moist
Oklahoma
warm/dry
Texas
4
N Intercomparison (Ps, Ls)
Longitude (° )
Kansas
F3
Oklahoma
Latitude (° )
+
Texas
5
N intercomparison (Ps)
Day (UTC)
Day (UTC)
6
N intercomparison (Ls)
Day (UTC)
Day (UTC)
7
Where does the bias come from?
8
Sources of Bias from Radar
Fixed range r
Ground targets
S-Pol radar
 N varies rapidly with height;
dN/dz = f(de/dz, dT/dz) varies with time.
 N measured is a function of the height of targets
9
Sources of Bias from Radar
Longitude (° )
Kansas
Oklahoma
F3
Latitude (° )
+
Texas
10
N intercomparison (Ls)
Day (UTC)
Day (UTC)
11
N intercomparison (Ls)
Day (UTC)
Day (UTC)
12
Sources of Bias from Radar
1. Areal Smoothing
Mean bias over 44 days.
Sites
(km from the radar)
H
(15.8)
Ls
(29.4)
Ps
(9.0)
Rs
(38.9)
a. Point value
(N units)
2.34
4.85
1.47
-
b. 5 km x 5 km
(N units)
2.44
3.95
1.30
2.54
b-a
(N units)
0.10
-0.90 -0.17
-
 improves the quality of data observed by radar
-> reduce the bias locally but not significantly.
13
Sources of Bias from Radar
2. N at night
 Radar beam bends toward the ground at night or
when dN/dZ < -40 N units/km.
-> causes different N depending on the target heights at
the same distance from the radar .
Mean bias over 44 days.
Sites
(km from the radar)
H
(15.8)
Ls
(29.4)
Ps
(9.0)
Rs
(38.9)
Night (N units)
(02:00~13:00 UTC)
2.38
4.53
1.64
3.27
Day (N units)
(13:00~02:00 UTC)
2.37
3.58
0.81
1.89
Night - Day (N units)
0.01
0.95
0.83
1.38
14
 From the radar measurements,
we can not fully explain the sources of N bias
 Let’s look at SURFACE STATION measurements.
15
Sources of Bias from Surface Station
P
e
N  77.6  373000 2
T
T
16
Sources of Bias from Surface Station
 Selecting
98 radiosonde soundings at Homestead
*
Sounding data up to 2 m
above the ground
Sounding data between
2 m and 10 m
17
Sources of Bias from Surface Station
Days after a lot of
rain: 05/24, 05/26, 05/28,
05/29, 06/04, 06/05, 06/06,
06/14, 06/15, 06/16
18
Discussions
 Measurement Height
Radar : up to 10~20 m
depending on the height of the ground targets
Surface stations: 2 m
19
Courtesy of Peggy Lemone
Discussions
Courtesy of Peggy Lemone
 Expected N differences
between 2 m and 0-10
m, and between 2 m and
0-20 m based on MoninObukhov similarity
theory using flux data
from the western track.
 Only applied to the
clear days.
20
Discussions
 The larger the surface moisture, the larger the bias with height
differences.
 Still, uncertainty remains because of
1. poor sampling of data at the lowest level
from sounding analysis .
2. applicability of the flux analysis to other times.
21
Discussions
N units
N units
300.0
299.5
299.0
Mean refractivity
station
radar
soundings
298.5
298.0
297.5
297.0
296.5
296.0
295.5
295.0
294.5
294.0
Sounding level: ~2 m
22
2~10 m
10~20 m
Conclusions
 Overall, N agreed pretty well between station and radar
measurements
 However, N Biases between Radar and Stations exist, especially
on wetter days.
- Radar-related biases are relatively small;
- Bias is humidity dependent. Possibilities:
. Height difference? Theory is sound; does not explain bias
of low-level sounding data vs. stations.
. Other explanations imply that at least one of the sensors
has biases to be characterized.
23
Acknowledgements
 FRÉDÉRIC FABRY, McGill University
 Tammy Weckwerth, NCAR
 Crystal Pettet, NCAR
 Peggy Lemone, NCAR
Detailed description is available at
http://www.radar.mcgill.ca/~shinju/shinju_MSthesis.pdf
24
N intercomparison (Rs)
Day (UTC)
Day (UTC)
25
N intercomparison (Rs)
Day (UTC)
Day (UTC)
26
N intercomparison (H)
Day (UTC)
Day (UTC)
27
28
29
30
31
32
The influence of areal smoothing at each site. over 44 days.
Sites
Es
(km from the radar )
a. At
H
F1
F2
F3
Ls
Ps
Rs
Vs
(2.2)
(15.8)
(18.4)
(14.6)
(35.8)
(29.4)
(9.0)
(38.9)
(25.2)
2.7
0
2.34
2.26
1.51
2.89
4.8
5
1.47
-
1.17
2.7
3
2.44
3.17
1.79
2.34
3.9
5
1.30
2.54
1.85
0.0
3
0.10
0.91
0.28
0.55
-0.9
0.17
-
0.68
point
(N units)
b. 5 km x 5 km
(N units)
b-a
(N units)
The influence of propagation delay over 44 days.
Sites
(km from the radar)
Es
H
F1
F2
F3
Ls
Ps
Rs
Vs
(2.2)
(15.8)
(18.4)
(14.6)
(35.8)
(29.4)
(9.0)
(38.9)
(25.2)
Night (N units)
3.23
2.38
2.95
1.47
3.18
4.53
1.64
3.27
2.78
Day (N units)
2.35
2.37
3.24
1.85
2.13
3.58
0.81
1.89
1.31
Night - Day (N units)
0.88
0.01
-0.29
-0.38
1.05
0.95
0.83
1.38
1.47
(02:00~13:00 UTC)
(13:00~02:00 UTC)
33
Mean bias in refractivity between radar and stations over 44 days
Sites
At point
1-km x 1km
3-km x 3km
5-km x 5km
8-km x 8km
Es (2.2 km)
2.70
2.69
2.99
2.73
2.51
H (15.8 km)
2.34
2.38
2.39
2.44
2.27
F1 (18.4km)
2.26
2.52
3.09
3.17
3.17
F2 (14.6km)
1.51
1.93
1.99
1.79
1.54
F3 (35.8km)
2.89
2.79
2.56
2.34
2.10
Ls (29.4km)
4.85
4.62
4.13
3.95
3.59
Ps (9.0km)
1.47
1.57
1.20
1.30
1.50
Rs (38.9 km)
-
1.76
3.15
2.54
2.36
Vs (25.2 km)
2.17
2.16
2.10
1.85
1.69
Spatial mean
2.52
2.49
2.62
2.46
2.30
34
Standard deviation of mean bias in refractivity
between radar and stations over 44 days.
Sites
At point
1-km x 1km
3-km x 3km
5-km x 5km
8-km x 8km
Es (2.2 km)
2.26
2.35
2.45
2.41
2.37
H (15.8 km)
2.96
2.84
2.46
2.42
2.32
F1 (18.4km)
2.46
2.31
2.14
2.04
2.04
F2 (14.6km)
2.39
2.08
1.80
1.79
1.60
F3 (35.8km)
2.93
2.83
2.87
2.74
2.57
Ls (29.4km)
3.78
3.25
2.82
2.85
2.72
Ps (9.0km)
2.18
2.09
1.86
1.89
1.95
Rs (38.9 km)
-
2.87
3.52
3.78
3.54
Vs (25.2 km)
2.99
3.02
2.86
2.73
2.71
Spatial mean
2.74
2.63
2.53
2.52
2.42
35
Calibration : Reliable surface targets (e.g., power poles)
and “Reference Time”
The refractivity from the radar and surface stations
during the period of the 14 May Calibration (from 20:10 UTC to 20:40 UTC).
The dotted lines are for the radar retrievals and
the solid lines are for the station retrievals.
The numbers in the upper part of each plot are one hour averages.
36
37
38
The differences between radar refractivity and station refractivity
at the calibration times
14 May Calibration
(20:00z - 21:00z)
Sites
21 May Reference
(08:30z - 09:30z)
N station
N radar
NstnNradar
N station
N radar
NstnNrada
r
(2.2km)*
263.013
263.483
-0.470
290.925
289.029
1.896
Home(15.8km)
263.846
263.318
0.528
291.123
288.061
3.062
F1
(18.4km)
264.024
263.700
0.324
291.677
288.333
3.344
F2
(14.6km)
263.543
263.445
0.098
290.226
288.320
1.906
F3
(35.8km)
263.098
263.287
-0.189
289.961
287.843
2.118
Ls
(29.4km)
263.660
263.848
-0.188
291.153
286.429
4.724
Ps
(9.0km)
263.169
263.150
0.019
290.484
288.408
2.076
Rs
(38.9km)
263.061
-
-
290.492
288.390
2.102
Vs
(25.2km)
263.500
263.231
0.269
290.167
288.217
1.950
Es
39
Radar Refractivity
d (1 (r , t )  0 (r )) 4f

(n1 (r , t )  n0 )
dr
c
1: current time, 0: reference time
 Assumptions for the algorithm
1) All targets are perfectly aligned in a given azimuth
2) Targets are stationary
3) n0 is known
40
41