Evaluation of Dropsonde Humidity and Temperature Sensors using IHOP and DYCOMS-II data
Junhong (June) WangJunhong (June) WangHal Cole Hal Cole
NCAR/ATDNCAR/ATD
Acknowledgement: Kate Young, Dean Lauritsen, Terry Hock, and Krista Laursen (all ATD), Matthew Coleman (PennState U.)
Wang (2004, submitted to JTECH)
MotivationsMotivations
1. Under-utilization of dropsonde humidity data in Hurricane forecasting,
2. Dry biases in dropsonde data suggested by previous studies,
3. Comparisons of dropsonde and LASE data during IHOP,
4. More field projects used dropsonde data to map moisture and validate remote sensors,
5. Our experiences with radiosonde humidity data.
Data courtesy Sim Aberson, HRD
Thanks to Thanks to
James Franklin, James Franklin,
NOAA/AOML/NHCNOAA/AOML/NHC
CAMEX-3
CAMEX-4
From Kooi et al. (2002)
% MR difference between LASE and dropsondeHumidity dry bias Humidity dry bias
from pervious from pervious studiesstudies
From Vance et al. (2004)
RD93-TWC
RD93-RS90
~8%
• Lear dropsondes were in good agreement overall (<5%), but Falcon dropsondes were consistently drier by ~8%.
LASE-Dropsonde Comparisons (<75 km & <75 min)
Courtesy Ed Browell, NASA/LARC
+
1
2
3
4
5
6 8 10 12 14
100.21W, 36.56N
June 9, 2002, 12:51 UT
H2O DIAL
Dropsonde
H2O Mixing Ratio [g/kg]
Alt
itu
de
[km
]
1
2
3
4
5
2 4 6 8 10 12 14
102.71W, 36.55N
June 9, 2002, 13:17 UT
H2O DIAL
Dropsonde
H2O Mixing Ratio [g/kg]
Alt
itu
de
[km
]
DLR-DIAL Comparisons with Dropsondes
Courtesy Gehard Ehret (DLR)
Errors/Biases in Dropsonde Humidity DataErrors/Biases in Dropsonde Humidity Data
1. Contamination dry bias due to outgassing from the sensor packaging material, sensor bulk head, the outer tube and others,
2. Humidity time lag error,
3. Sensor wetting or icing.
Data from two field experimentsData from two field experiments
1.1. IHOP_2002 (SGP, May-June 2002): IHOP_2002 (SGP, May-June 2002):
• 71 pairs of co-incident dropsonde and 71 pairs of co-incident dropsonde and radiosonde soundings for intercomparisons,radiosonde soundings for intercomparisons,
• Comparisons of old and young sensors. Comparisons of old and young sensors.
2.2. DYCOMS-II (NE Pacific, July 2001): DYCOMS-II (NE Pacific, July 2001):
• All 63 dropsondes into marine stratocumulus All 63 dropsondes into marine stratocumulus clouds,clouds,
• Comparisons with co-incident airborne Comparisons with co-incident airborne ascending and descending data.ascending and descending data.
DYCOMS-IIDYCOMS-II
Comparisons with radiosonde data (IHOP)Comparisons with radiosonde data (IHOP)• Total 420 dropsondes from two aircrafts and for four types of missions
• Total 2879 radiosondes from 19 fixed stations and three mobile systems
• Total 158 pairs within 50 km and half hour, and 71 sampled the same air masses based on visual examination.
Norman 1996-2002
0
10
20
30
40
50
60
70
80
90
100
9601 9701 9801 9901 0001 0101 0201
Year/Month
Nu
mb
er o
f S
ou
nd
ing
s
Vaisala RS80-HVIZ-B VIZ-B2
ARM-B6 1996-2002
0
25
50
75
100
125
150
175
200
225
250
9601 9701 9801 9901 0001 0101 0201
Year/Month
Nu
mb
er o
f S
ou
nd
ing
s
Vaisala RS80-H Vaisala RS90
June 9, 18 UTCJune 9, 18 UTC
RHT Q
Mean Differences (Dropsonde-Radiosonde)Mean Differences (Dropsonde-Radiosonde)
RH T Q
Heat conduction Heat conduction to explain the cold biasto explain the cold bias
1. Inside
2. outside
3. reach equilibrium
4. in the flight
The bulk-head and sensor boom are warmer than the environment, so conduct heat to the sensors:
Tm > Ta and RH2 < RH1
Sensors come from colder to warmer air, so sensors lose heat to the BH/SB :
Tm < Ta and RH1-RH2
Colder dropsonde T than radiosonde in IHOP (~0.4C)
RH2
RH1
T
Ages of PTU Ages of PTU sensors for IHOPsensors for IHOP
IHOP Lear dropsonde PTU sensors (402)
48 49
8
297
0
50
100
150
200
250
300
350
1999 2000 2001 2002
Manufacture Year
Nu
mb
er o
f so
nd
es
IHOP Falcon dropsonde PTU sensors (91)
1 0 2
88
0
10
20
30
40
50
60
70
80
90
100
1999 2000 2001 2002
Manufacture Year
Nu
mb
er
of
so
nd
es
Sonde built dates:
Feb-Apr 2002
Comparisons of old Comparisons of old and new dropsondesand new dropsondes
<20 km, < 40 min<20 km, < 40 min
Performance in Clouds (Dycoms-II)Performance in Clouds (Dycoms-II)
Marine Stratus Cumulus clouds
Instrument Variables Range Precision Accuracy
Vaisala Dropsonde RD93:H-HUMICAP thin film capacitorBAROCAP silicon sensorTHERMOCAP capacitive beadCodeless GPS receiver GPS 121
RH
pressuretemperaturewind
0-100%
1080-3 hPa-90C to 60C0-200 m/s
1%
0.1 hPa0.1 C0.1 m/s
2%*
0.4 hPa*0.2 C*+0.5 m/s
NCAR Lyman-alpha hygrometer (“stub” and cross-flow)
mixing ratio 0.1-25 g/m3 0.2% 5%
GE 1011B Dew Point Hygrometer
dew point temperature
-65C to 50C 0.006C 0.5C (>0 C)1.0 C (<0 C)
Rosemount temperature sensor temperature -60C to 40C 0.006C 1.0C
PMS Liquid Water Sensor liquid water content
0-5 g/ m3 0.001 g/m3 0.02 g/m3
Specifications of different sensors Specifications of different sensors during DYCOMS-IIduring DYCOMS-II
Matching dropsonde with C-130 Matching dropsonde with C-130 ascending/descending profileascending/descending profile
DescendingAscending
Overshooting
Time-lag ErrorTime-lag Error
Mean estimated time constant of ~5 s is larger than 0.5 s given by the manufacture.
Sensor WettingSensor Wetting
Introduce alternative heating of twin humidity sensors to speed up the evaporation
Performance of the Temperature Sensor: Wetting Error?Performance of the Temperature Sensor: Wetting Error?
Wetting error in airborne in-situ T sensors (e.g. Eastin 2002): ~1-3C for Rosemount.
Summary on Dropsonde EvaluationSummary on Dropsonde Evaluation1.1. Dry Bias:Dry Bias: No systematic dry bias is found in dropsonde humidity data as
suggested by previous studies.
2.2. In Clouds:In Clouds: The maximum RH inside clouds does not show 100% all the time, but is within the sensor accuracy range (95-100%).
3.3. Time Lag Errors:Time Lag Errors: The dropsonde humidity sensor experienced large time-lag errors when it descended from a very dry environment above clouds into clouds. Mean estimated time-constant of the sensor is 5 s at 15C, which is much larger than 0.5 s at 20C given by the manufacture.
4.4. Sensor Wetting:Sensor Wetting: The dropsonde humidity sensor still reported near-saturation RH after it exited clouds because of water on the sensor. The alternative sensor heating for twin humidity sensor (not currently implemented) might help speeding up evaporation of the water.
5. Temperature: Another sensor wetting effect is on temperature data. The DYCOMS-II comparison show colder dropsonde temperatures inside and below clouds by 0.21C and 0.93 C, respectively. The IHOP data also show ~0.4 C colder dropsonde data, which might be due to the heat conduction between sensors and the bulk head and sensor boom.