You are on page 1of 129

WORL METEOROLOGICAL ORGANIZATION

INSTRUMENTS AND OBSERVING METHODS


REPORT No. 29

WMO INTERNATIONAL RADIOSONDE INTERCOMPARISON PHASE II, L985

by

f.J. SCHMIDLIN

1988

WMO/TD-No. 312

t
W0 R L D ME T E 0 R 0 L 0 G I C A L 0 RGANI Z AT I 0 N

INSTRUMENTS AND OBSERVING METHODS

REPORT No. 29

WMO INTERNATIONAL RADIOSONDE INTERCOMPARISON PHASE II, 1985


Wallops Island, Virginia USA

4 February 1985 - 15 March 1985

F.J. SCHMIDLIN

National Aeronautics and Space Administration


Goddard Space Flight Center
Wallops Flight Facility
Wallops Island, Virginia 23337

1988
FOREWORD

The WMO International Radiosonde Intercomparison 1984-85 was carried


out in two phases, Phase I in Beaufort Park, U.K., from 18 June to
27 July 1984 and Phase II in Wallops Island, Virginia, USA, from 4 February to
15 March 1985. The final report of Phase I, written by Mr. A.H. Hooper (U.K.),
was published in 1986 as No. 28 in the Instruments and Observing Methods
Report series and the final report of both phases, written by Dr. J. Nash
(U.K.) and Mr. F.J. Schmidlin (USA) in 1987 as No. 30 in the same series. The
present publication which has been prepared by Mr. F.J. Schmidlin concludes
the Intercomparison by giving a detailed description of the proceedings of
Phase II.

I am very pleased to have this opportunity to thank Mr. Schmidlin as


well as the management and staff members of the Wallops Flight Facility for
the excellent arrangements made for and the . conduct of Phase II and for the
laborious effort in preparing this Final Report. The meteorological community
has already found several conclusions of the WMO International Radiosonde
Intercomparison both useful and challenging.

(Seppo Houvila)
President of CIMO
Contents

~
Foreword

Contents

Table· Legends . . . . . . . . . . . . . . . . . . . . . . . V

Figure Legends . . . . . . . . . . . . . . . . . . . . . . . V~

1. Introduction, Purpose, and Overvie.w 1

1.1 Introduction . . . . . . . . . 1

1.2 Background, Requirements, and Pha.se II Plan . . . . . . . 2

1.3 Participating Radiosonde Instruments 3

1.4 Preparations for the Intercomparison


5
1.4.1 Host Organization . . . . . . . . . . . . . . . . .
5
1.4.2 Composition of Participating Teams . . . . . . . . 6

1. 4. 3 Responsibilities of Host and Participants . . . . • 6

1.5 Visitors . . . . . . .
6

2. Flight and Ground Equipment . . . . . . . . . . . . . . . 8

2.1 Australia (Philips RS4 MK3) . . . . . . . . . . . . . . 8

2.2 Finland (Vaisala RS80-15N) 9

2. 3 .India (Mk-III) 9

2. 4 United States (VIZ Model 1392) 9

2.5 Graw M6.0 10


2.6 Balloons and Ancillary Flight Train Equipment . . . . . . 10

2. 6.. 1 Ba.lloon Type and Performance. . . . . . . . . . . . 10

2.6.2 Radiosonde Flight Train . . . . . . . . . . . . . • 11


2. 7 Radar ll

2.8 Instrument Preparation and Release 12


2.8.1 Intercomparison Flight Schedules
12
2.8.2 Coordination Procedures . . . . . . . . . . .
13
- ~~~ -
2.9 Surface Weather Observation~ and Applications . . .
13

3. Data Acquisition and Quality Control . . . . . . . . • 16

3.1 Time Synchronization . . . . . .. . . . . . . . . . 16

3.2 Data Format . 16

3.3 Data Transfer 17

3.4 Data Verification and Editing 17

21
4. Discussion of Selected Observations

4.1 Sensor Characteristics of the Systems 21

4.2 Comparison of Selected Profiles . . . . . . . . . . . 22

5. Data Analysis . . . . . . . . . . . . . . . . . . . . . 33

5.1 Statistical Methods 34

5.2 Temperature 34
5.2.1 Standard Level Temperatures 34

5.2.2 Simultaneous Temperature Measurements 36

5.3 Relative Humidity . . . . , . 38

5.3.1 Standard Level Relative Humidity Measurements 38

5.3.2 Simultaneous Relative Humidity Measurements . 39

5.4 Geopotential 40
5.5 Pressure . . 41
5.5.1 Simultaneous Pressure . . . . . . . . . . . . .
42

6. Discussion of Radar Results 68

6.1 Geopotential 68
6.2 Wind . . . .
70

7. Specific Analysis . . . . . . . . . . . .. 81

7.1 Comparison of Phase II, Australian Data After Correction 82

7 .1.1 Pressure . . . . . . . . . . 82

7. 1. 2 Temperature . . . . . . . . . . . . . . . . . . . 82
iv
7.1.3 Geopotential . . . . .
!
83

7.2 Indian Radiosonde Reduction Technique . . . . . . 84

7. 2.1 Compa_rison of Indian MK III Radiosonde Data 84

8. Instrument Precision Estimates . . 96

8.1 Method/Rationale . . . . 96

8.2 Constant Pressure Levels 97


8.2.1 Temperature 97

8.2.2 Geopotential . 98

8.3 Instrument Precision 99

106
9. Conclusions . . . . . . . . . . . . . . . . . . . . . .

9.1 Temperature . . . 107

9. 2 Relative Humidity 107

9.3 Geopotential 108

9. 4 Pressure . . 108

9.5 Other Comments 108

10. References . . . . . . . . . . . . . . . . . . 110

Appendix

A •• 112
- V -

Table Legends

Table. 1.1 Resources Required to Obtain the Number of Matched


Pairs

Table 1.2 Phase II Radiosondes and Frequency Assignment

Table 2.1 Radiosonde Instruments and Sensors Used in Phase II

Table 2.2 Number of Instruments Reaching the Pressure Surface

Table 5.1 Pressure Layers for Comparison of Simultaneous Measure-


ments

Table 5.2 Daytime Humidity Errors Due to Thermal Lag and Insola-
tion
vi

Figure Legends

Figure 2.1 Lightweight styrofoam launch rig used in Phase II that


was designed to carry five radiosondes.

Figure 2.2 Typical Phase II configuration balloon, parachute,


radar reflector, launch rig, and radiosondes.

Figure 3.1 Data entry format for Standard Pressure Level data.

Figure 3.2 Data entry format for one-minute interval data.

Figure 4.1 Graphical comparison of Australia (AUS), US (USA),


Finland (Fil), and Indian (IND) relative humidity measurements at
one-minute intervals. Ascension no. 193 data shown (4 March 1985,
1411+18UT) .

Figure 4.2 Graphical comparison of Australia (AUS), US (USA),


Finland (Fil), and Indian (IND) temperature measurements at one-
minute intervals. Ascension no. 193 data shown (4 March 1985,
1411+18UT) .

Figure 4.3 Same as Fig. 4.1, except data are for Ascension no.
197 (5 March 1985, 1345+50UT).

Figure 4.4 Same as Fig. 4.2, except data are for Ascension no.
197 (5 March 1985, 1345+50UT).

Figure 4.5 Same as Fig. 4.1, except the Graw M60 radiosonde data
are included. Data are for Ascension no. 196 (4 March 1985,
2245+51UT) .

Figure 4.6 Same as Fig. 4.2, except India Mk III radiosonde data
are included. Data are for Ascension no. 196 (4 March 1985,
2245+51UT)>

Figure 5.1_a) Mean temperature differences in degrees C at con-


stant pressure levels for 1400UT observation set. b) Standard
deviation of the mean differences in degrees C. Numbers in
parenthesis on the right side of the figure represent the number
of observations.

Figure 5.2 Same as Fig. 5.1, except for 1700UT observation set.

Figure 5.3 Same as Fig. 5.1, except for 2000UT observation set.

Figure 5.4 Same as Fig. 5.1, except for 2300UT observation set.

Figure 5.5 a) Simulataneous comparison of mean temperature


differences in degrees C for the 1400UT observations. b) Stan-
dard deviation of the simultaneous mean differences in degrees C.
Numbers in Parenthesis on the right of the figure represent the
number of observations.
vii
Figure 5.6 Same as Fig. 5.5, except for 1700UT observation set.

Figure 5.7 Same as Fig. 5.5, except for 2000UT observation set.

Figure 5.8 Same as .Fig •. 5.5, except for 2300UT observation set.

Figure 5. 9 a) Mean relative humidi-ty differences in percent at


constant pressure levels for 1400 UT. b) Standard deviations of
the differences. Numbers in parenthesis on the right side of the
figure represent the number of observations.

Figure 5.10 Same as Fig. 5.9, except for 1700 UT.

Figure 5.11 Same as Fig. 5.9, except for 2000 UT.

Figure 5.12 Same as Fig. 5.9, except for 2300 UT.

Figure 5.13 a) Simultaneous comparison of mean relative humidity


differences in percent for the 1400UT observation set. b) Stan-
dard deviation of the simultaneous mean differences in percent.
Numbers in parenthesis on the right of the figure represent the
number of observations.

Figure 5.14 Same as Fig. 5.13, except for 1700UT observation set.

Figure 5.15 Same as Fig. 5.14, except for 2000UT observation set.

Figure 5.16 Same as Fig. 5.15, except for 2300UT observation set.

Figure 5.17 a) Mean geopotential differences in meters for con-


stant pressure levels for 1400UT. b) Standard deviation of mean
geopotential difference sin meters. Numbers in parenthesis on the
right of the figure represent the number of observations.

Figure 5.18 Same as Fig. 5.17, except for 1700UT observation set.

Figure 5.19 Same as Fig. 5.17, except of 2000UT observation set.

Figure 5.20 Same as Fig. 5.17, except for 2300UT observation set.

Figure 5.21 a) Simultaneous comparison of mean pressure differ-


ences in.hPa for 1400UT. b) Standard deviation of the simultane-
ous pressur~differences. Numbers in parenthesis on the right of
the figure represent the number of observations.

Figure 5.22 Same as Fig. 5.21, except for 1700UT observation set.

Figure 5.23 Same as Fig. 5.21, except for 2000UT observation set.

Figure 5.24 Same as Fig. 5.21, except for 2300UT observation set.

Figure 6.1 a) Simultaneous comparison of mean geopotential


differences in meters between radar and radiosondes for 1400UT.
b) Standard deviation of geopotential differences, in meters,
between radar and radiosondes. Numbers in parenthesis represent
the number of observations.

Figure 6.2 Same as Fig. 6.1, except for 1700UT observation set.
Sample size between radar and Graw M60 radiosonde is smaller.
viii
Figure 6.3 Same as Fig. 6 .1, except for 2000UT observation set.
Sample size between radar and Graw M60 radiosonde is smaller.

Figure 6.4 Same as Fig. 6.1, except for 2300UT observation set.
Sample size between radar and Graw M60 radiosonde is smaller.

Figure 6.5 Comparison of the US GMD derived winds with radar


derived winds for Ascension No. 166 made on 25 February 1985. The
GMD results are considerably noiser than the radar's..

Figure 6.6 Similar to Fig. 6.5, except for Ascension No. 181 made
on 1 March 1985. GMD results compare well with radar, however,
the radiosonde pressure cell was leaking resulting in radiosonde
heights higher than the radar's.

Figure 6.7 Comparison of Omega navaid derived winds with radar


for ascension no. 119 on 7 February 1985 during a very high wind
situation.

Figure 6.8 Similar to Fig. 6.7, except for Ascension No. 129 made
on 13 February 1985 during light wind conditions.

Figure 6.9 Comparison of winds from Finnish MicroCora and


DigiCora systems for Ascension No. 181 made on 1 March 1985.

Figure 7.1 a) Simultaneous comparison of mean pressure differ-


ences in hPa at 1700 UT between US (USA) and Australian
radiosondes before (AUS) and after (AUC) the pressure sensor
correction was applied.

Figure 7.2 Same as Figure 7.1, except for 2300UT observation set.

Figure 7.3 a) Mean temperature differences at the constant pres-


sure levels for 1700 UT between US (USA) and Australian
radiosondes before (AUS) and after (AUC) the pressure sensor
correction was applied to the Australian radiosonde. b) Standard
deviation of pressure sensor differences. Numbers in parenthesis
on the right of the figure represent the number of observations.

Figure 7.4 Same as Fig. 7.3, except for 2300UT observation set.

Figure 7 .. 5 a) Mean geopotential differences at the constant


pressure levels for 1700 UT between US (USA) and Australian
radiosondes before (AUS) and after (AUC) the pressure sensor
corrections were applied to the Australian radiosonde. b) Stan-
dard deviations of the geopotential differences. Numbers in
parenthesis on the right of the figure represent the number of
observations.

Figure 7.6 Same as Fig. 7.5, except for 2300UT observation set.

Figure 7.7 a) Mean temperature differences at the constant pres-


sure levels for 1700 UT betwen US (USA) and India. OLD indicates
the original Indian data reduced by Indian participants; IND
indicates the same observations after reprocessing by the Indian
participants. b) Standard deviations of the temperature differ-
ences. Numbers in parenthesis on the right of the figure
represent the number of observations.
Figure 7.8 Same as Fig. 7.7, except for 2300UT observation set.

Figure 7.9 Same as Fig. 7.7, except for mean geopotential differ-
ences.

Figure 7.10 Same as Fig. 7.8, except for.mean geopotential


differences.

Figure 8.1 Temperature sensor precision at reported constant


pressure levels for four radiosonde instruments. The reprocessed
Indian data (IND) are included.

Figure 8.2 Same as Fig. 8.1, except for geopotential heights.

Figure 8.3 Same as Figure 8.1, except for simultaneous measure-


ments of pressure.

Figure 8.4 Same as Fig. 8.3, except for temperature.

Figure 8.5 Geopotential precision of radiosondes and radar based


on simultaneous measurements.
WMO International Radiosonde Comparison, Phase II, Wallops Island, USA, 1985

Members of the International Organizing Committee and team leaders of the


participating countries Australia, Finland, India, UK and USA
1. Introduction, Purpose, and Overview

1.1 Introduction

Debates on radiosonde comparability have gone on for many years


within the World Meteorological Organization (WMO) and within various
national meteorological services. Radiosonde comparisons designed to
improve instrument comparability have been conducted a number of times.
For example: internationally, in Payerne, Switzerland, (Nyberg, 1952;
Delver, 1956); regionally, in preparation for the Alpine Experiment
(ALPEX) (Richner and Phillips, 1981); and nationally, in the United
States (Hoehne, 1980). Other radiosonde intercomparions and analyses
have been carried out seeking to derive temperature and geopotential
errors (Hooper, 1975; Finger and Mcinturff, 1978; Spackman, 1978;
Schmidlin et al, 1982). Each of these efforts provided new information
of some type, but none fully satisfied the search for comparability
among the major radiosonde instruments used globally.

The concept of radiosonde compatibility assumes, incorrectly,


that similar procedures and identical algorithms are used by all Members
making radiosonde observations. The WMO "green-stripe" report entitled
"Catalogue of Radiosondes in Use by Members, 1986" prepared by the Com-
mission on Instruments and Methods of observation (WMO Report No. 27,
1986) is quite useful in identifying the error of this assumption. This
report lists 40 different radiose>nde types available from 16 manufactur-
ers in use worldwide and notes that not all Members apply corrections to
the temperature measurements. Even within a single country, radiosondes
in use differ, or where they are the same, two and sometimes three dif-
ferent types of_radiosondes of the same manufacturer are in use. Furth-
ermore, while some radiosonde temperature measurements are corrected,
this same instrument when used in a different country may not have
corrections applied. Thus, the many different radiosondes in use, and
the different ways in which they are operated, make it difficult to
determine t~ue measurement compatibility.

In November 1982, plans were initiated within WMO for an inter-


comparison of radiosonde instruments. Because of the complexity and the
enormous burden of testing a large number of radiosonde types at a sin-
gle site, it was felt that two or more sites were necessary. The United
Kingdom (UK) Meteorological Office agreed to host and carry out the
first part, or Phase I of the International Radiosonde Intercomparison
during June and July 1984 at its test facility located in Beaufort Park,
UK (Hooper, 1986). In the United States (US) the National Aeronautics
and Space Administration (NASA) agreed to host and conduct Phase II of
the intercomparison during February and March 1985 at its Wallops Flight
Facility, located at Wallops Island, Virginia. The availability of pre-
cision tracking radars located at Wallops Island made the carrying out
of Phase II particularly desirable since precise height and wind infor-
mation could be obtained for comparison with radiosonde data.
page 2

1.2 Background, Requirements, and Phase II Plan

The present day upper-air data bas~ contains considerably more


information then was available just a few years ago. The larger data
base is necessary for the successful utilization of the newer, more
sophisticated numerical prediction systems, which are highly dependent
on timely, accurate, and well distributed data. Requirements for
upper-air data are expected to increase with emphasis on coverage for
mesoscale and even smaller meteorological events. ·

The fact that incompatibility exists between upper-air instru-


ments may not be of great surprise to some, since a given atmospheric
parameter may be measured in different ways by the various systems. For
example, the technology employed to obtain in situ measurements from
ascending balloon-borne radiosondes is different from that used for
remote techniques such as satellite instruments. Each of the techniques
(balloon-borne radiosonde, satellite instruments, rocket-borne devices,
aircraft, and ground-based remote methods such as lidar, etc.) has its
unique measurement problem. However, the one basic system in use for
over 50 years, and the one which has provided upper air information con-
sidered to be the standard by which the results of the other techniques
are judged, is the radiosonde.

A number of radiosonde types functioning on different principles


have been developed that inevitably have led to some of the inconsisten-
cies. Other shortcomings are functions of instrumental behavior and
possibly design limitations. Poorly, or ill-defined data reduction pro-
cedures, such as inadequate algorithms, can have a significant negative
impact on the quality of the reported data. These shortcomings, when
considered collectively, make it difficult to determine whether the
problems result from instrumental misbehavior or procedures.

The Commission for Instruments and Methods of Observation (CIMO)


Working Group on Upper Air Technology Basic to User Needs (F.G .. Finger,
Chairman) met in late 1982 and recommended an intercomparison of
radiosonde instruments. An organizing committee under the chairmanship
of Dr. S. Huovila, President of CIMO was given responsibility for
arranging the intercomparison. This committee met at WMO Headquarters,
in Geneva., Switzerland, from 28 November to 2 December 1983. The pro-
cedures anQ_ results of previous intercomparisons were reviewed. ~h&
committee agreed that the plans and procedures for carrying out the
intercomparison needed to be very carefully prepared. It was also
agreed that the radiosonde intercomparison should include all of the .
variables that are usually measured or calculated, i.e., temperature,
relative humidity, pressure, geopotential, and wind, as well as the com-
parability of data reported over the Global Telecommunications Network
(GTS) .

The organizing committee established guidelines and requirements


designed to insure a successful comparison. It was agreed that one site
would have difficulty in obtaining a satisfactory set of measurements if
too many radiosonde types participated. Thus, a second intercomparison
site was considered necessary. Plans were formulated to carry out the
intercomparison at two sites in two phases, with each phase scheduled
one year apart. Phase I was scheduled in the UK during June-July 1984
(Hooper, 1986) and Phase II in the US during February-March 1985.
Although atmospheric conditions most favorable for conducting
comparisons
page 3

generally occur during the summer, the second phase scheduled during the
winter permitted instrument comparability to be studied under different
atmospheric conditions, such as contine~tal air masses and frontal
weather events.

Phase II took place between 4 February and 15 March 1985.


National teams from Australia, Finland, India, and the United States
arrived at Wallops Island a number of days prior to the start of the
flight tests. Invitations also had been sent to Canada, the Federal
Republic of Germany (FRG), France, Japan, the Peoples Republic of China,
and the Union of Soviet Socialist Republics. These Members could not
participate because of organizational, financial, or other reasons. In
addition to the four Members participating, the UK offered to partici-
pate using the Graw M60 radiosonde. This instrument is used at the UK
overseas site located at Gibraltar. Since the Graw M60 instrument is
manufactured and used in the FRG, permission to include it in the inter-
comparison was obtained from the FRG meteorological authorities.

Requirements for a successful intercomparison called for the host


country to provide a number of facilities for use of the participants.
The participants also had to provide ground and ancillary equipment
unique to their radiosonde operation. The organizing committee agreed
further that a minimum of 72 balloon flights should reach the 50-hPa
pressure level within each phase for a successful intercomparison. To
insure that this requirement was achieved each participant was requested
to bring at least 100 instruments to Wallops.

The Phase II intercomparison plan identified the observation


schedule, flight restrictions, criteria for success, procedures for data
acquisition and flow, logistics, and quality control methods. The plan
listed the participants and the team leaders from each country, the host
site managers, the key personnel, and identified the responsibilities of
the host and participants.

This report is concerned only with results of Phase II of the


intercomparison and is based on data provided by the participants after
it was suitably quality checked by them. Objective examination of the
data records by the Project Leader and the data management team provided
a secondary quality check.

1.3 Participating Radiosonde Instruments

One concern of the organizing committee was the method by which


the many different instruments in use globally could be compared at the
same time. The International Organizing Committee recommended that only
widely used operational radiosondes be flown in the intercomparison.
The committee noted that many Members may use the same radiosonde type
and suggested that only the country where that particular radiosonde was
manufactured should participate in the intercomparison. Radiosondes
from 17 manufacturers were known to be in use at the time the intercom-
parison was being planned.

The offer by the UK and the US to each host one phase of the
intercomparison mitigated the burden that a single site would have in
accommodating all of the participants. In order that the success cri-
teria be met (i.e., that at least 72 instruments of each participant
should
page 4

reach the 50-hPa pressure level) a limit was imposed on the number of
participants that either-host could reasonably and successfully accommo-
date at their sites. The UK established.their limit at five partici-
pants (Hooper, 1986), while the US was willing to accept up to eight
participants.

Preliminary considerations indicated that, given the available


meteorolgical balloon capability, the weight of the instruments needing
to be lifted, and the potential confusion that might have been encoun-
tered from trying to receive signals from too many instruments at the
same time, only five instruments would be flown at the same time. A
scheme was devised to provide a cyclic rotation of instruments in the
event there were more than five participants at one site. For example,
given six participants the first balloon ascension would carry five
instruments, omitting the sixth. The second balloon flight would omit
the first instrument but would include the sixth. The second instrument
would be omitted from the third balloon flight, and so forth until the
desired number of pairs were obtained. Table 1.1 shows the required
number of instruments and balloons needing to be expended to accommodate
the number of participants shown in order to obtain the number of
matched pairs listed in the last column. Obviously, the large number of
additional flights required in order to obtain the desired 72 successful
flights would increase cost, time, manpower, etc., and would seriously
impact available resources. Since no more than five participants were
involved in either phase the task of designing the flight schedule was
greatly simplified. Table 1.1 is useful, however, for planning future
intercomparisons.

Table 1.1 Resources Required to Obtain the Number of Matched Pairs

Number of
Participants RS/Balloon Instr Balloons Matched Pairs

6 3 10 20 4
3 20 40 8
3 40 80 16
3 80 160 32
3 160 320 64
3 180 360 72

7 5 5 6 4
5 10 12 8
5 20 24 16
5 40 48 32
5 80 96 64
5 90 108 72

8 5 35 56 20
5 70 112 40
5 126 202 72

Finland originally proposed to bring the Vaisala RS-80 radiosonde


and the MicroCora Ground Station to Wallops Island. Finland also
brought a second RS-80 radiosonde design for use with the DigiCora
ground
page 5

station. Although the thermodynamic measurements from this second


radiosonde were not part.of Phase II, the organizing committee agreed to
include its wind measurements. India proposed to bring the India
Meteorological Department MK-III radiosonde; Australia the Philips-TMC
RS4 MK3; and, the United States the VIZ Model 1392 radiosonde. These
instruments all operate nominally at 403 MHz except the US VIZ instru-
ment which operates at 1680 MHz. Because of potential radio frequency
interference with the TIROS Operational Vertical Sounder(TOVS) Command
and Data Acquisition station located very near to Wallops Island, per-
mission to use a frequency of 1650 MHz was applied for and subsequently
used when the VIZ radiosonde was flown.

Table 1.2 Phase II Radiosondes and Frequency Assignment

ParticiEant R/S TyEe Nominal Freg. Phase II Freg.

Australia Phi lips RS4 MK3 403 403.5

Finland Vaisala RS 80-15N 403 406.6

India Indian MK III 403 398

United States VIZ Model 1392 1680 1650

Graw M60 Graw M60 28 28

Tests of radio frequency interference at 403 MHz were conducted


prior to the first ascension resulting in the establishment of the
assigned operational frequencies shown in Table 1.2. Even when the fre-
quencies were sufficiently separated there were occasions when interfer-
ence terminated data reception from one or more instruments.

1.4 Preparations for the Intercomparison

Preparation began months prior to the start of the actual inter-


comparison._Consideration needed to be given to obtaining and maintain-
ing a sufficient supply of inflation gas (helium was used), balloons,
radar reflectors, parachutes, and ancillary items necessary for carrying
out an uninterrupted succession of comparison flights. Arrangements had_
to be made for air space clearances, radio frequency clearance, and for
coordination with other US agencies that might influence the success of
Phase II. The organizational structure and roles of the host and parti-
cipant personnel had to be identified as did the responsibilities of
each.

1.4.1 Host Organization

The organization of the intercomparison was consistent with the


long-standing procedures established for the NASA Wallops Flight Facil-
ity rocket launch range and research site. This involved four organiza-
tional elements: scientific, operations, engineering, and logistics.
The scientific group consisted of the Project Leader, Data Manager, and
the
page 6

computer group for ensuring data flow, processing, and quality control.

The operations group consisted of the meteorological launch site


manager, radar personnel, balloon preparation and launch personnel, and
telemetry personnel. The engineering group consisted of support person-
nel to provide, among other things, special electrical service needed to
ope.rate equipment at 50 hertz (e.g., Australia's and India's ground sta-
tions) . And finally, the logistics group consisted of transportation
support, supplies, and domestic facilities.

1.4.2 Composition of Participating Teams

The participating teams consisted of a team leader and at least


three additional members. The recommendation that each team consist of
no fewer than four members was made by the Organizing Committee. The
availability of four members would permit the maximum number of
scheduled balloon flights to be made each day without diminishing the
effectiveness of the team. The team from the US National Weather Ser-
vice consisted of eight members, while Australia sent two members.
Australia, however, managed to supplement their team with two additional
observers recruited in the US.

1.4.3 Responsibilities of Host and Participants

The Project Leader was responsible for the management of the


intercomparison, the day-to-day decisions affecting the conduct of the
intercomparison after consulting with the team leaders, and for monitor-
ing progress and reporting results.

Generally, the host was responsible for: providing support to


insure successful launching of each balloon; providing transportation,
both locally and between the balloon launch site and lodging and/or res-
taurants; providing domestic facilities and supplies; scheduling air-
space and rf interference clearances; preparing the balloon and
radiosonde train for flight; releasing balloons and providing radar
tracking; providing technical assistance to the teams, as requested;
providing computer terminal interfaces between each participant's ground
station and_j::.he host's computer; arranging for "quick-look" data -fop
quality control by participants, e.g., plots, listings, etc.; and, pro-
viding final data listings to WMO and to each participant.

The participating teams were responsible for: setting up their


unique ground equipment; preparing their own radiosondes for flight;
reducing the measurements from each observation; preparing and providing
their data for entry into the host's computer, either by direct transfer
through the host-provided terminals and modems or by direct keying into
the computer; quality control of their data; and, preparation of their
equipment for shipment back to their respective countries.

1.5 Visitors

There were many visitors to Wallops Flight Facility during Phase


II, including the President of CIMO, Dr. S. Huovila, who was also Chair-
man of the Third Session of the International Organizing Committee held
at
page 7

Wallops Island from February 15 to February 19, 1985. Others included


the Project Leaders ~~om Phases I and II, Dr. J. Nash and Mr. F.J.
Schmidlin, respectively; the Chairman of the Working Group on Upper Air
Technology Basic to Users Needs, Mr. F.G~ Finger; and Dr. S. Klemm, WMO
Secretariat. A briefing was given by the team leaders and a discussion
held concerning the problems encountered in Phase II.

Personnel from the. US National Weather Service and from the US


Office of the Federal Coordinator for Meteorology, and others with a
scientific and technical interest in the work also visited the site dur-
ing the intercomparison to appraise the quality of the results being
achieved. Appendix A lists the names of the participants, key host per-
sonnel, and visitors.
page 8

2.0 Flight and Ground Equipment

The radiosondes flown in Phase II were operational instruments


manufactured and used in the countries participating. The exception to
this was the Graw M60 radiosonde provided by the UK and a second Finnish
radiosonde design, the Vaisala RS80-15P. The RS80-15N data were pro-
cessed with the MicroCora ground station and the RS80-15P data with the
DigiCora ground station. Only wind data from the DigiCora are included
in the intercomparison results. A list of radiosonde designs and sen-
sors is given in Table 2.1

Table 2.1 Radiosonde Instruments and Sensors Used in Phase II

Country Inst. Pres. Sens. Temp. Sens. RH Sens.

Australia Philips Aneroid Rod Thermistor Carbon


RS4 MK3 Mechanical (lead Carbonate)

Finland Vaisala Aneroid Thin-film Thin-film


RS 80 Capacitive Capacitive Capacitive

India Ind. Met. Aneroid Rod Thermistor. Lithium


Dept. Mechanical (titan. Diox.)

us VIZ 1392 Aneroid Rod Thermistor Carbon


Mechanical (lead Carbonate)

Graw Graw M60 Dbl Aneroid Bi-metal Hair


Mechanical

2.1 Australia (Philips RS4 MK3)

The Philips RS4 MK3 radiosonde was developed for routine use at
upper air stations of the Australian Bureau of Meteorology. The
radiosonde uses an aneroid cell pressure switch arrangement that con-
trols the sequential switching of the temperature and relative humidity
sensors, as well as reference resistors used for monitoring telemetry.
The baroswitch is made of Ni-span-C, with a measurement range of 1060 to
5 hPa and a specified accuracy of +/- 1 hPa rms. The temperature sensor
is the white-coated rod thermistor manufactured by VIZ Manufacturing
Company (the same rod thermistor is used with the US radiosonde) . The
measurement range is +50 C to -90 C with a stated accuracy of +/- 0.5 C
rms. The relative humidity sensor is identical to the hygristor used
with the US radiosonde and also is manufactured by VIZ Manufacturing
Company. This sensor is a carbon-coated polymer flat plate with a meas-
urement range of 20 to 100 percent and a stated accuracy of +/- 5 pe:.-
cent. The carbon hygristor is capable of measuring values lower than 20
percent,
page 9

but the range is limited by the transfer equation. The Australian sys-
tem required manual §election of constant pressure surface and one-
minute interval data from a strip chart ~ecording. Processing was via a
micro-computer.

2.2 Finland (Vaisala RS80-15N)

The Vaisala RS80-15N radiosonde uses capacitance type pressure,


temperature, and relative humidity sensors. The measuring range of the
pressure sensor is 1060 to 3 hPa with a 0.1 hPa resolution and a stated
accuracy of +/- 0.5 hPa. The temperature sensor measures between +60 C
and -90 C with an accuracy of +/- 0.2 C rms, and the humidity measure-
ment ranges between 10 to 100 percent with a +/- 2 percent rms accuracy.
The accuracies stated above were provided by the manufacturer. Radia-
tion corrections, applied to the temperature measurements, were identi-
cal to those applied during Phase I. The wind information depends on
the Omega navigation network. Wind direction and speed is calculated by
measuring phase differences of the signals from the eight Omega stations
(Lange, 1985). The calculated wind direction and speed is derived from
a four-minute integration of the radiosonde's position.

2.3 India (MK-III)

The Indian radiosonde is audio-modulated by varying the frequency


which is controlled by changes in the resistances of the meteorological
sensors. The radiosonde includes an aneroid cell constructed of Ni-
span-C, a rod thermistor, and a humidity sensor. All of the components
are manufactured by the India Meteorological Agency. The pressure meas-
urement range is stated as being from 1040 to 5 hPa, but instruments
flown at Wallops were calibrated only to pressures no higher than 1020
hPa. The reason for this was simply that the operational instruments
released in India do not require calibration at higher pressure.- How-
ever, difficulty was encountered at Wallops because of the much higher
atmospheric pressures present in mid-latitudes, especially when under
the influence of winter anticyclones. The rod thermistor is capable of
measurement between +50 C and -90 C, and the humidity sensor
measurementsranges between 10-90 percent. There were no accuracy state-
ments made available for the sensors of this instrument.

2.4 United States (VIZ Model 1392)

The VIZ Manufacturing Company model 1392 is the standard


radiosonde used operationally by the National Weather Service in the
United States. This instrument transmitted on 1680 MHz and was tracked
with a ground meteorological device (GMD-1b). The radiosonde uses a
Ni-span-C aneroid pressure cell and a temperature compensated baroswitch
linkage. The baroswith sequentially switches the meteorological sensors
as the radiosonde is ascending. The switching rate is controlled by the
rise rate of the balloon. Data were selected manually from a strip
chart recorder and entered into a microcomputer for processing. Azimuth
and elevation angles to the instrument were obtained by tracking the
1680 MHz transmitter with the GMD tracking system. Azimuth and eleva-
tion angles in conjunction with calculated altitudes were used to deter-
mine wind direction and speed at each minute of the flight.
page 10

2.5 Graw M60

This instrument has a morse code-sending transmitter operating on


a frequency of 28 MHz. The Graw M60 radiosonde uses an aneroid cell for
pressure measurements between 1050 and 5 hPa and a bimetal temperature
element designed to measure between +35 C and -70 C. For relative humi-
dity, a hair element is used which is designed to measure between 15 and
100 percent.

2.6 Balloons and Ancillary Flight Train Equipment

It originally was planned that a minimum of four instruments


would be flown on any given balloon and that five instruments probably
would be available for testing. This made it necessary to consider the
type and size balloon that would satisfy the release and flight require-
ments: i.e., nominal rise rate, a high burst altitude, relatively small
diameter to make handling easy at release. Additionally, it was desir-
able that the radiosondes should be flown at the same level relative to
each other. A carrier rig was designed that would enable this to be
accomplished. It was necessary, for safety reasons, that a lightweight
material be used that would be rigid enough to withstand bending moments
resulting from the stress of balloon release, and also capable of carry-
ing the weights of the combined instruments. The flight train also was
required to carry a radar target and because of the test site proximity
to populated areas a parachute was required on each flight in order to
inhibit the descent speed of the instrument package after balloon burst.

2.6.1 Balloon Type and Performance

The primary balloon adopted for use in Phase II was the 3000 gram
natural rubber balloon manufactured by Totex. Ninety-nine ascensions
were made with these balloons. Three balloons burst at or prior to
release, one leaked, and one iced after release. Ninety-four successful
balloon releases reached an average burst altitude of 32.4 km.
Seventy-one of these balloons reached the 10-hPa pressure level. This
was fewer than anticipated and rather disappointing, nevertheless, all
flights pid reach the 50 hPa level or higher except for the leaking and
icing balloons. Table 2.2 gives the number of instruments reaching tp~
pressure surface shown. From this table it is seen that a large number
of instruments were still operating at the 10 hPa level. Table 2.2 does
not specify whether the balloon or the radiosonde failed to reach the
indicated pressure, however, 77 percent of the balloons did reach the 10
hPa level.

A small number of 2400 gram Kaysam balloons (Type 130G) also were
released, however, this balloon: was large, was flaccid, and was quite
difficult to maneuver through the balloon shelter door during release.
This latter primarily was because of the over-inflation required to lift
five radiosondes. Nevertheless, the successfully released balloons per-
formed well, in fact, one of these balloons reached the highest altitude
in Phase II (37.6 km).
page 11

Table 2.2 Number of Instruments Reaching the Pressure Surface


Indicated,
Participant 50 hPa 30 hPa 20 hPa 10 hPa 7 hP a
No. % No. % No. % No. % No. %

Australia (99) 94 95 87 88 77 78 50 51 18 18

Finland(102) 100 98 92 90 84 82 60 59 48 47

India (94) 79 84 67 71 58 62 48 61 11 14

USA (lOO) 97 97 88 88 82 82 70 70 57 57

Graw M60(44) 40 91 37 84 33 75 24 55 20 45

2.6.2 Radiosonde Flight Train

Because the density of the instrument carrying rig had to be low


in the unlikely event of encounters with aircraft, various designs using
lightweight styrofoam were considered. The final design is shown in
Figure 2.1. The cross members were securely held in place by thin
polystyrene disks and the ends of each leg were tied together to prevent
excessive bending, a major consideration during release. This design
allowed four instruments (one tied to each leg) to be flown with a fifth
instrument suspended from the center. The flight configuration is shown
in Figure 2.2. The relative positions of the various flight train
members can be seen. The standard parachute used in the US decreased
the fall rate of the instruments significantly after balloon burst. The
parachute was located directly below the balloon so as not to interfere
with the airflow around the instruments.

2.7 Radar

The Wallops Flight Facility radar complex is composed of several


different radar installations and comprises a unique research capability
found at very few locations. During the intercomparison one radar ~a~
designated -for prime use for tracking every balloon released; a second
radar also was available during most of the test. The comparisons dis-
cussed in this report are with reference to the primary radar. The
prime tracking radar was the FPQ-6, a C-band monopulse system designed
for long-range tracking. This radar is capable of providing continuous,
accurate spherical coordinate information on targets to ranges of 60,000
kilometers. The FPQ-6 radar's beam width is 0.39 degrees and provides
range measurement precision of +/- 3 meters rms and angular measurement
precision of +/- 0.0028 degrees rms. The second radar used was the
FPS-16, another C-band system with a beamwidth of 1.23 degrees and angu-
lar precision of +/- 0.0056 degrees rms.

The precise nature of the tracking radars and the excellent com-
puter facilities at Wallops Island permitted detailed wind data to be
calculated for comparison with the wind information made available from
other tracking methods, e.g., GMD-lb and navaid techniques. The very
good
page 12

quality information on the heights of each balloon has been compared


with computed heights._.. Differences were examined to determine whether
the radar does indeed fulfill the need as a standard. The availability
of a second radar confirmed the measurements of the primary radar lead-
ing to conclusions that all of the radiosonde types had either a tem-
perature or pressure error which resulted in a bias when calculating
geopotentials.

Radar measurements provided angles and slant ranges from the


radar location to the target location as a function of time. All meas-
urements are referenced to the tangent plane through the radar axis
(i.e., relative to a flat earth). These data must be converted to a
spherical earth and the resulting computations corrected for refraction
through the atmosphere. Finally, the height information (i.e., English
Engineering Units) was put into the same units as height information
provided by the radiosonde computations (meter-kilogram-sec units) .

After the radar's elevation angles were corrected for the earth's
curvature and refractive index the east-west (x), north-south (y), and
altitude (z) position components were calculated. The altitude (z)
information was converted to geopotential altitudes (h). Complete pro-
files of time vs wind and time vs geopotential altitude are available.

2.8 Instrument Preparation and Release

Each radiosonde was prepared for flight by the responsible parti-


cipant. Balloons were inflated and prepared for flight and the instru-
ment carying rig assembled by the host site support team. This multi-
faceted effort required extremely close working relationships among all
of the participants and host members.

2.8.1 Intercomparison Flight schedules

The ·observation schedule was designed to obtain the maximum


number of comparisons during daylight. This plan resulted in flights
being sch~duled at nominal local times of 0900, 1200, 1500, and 1800.
The 1800 local time observation corresponded to the routine upper-~iL
observationtime for the US East Coast radiosonde sites and was carried
out in darkness. During the last 8-10 days of the intercomparison the
sun's elevation angle at 1800 UT was approximately 3 degrees above the
horizon. Although this small amount of solar radiation was bothersome
it only affected the initial minutes of the flights, the upper level
measurements, i.e., above about 10 km, were still obtained in darkness.

The scheduled launch times permitted data to be obtained at three


solar angles during daytime and in darkness. This permitted some infor-
mation to be obtained on measurement differences in the presence of
solar radiation and should result in the development of instrument
adjustments. Additionally, NOAA-7 and NOAA-9 passed over or near Wal-
lops Island twice a day; these satellites should provide independent
temperature profiles for additional analysis.
page 13

2.8.2 Coordination Procedures

Because four or five radiosondes were scheduled to be launched on


the same balloon, numerous tasks needed to be accomplished in an effi-
cient and timely manner. The Meteorological Requirements Coordinator's
responsibility was to ensure that the balloons were properly inflated,
to see to it that the length of the flight train was kept standard with
a parachute and radar reflector located the same distance from the bal-
loon for each flight, and that the flight trains were prepared early
enough to allow each participant's radiosonde to be attached to the
instrument carrying rig.. Monitoring of the radiosondes' transmitter
frequencies prior to flight to ensure that they did not interfere with
each other was also the responsibility of the host site. Additionally,
surface weather observations were provided prior to balloon release.
Technical assistance to the participating teams was provided, when
requested.

Coordination of the personnel, and activation of the equipment at


the moment of balloon release was singularly important. Each ground
recorder was given a time pulse at release and at one-minute intervals
thereafter. The interval clock, although electronically controlled and
sequenced with Wallops Island's master clock, was started manually at
balloon release, and each successive time pulse was controlled electron-
ically. Two releases with timing offsets of one minute and four minutes
each were experienced that were traced to mechanical problems in the
start switch. Data records for these two flights were adjusted for the
time fault.

2.9 Surface Weather Observations and Applications

Since each team would follow somewhat different procedures and


would require more or less time in preparing their instruments a series
of surface weather observations beginning twenty minutes prior to the
scheduled time of balloon release was made. Observations were made at
five-minute intervals through release time. When the balloon release
was delayed, surface observations continued at five-minute intervals.
The surface observations provided the basic information needed for set-
ting the_ radiosondes' pressure sensors and for initializing the observa-
tions. The frequency of the observations permitted continuity of th&
observations such that errors were reduced.
page 14

Figure 2.1 Lightweight styrofoam launch rig used in Phase II that was
designed to carry five radiosondes.
page 15

Figure 2. 2 Typical ·phase II configuration b.al.lo:on, parachute, radar


re.flect.or, launch rig., and radiosondes.
page 16

3.0 Data Acquisition and Quality Control

The efficient acquisition of data was of considerable importance


to the Phase II effort. The participants received, recorded, and
reduced their own radiosonde data. Depending on the type of ground sta-
tion in use, each participant insured that its data were received with
the least possible interference. They also took care that the one-
minute interval and constant pressure surface data were properly
reduced, and that the data were well checked prior to delivering it to
the NASA computing facility for comparison processing. The Graw M60
radiosonde data were received at NASA's telemetry station and the
morse-coded data recorded. Recordings of the data (on cassette tape)
were decoded, reduced, and analyzed by the UK Meteorological Office.
Data were returned to Wallops on digital magnetic tape in the recom-
mended format (Figures 3.1 and 3.2).

3.1 Time Synchronization

The elimination of measurement bias between two or more


radiosondes is highly dependent upon recording the received signals of
the radiosondes at the same exact time. A timing system was fabricated
that transmitted a de pulse to each participant at the moment of balloon
release and at each minute of the observation. The participants were
responsible for starting their own recorders when the balloon release
signal was received. It is possible that this method could have time
offsets between each radiosonde of 1-2 seconds due to different r~action
times of the operators. In general, corrections for time offsets of
this type were minimal and when corrections were necessary, they were
derived by examination of the subsequent one-minute time pulses to
ensure time agreement throughout the flight. Strip-chart recorder times
have a t~ndency to disagree, probably due to varying power supply fre-
quency and drag inherent in the mechanical drive systems. Time corre~a~
tions are important when it is considered that given a nominal ball~on
rise rate with the usual tropospheric temperature lapse rate a two-
second offset in time could create approximately a 0.1-0.2 C bias error
in temperature.

3.2 Data Format

Standardized formats were designed to ensure uniformity of the


data sets for analysis and archiving. Separate formats were designed
for standard pressure level and for one-minute interval entries. Both
formats are shown in Figures 3.1 and 3.2.

The intention of the standard forms simply was to prevent having


to use data entries requiring different computer read-in routines. Each
team provided tabulations of their data either by manually completing
and submitting the forms provided, or by submitting their computer
printout
page 17

of the data. These records served as a "first look" at the data and
provided back-up in_ the.. event direct transfer of data between computers
(see Section 3.3) could not be achieved. A more efficient computer-to-
computer data transfer method was used whlch reduced the manual handling
of data records to a minimum. Only the Indian data were handled manu-
ally since their mini-computer would not successfully interface with the
Wallops system. After successful data entry into the Wallops computer
was achieved, quality checking and pre-editing of formats were carried
out to enhance record storage, future data transfers, and analysis.

It was planned that the standard pressure level data be


transferred to the NASA computer not later then 30 minutes following
balloon burst. The one-minute interval data were to be entered into the
computer within one and one-half hours following balloon burst. Gen-
erally, the participants delivered reduced data to the Wallops computer
facility within a few minutes of balloon burst because of the efficiency
of the computer-to-computer data transfer. In order to comply with the
Organizing Committee's requirement that each participant provide prop-
erly completed one-minute and standard pressure level data, the Indian
participants withheld final delivery of data until after the completion
of Phase II.

3.3 Data Transfer

The procedure adopted for the transfer of data was direct


transfer from the participants' mini-computers to the NASA/Wallops
Flight Facility's research computer system Eclipse 330. Dedicated tele-
phone lines were used between computers and each participant was
assigned a password and data transfer codes. This method eliminated the
need for manually completing the data input forms and subsequent keying
of data into the computer. Thus, two possibilities of making transcrip-
tion errors were removed, i.e., transcribing the data onto the forms and
key-in errors. The computer-to-computer transfer of data also included
the capability for each participant to retrieve their data files at any
time in order to make corrections, if necessary.

3. 4 Da-ta Verification and Editing

The results of each flight were made available in standard pres-


sure level, one-minute interval, and in TEMP message formats. As a
method of determining whether gross problems existed with any of the_
reduced radiosonde data, all flights were plotted on Skew T Log P charts
along with the routine radiosonde message from Washington's Dulles
International Airport (located about 120 kilometers west of Wallops
Island). The TEMP message's were prepared manually and given to a
member of the NASA meteorological team for plotting. The exception was
the Finnish team submitted computer produced TEMP messages. The
"quick-look" analysis provided by the Skew T Log P charts was valuable
in detecting gross errors, which subsequently were corrected.

Machine produced plots of all of the radiosonde data profiles


were available within one to two days following each flight. These
plots provided a secondary quality control mechanism for the partici-
pants. Mistakes found by the teams were corrected immediately.
page 18

Final quality control was carried out by the NASA Data Management
Team personnel using objective methods designed for the intercomparison.
All mistakes were not necessarily uncovered. During the data analysis
phase, however, unusually large variances' were observed at some pressure
levels which highlighted previously undetected errors. These were
either removed as being outlandish and uncorrectable, or were corrected
by referring to the original data recordings. The latter was possible
using the manually reduced strip charts maintained in a file at Wallops
Flight Facility.
page 19

WM0
INTERNATIONAL RADIOSONDE
COMPARISON - 1985
GOODARD SPACE fLIG«T CENTER/WALLOrS fLIG«T fACILITY

CONTROL SCREEN DATA SURFACE DATA

f::?~.J~:::. .
1

STANDARD LEVELS
9

m!. 1~

1 .....
~~ ~

e:;~·-· 8P: . m!.~: m~~


4~ ~~ 6~ 2;

NOTES:
t. FILL IN UNSHADED RREAS ONLY. KEEP WITHIN BOXES.
2. ENTER FIELDS RIGHT JUSTIFIED.
3. If DRTA NIS31No ENTER X IN FIELD. RIOHT JUSTIFIED.
4. If NO HORE DATA fOR AN ELEMENT. LEAVE REHAININO ENTRIES BLANK.

Figure 3.1 Data entry format for Standard Pressure Level data.
page 20

WM 0
INTERNRTIONRL RADIOSONDE
COMPARISON - 1985
COODRRO SPAC.E fLIGHT CENTER/WALLOPS FLIGHT FACILITY

MINUTE DATA

NOTES:
1. FILL IN UNSHAOEO AREAS oNLY, KEEP WITHIN BOXES.
2. ENTER fiELDS RIGHT JUSTIFIED.
3. IF DATA MISSINO, REPLACE FIELD BY X, RIGHT JUSTifiED·
4. If NO MORE DATA FOR RN ELEMENT, LEAVE REMAINING ENTRIES BLANK.
6. TERMINATE DATA BY ENTERING 999 IN MM fiELD.

Figure 3.2 Data entry format for one-minute interval data.

·'
page 21

4.0 Discussion of Selected Observations

The question of whether radiosonde observations provide a true


representation of the atmospheric parameters being measured is an impor-
tant one. All radiosonde systems, however, do not perform in a manner
which provide similar information. Selected intercomparison observa-
tions show that the five instruments, when intercompared simultaneously
in time and space, gave different values of similar atmospheric
features. Inversions and humidity gradients are used as examples in the
following discussion.

4.1 Sensor Characteristics of the Systems

The temperature and relative humidity sensors of the Australian


and US radiosondes were identical and, in fact, were manufactured by the
same company. The Indian radiosonde's sensors, fabricated by the India
Meteorological Department, look similar to the Australian and US sen-
sors. The Indian rod thermistor is coated with titanium dioxide, while
lead carbonate is used to coat the Australian and US thermistors. While
looking similar, the lead carbonate coating responds differently to
long- and short-wave radiation than does the titanium dioxide coating.
The rod thermistors are the resistive type having a range of resistance
between approximately 5000 to 1,500,000 ohms with a negative temperatqre
coefficient, i.e., as the temperature decreases the resistance
increases. The Graw M-60 instrument uses a resistive type temperature
sensor but its measurement range of resistance is considerably less than
that of the rod thermistors, being approximately 5000 ohms full scale.

The Indian relative humidity element is coated with lithium


cloride as compared with the carbon coatings of the Australian and US
humidity sensor. These sensors are the resistive type and appear simi-
lar, but their responses are different. The difference between the
lithium cloride-coated and the carbon-coated hygristors, other than in
performance, is the method in which the resistance changes with moisture
change. The carbon-coated element's resistance increases as the mois-
ture increases while the lithium cloride-coated element's resistance
increases with decreasing moisture. The Graw M60 uses human hair for
the humidity sensing element. Although the hair hygristor has been
around for a long time it has certain drawbacks. At cold temperatures
this sensor's lag can be 3-4 times longer than when it is exposed at
warm temperatures, and may even be infinite at temperatures of -40 C and
below. Hair elements also are subject to large errors, especially if
not stored in a clean environment, and have been known to change cali-
bration if stored under extremely dry or wet conditions.

The temperature and humidity sensors of the Finnish radiosonde


are of the capacitive type and are smaller than the rod thermistor and
hygristor. Nonetheless, these sensors respond to temperature and humi-
dity changes at about the same rate as the rod thermistors and the
carbon
page 22

hygristors. The thermistor is coated with aluminum to aid in reflecting


solar radiation and should have a smaller emissivity than the white rod
thermistors, thus a ·smaller infrared effect. Finland's relative humi-
dity sensor is a capacitive thin-film sensor shielded with an aluminum-
coated cap to ensure uncontaminated performance in the presence of pre-
cipitation and heavy moisture-laden clouds.

The carbon-film hygristor's time constant is extremely fast, how-


ever, because of its heavy plastic substrate, the thermal lag of this
sensor is quite long (up to 15 seconds at sea level) . Although the
humidity sensing capability of this sensor is excellent, the ambient
measured temperature is used in the transfer equation to determine rela-
tive humidity, rather than the actual substrate temperature. A similar
temperature lag problem exists with the Indian lithium cloride sensor.
The capacitive thin-film sensor used with the Finnish radiosonde is
quite small, nevertheless 1 its time constant is similar to the carbon-
coated sensor (WMO Guide to Meteorological Instruments and Methods of
Observation, 1983, see Table 13.4).

The third measurement of concern is pressure. The Australian,


US, and Indian radiosondes use an aneroid cell-baroswitch arrangement.
The baroswitch performs two functions: (1) it provides an indication of
pressure during the flight and, (2) it switches the temperature, rela-
tive humidity, and reference sensors alternately into the circuit for
periodic transmission. The design of the baroswitch may or may not be
affected by temperature changes and could require temperature compensa-
tion. The baroswitch used with the US instrument employs a linkage
design that is self-compensating. The baroswitch used with the Aus-
tralian radiosonde is designed in a similar manner, however, during
Phase II it was discovered that its measured pressures were considerably
lower than those of the other instruments. The method derived by Aus-
tralia to correct these pressures is discussed in Section 7.0. The
Indian baroswitch has temperature compensation applied during the obser-
vation and is based on the ambient temperature profile. The ~ressure
measurements of the Graw M-60 are based on an aneroid cell, however, a
clock mechanism rotates a series of electric contacts into position
against the aneroid contact arm to effect the transmission of pressure.
The temperature and humidity data are switched in a similar manner.
Values are transmitted using morse code. The Finnish pressure sensor is
made of encapsulated capacitance plates which move as the pres~ur~
changes, thi-s arrangement permits continuous pressure measurement. Tem-
perature and relative humidity data transmission are independent of the
pressure sensor and are electronically switched.

From the above it should be obvious that all of the radiosondes


operate on different principles and that measurement differences should
be expected. Whenever data losses occurred during Phase II they were
primarily the result of rf signal loses and none were due to sensor mal-
functions.

4 .2 Comparison of Selected' Profi.les

Although a reasonably extensive discussion of the data is given


late.r in this report it. may be instructive- at this point to examine the
measurement profi.les in some detail. Simultaneously measured. data only
are. considered which represents a true comparis;on of sensors. Copies of
the
page 23

original records, retained at Wallops Island, permitted us to recon-


struct measurement profiles at 10-second and 30-second time intervals.
The examples selected were mostly associated with frontal passages.
Generally, the examples correspond to the occurrence of rain or drizzle
at the times of the flights. These examples describe differences
between the measurements made with the three types of hygristors (flat
plate, thin-film, and hair) and the different thermistors.

Figure 4.1 depicts relative humidity measurements from ascension


number 193 released on 4 March 1985 at 1411+18 UT (times are in hours,
minutes, and seconds) . Rain and drizzle were occurring during this
flight with an overcast of stratocumulus. Measurements are presented at
10-second intervals for the first six minutes and at 30-second intervals
thereafter for the Australian, Finnish, and US instruments and at one-
minute intervals for the Indian instrument. Surface relative humidity
was 88 percent, however, 10 seconds after balloon release the Aus-
tralian, Finnish, and US instruments were observing different values,
e.g., Finnish values decreased while Australian and US values increased
slightly. Nonetheless, between minutes 2 and 10, these sensors report
within 2-4 percent of each other. At 9 1/2 minutes after release a
rapid decrease in the amount of moisture present occurred, characteris-
tic of conditions existing above frontal clouds. The lowest humidity
recorded in the clear air was observed at 13 percent with the carbon-
coated hygristors, while the capacitive thin-film sensor observed 23
percent. The lithium cloride-coated sensor used by India was observed
to have a sluggish response and poor agreement with the other sensors.
It is not known whether this is from a long time constant, the type of
coating employed, or from some other reason. Figure 4.1 also shows that
the maximum humidity value given by all of the hygristors was approxi-
mately 96 percent, whereas India's gaves a maximum reading of 84 per-
cent. At minute 14 and above, tne Indian sensor's measurement is 30~40
percent lower than the other sensors' measurements and apparently can no
longer follow detailed structure. In this same region the capacitive
thin-film sensor of Finland indicates about a 14 percent higher humidity
value than the carbon-coated sensors of Australia and the US.

A change in the temperature structure corresponding to the above


described humidity structure also was observed. This is shown in Figure
4.2. At minute 9 1/2 the US measured temperature cools dramatically in
response to entering the dry air above the cloud top after b~jng
immersed in-very humid air. Not surprisingly, the general temperature
structure observed with the Australian, Indian, and US sensors, is simi-
lar, though not exactly the same. More importantly, at this same time
the Finnish capacitive sensor is indicating an inversion.

The question of whether or not the temperature structure above a


cloud is super adiabatic or inverted, is important. Historically, the
super-adiabatic temperature lapse rate observed in the dry air above the
cloud has been thought to be due to rapid evaporative cooling of water
deposited on the thermistor's surface from its flight inside the cloud,
and is known as the "wet Bulb" effect. The Finnish capacitive thin film.
sensor indicated an inversion in this same layer. During Phase I, held
in Bracknell, U.K., during the summer of 1984 (Hooper, 1986), the wet-
bulb effect was obser~ed by all of the sensors, i.e., both the resistive
rod and capacitive thin film sensors. The inversion observed at the
cloud top by the Finnish radiosonde in Phase II made it suspect that the
Finnish instrument (sensor or software) might be different from that
used
page 2.4

in Phase I. It was lea.rned subsequent to the completion of Phase. II'


that Finland had applied a water repe.lling coating to the sensors used
in Phase II that was not. applied during Phase I. Clearly I a more
detailed investigation and analysis of all of the temperature sensors'
behavior after exiting a saturated cloud is needed.

Figure 4.3 depicts the observation near 0900 local time on 5


March and indicates that the Indian measurements are in relative agree-
ment with the other measurements. However, the poorer response of the
Indian sensor is still obvious between the surface and minute 12.
Specifically, this sensor did not report the high humidity values (-97
percent) reported by the other sensors near minutes 2 and 3, nor did it
decrease as fast as the other sensors to report the value of the drier
layer (-75 percent) at minute 5. Between minutes 5 and 12 the lithium
cloride sensor's reported values are approximately the mean of the other
sensors' measurements. Compared with Figure 4.1, which showed poor
agreement and poor response of the lithium cloride sensor relative to
the carbon and capacitive sensors, Figure 4.3 indicates that better
agreement exists between the sensors and in general the data are accept-
able in spite of the lithium sensor's lag. The measurements of the
capacitive sensor relative to the carbon-coated sensor measurements are
different from those observed on 4 March (Figure 4.1). The capacitive
sensor seems to read higher than the carbon sensor one time (Figure 4.1)
and lower another time (Figure 4.3). This characteristic needs further
investigation. The Finnish capacitive thin-film sensor indicated a
drier atmosphere than that measured with the carbon elements at minute
12 and later. The carbon-coated sensor measurements are, once again, in
agreement with each other. One peculiarity of the carbon hygristor is
its failure to report relative humidity values lower than 15-20 percent.
This measurement is limited by the transfer equation used in the data
reduction process.

Associated with the rapid change from the moist to dry atmosphere
at minute 8 in Figure 4.3 is the super-adiabatic temperature structure
observed in Figure 4.4. The Australian and US instruments show the
existence of a super adiabatic temperature lapse rate followed by an
inversion. The temperature structure observed with the Finnish capaci-
tive thermistor does not indicate the super adiabatic lapse rate nor, in
this cas~ did it observe an inversion. At minute 11 the temperature
agreement among the sensors becomes better. The Indian tempera~~r~
measurement-appears to have a bias duing this observation.

This raises the question, what is the characteristic temperature


structure near the freezing level? In Figure 4.2 the inversion reported
by the Finnish capacitive temperature sensor occurs at -5 C, however, in
Figure 4.4 the temperature is above freezing and an inversion is not
observed. In this same situation the rod thermistor used by Australia
and the US indicated a super adiabatic lapse rate at minute 8. The rea-
son might be hypothesized thus1y: during the instruments' flight through
the cloud, water accumulated on the rod thermistor, upon reaching the
dry layer immediately above the cloud the moisture evaporated rapidly
leading to the characteristic super adiabatic temperature structure.
Because of the large mass of the rod thermistor its heat capacity is
relatively high and the moisture remains in liquid form for a relatively
long time even at temperatures below 0 C, thus, setting up conditions
for evaporative cooling. On the other hand, the capacitive thermistor
used with the Finnish instrument is smaller, is coated with aluminum
and,
page 25

presumably has a smaller heat capacity (although the actual diameter of


the capacitive bead ..is larger than the rod's). Thus, at temperatures
above freezing the moisture may remain in liquid form and subsequently
evaporate, but at temperatures .below freezing the liquid freezes
(because of the smaller heat capacity) and releases the heat of fusion.

Brousaides (personnel communication) performed a simple test. He


weighed a rod thermistor, dipped it in water, and then weighed the
thermistor and the water adhering due to surface tension. Eleven milli-
grams of water was retained. Instantaneous evaporation of this amount
of water would result in the release of a significant number of calories
giving an instantaneous temperature drop of approximately 144 C. If a
similar volume of water is evaporated from the rod thermistor in the
atmosphere over a finite time interval the temperature drop could be
somewhat similar to the values observed in Figures 4.2 and 4.4. On the
other hand, if the same volume of water freezes, the amount of caloric
heat absorbed due to releasing the heat of fusion could result in an
inversion of approximately 2-3 C. This may explain, in part, the
differences observed in the temperature structure at the cloud top by
the different sensors, and may be a possible reason for the temperature
lapse-rate differences observed between two different-type thermistors.

The discussion of relative humidity and temperature profiles so


far has been keyed to morning observations. It is noted that not only
did the three types of humidity sensors respond differently within a
given ascension but in the three examples cited, the lithium cloride-
coated, capacitive thin-film, and carbon-coated sensors showed incon-
sistent bias among their measurements.

A third example of humidity sensor comparison is given for the


nighttime ascension at 2300 UT on 4 March in Figure 4.5 and includes
humidity data obtained with the hair hygristor of the Graw M60
radiosonde. The hair element indicates the levels where humidity
saturation and minimums occur. The temperature structure for this
flight on 4 March is extremely important since the various thermistors
show different responses in the inversion layer. The inversion indi-
cated by the Finnish thermistor beginning near minute 9 occurs at least
30 seconds before the capacitive humidity sensor signals a drop in humi-
dity. The base of the inversion indicated by the US radiosonde occurs a
full one-minute after being reported by the Finnish radiosonde. ~The
atmospheric- temperature in this observation stayed above freezing
through minute 13, while the temperature discussed in Figure 4.2 reached
freezing at minute 7, and the temperature on 5 March (Figure 4.4)
reached freezing at minute 10 before an inversion brought the tempera-
ture to above freezing again. Hence, it is possible that these environ-
mental temperatures might be connected to the responses of the different
thermistors and the different inversion structures.

In summary, the carbon element used by Australia and the US


appears capable of responding rapidly to humidity changes, while the
sensor substrate's thermal lag is known to be a detriment and inhibits
accurate humidity measurement through the transfer equation; the effect
of insolation worsens the overall error during daylight. Given that the
sensor 1s at a different temperature than ambient (in the case of
decreasing temperature with increasing altitude the sensor would be
expected to be warmer than ambient) the use of the wrong temperature
value in the transfer equation will produce an error in relative
humidity
page 26

and mixing ratio (Brousaides and Morrissey, 1974). Temperature differ-


ences of only a few deg~ees could produce a 7-10 percent humidity error.
This is an amount much less than the differences observed between the
carbon-coated sensor and the other .. sensors. Furthermore, the flow effi-
ciency surrounding this sensor is not well studied and if poor may also
be responsible for poor measurements.

The poor agreement between temperature and humidity measurements


of the nighttime ascension on the 4th of March (Figure 4.5 and 4.6) upon
exiting the clouds suggests poor response, or possibly sensor contamina-
tion. Unlike the carbon-coated sensor with its large thermal lag the
capacitive thin-film sensor is expected to respond faster, however,
information published by WMO (1983) indicates that it has a time con-
stant similar to that of the carbon element. Furthermore, the addition
of the aluminum cap that serves to protect the thin-film sensor from
liquid water contamination may have an unexpected thermal effect on the
measurements. Further investigation in this area is needed. The Indian
lithium cloride-coated sensor gave the poorest response to humidity
changes. The cause for this requires further detailed examination by
the Indian Meteorological Department. Finally, although only one exam-
ple is presented here, the hair element used with the Graw M60
radiosonde responds too slowly. Nonetheless,this sensor did follow the
general profiles of the other sensors.
page 27

Fig 4. 1

ASCENSION NO. 193 ON 04-MAR-85 AT 1 4: 11: 18 GMT

0 10 20 30 40 so 60 70 80 90 100
24 24

22 I< 22

/
f
20 )1: 20
I

18 "'~ 18
'x.
16 ' 'X\ 16

~
14 )< 14
w
1-
:::)
12
z: 12
.......
L:
10 10
--:..-=-·-·-------
--- -----7(..._ ..._
..._ ..._
8 --~ 8
\
~
6
(9 USA
E!)

AUS
&--·-·-.6
VIZ-1392
RS4-MK3
\
Jr:
l 6
+----+ FI1 RS80-15N .J<
4
x----7( IND MK-III /
/
4-
.-.X
_..x-
2 -- --- 2

0 0
0 10 20 30 40 so 60 70 80 90 100

HUMIDITY

Figure 4.1- Graphical comparison of Australia (AUS), US (USA), Finland


(Fil), and Indian (IND) relative humidity measurements at one-minute
intervals. Ascension no. 193 data shown (4 March 1985, 1411+18UT).
page 28

F1g 4.2

ASCENSION NO. 193 ON 04-MAR-85 AT 14: 11 : 18 UT

-40 -30 -20 -10 0 10 20


24 24

2.2 22
"\
><
\
20 20

18 18

16 16

14 14
w
1-
:::::> 12 12
z
.......
:L
10 10

8 8

(9 el USA VIZ-1392
6 4or--·-·- AUS RS4-MK3 6
+---+ FI 1 RS80-15N
4
>E- --=--~ IND MK-III
4
-

2 2

0 0
-40 -30 -20 -10 0 10 20

TEMPERATURE

Figure 4.2 Graphical comparison of Australia (AUS), US (USA), Finland


(Fil), and Indian (IND) temperature measurements at one-minute inter-
vals. Ascension no. 193 data shown (4 March 1985, 1411+18UT).
page 29

Ftg 4-J

ASCENSION NO. 197 ON 05-MAR-85 AT 13=45=50 GMT

0 10 20 30 40 50 60 70 80 90 100
24 24

22 22

20 20

18 18

16 16

14 14
w
1-
t
:::::>
z
......
l:
12
--- 12

10 10

8 8

(!) e> USA VIZ-1392


6 ....._·-·-A AUS RS4-MK3 6
t - - - + F I 1 RS80-15N
*---~ INO NK-III
4

2 2

0 0
0 10 20 30 40 50 60 70 80 90 100

HUMIDITY

Figure 4.3 Same as Fig. 4.1, except data are for Ascension no. 197 (5
March 1985, 1345+50UT) .
page 30

Fig 4-4

ASCENSION NO. 197 ON 05-MAR-85 AT 13=45=50 GMT

-40 -JO -20 '-10 0 10 20


24 24
\
x,
22 'X 22
' x_,_
20 ' 20

18 18

16 16

14 14
w
1-
::::>
:z 12 12
......
:L
10 10

8 8

(9 e> USA VIZ-1392


6 Ar-·-·-A AUS RS4-MK3 6
+----+ FI1 RS80-15N
>E--=--K IND MK-1 II
4 4

2 2

0 0
-40 -JO -20 -10 0 10 20

TEMPERATURE

Figure 4.4 Same as Fig. 4.2, except data are for Ascension no. 197 (5
March 1985, 1345+50UT).
page 31

Ftg 4.5

ASCENSION NO. 196 ON 04-MAR-85 AT 22~45:51 UT

0 10 20 30 40 50 60 70 80 90 100
24 ~~--~----.------.-----.-----.----~r-----~----.------r-----,24

22
22 ...........
-.....
20 20

18 18

16 16

14 14
w
1-
::J 12
z 12
......
L:
10 10

8 8

<9 e~ USA VI Z-1392


6 4-·-·--4 AUS RS4-MK3 6
+---+ FIt RS80-15N
~···-··~ UK GRAW-M60
4 4

2 2.

0
0 10 20 30 40 50 60 70 80 90

HUMIDITY

Figure 4.5 Same as Fig. 4.1, except the Graw M60. radiosonde data are
included. Data are for Ascension no. 196 (4 March 1985, 2245+51UT).
page 32

Figure 4.6 Same as Fig. 4.2, except India Mk III radiosonde data are
included. Data are for Ascension no. 196 (4 March 1985, 2245+51UT)>
page 33

5.0 Data Analysis

Approximately 500 radiosonde profiles were obtained from the 100


balloon flights of Phase II. These were carefully checked to insure
that the data did not contain undetected errors. Outliers, if present,
were removed.

Studying reported measurement differences at the standard pres-


sure levels provides a unique perspective on the over-all agreement to
be expected from routinely reported radiosonde data transmitted over the
GTS. In order to determine the performance and errors inherent in the
various systems, the one-minute interval measurements also were studied.
Thus, time-based or simultaneous measurements are important if instru-
mental accuracy and precision are to be determined.

Wind information was to be obtained from those systems capable of


producing it. During Phase II, only Finland (using navaid methods) and
the United States (using the GMD tracking system) provided wind data.
The Wallops Island precision radars tracked every balloon released, and
provided position information (x, y, and z) for every second of flight.
Radar tracking yields more accurate and precise wind and altitude meas-
urements and, therefore, serves as an excellent comparison standard for
these parameters. Radar results are discussed in Section 6.0

Simple statistics were used in carrying out the data analysis.


Comparison of the measurements of the standard pressure levels are
straight forward; however, comparison of the one-minute interval data
demanded special handling. Because the data of Phase I and Phase II
were to be merged for final analysis the method used in Phase I (Hooper,
1986) for analyzing minute data was followed. Simultaneous comparisons
was obtained by grouping the data from each instrument in the vertical
according to pressure layers. The mean of the measurements from each
radiosonde observing within the pressure layer was calculated. Diff~r­
ences between the radiosonde's layer mean values were obtained and
finally the mean of the differences and standard deviations for each
pressure layer were derived.

The analysis of these data was carried out independently for each
of the observation times. At first look, it seemed that the measure-
ments for all time groups should be combined and averaged. This could
not be done however, since the solar influence affects each sensor dif-
ferently. Furthermore, the cooler nighttime observations would offset
the influence of the solar heating bias of the daytime observations.

For purposes of consistency throughout this Phase II report all


comparisons are with respect to the US radiosonde.
page 34

5.1 Statistical Methods

Simultaneous observations have an advantage in that the measure-


ments are made under identical environmental conditions while eliminat-
ing differences due to space and time. This reduces measurement varia-
bility resulting from atmospheric effects and increases the validity of
real differences between instruments. Additionally, paired observations
may be carried out under a wider range of. operating conditions that are
more representative of actual use and assures wider applicability of the
conclusions.

The analyses called for comparison of each of the time-based data


sets. It was important, therefore, to determine whether the data sets
for each of the time groups could legitimately be subjected to the sta-
tistical testing proposed for Phase II, i.e., Student's t for small sam-
ples. In this regard, both the Fisher's test of the variances and the
Kolmogorov-Smirnoff test for distributions were employed. The signifi-
cance level of both tests was 95 percent. Variances of the radiosonde
measurements from each pair were calculated, and the ratios v /v when
compared against the F distribution were found to be acceptable. Furth-
ermore, the Kolmogorov-Smirnoff test, generally a more powerful test of
the distributions, will reject more oftener than other tests when the
distributions of each data set are different. This test, however, indi-
cated that the measurements from each radiosonde data set were from the
same distribution. Thus, there was no reason to reject the use of
Student's-t test for small samples.

The mean of the differences and the difference of the means of


the pairs were calculated to determine the relative biases between the
instruments. The standard deviations also were calculated and the vari-
ance for each instrument was used to estimate the measurement precision
based on a procedure developed by Grubbs (1948). This procedure is dis-
cussed in Section 8.0. The statistical analysis techniques employed
equal sample sizes, i.e., only those observations that included measure-
ments from all instruments were used. It was considered undesirable to
use unequal sample sizes simply because of the uncertainty that may
ensue when ~rying to interpret results.

5.2 Temperature

Comparison of radiosonde mean temperatures are discussed for the


standard pressure levels and for the simultaneous measurements. The
Graw M60 data are not included because of the smaller number of samples
available for this instrument. For example, 25 balloon ascensions were
made at 1700 UT but when the Graw M60 data are included only 14 samples
are available, which significantly reduced the number of equal pairs
that could be analyzed. The Graw M60 data are discussed in Section 7.0.

5.2.1 Standard Level Temperatures

The US radiosonde was arbitrarily selected as the comparison


standard for this discussion of the Phase II data. For pressure levels
between
page 35

1000 hPa and 100 hPa (Figures 5.1 to 5.3), the range of the mean of the
temperature differences-for all pairs was observed to be 0.5 Cor less;
however, the data set of 2300 UT (Figure 5.4), showed a larger range of
up to 1.0 C. While this range is not necessarily large, it translates
into altitude differences between the radiosondes of approximately 35-40
meters at 100 hPa. Between 100 and 10 hPa, larger mean differences were
observed between the instruments. The mean temperature difference
between the Australian and US radiosondes exceeded 1 C at 20 hPa and
reached approximately 4 C at 10 hPa. Such a large difference between
similar sensors is difficult to accept as valid since, as will be dis-
cussed later, the Australian radiosonde contained a serious pressure
sensor calibration problem. The differences between the Indian and us
radiosondes were less than 1 C up to the 20 hPa level but increased to 2
C at 10 hPa in the 2000 UT observation group. Generally, daytime US
radiosonde temperatures in the mid-stratosphere above about 100 hPa
(Figure 5.1 - 5.3) were lower in the mean than the Australian and Indian
radiosonde temperatures, but were higher than temperatures reported by
the Finnish radiosonde. At night (Figure 5.4), the US radiosonde
reports lower temperatures than those of the other radiosonde types.
The mean nighttime temperatures reported by the US radiosonde were lower
than the temperatures reported by the Finnish radiosonde by approxi-
mately 1 C at 50 hPa and 3.5 C at 10 hPa.

Considering that the rod thermistors used with the Australian and
US radiosondes were identical it should be expected that there would be
little, if any, difference between temperature measurements these
instruments. Figures 5.1 through 5.4 show this not to be the case.
Although the differences were about 0.5 Cor less up to 100 hPa, they
increased rapidly to almost 4 C above this level. A further point to be
noted is that temperatures measured with the Australian instrument were
not colder at night relative to the day, such as experienced between the
Finnish and US radiosonde measurements. The differences between India
and US temperatures were similar to those between Australia and the US,
except at the upper levels where the differences were smaller. The
Indian rod thermistor, though similar to that of the Australian and US
thermistors, is slightly larger in length and diameter and is coated
with titanium dioxide instead of lead carbonate. It is expected that
the titanium dioxide coating responds differently to radiation than the
lead carbonate coating, which may explain some of the differences
observed. blthough the actual cause of these temperature differen~e~
were not determined, it is clear that the four instruments responded
differently to ambient temperature. This strongly suggests that correc-
tions should be derived for each system.

The magnitude of the standard deviations of the temperature


differences (Figures 5.1 through 5.4) suggests considerable variation.
This behavior may arise from: (1) unstable, or poorly performing, tem-
perature sensors or, (2) reading the temperatures at some level other
than the indicated level. Cause 1 is not considered to be a valid rea-
son, for when comparisons are made at simultaneous times as shown in
Figures 5.5. through 5.8 the observed bias is small and the variance
diminishes considerably. Cause 2, on the other hand, arises when errors
in the pressure measurements occur resulting in temperature measurements
being reported for a level other than the assigned presssure. Cause 2
is probably responsible for the large variances and may also contribute
substantially to the character of the biases observed at the standard
pressure levels. As discussed in Section 3.0, great effort was expended
to
page 36

provide each participant with an exact indication of time at each minute


after release of the balloon. Marks were automatically recorded on each
participant's strip chart recording, or as in the case of Finland the
time of the pulse was compared internally in their ground system com-
puter. These time pulses were used to prevent errors in reading data
caused by strip chart recorder speed differences. It is concluded that
large temperature biases accompanied by large variations are not from
recorder speed differences nor time offsets, but rather from errors and
variations in the pressure measurements of the individual radiosondes.
This raises a question, how bad are pressure measurements?

5.2.2 Simultaneous Temperature Measurements

Comparison of temperature measurements made at simultaneous times


provides different results from those obtained from comparisons at the
standard pressure levels. It is impossible to compare measurements made
at the same indicated minute over a large number of ascensions, simply
because the measurements at each minute of one ascension will not occur
at identical altitudes or pressures for the same minutes of subsequent
observations. To overcome this problem measurements occurring within
the bounds of fixed pressure layers were averaged. The differences
within each layer were then averaged over time to provide mean differ-
ences and standard deviations for each observational period. This
method required 13 pressure layers; the layers are listed in Table 5.1.

Table 5.1 Pressure Layers for Comparison of Simultaneous Measurements

Layer-hP a Center of Layer-hPa

975.0-840.0 900
840.0-589.0 700
589.0-415.0 500
415.0-245.0 320
245.0-164.0 200
164.0-119.0 140
119.0-084.0 100
084.0-058.9 70
058.9-041.5 50
04.1. 5-024.0 32
024.5-016.4 20
016.4-011.0 14
011.9-008.4 10

Figures 5.5 through 5.8 indicate that the Australian and US tem-
perature measurement differences are less than a few tenths of a degree
between the surface and the layer centered at 10 hPa. This confirms
that the Australian and US temperature measurements are similar and that
the large differences obtained at the standard pressure levels resulted
from errors from other sources, probably, as indicated, in the pressure
measurements. Differences between the Finnish and US radiosonde tem-
peratures are of similar magnitude to those presented for the standard
pressure
page 37

levels. This suggests that pressure errors in either system do not sig-
nificantly affect temperature measurements in the mean. US temperature
measurements are also warmer during day and colder during night than the
Finnish temperatures. Simultaneous comparisons also confirm that the
differences between Indian and US temperatures are smoother in the vert-
ical than noted with the constant pressure level comparison and are of
nearly equal magnitude. This suggests that variations exist in the
pressure measurements. US temperatures also are colder than Indian tem-
peratures for all the observation times suggesting that the thermistor
coatings of both radiosondes are consistent during day and night but
that the US lead carbonate coating has a larger emissivity or absorp-
tivity at short-wavelengths is larger for titanium dioxide.

The standard deviations are considerably smaller than those cal-


culated for the standard pressure levels. This suggests once again,
that a considerable part of the variance observed at the standard pres-
sure level temperatures comes from variations in the pressure measure-
ments. The standard deviations confirm that the behavior of the Aus-
tralian and US temperature sensors are similar. Since these are identi-
cal sensors the magnitude of the standard deviation can also be con-
sidered to be a measure of the r.m.s. error of the rod thermistor, i.e.,
about 0.2-0.3 C. This is consistent with results of Schmidlin, et al
(1982), Lenhard (1973) and others. The standard deviation between the
Indian and US temperature sensors, however, are larger than those shown
for the other instrument pairs.

The interpretation of Figures 5.1 through 5.8 can not be stated


simply since a number of factors may be operating at the same time. The
agreement between temperatures measured with the Australian and US
radiosondes shown in Figures 5.5 through 5.8 is excellent. These sen-
sors are identical in size, shape, temperature vs resistance charac-
teristics, and were manufactured by the same company. With such simi-
larities it would be unusual if they didn't agree when compared simul-
taneously. On the other hand, the disagreement between these same two
sensors at the standard pressure levels (Figure 5.1 through 5.4), obvi-
ously can only result from poorly resolved pressure measurements since
timing offsets in the data are known to be not more than one second. As
will be discussed later in Sections 5.5 and 7.0, a serious calibration
offset existed with the Australian pressure cell/baroswitch assembly.

Although the Indian temperature sensor looks very similar to the


Australian and US temperature sensors it is different in size. It also
uses a white reflective coating, however, the coating material (titanium.
dioxide) has a different spectral response in the infrared than the
coating of the US thermistor. As shown in Figures 5.1 through 5.8 the
Indian temperature measurements during both daytime and nighttime are
higher in the mean than those of the US radiosonde. Daytime measure-
ments are affected by incident solar radiation and the Indian thermistor
responds differently to short-wave radiation than does the US tempera-
ture sensor. More specific tests are needed to determine the magnitude
of the difference.

Differences between temperatures measured by the Finnish and the


US radiosondes show that the US temperatures are higher than the Finnish
temperatures during the day and lower during the night. It should be
kept in mind that the Finnish thermistor temperatures have had correc-
tions applied (Antikainen and Hyvonen, 1983) while, at the time of the
intercomparison
page 38

there was little to no information on corrections that might be applied


to the US thermistor measurements. A study of the radiation error of
the rod thermistor has been in progress at Wallops Island since the com-
pletion of the intercomparison. (Schmidlin and Luers, 1987). Tests have
shown that when these preliminary corrections are applied, the results
of the intercomparison temperatures and geopotentials change signifi-
cantly, i.e., the agreement between the Finnish and US instruments
becomes better.

Finally, the very obvious and large decrease in the standard


deviations resulting when the comparison was made at simultaneous times
is a positive indication that the large temperature variations shown at
the standards pressure levels are caused by other then the temperature
sensors, and probably were caused by poor pressure measurement resolu-
tion, randomness in the pressure measurement, or pressure sensor cali-
bration error.

5.3 Relative Humidity

Statistical analysis of the relative humidity measurements are


presented in a manner similar to that used in the discussion of tempera-
ture. Relative humidity measurements generally extend to levels higher
than 400 hPa, but for analysis purposes are truncated at 400 hPa for
standard level analysis and at the 500 hPa layer for simultaneous meas-
urement analysis.

5.3.1 Standard Level Relative Humidity Measurements

Figures 5.9 through 5.12 indicate that very little difference


exists between relative humidity measurements made with the carbon-
coated hygristors (i.e., between Australia and the US). The largest
mean difference (i.e., -5 percent) between Australia and US is found
near the surface for the 2300 UT observations as shown in Figure 5.12,
otherwise the differences between the carbon hygristor measurements are
generally no more than 2-3 percent. Observed mean differences between
the instruments of Finland and the US are also relatively small,
although -the capacitive thin-film sensors tended to record drier than
the carbon~coated sensors at all pressure levels. Measurements made
with the Indian lithium cloride-coated sensor also recorded a drier
atmosphere than the carbon elements. The differences between India and
US humidities are large but consistent considering the large time lag of
the Indian sensor.

nighttime observations exhibited little noticeable difference


from daytime observations. Brousaides and Morrissey (1974), found that
in the absence of insolation the carbon-coated element is subject to
error mainly induced by thermal lag of the plastic strip. Table 5.2
reproduced from Brousaides and Morrissey (1974), shows that the error
induced in sunlight is larger than the thermal lag error. Figure 5.12
indicates a nighttime bias at 400 hPa of approximately 8 percent between
the carbon element and the capacitive element. The decrease in relative
humidity difference with increasing pressure (decreasing altitude) shown
in Figure 5.12 follows the values given for thermal lag by Brousaides
and Morrissey. On the other hand, daytime differences should be larger.
According to Brousaides and Morrissey (1974) a total relative humidity
measurement
page 39

error resulting from insolation and lag of 15 percent in the 500-351 hPa
layer exists (Table 5.2). Figures 5.9, 5.10, and 5.11 do not show these
large differences between the carbon and thin-film sensors. The smaller
measurement differences suggest that the error due to the insolation
effect on each sensor are not very different from each other.

Table 5.2 Daytime Humidity Errors Due to Thermal Lag and Insolation
(From Brousaides and Morrissey, 1974)

Level, hP a Insolation(%) Thermal Lag(%)

1013-701 3 3
700-501 6 4
500-351 9 6
350-250 14 9

The variability is shown by the standard deviation profiles of


Figures 5.9 through 5.12. The variance between the Australian and US
instruments is the smallest. This should be expected since both instru-
ments use the same carbon-coated hygristor and essentially a similar
mounting configuration. The variation observed between Finland and the
US is slightly larger than that between Australia and the US while a
considerably larger variation is noted with the lithium cloride element
employed with the Indian MK III radiosonde.

5.3.2 Simultaneous Relative Humidity Measurements

Simultaneous relative humidity measurements were limited to the


layers between the surface and 500 hPa (see Table 5.1). The criterion
used to screen data for all layers required that a measurement existed
for each of the one-minute points within a layer. Thus, the layer 415-
245 hPa could not be used because of the lack of humidity measurements
above 300 hPa. Nonetheless, the comparison information are consistent
with the-results discussed previously for the standard level data. If
anything, moie precise differences are obtained since any errors indttGed
from poor pressure resolution are excluded.

The profiles shown in Figures 5.13 through 5.16 lead to a number_


of conclusions. The first is that the carbon hygristor used with the
Australian and US radiosondes indicate a repeatability of approximately
2 percent. Since errors resulting from poorly resolved pressure meas-
urements are not a factor in these figures, the mean difference between
Australian and US instruments is small and, since identical sensors are
being used the standard deviation is an indication of the carbon
hygristor's repeatability. Secondly, based on acceptance of two percent
repeatability for the carbon element, the Indian and Finnish hygristor
repeatability may be determined. Subtracting the variance of the carbon
element from the total variance between the Indian and US sensors, it
appears that the Indian humidity sensor's repeatability is of the order
of 10 percent. Using a similar analysis the Finnish capacitive sensor's
repeatability is about 4-6 percent.
page 40

5.4 Geopotential

Average differences of the calculated geopotential at standard


pressure levels were derived only from matched pairs. Comparison of
geopotential heights from simultaneous measurements are discussed in
Section 6.0 where comparison with radar is also discussed. The profiles
of the daytime geopotential differences (i.e., 1400, 1700, and 2000 UT)
shown in Figures 5.17, 5.18, and 5.19 are similar and will be discussed
as a group.

The differences between Finland and US geopotential heights are


approximately 15 meters at 100 hPa, 30 meters at 50 hPa, and 94 meters
at 10 hPa. These differences represent a significant error in either
pressure or temperature measurements from these instruments. Reexamin-
ing Figures 5.1, 5.2, and 5.3, one is confronted with a bias in tempera-
ture between the Finnish and US radiosondes with the US instrument
recording higher temperatures up to as much as 1.5 C. Differences in
reported temperatures lead to geopotential differences that can be
determined from the hypsometric expression

H R/g( T)ln(p1/p2)

where H is geopotential difference, R is the gas constant for dry air


(287.05J/kg/ K), g is the constant of gravity (9.80665 m/sec2), T is
mean temperature difference in the layer p1/p2, with p1 and p2 being the
pressures at the lower and upper bounds of the layer, respectively.

Using the mean temperature differences derived from simultaneous


measurements (Figures 5.5- 5.7) it is possible to estimate the differ-
ences in the thickness H, between the standard pressure levels.
Integrated height differences between the Finnish and US radiosondes to
be expected from this method are approximately 7 meters at 100 hPa, 17
meters at 50 hPa, and 71 meters at 10 hPa. These values are slightly
less than the reported measurement differences noted above in Figures
5.17-5.19. Bias in pressure measurements also contributes to differ-
ences in geopotentials, however, the magnitude of the difference is
dependent on the mean temperature lapse rate with respect to pressure (
T/ p) and _the average pressure difference ( p). These generally -are
small (<10 meters). Thus, it is highly probable that the geopotential
difference between Finland and the US is caused almost entirely by the
reported temperature differences between the two instruments.

The mean difference between Australia and the US is considerably


smaller than that between Finland and the US. Australia vs US simul-
taneous temperature profiles (Figure 5.5 - 5.8) are virtually the same
indicating that differences in temperature should not be a major contri-
butor to the differences observed in the reported geopotentials at stan-
dard pressure levels. Actual height differences however, were not con-
sistent with height differences expected from the actual temperature
differences. For example, using the hypsometric procedure, the tempera-
ture difference between Australia and US radiosondes would result in
geopotential differences of approximately -11 meters at 100 hPa, -13
meters at 50 hPa, and -10 meters at 10 hPa. Geopotential differences at
standard levels for the nominal time of 1700 UT, shown in Figure 5.18
are 20 meters at 100 hPa, 17 meters at 50 hPa, and -42 meters at 10 hPa.
Since
page 41

the temperature difference shown by the simultaneous comparison indi-


cates that the true geopotential differences should be smaller, it is
concluded that something other than temperature disagreement is causing
the large geopotential differences. As will be seen in Sections 5.5 and
7.0 a large pressure error was present in the Australian measurements;
no doubt the erroneous pressure data were causing the differences as
well as the nonrepresentative comparison of temperature (Figures 5.1
5.4). (The source of this error was isolated by the Australian Bureau
of Meteorology soon after the completion of Phase ii, and adjustments
were recommended. See Section 7.1).

The difference in geopotential between India and the US is con-


sistently in the same direction day and night. The US geopotentials for
the 1700 UT observations, are lower on average, than those of India by
approximately -8 meters at 100 hPa, -11 meters at 50 hPa, and -10 meters
at 10 hPa. However, differences found in the observations of 1400,
2000, and 2300 UT are considerably larger. Since the simultaneous com-
parison of temperatures (Figures 5.5 - 5.8) indicates that the US tem-
perature sensor reports lower temperatures than those of the Indian sen-
sor, the direction of this height bias was to be expected.

Figures 5.17 through 5.20 also show the geopotential standard


deviations reported by the radiosonde observations. The standard devia-
tions reveal that randomness and/or error in reading the temperatures
and pressures can result in large variability in the geopotential
heights. The magnitudes of the standard deviations suggest that the US
and Finnish paired measurements are least variable, followed by the Aus-
tralian and US measurements, and then the Indian and US measurements.

During preparation of the.data for the above analysis it was


found that significant differences in the value of the constant g was
used for Phase II geopotential calculations. For example, the US used
9.8 m/sec2, Finland and Australia used 9.80665 m/sec2, and the value
used by India is unknown. However, India used the value of 29.3 for the
ratio R/g as opposed to the more precise value of 29.27095389. These
differences lead to geopotential biases of as much as 20 to 30 meters.

5.5 Pressure

The pressure sensors described in Section 2.0 all operate on the


same basic principal, i.e., deflection of the plates of an aneroid cell.
The method of transferring pressure change with respect to altitude
change is different for the Finnish radiosonde relative to the Aus-
tralian, Indian, US, and Graw M60 radiosondes. The latter instruments
use aneroid cells connected by a mechanical linkage that moves a contact
over a segmented electronic switch. Such an arrangement is susceptible
to errors because of friction within the mechanical linkage and uncom-
pensated temperature effects. Furthermore, unseasoned aneroid capsules
have been noted to change their sensitivity to pressure change with
time. The Australian, Indian, and US pressure sensor capsules are made
of Ni-span-C, but only the US baroswitch linkage is designed to be self
compensating for temperature effects. Australia compensates for tem-
perature effects during pressure cell calibration while the India
Meteorological Department applies corrections to the measured pressures
during the observation. The Finnish pressure capsule is a friction-free
aneroid
page 42

cell with capacitor plates inside of the capsule. Temperature compensa-


tion is applied during observation by measuring the temperature of the
pressure cell enclosure with a small bead-type thermistor. The Graw M60
radiosonde pressure sensor uses a double aneroid cell. This capsule
also uses a mechanical linkage to transmit measurements and is tempera-
ture compensated by adjusting the pressure readings according to a table
developed for that purpose. A more complete explanation of pressure
sensor operation can be found in the WMO Guide to Meteorological Instru~
ments and Methods of Observation (1983}.

It should be remembered that errors in geopotential are esta-


blished by the errors that occur in temperature and pressure measure-
ments. These errors differ according to whether the geopotential in
question is that for a given pressure level or a given point. However,
the standard pressure level is given by definition and, thus, does not
have a pressure error associated with it, while significant levels (or
in the case of the intercomparison data, one-minute interval data) do
contain errors of observation. As a result, the geopotential for the
significant levels degrade by comparison to standard levels in propor-
tion to the pressure measurement error. It should be pointed out that
the pressure measurements were not read to the same degree of precision
by all participants. For example, Australia and India recorded pressure
to the tenth of a hecto-pascal, Finland to hundreths, while the US read
pressure to whole hecto-pascals between the surface and 20 hPa, to one-
half of a hecto-Pascal between 20 and 10 hPa, and to a · tenth of a
hecto-Pascal above 10 hPa. The effect of this can be noted in the accu-
mulation of geopotential error with height. For this reason the discus-
sion of the comparisons of temperature given in Sections 5.2.1 and 5.2.2
and the discussion of the comparison of pressure measurements to follow
are important.

5.5.1 Simultaneous Pressure

Figures 5.21 through 5.24 are derived from pressure measurements


at one-minute intervals. The method of simultaneous comparisons was
discussed in Section 5.0. Data shown in these figures do not include
pressure measurements from the Graw M60 radiosonde. Discussion of Graw
M60 pressure measurements is given in Section 7.0.

At 1400 UT (Figure 5.21) there is a relatively large mean pres-


sure difference between Finland and the US at levels below about 300
hPa. This difference decreases to less than 0.2 hPa between 300 and 20.
hPa, and then increases to 0.5 hPa in the layer centered at 10 hPa. The
abrupt change in pressure differences observed between India and the US
between the surface and about 500 hPa requires explanation.

Above 500 hPa, comparisons at the four observation times (Figures


5.21 to 5.24) show the mean pressure difference to be less than 0.5 hPa.
The standard deviation of the differences between India and the US pres-
sures is less than 2 hPa. These magnitudes, or tolerances, are similar
for all of the instruments, however, the Indian vs US pairs show
slightly more variability level to level. In contrast, the Finnish vs
US pairs indicate less variability and a smooth transition level to
level.

The fact that the temperature compensation applied to the


measurements
page 43

of the India Meteorological Department MK III radiosonde were accom-


plished under very different environmental conditions than those nor-
mally experienced in India speaks well for the adaptability of the
instrument and the operators. However, it is also obvious that to apply
the required procedures experienced operators must be employed.

The mean pressure differences observed between Australian and US


measurements reach nearly 5 hPa and the standard deviation is the larg-
est of all of the instrument pairs. Because the Australian pressure
measurements are lower than the other instrument's pressures by such a
large amount the geopotentials of the significant levels and the one-
minute interval levels have very large errors as shown in Section 6.0.
Furthermore, the use of these temperature measurements to calculate the
geopotentials are at risk simply because the temperatures assigned to
the standard pressure levels are actually measured at higher pressures,
e.g., temperatures assigned to the lOO hPa pressure surface should actu-
ally be assigned to the 96 hPa surface, or approximately 300 meters
higher then the height of the 100 hPa level.

Examination of Figures 5.21 through 5.24 also shows that instru-


mental differences over the four observation times are consistent. For
example, Finland's pressures relative to the US pressures are larger in
the troposphere and middle stratosphere, and smaller in the lower stra-
tosphere. The mean difference between India and the US and between Aus-
tralia and the US also are similar. Although there is variation in the
temperature and geopotential between daytime and nighttime measurements,
there is no such change apparent in the pressure differences. This sug-
gests that the temperature compensation values are consistent between
day and night.

The results of this chapter conclude that: (1) temperature


corrections should be developed for those radiosondes for which none
currently exist, (2) pressure measurements should be resolved to a
higher degree of precision, and, (3) the accuracy of the pressure meas-
urement should be improved.
page 44

FI 9 5.1

PHASE-2 CPL
1400 UT

(N)
5

+ (61
10 + 4.

I I !
J/ 1
-· - - "'
(121
"&
!"' ~-
20 -~ 1 ¥1-v (121

(IJ
\\
~
\
\\.
I
I!
(1 31

a..
...c 50 t ! (161

I I \
w
a:::
:::::>
t \ I
1- (17J

~
w
0:::
a..
100
\ t
I li
t (17J

(1 71
<9-------el US A- F I 1
A-·-.!> USA-AUS (18)
200 +-----+ USA- I NO
(181
(181

(181
US COLDER US WARMER
(191
500
(1 81
(201
(201
(201
(201
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF CDEG) STD DEV CDEG)

Figure 5.1 a) Mean temperature differences in degrees C at constant


pressure levels for 1400UT observation set. b) Standard deviation of
the mean differences in degrees C. Numbers in parenthesis on the right
side of the figure represent the number of observations.
page 45

Fig S-2

PHASE-2 CPL
1700 UT

(NJ
5

(9)
10
t
20
t+ (13J

(16)

" I
"-~ (18)

-~
lO
0....
..c.

w
0:::
50
\ ~-
(18)

(19)

::::>
~ 100
(21)
w
0:::
0.... (21)
<9----e> USA-F I 1
A-· --.~t. USA- AUS
(23)
200 +--+ USA-IND
(23)
(22)

(23)
US COLDER US WARMER
(23)
500
(23)
(23)
(23)
(22)
1000 ' - - - - - - - ' - - - - - - f < t ! & - - - - - . . . . . . . . __ _ ____J ~~--L-------l (21)
-4 -2 0 2 4 0 2 4

TEMP D I FF WEG) S T0 DEV WEG)

Figure 5.2 Same as Fig. 5.1, except for 1700UT observation set.
pa.ge 46

F t g 5. 3

PHASE-2 CPL
2000 UT

(N)
5

(I OJ
10

(13)

(13)
20

(14)

(19)

(19)
w
0::::
::)
(/) (19)
(/) 100
w
0::::
a.. C9----el USA- F I 1 (20J

.&-·-..t.. USA-AUS
(18)
200 +--+USA-I NO
(20)
(20)

(20)
US COLDER US WARMER
(20)
500
(20)
(20J
(20J
·(20J
1 000 L __ _ _---l.----+1151------...L------' '-e)-Bit---~---~ <20 '
-4 -2 0 2 4 0 2 4

TEMP OIFF COEG> S T0 OEV COEG>

Figure 5.3 Same as Fig. 5.1, except for 2000UT observation set.
page 47

Fig S-4

PHASE-2 CPL
2300 UT

(N)
5

(5)

(9)

(12)
20

(14)

Ill
a... (20)
..t: 50

w (23)
0:::
:::::>
~ 100 \ (23)
w
0:::
a... (24)
C9-----e> USA- F I 1
.-.-..t. USA-AUS (24)
200 +--+ USA-IND
(24)

(25)

(25)
US COLOER US WARMER
(25)
500
(25)
(25)
(25)
(25)
1000 '-------L~---(!~=-------'---------' ~~--.1.....------' (25)
-4 -2 0 2 4 0 2 4

TEMP 0 I FF COEG) STD DEV COEG)

Figure 5.4 Same as Fig. 5.1, except for 2300UT observation set.
_page 4'8

FI-g s.s

PHASE-2 StMULTANEOUS
'1400 UT

(NJ

(9)
10
\ + t
I
\
t t { (11)

I I
~
20
\\ "'I
I I
I
(12)

(13)
t t -f
IU l I
Q_
..c 50 ~ ~ (14)

\ ·I
w
0:::
::J
+\ (16)

~
(16)

w
100 t
0:::
Q_
I (16)
&---e> USA-FI 1
a.----.t.. USA-AUS
(17)
200 +----+ USA- I ND

(17)

US COLDER US WARMER (17)


500

(18)

(20)

1000
-4 -2 0 2 4 0 2 4

TEMP DIFF <DE G) STD DEV <DEG>

Figure 5.5 a) Simulataneous comparison of mean temperature differences


in degrees C for the 1400UT observations. b) Standard deviation of the
simultaneous mean differences in degrees C. Numbers in Parenthesis on
the right of the figure represent the number of observations.
page 49

Fig 5.6

PHASE-2 SIMULTANEOUS
1700 UT

5 <N>

10
\ ... t (1 0)

I
\
t t I (12)

20
~ \f (1 J)

\ I (17)
t
ID I
~
a_ (18)
..c 50
I· (17)
w
~
::::>
~ 100 (21)

w
~
a_ (21)
~USA-FI1
.&--·--..e. USA- AUS
(22!
200 +---+ USA- I NO

(22!

US COLDER US WARMER (23)


500

(2JJ

(2JJ
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF <DEGJ STD DEV <DEGJ

Figure 5.6 Same as Fig. 5.5, except for 1700UT observation set.
page 50

Fig 5.7

PKASE-2 SlMULTAKEOUS
2000 UT
(N)
5

(1 0}
10 "t
\ (11)

t
20
1\ (12}

(1 J)
t
I1J \ I
CL
...c. 50 ~ t (19)

\
w
0::: + (19)

::::> I
~ 100
(19)

w
0::: (19)
CL
~USA-Flt
A-·-.t. USA-AUS (18)
200 ~-+ USA-IND

(19)

US COLDER US WARMER (20)


500

(20)

(20)
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF CDEG> STD DEV CDEG>

Figure 5.7 Same as Fig. 5.5, except for 2000UT observation set.
page 51

Fig s.a

PHASE-2 SIMULTANEOUS
2300 UT

5 CN>

(5)
10 4>
I
t
\
t (9)

20 { (Ill

iI (14)

IU
a_ (19)
..s::. 50

w (23)
0:::
:::::>
~ 100
(21)

w
0:::
a_ (24)
~USA-FI1
......_ . ......,.. USA-AUS
(24)
200 +--+ USA- I NO

(24)

US COLDER US WARMER (25)


500

(25)

(25)
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF CDEGJ STD DEV CDEGJ

Figure 5.8 Same as Fig. 5.5, except for 2300UT observation set.
page 52

Fig 5.9

PHASE-2 CPL
1400 UT

300

(f 0)
A
t
t \ I
I \
I I
I \ I
I \ I
+\
CO
0.... 500
...c.
I
f + (1 Q)

I \
w I I \ \
0::
:J I I \
\
I
(/)
(1 Q)
(/)
4
w
0:: \
I
r
0....
e-----e> USA-F I 1 \ I
A-·--4 USA-AUS I I (12)
+--+ USA-IND

I
)/. *\
\
I \
US DRIER
I MOISTER \ (15)
;+
/
r/ (19)

1 000 '------J..._--~-~f-----'------' L-~....&.----1!-----~ (19)


-20 -1 0 0 I0 20 0 10 20

HUMIDITY DIFF (%) STD DEV (%)

Figure 5.9 a) Mean relative humidity differences in percent at constant


pressure levels for 1400 UT. b) Standard deviations of the differences.
Numbers in parenthesis on the right side of the figure represent the
number of observations.
page 53

F 1 g 5.10

PHASE-2 CPL
1700 UT

300 <N>

(9)
+ 't "\
i I \
I I \
\
\
I \
I \ \
10
CL 500
..c.

w
l {
I
l
\
\ I
r (1 J)

0::::
:::>
I \ I
_,.....I I
(/)
(/)
w /
;- (1 J)

0::::
CL / I
<!r---e USA-FI 1 I / I
USA-AUS
&-·-.t. / I
+--+USA-I NO t/ (12)

\
\
US DRIER US HOlSTER (20)

(20)

1000 (20)
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.10 Same as Fig. 5.9, except for 1700 UT.


page 54

FIg 5.11

PHASE-2 CPL
2000 UT

300
<N>

(8)
+ t +
I I \
I I \
I \
I I \
I
{ /~
10
(12)
0... 500
...c.
/
w
I /
a:::
:::)
I /
(f) I /
t,· /
(f) (11)
w
a::: I
t
0... I !I
(9------(!) USA- F I 1 \ I
.t.---.A USA-AUS
\ !I
+--+ USA- I NO A. ! t "\
(9)
I I
I \
I \
I \
I /US HOlSTER
US DRIER
\t (16)

(19)

f
(19)
1000
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.11 Same as Fig. 5.9, except for 2000 UT.


page 55

Fig 5.12

PHASE-2 CPL
2300 UT

(NJ
JOO

...
r
(11)

\
t t
I I \
\ \
I I
\ I \
I
IU \ I \
(L
..c
500
t ~ t (121
._, I
I I
w I I I
a:::
:::::>
en I I I
en
w
I + (141
I ""
a:::
(L
~USA-Fll I
\
\ ""
......_.__. USA-AUS
+--+ USA- I NO I
I
l "~~
I I I
I I I
I I
I I
US DRIER I US HOlSTER
.l I (211
\
/
~ i\.
I
J (21)

1000 (251
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.12 Same as Fig. 5.9, except for 2300 UT.


page 56

FIg 5.13

PHASE-2 SIMULTANEOUS
1400 UT
(N)
300 -

.......
m (7)
(l.. 500 t- A
t
..J::.
I I I
w
0::
I I I
:::::> I I I
(f)
(f) I I I
w I
0::::
I I
(l..
(!)---e) USA-FI 1 I I
I I
.-._. USA-AUS I
+--+ USA- I ND f + t
(7)

I \ I
I I \
I I
I \ I
US DRIER tls HOlSTER
I I \ I
1 I l ! (16)

1000
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.13 a) Simultaneous comparison of mean relative humidity


differences in percent for the 1400UT observation set. b) Standard
deviation of the simultaneous mean differences in percent. Numbers in
parenthesis on the right of the figure represent the number of observa-
tions.
page 57

Fig 5.14

PHASE~2 SIMULTANEOUS
1700 UT
(N)
JOO

«< (9)
a_ 500 A
I+ "!'- +
..c.
I I I
I I
w I I I
0::::
:::> I I
I
Cl)
I I
Cl)
I I I
w I I
0::::
a_ I I I
USA- F I 1
(1)-----el
.t.-._,. USA-AUS
I I I
{ J.
~
(1 0)
+---+ USA- I NO
I I
I \
I I I \
I I I \
US DRIER US MOISTER
I I I \
l t .1 \ (20)

1000
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (i.) STD DEV (i.)

Figure 5.14 Same as Fig. 5 .13, except for 1700UT ·observation set.
page 58

F 1 g. 5.15

PHASE-2 SIMULTANEOUS
2000 UT

(N)
300

10 (9)
()_ 500 +
...c. t I
I I
w
~ I I
::::>
(/) I I
(/)
w I I I
~
()_ I I I
~USA-FI1
I I I
&-·-.A USA-AUS I
+--+ USA-INO +
I
fI t
I
(6)

I I
US DRIER
I
I us MOISTER I
I
1 J t (17)

1000
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.15 Same as Fig. 5.14, except for 2000UT observation set.
page 59

Ftu 5.16

PHASE-2 SIMULTANEOUS
2300 UT

JOO <N>

10 (9)
a..
..c.
500 ~
t ~
I
"t
\
w I I
a::: I \
;:)
en I I \
en I I \
w
a:::
a.. I I \
<9-----e> USA-FI 1 I I \
tI
._ · _,. USA- AUS
I ... (9)
-+---+ USA-IND
I
t I
I
I I I I
I I I I
US DRIER US HOlSTER
I I I I
1 ! 1 J (21)

1000
-20 -10 0 10 20 0 10 20

HUMIDITY DIFF (/.) STD DEV (/.)

Figure 5.16 Same as Fig. 5.15, except for 2300UT observation set.
page 60

Flu 5-17

PHASE-2 CPL
1400 UT

5
<N>

+ ....
(6)
10
\
t t
\"
I I (12)
t f
20 \
\
+
I I
I (12)

~ t t
(1 J)

\ I
ill
\ .I I
(L
...c 50 t I
+t (16)

w
0:::
~
t t I
tf1! (17)

~
w
0:::
100
t 41.
I !I
(17)

(L

t+ 1 I
(9------{!)USA-F I 1
...__.___. USA-AUS ff
ff
(17)

(18)
200
t +--+ USA-IND
(18)

+ (18)

t
(1 !:J)
US LOWER US HIGHER
(19)
500
(18)
(20)
(20)
(20)
(20)
1000
-200 -100 0 100 200 0 50 100

GEOP DIFF (M) STD DEV (M)

Figure 5.17 a) Mean geopotential differences in meters for constant


pressure levels for 1400UT. b) Standard deviation of mean geopotential
difference sin meters. Numbers in parenthesis on the right of the fig-
ure represent the number of observations.
page 61

FIg 5.18

PHASE-2 CPL
1700 UT

5 CNl

10
'\ r i I
(9)

-~ 1;I (1 J)

//
20 r H (16)

\ \
~ .
if (17)

I ~ ++
.I
.......
IQ I \ !I
CL
..s::. 50 t +t (18)

w
0::
::::>
t ff (20)

~·too
w
0::
t 1J
!I
(21)

CL

200
t ~USA-FI1
...__.-.~~.

+--+
USA-AUS
USA-IND
#
1f
(21)

(2J)

(2J)
(2JJ

(2J)
US LOWER US HIGHER
(2J)
500
(2J)
(23)
(2J)
(22)
(21)
1000
-200 -100 0 100 200 0 50 100

GEOP DIFF CM) STD DEV CM)

Figure 5.18 Same as Fig. 5.17, except for 1700UT observation set.
page 62

Fig 5.19

PHASE-2 CPL
2000 UT

5 <N>

(9)
10 t
I
I- (11)

I (1 J)
20 t
I
(14)

10
Q_ (19)
..c 50

(19)
w
a:::
:::>
~
(19)
100
w
a:::
a... (19)
<:9--e:l USA-Fl 1
.&.--- __,. USA- AUS
(19)
200 +--+ USA-IND
(20)

(20)

(20)
US LOWER US HIGHER
. (20)
500
(20)
(20)
(20)
(20)
(20)
1 000 L - - - - - - L - - - - . - - - - - - - ' - - - - - - - - '
-200 -100 0 100 200 0 50 100

GEOP DIFF CM) STD DEV CM)

Figure 5.19 Same as Fig. 5.17, except of 2000UT observation set.


page 63

Ftu 5.20

PHASE-2 CPL
2300 UT

(N)
5

(5)
10 ~ r
\ I
\ J (9)

I (12)
20 -\ t
\ \I
<t tt (14)

IU
\ i\
. \
a..
..c. 50 ~ ~ t (20)

I
w t (2Jl
a:::
::>
; f
~ 100 A
I l~ (24)
w
a:::
I !I
a.. 1 <9--€.l USA-F I 1
(24)

I .t.- ·--.e. USA- AUS ff


200
t +---+ USA- I NO tt
"f
(24)

(24)
t (25)
t f
US lOWER US HIGHER .f (25)

500
.t (25)
(25)
(25)
(25)
(25)
1000 (25J
-200 -100 0 100 200 0 50 100

GEOP DIFF <M> STD DEV (M)

Figure 5.20 Same as Fig. 5.17, except for 2300UT observation set.
page 64

Fig 5.21

PHASE-2 SIMULTANEOUS
1400 UT

5 <N>

10 ... (9)

I (11)
~
20 l (12}

I
A
I (1 J)

I
10
a_ I (14)
..c. 50
t
w
0:::
~ ~ (16)

:::> I I
(16)
~ 100 + +
w
0:::
a_
I I
A (16)
\ (9------e) USA-FI 1
..- ·-.~>. USA- AUS
\
)..
200 ~ +--+USA-I NO
(17)

I \
1 l (17)

I \
USj HIGHER US LOWER \ (17)
A A
500
\ ~
+ ~'*
/
(18)

{20)

1000
-6 -4 -2 0 2 4 6 0 2 4 6

PRESSURE DIFF (h Pa) STD DEV (h Pa)

Figure 5.21 a) Simultaneous comparison of mean pressure differences in


hPa for 1400UT. b) Standard deviation of the simultaneous pressure
differences. Numbers in parenthesis on the right of the figure
represent the number of observations.
page 65

Fig 5-22

PHASE-2 SIMULTANEOUS
1700 UT
(N)
5

10 +A (1 Ol
t \i
J.
\
tt\. (12)

(1 J)
20 t

.......
ID
a..
..c 50 J..
J.
I
I

l/I
t
(1 8)

(19)

I
w t
(19)
0:::
::>
I
~ (21)

w
0:::
100
t
a.. J. &---e> USA-FI 1
(21)

I ........_ · _, USA- AUS


(22)
200 t +---+ USA-IND
I
(23)
- -\
US ~GHER US LOWER (23)
500
\
(23)
~
(2Jl
1000
-6 -4 -2 0 2 4 6 0 2 4 6

PRESSURE DIFF (h Pa) STD DEV (h Pa)

Figure 5.22 Same as Fig. 5.21, except for 1700UT observation set.
page 66

Flu 5.23

PHASE-2 SIMULTANEOUS
2000 UT
(N)
5

(10)
10
t
l (12)

I
(1 J)
20 .f.
I
l (15)

ltl
I
a_
..c 50 J (20)

I
(20)
w f
0::
:::::> I
~
(20)

w
100
f
0::
a_ ~ G---E> USA-FI 1
(19)

I ...._ • __,. USA- AUS


200 + +--+ USA- I NO
(20)

I
.l (20)

I
US fiGHER US LOWER (20)
500
\
(20)

1000
" \.
L-----L-----L-----~----~----~--~
-2 0 2 4 6
(20)

-6 -4 0 2 4 6

PRESSURE DIFF (hPa) STD DEV (hPa)

Figure 5.23 Same as Fig. 5.21, except for 2000UT observation set.
page 67

F1g 5.24

PHASE-2 SIMULTANEOUS
2300 UT

5 <N>

10 ... (5)

\
(9)
~
I (11)
20
t
I
{14)
t
........
CO I
a..
...c 50 ;. (19)

I (2Jl
w
0:::
::::>
f
~ 100 ~ (23)
w I
0:::
(24)
a..
G----el USA-FI 1
"'\ ...._.--. USA-AUS
(24)
200
t +---+ USA-IND
I
(24)
~
US H'tGHER US LOWER (25)
500 ~
\
(25)
l\
'A (25)
1000
-6 -4 -2 0 2 4 6 0 2 4 6

PRESSURE DIFF (h Pa) STD DEV (h Pa)

Figure 5.24 Same as Fig. 5.21, except for 2300UT observation set.
page 68

6.0 Discussion of Radar Results

The Organizing Conunittee fel.t that the advantage in holding Phase


II at the NASA Wallops Island facility was the independent source of
height and wind data provided by the precision tracking radars located
there. All balloon flights were tracked using the FPQ-6 C-Band radar
described in Section 2.7. A second precision radar (FPS-16) was also
used and tracked about 90 percent of the ascensions. This redundancy
was important because height differences that might occur between the
radiosonde systems and the FPQ-6 radar could be confirmed using the
FPS-16 radar.

6.1 Geopotential

Radar position data (x,y,z) are calculated relative to the


tangent plane through the center of the radar antenna. Corrections for
the antenna height above sea level are made to the calculated heights to
obtain true height. To achieve compatibility of radar heights with
radiosonde heights the radar angles and slant ranges at every sample
(one per second) were corrected for the curvature of the earth and the
refractivity of the atmosphere. Radar data are referenced to the earth
(geometric altitude), and for comparison with radiosonde heights must be
converted to geopotentia1 units. The excellent approximation given in
the Smithsonian Tables (List, 1951) was used to convert from geometric
to geopotential heights.

Results of simultaneous comparisons of mean geopotential height


differences between radar measured geopotentials and the Finnish,
Indian, and US radiosonde calculated geopotentials for the 1400 UT
observations (Figure 6.1) were similar up to levels near 200 hPa. At
levels above 200 hPa the mean differences became larger. The geopoten-
tial differences between the Finnish radiosonde and radar are relatively
small, and do not exceed 35 meters except at 10 hPa. Furthermore, the
standard deviation of the differences between radar and the radiosondes
also were relately small up to to about 70 hPa. At levels above 70 hPa
the standard deviation between the radar and the US radiosondes, and the
radar and the Indian radiosondes became very large, e.g., at 10 hPa the
radar minus US and radar minus India exceed 500 meters.

It is obvious from Figure 6.1 that the height differences between


radar and the Australian instrument are excessive and of a different
sign (positive vs negative) relative to the other radiosondes. As
al.ready mentioned, the Australian radiosonde suffered a pressure sensor
calibration problem. The effect of this is characterized by lower gee-
potential heights. This error in geopotentials clearly is attributable
to the pressure measurement error since it was previously shown that the
Australian temperature sensor is identical to the US temperature sensor
(Section 5.2.2) and that these report similar values. The Australian
height discrepancy seen in Figure 6.1, and also evident in Figures 6.2,
6.3,
page 69

and 6.4, arise from the same reason and will not be discussed further in
this section.

The local noon observations {1700 UT) illustrated in Figure 6.2,


show that the geopotential differences between radar and Finland, and
between .radar and the US were similar to each other, as were the magni-
tudes of the standard deviations. The Finnish temperature measurements
were corrected (Antikainen and Hyvonen, 1983), while the US temperature
data were not corrected. Thus, the nearly similar height differences
between these radiosondes and radar, seen in Figure 6.2, were unex-
pected. This similarity is of concern since it is difficult to accept
such agreement with radar when the temperature measurements from these
two radiosondes disagree by 1-2 C (Figure 5.6). These results lead one
to ask whether the Finnish temperature corrections might need reevalua-
tion, or whether the pressure sensors' measurements are seriously in
error. Figures 6.1 and 6.3 are very similar, but are different than
Figure 6.2. Thus, one wonders even more why such discrepancies occur.
A large difference between radar and Indian geopotentials was found and,
although obtained from a smaller sample size, the Graw M60 calculated
geopotential heights were considerably higher than the radar heights.
The standard deviations of the differences for this instrument also were
large and might be considered an indicator of its precision.

Mean geopotential differences derived for the 2000 UT (1500 LST)


observations are presented in Figure 6.3. At levels above 50 hPa the US
radiosonde heights were considerably higher than those measured by
radar. This was expected because of the warmer radiosonde temperatures.
Comparison of Finnish, and Indian radiosonde heights with radar heights
indicated excellent agreement. However, the standard deviation of the
radar vs Indian radiosonde height differences are much larger then those
of radar vs Finland radiosonde height differences. The comparison with
the Graw M60 indicates that at levels above 200 hPa this radiosonde's
heights exceeded radar's by a considerable amount, and in fact suggests
the existence of a serious instrument problem. The Graw M60 measurement
data are adjusted for radiation and lag effects using procedures stan-
dard for this instrument.

The nighttime data at 2300 UT (1800 LST) shown in Figure 6.4,


reveal a uifferent characteristic between radar and radiosonde geopoten-
tials. First, the standard deviations are relatively consistent with
each other, whereas each of the daylight observation data sets, examined
above, showed a larger dispersion. The more consistent variance between
the instruments at night is likely due to the absence of sunlight on the
sensors. The difference in mean geopotential between radar and each of
the Finnish, Indian, and US instruments was about 25 meters at 500 hPa
with the radar heights higher than the radiosondes. Differences between
the Indian and US heights relative to radar heights are about 10 meters
at 70 hPa; while the radar minus Finland height differences reach a
maximum of -45 meters at 100 hPa. At levels above 70 hPa the radar
minus US differences are larger, e.g., almost 75 meters near 30 hPa.
The Finnish and US radiosondes show similar trends at most levels, i.e.,
being nearly the same magnitude at 20 hPa and at higher levels. This
example is similar to that shown in Figure 6.2, i.e., the differences
being similar in spite of corrections being applied to one radiosonde
and not the other. It should be noted however, that the differences
between radiosonde and radar heights at 1700 UT are negative (radar geo-
potentials lower) while at 2300 UT the differences are positive (radar
geopotentials
page 70

higher) . The source of this day-night difference between radar and US


radiosondes lies in the radiative influence on the US radiosonde's tem-
perature sensors.

6.2 Wind

Generally, the comparison of GMD derived winds with radar derived


winds are inherently noisy, with the GMD winds showing greater variation
than radar derived winds. Direct comparison showed that the wind per-
turbations derived with each of the techniques were found to be out of
phase and, at least for the intercomparison results, the GMD wind speeds
also were lower, on average, than the radar wind speeds. Below 10
kilometers the agreement between GMD and radar winds was found to be
good with many of the perturbations in phase. Apparently, during the
early stages of the observations the GMD tracking capability was quite
good, probably due to the higher antenna elevation angles and unambigu-
ous reception of the transmission signal (i.e., without having encoun-
tered previous reflections) . At long ranges and low elevation angles,
even small antenna motions, whether due to tracking ambiguities or poor
resolution of the antenna angle recordings result in large wind errors,
resulting n increasingly noisier GMD winds.

Examples of GMD performance relative to radar is given for the


component winds in Figure 6.5 for ascension number 166 made on 25 Febru-
ary 1985. Below 5 kilometers altitude good agreement exists, however,
above 5 kilometers the GMD winds were not only noisy, but the East-West
component is in serious disagreement. This example should be contrasted
to Figure 6.6 for ascension number 181 made on 1 March 1985, which indi-
cates relatively good agreement in the lower layers and much less
noisier data in the higher layers, however, the radiosonde heights of
this observation were much higher than that of the radar's due to a
leaking baroswitch cell which becomes apparent near 23 kilometers.

Omega derived winds from the Finnish MicroCora system were in


good agreement with the radar derived winds as seen in ascension number
119 data of 7 February 1985 shown in Figure 6.7. Ascension number 119
represents an extremely high wind speed situation. On the other hand
however, the low wind speeds of ascension number 129 made on 13 February
1985 shown in Figure 6.8, also indicate that omega derived winds com-
pared with the radar winds were quite comparable in many ways. It is
especially important to note that the characteristic of the Omega
derived wind to emphasize the small perturbations not seen in the radar
derived wind profile. Larger oscillations (long wavelength) are not
affected, however. Winds derived from GMD tracking and Omega navaid
methods when compared directly indicate biases similar to those noted
between GMD and radar.

A comparison of Omega navaid winds using the MicroCora and


DigiCora ground systems is shown in Figure 6.9 for ascension number 181
made on 1 March 1985. The DigiCora wind reduction used a two and one-
half minute integration (15 Omega transmissions) and the MicroCora winds
were integrated over four minutes (25 Omega transmissions) . Except for
the peaks that occur in some of the shorter vertical perturbations the
agreement
page 71

between the two systems is good. These same two profiles shown in Fig-
ure 6.9 can be compared with Figure 6.6. It can be seen that the navaid
derived winds are comparable to the radar derived winds.
page 72

fl g 6.1

PHASE-2 SIMULTANEOUS
1400 UT

5 (Nl

(7)
10 r tXI"
. Il(
"
'1-. I I )',
I t )< (I Ol
X\ " . I I
\ \
I ...I ~I (1 0)
~\
20
I . I I I
. I
I \ I II
x· f f:1: (11)

\ . . I
10 I I I I'I
a... \.
J (13)
.c. 50
I
w
a::: .f. (15)

::> I
~ 100 f (16)

w
a:::
a...
I (16)
t ~RDR-FI1
..- . .-. RDR-USA
200
fI +--+ RDR- AUS
*-~ RDR-IND
(16)

(16)

RDR LOWER RDR HIGHER (17)


500

(18)

(19)
1000
-2000 -1000 0 1000 2000 0 1000 2000

GEOP DIFF <M> STD OEV (M)

Figure 6.1 a) Simultaneous comparison of mean geopotential differences


in meters between radar and radiosondes for 1400UT. b) Standard devia-
tion of geopotential differences, in meters, between radar and
radiosondes. Numbers in parenthesis represent the number of observa-
tions.
page 73

Fig 6-2

PHASE-2 SIMULTANEOUS
1700 UT

5 <NJ

(9)
10

/
/
Jr
1···,·. (12}

I ti /)'>
20 /I IfI' /.~ (14)

I
f .J.il
I :
(17)

«< I /I .

~ 50 J t ~·' (18)

I I
w .J. (18)
0::
:::::> I
~ 100 t (21)

w
0::
a...
I (21)
.t
I
(9-----e> RDR-FI 1
........_._.. RDR-USA
(22)
200 +--+ RDR- AUS
*-~ RDR-IND
~ ... _. RDR-M60 (23)

RDR LOWER RDR HIGHER (23)


500

(23)

(23)
lO~~OLO_O_______-l~OL00--------~0~-------1~00-0------~2000 0
1000 2000

GEOP 0 I FF (M) STD DEV (M)

Figure 6.2 Same as Fig. 6.1, except for 1700UT observation set. Sample
size between radar and Graw M60 radiosonde is smaller.
page 74

Ftg 6.3

PHASE-2 SIMULTANEOUS
2000 UT
5 <Nl

10 '\
/
..... t r- / • (8)

\
/j~·
\ I (1 0)

~~
\A
~

.....
.....
\
\
II I/ ~
(12}

«<
~

' I
f
f.l
I
( 1 4)

a_
..c 50 ~ J (19)

I
w ' ·~ .j. (19)
0:::
::::> I
~ 100 (19)

w
~ f
a:::
a_
I (18)
t (!)-----(!)RDR-F I 1
.__ • __.. RDR- USA
200
fI +---+ RDR- AUS
;x.,.--x RDR-IND
(19)

~--·~ RDR-M60 (20)

RDR LOWER
I RDR HIGHER (20)
500

(20)

(20)
1000
-2000 -1000 0 1000 2000 0 1000 2000

GEOP DIFF (M) STD DEV <Ml

Figure 6.3 Same as Fig. 6.1, except for 2000UT observation set. Sample
size between radar and Graw M60 radiosonde is smaller.
page 75

Fig 6.4

PHASE-2 SIMULTANEOUS
2300 UT

5
<N>

10 ~
,.... ~
(5)

\ / I
/ ~ (8)
~
I
/ I
I
20 ~
I
I •
:
(11)

\
I (I 3)
~ t
.......
«< I
0....
..c. 50 ./. (18)

I
w
0:::
t (22)

~ I
~ 100 f <23)
w
0::: I (23)
0....
.t
I
G---e> RDR-FI 1
,._ · _.,. RDR- USA
(23)
200 +--+ RDR- AUS
>fo--x RDR-IND
._ ... ~ RDR-M60 (23)

RDR LOWER RDR HIGHER (24)


500

(24)

(24)
1000
-2000 -1000 0 1000 2000 0 1000 2000

GEOP DIFF (M) STD DEV (M)

Figure 6.4 Same as Fig. 6.1, except for 2300UT observation set. Sample
size between radar and Graw M60 radiosonde is smaller.
page 76

fig 6.5

ASCENSION NO. 166 ON 25-FEB-85 AT 16•45•12 UT

-20 0 20 40 60 80 -20 0 20 40 60 80
40 r-------.-------,-------,--------r-------, .-------r------~-------,------~r-------,40

35 35

30 30

USA GMD
RADAR
25 25
i:
~

w 20
0 20
::>
I-

I-
_J
<(,
15 -, .....

\..._
15
'\
..... .,
I
/
10 < 10
.I
/

--
/
(

I
.
--
)
5 5
/

0 0
-20 0 20 40 60 80 -20 0 20 40 60 80

EW WIND (M/S) NS WIND (M/Sl

Figure 6.5 Comparison of the US GMD derived winds with radar derived
winds for Ascension No. 166 made on 25 February 1985. The GMD results
are considerably noiser than the radar's.
page 77

ASCENSION NO. 181 ON 01-MAR-85 AT 14•00•19 UT

35

30

USA GMD
RADAR
25

1--
....J
< 15

10

EW WINO (M/S) NS WINO (M/S)

Figure 6.6 Similar to Fig. 6.5, except for Ascension No. 181 made on 1
March 1985. GMD results compare well with radar, however, the
radiosonde pressure cell was leaking resulting in radiosonde heights
higher than the radar's.
page 78

Fig 6.7

ASCENSIO~ NO. 119 ON 07-FEB-85 AT 22•40•33 UT

35 35

30 30

USA GMD
RADAR
25 25
MICROCORA
i:
~

w 20
Cl
.._
;::)
~-l
.._ ·-~
...J _.-;;:>
< 15 15

10 10

EW WINO (M/S) NS WINO (M/S)

Figure 6.7 Comparison of Omega navaid derived winds with radar for
ascension no. 119 on 7 February 1985 during a very high wind situation.
page 79

fig 6.8

ASCENSION NO. 129 ON 13-FEB-85 AT 18•51 •07 UT

35
-/

30

MICROCORA
RADAR 25
25
~

::c
:><::

w 20
Cl 20
:::>
1-

1-
_J
< 15 15

EW WINO <MIS) NS WINO (M/S)

Figure 6.8 Similar to Fig. 6.7, except for Ascension No. 129 made on 13
February 1985 during light wind conditions.
page 80

Ftu 6-fl

ASCENSION NO. 181 ON 01-MAR-85 AT 14•00•19 UT

EW Wl NO (M/SJ NS WINO (M/SJ

Figure 6.9 Comparison of winds from Finnish MicroCora and DigiCora sys-
tems for Ascension No. 181 made on 1 March 1985.
page 81

7.0 Specific Analysis

Early in the observational period it was noted that pressure


measurements made with the Australian radiosonde were consistently lower
than pressure measured by the other instruments. Since this anomaly
required an answer as rapidly as possible a different lot of radiosondes
were air-shipped from Australia to Wallops Island. Subsequent flights
with the new instruments revealed that these radiosondes also exhibited
similar pressure characteristics. Since it was apparent that the prob-
lem could not be fixed in the field, the Australian Team Leader elected
to continue making observations with the strong possibility of resolving
the problem after completion of the intercomparison. Since the error in
pressure had a definite bias (always lower), this decision appeared rea-
sonable.

The Australian Bureau of Meteorology and Philips Communication


Systems (PCS) carried out a number of tests on the RS4-MK3 radiosonde
(Stickland, personal communication). These tests were designed to iso-
late possible causes of the pressure bias. PCS had considered that a
possible cause could be friction and inertia effects in the
baroswitch/aneroid cell linkages. Pressure biases similar to those seen
in the intercomparison flights were not observed during the tests and
further analysis indicated that friction and inertia effects did not
appear to be significant error sources.

Varying the friction pressure of the commutator contact arm of


the baroswitch had little effect on the calibration readings at ambient
pressures, and virtually no effect on low pressure readings. The cali-
bration procedure was altered so that calibrations were obtained at
ambient temperature and again at about -40 C temperatures at identical
pressure exposures. This change did not significantly improve calibra-
tion accuracies, nor did the differences observed correlate with the
magnitude and sign of the differences found in the flights of Phase II.

Further tests revealed that the temperature of the working stan-


dard pressure transducer used by PCS for baroswitch calibrations could
differ significantly from the ambient air temperature, particularly in
early morning. It also was discovered that the temperature coefficient
of the transducer had increased in value by up to 50 percent from the
original value. These factors could result in pressure errors of up to
+4 hPa for some batches, but the average error of all batches calibrated
was estimated to be only +0.8 hPa.

Analysis of residual differences resulting from this test indi-


cate that there is an apparent temperature coefficient of -0.033 hPa/ C
existing during simulated and actual flight. The temperature is that of
the air surrounding the radiosonde during flight. The derivation and
application of corrections using this coefficient reduced the difference
in the flight similitude runs. The application of this correction to
the Phase II flights (ascension numbers 139 to 201) of the
intercomparison
page 8.2

also reduced the differences to within +/- 1.0 hPa. Differences


obtained with the earlier flights (ascension numbers 119 to 138) were
reduced only to +3.3 and +2.9 hPa for the 700 and 500 hPa levels,
respectively. The cause, or source, of these differences remains unex-
plained. Probably they arise from a combination of errors inherent in
the particular batch of radiosondes used (e.g., possibly calibrated
early in the day before the pressure transducer temperature stabilized),
and the need to modify release techniques during Phase II.

In Australia, PCS applied corrections to remove the error bias


resulting from the dynamic performance of their radiosondes by utilizing
the -0.033 hPa/ C temperature coefficient to correct measured pressures.
PCS also have arranged for temperature stabilization of their working
standard pressure transducer and direct monitoring of its temperature
during calibration operations to effectively remove the contribution to
pressure measuring errors resulting from this source. The Australian
Bureau of Meteorology has taken steps to monitor flight similitude
results to check that the magnitude of the temperature coefficient
remains constant, and is to revise the corrections if this is not found
to be the case.

7.1 Comparison of Phase II, Australian Data After Correction

The temperature coefficient (-0.0323hPa/ C) developed by The


Bureau of Meteorology and PCS to correct measured pressures was applied
to all the Australian radiosonde measurements obtained in Phase II. The
improvement was quite noticeable as described in Figures 7.1 to 7.6.

7 .1.1 Pressure

The Australian observations made at one-minute intervals were


available at Wallops Island making it relatively simple to adjust all
pressures using the PCS temperature coefficient. After correcting the
pressures new geopotentials were calculated since the pressure-
temperature relationship was now different from the original data set.
Techniques for comparing mean data are similar to those discussed ear-
lier. A discussion of only the daytime (1700 UT) and nighttime (2300
UT) differences between the Australian and US radiosondes is given along
with the standard deviation of the differences in Figures 7.1 and 7.2,
respectively. The solid curves represent the difference before the tem-
perature coefficient was applied. These curves are identical to those
shown in Figure 5.22 and 5.24.

After applying the temperature coefficient the Australian pres-


sures were then compared to the US pressures. The resulting improvement
in agreement is shown by the dashed curves of Figures 7.1 and 7.2.
Clearly, the adjustment of the pressures removed a large part of the
measurement bias and did not affect the measurement variability as indi-
cated by the standard deviations.

7 .1.2 Temperature

The constant pressure level temperature comparison between the US


and Australian radiosondes previously shown in Figure 5.2 for 1700 UT
are
page 83

shown again in Figure 7.3 (solid curve). After pressure adjustments


were made to the Australian instrument the difference with the US
radiosonde's temperature data were notably improved, as shown by the
dashed curve. The improvement begins near the 500 hPa level; differ-
ences between US and Australian temperature measurements are now no
larger than 0.4 C, except at the 10 hPa level. Thus, temperatures meas-
ured with the Australian radiosonde correspond more nearly to the
correct pressure. A similar temperature difference (US-Australia) also
is seen for the nighttime data in Figure 7.4. Although, in this case,
the 15 hPa and 10 hPa levels show larger differences relative to those
at the lower altitudes.

Corresponding to the improvement of the mean pressure-temperature


relationship is the improvement in the standard deviation of the differ-
ences. The US temperatures were not changed relative to the pressure
surfaces, therefore, the standard deviation of these data did not
change; however, since the temperatures of the Australian instrument are
now reported at more nearly the correct pressures, the standard devia-
tion of these measurements will change, as can seen in Figures 7.3 and
7.4. The smaller variance also indicates better agreement between the
two radiosondes since the simultaneous readings suppresses variability
coming from the atmosphere. The improved standard deviations of the
differences are now less than 1 C to ievels near the 20 hPa level for
the daytime and nighttime sets.

Other methods of determining precision of radiosonde parameters


are discussed in Section 8.0. The root-mean-square differences of the
US-Australian pairs are very similar to the estimated precision derived
in Section 8.0; either of the methods discussed excludes any variation
that might arise from atmospheric variability.

7.1.3 Geopotential

Comparison of daytime geopotentials between Australia and the US


after the Australian pressure measurements were corrected indicates a
more consistent relationship in the vertical (Figure 7.5). The earlier
(original) US-Australia differences indicated that the US geopotentials
were higher between the surface and about 20 hPa. At levels above 20
hPa the US geopotentials become lower than those of Australia. This
was, again, symptomatic of the pressure error. After calculation of new
geopotentials the differences in the vertical became consistent and
smaller, overall.

The same result is to be noted for the nighttime observations


(Figure 7.6). In spite of Australia and the US using identical thermis-
tors, there remain differences in geopotential. In all probability this
is caused by residual errors in the pressure measurements. It is impor-
tant to note the root-mean-square-differences (rmsd) of the rod-
thermistor temperatures of the two radiosondes. The values at the con-
stant pressure surfaces were calculated and compared with other methods
of determining precision (see Section 8.0). It is concluded that the
small rmsd calculated between Australia and US temperature measurements
is equivalent to the rmsd of dual radiosonde flights of the US instru-
ment as reported by Schmidlin, et al, (1982) and by other techniques
utilized in section 8.0.
page 84

7.2 Indian Radiosonde Reduction Technique

During Phase I of the intercomparison test, three radiosonde sys-


tems used automated data reduction methods and one relied on strip
charts and manual reduction methods. The number of mistakes clearly was
less and measurement consistency was better for the automated systems
(Hooper, 1986; Nash and Schmidlin, 1987). During Pha.se II three
radiosondes systems used manual reduction techniques (India was one) and
one system was totally automated. It was apparent that a larger number
of observer mistakes existed in the manual data that required consider-
ably more editing and quality control checks.

Two of the participants using manual techniques were able to meet


the requirements of the Organizing Committee, i.e., they provided con-
stant pressure and one-minute level data. India followed somewhat dif-
ferent radiosonde data selection criteria. The operator selected only
significant level data which then were interpolated at constant pressure
and one-minute levels. Additionally, the computer processing equipment
brought to Phase II by the Indian participants was limited by the
software design to 39 input levels. These limitations forced interpola-
tion of the variables over relatively deep layers.

Inspection of the Indian strip chart recordings showed that


interpolation between significant levels quite often distorted important
temperature lapse rate changes. The effect of this was excessive varia-
bility at standard levels and even at the one-minute levels. The Indian
participants agreed to change their procedure and re-submitted their
data at one-minute intervals. Interpolation at the constant pressure
levels was still required, but the results were considerably improved.

It must be borne in mind that the Organizing Committee was


interested first, in the data used by the worldwide community that nor-
mally are reported over GTS; second, in the quality of the various
instrument designs. The intercomparison was designed with these two
objectives in mind. Thus, the analysis of the Indian data serves as a
focus on observational procedure vs instrumental quality and the effects
caused by each.

7.2.1 Comparison of Indian MK III Radiosonde Data

Differences between the initially reported Indian temperature


measurements (i.e., those interpolated between significant levels) and
US temperature measurements are shown by the solid curve in Figure 7.7.
Comparison of the reprocessed Indian temperature measurements interpo-
lated from one-minute interval levels and the US temperature measure-
ments similarly obtained is indicated by the dashed curve. The original
data set indicates that the US-Indian mean differences were large at
levels above 50 hPa, as were the standard deviations. This suggests
large variability in one of the radiosonde systems. After analysis of
the new data, the agreement in the mean difference improved and there
was a reduction in the standard deviation from about 1.0 C to about 0.5
C up to the 50 hPa level. Since the US data were not changed, the
reduction in the magnitude of the standard deviations obviously . came
from the improvement in measurement consistency brought about by better
selection of data by the Indian participants. Comparison of the data at
2300 UT (Figure 7.8) showed little difference in the size of the mean
difference,
page 85

but did show a reduction in the size of the standard deviation.

Similarly, the geopotential differences, shown in Figures 7.9 and


7.10 were also improved. At 1700 UT (Figure 7.9) the mean differences
were found to improve by almost a factor of 2 beginning near the 200 hPa
level. The standard deviation was considerably smaller indicating much
less variability at levels as low as the 500 hPa surface. The standard
deviation at 2300 UT (Figure 7.10) is smaller and is indicative of more
consistent measurements.

The figures are important, not only because they demonstrate com-
patibility of the two radiosondes but because they raise the question of
whether the Indian instrument is a bad instrument, or whether the data
were basically mishandled through the reduction procedure used.

The answers to these important questions lie in a comparison of


US-Indian pressure measurements at simultaneous times. Comparison of
US-Indian pressure differences indicated large mean differences and
large standard deviations. It is believed that these large differences
stem from interpolation of one-minute interval data from the significant
level data. After data were read from the Indian strip chart recordings
at one-minute intervals a notable improvement occurred. This improved
agreement of mean pressure differences suggests that: the Indian pres-
sure measurements are not bad, as the original data set would imply,
and, when properly selected are then comparable to US pressure measure-
ments, in the mean. This improvement results from a more precise read-
ing of pressure data forced by using one-minute data points. It is con-
cluded that interpolating between significant levels provides a poor
representation of pressure and, consequently temperature and should
never have been used.
page 86

FIg 7.1

PHASE-2 SIMULTANEOUS
1700 UT

5
CN>

(19)
10 't
...\ (21)
\
(21)
20

(25)

ID
Q_
...c 50 1 (25)

I
(25)
w
0:::
J.
::::> I
~
(25)

w
100 t
0:::
CL 1 (9----(!) USA-AUS
(25)

\ ...._·---.A USA- AUC


(25)
200 ~
I
1 (25)

I
US LOWER (25)
500

(25!

1 __j (25)
1000
-6 -4 -2 0 2 4 6 0 2 4 6

PRESSURE DIFF (hP a) STD DEV (hP a)

Figure 7.1 a) Simultaneous comparison of mean pressure differences in


hPa at 1700 UT between US (USA) and Australian radiosondes before (AUS)
and after (AUC) the pressure sensor correction was applied.
page 87

Frg 7.2

PHASE-2 SIMULTANEOUS
2300 UT
5 (Nl

(9)
10 ~
\... (12)

I
20 ! (15)

I
l (17)

,...., I
ID
a....
..c 50 l (23)

I
w f. (25)
0:::
::::> I
~ 100 f
(25)

w
0:::
a.... ! <9-----0 USA- AUS
(26)

\ .s.- ·-. USA- AUC


~
(26)
200
I
J (26)

I
US LOWER (26)
500

(26)

(26)

1000
-6 -4 -2 0 2 4 6 0 2 4 6

PRESSURE DIFF (hP a) STD DEV (h Pa)

Figure 7.2 Same as Figure 7.1, except for 2300UT. observation set.
page 88
Ftg 7.3

PHASE-2 CPL
1700 UT

<Nl

(14)

"t
\
~
/
(21)

I
~ f
(23)

\
(25)

10
a_ (25)
..c.

(25)
w
a:::
:::::>
~
(25)
100
w
a:::
a_ (25)
&---e:~ USA- AUS
.t.-·.......A USA-AUC

(251
US COLDER US WARMER
(25)
(25)
(25)
(25)
(24)
<23)
-2 0 2 4 0 2 4

TEMP DI FF <DE G) STD DEV <DEGJ

Figure 7.3 a) Mean temperature differences at the constant pressure


levels for 1700 UT between US (USA) and Australian radiosondes before
(AUS) and after (AUC) the pressure sensor correction was applied to the
Australian radiosonde. b) Standard deviation of pressure sensor
differences. Numbers in parenthesis on the right of the figure
represent the number of observations.
page 89

Ftu 7.4

PHASE-2 CPL
2300 UT

(N)

(6)

L' (12)
/
.1(' {14)

I
J. (17)

.-. I
ID
a... A
I (23)
..s::.
\
w ~ (25)
0::
:::::> I
~ 100 t
(26)
w
0::
a... I (26)
USA-AUS
(9-----(!)

.t.- ·--.~~. USA- AUC


~
(26)

(26)
(26)

(26)
US COLDER US WARMER
(26)
(26)
(26)
(26)
(26)
(26)
-2 0 2 4 0 2 4

TEMP DIFF <DEGJ STD DEV <DEG)

Figure 7.4 Same as Fig. 7.3, except for 2300UT observation set.
page 90

Fto 7-5

PHASE-2 CPL
1700 UT

5 CNJ

(14)
10

(21)

(2J)
20

(25)

10
a.. (25)
..c 50
.......
w (25)
0:::
:::::>
~ 100
(25)
w
0:::
a.. (25)
G---e> USA- AUS
._.___... USA-AUC
(25)
200
(25)
(25)

(25)
US LOWER US HIGHER
(25)
500
(25)
(25)
(25)
(24)
1000 (2Jl
-200 -100 0 100 200 0 200 400

GEOP DI FF <M> STD DEV <M>

Figure 7.5 a) Mean geopotential differences at the constant pressure


levels for 1700 UT between US (USA) and Australian radiosondes before
(AUS) and after (AUC) the pressure sensor corrections were applied to
the Australian radiosonde. b) Standard deviations of the geopotential
differences. Numbers in parenthesis on the right of the figure
represent the number of observations.
page 91

Ftg 7.6

PHASE-2 CPL
2300 UT

5 <N>

(6)
10

(12)

(14)
20

(17)

.-..
11
a.. (2J)
J: 50

w (25)
0:::
:::>
~ 100 (26)
w
a:::
a.. ~USA-AUS
(26)

..-.___. USA-AUC
(26)
200
(26)
(26)

(26)
US LOWER US HIGHER
(26)
500
(26)
(26)
(26)
(26)
(26)
1000 ~--------~------------------~--------~
-200 -100 0 100 200 0 200 400

GEOP 0 I FF CM> STD DEV <M>

Figure 7.6 Same as Fig. 7.5, except for 2300UT observation set.
page 92

FIJJ 7.7

PHASE-2 CPL
1700 UT

(NJ
5

(6.)
10
I I'
1
I (12)
20 t
(13)

ID
a_ (14)

-
..s::.

w
50
(14)

0:::
::::>
en (14)
en 10.0
w
0:::
a_ (14)
<9----E> USA-OLD
.t.--_. USA-IND (15)
200
(16)

(15)

(16)
US COLOER US WARMER
(16)
500

(16)
(16)
(16)
(16)
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF COEG> STD OEV <DEG>

Figure 7.7 a) Mean temperature differences at the constant pressure


levels for 1700 UT betwen US (USA) and India. OLD indicates the origi-
nal Indian data reduced by Indian participants; IND indicates the same
observations after reprocessing by the Indian participants. b) Stan-
dard deviations of the temperature differences. Numbers in parenthesis
on the right of the figure represent the number of observations.
page 93

Ftg 7.8

PHASE-2 CPL
2300 UT

5 <Nl

10 ~
... (5)
I
I
I
20 L (10)

/
/
(11)
f'
"""'
ID
I
a_
.s=
...., so J (18)

I
w 't
(18)
a::
::::>
\
~ 100
(18)

w
a::
a_ (19)
G--(1) USA-OLD
.t.-·......t. USA-IND (19)
200
(19)

(19)

(19)
US COLDER US WARMER
(19)
500

(19)
(19)
(19)
(19)
1000
-4 -2 0 2 4 0 2 4

TEMP DIFF <DE G) STD DEV CDEG>

Figure 7.8 Same as Fig. 7.7, except for 2300UT observation set.
page 94

Ft.g 7.9

PHASE-2 CPL
1700 UT

5
<N>

(6)
10

(12)
2D

(13)

..,
~. 50
(14)

(14)
w
et:
::::>
~ 100
w
et:
a_ (14)
G----0 USA-OLD
......_ · --.r. USA- I ND
(15)
200
(16)
(15)

(16)
US LOWER US HIGHER
(16)
500

(16)
(16}
(15)
1000 L------'-------e------'-----~ ·f------1-------' (16)
-200 -100 0 100 200 0 200 400

GEOP 0 I FF <M> ST 0 OEV (M)

Figure 7.9 Same as Fig. 7.7, except for mean geopotential differences.
page 95

Fig 7.10

PHASE-2 CPL
2300 UT

5
(Nl

(5)
10 t
I
I
20 ./. (1 0)

\ I
t + (11)

......., \ I
a..
..s::. 50
......, l ! (18)

\ I
w t t (18)
a:::
::> I
I
~ 100
(19)

w \ I
a:::
a.. (19)
~USA-OLD
..._.___.. USA-IND
(19)
200
(19)
(19)

(19)
US LOWER US HIGHER
(19)
500

(191
(19)
(19)
(19)
1000
-200 -100 0 100 200 0 200 400

GEOP DIFF (M) STD DEV <MJ

Figure 7.10 Same as Fig. 7.8, except for mean geopotential differences.
page 96

8.0 Instrument Precision Estimates

In spite of the previous discussion concerning measurement compa-


tibility of the radiosondes flown in Phase II nothing has been suggested
about instrumental measurement precision. All participants were
requested to provide info.rmation on their i.nstrument' s precision prior
to the start of the intercomparison. Finland, Australia, and the United
States replied to this request. Much of the information received from
Finland and the US was already available, either in published reports or
as part of the manufacturers' specifications. The information for the
Australian radiosonde RS4-MK3 (Stickland, personnel communication) was
provided for 18 pressure levels.. The values given by this information,
however, .we.re obscure as to whether they actually represented repeata-
bility~ accuracy, or precision.

8.1 Method/Rationale

While significant information about measurement differences and


reliability have been derived from the intercomparison data, questions
still exist about instrument precision. Different approaches to deter-
mine instrumental precision have been employed from time to time. In
this section, a technique to separate instrumental measurement errors
from true variations of the measured parameter is employed. This tech-
nique was originally developed for testing of timers used to measure
artillery shell fuse delay times (Grubbs, 1948). The method is best
employed when multiple measurements are made of a single parameter with
several instruments simultaneously. This technique was extended to
atmospheric measurements of the intercomparison. The meteorological
parameters to be measured are variables and the instruments can only be
used once, so that repeated measurements with the same instrument are
not possible. The example outlined by Grubbs in his 1948 paper was used
as an analog for measurements made with the four radiosonde instruments.

Both the constant pressure level and at one-minute interval data


sets were evaluated. Grubbs method assumes that a measurement is com-
posed of the true value and an instrumental error. In the case of the
atmosphere, the true value varies over time at the same location, or
with location at the same time; instrumental error is present for a
given instrument design whenever measurements are being made. If a
series of measurements is made with the same instrument the variability
is composed of the parameter variation and the instrument variation. In
this example, individual variances can not be separated. However, if
several instruments are used to independently measure the same parameter
simultaneously the parameter and instrument variability can be
separated. This permitted about 70 samples of temperature, pressure,
relative humidity, and geopotential measurements to be included in the
analysis. If radar measurements are included to examine the estimate of
the precision of geopotential than the test must be applied to five
instruments. (The Graw M60 radiosonde was not included simply because
of its smaller sample size) . The general case used to estimate the
variances,
page 97

or precision, is

N-I

where N is the number of instruments and must be equal to 3 or more, and


< S~1.-e" S€t·-e1. )
are the sample variances of the differences of the meas-
urementsJJ 1'-
o the •
f1rst •
1nstrument (1) through k. •
Rotat1on of t h e su b -
scripts provides the estimated variance of each of the radiosondes.
Lack of space inhibits a more detailed discussion and the reader is
referred to Grubbs (1948, 1973) for a more comprehensive explanation.

8.2 Constant Pressure Levels

The test for precision was applied to the constant pressure level
parameters of temperature and geopotential. The estimates derived are
representative of operational radiosonde reports, i.e., data that are
reported over GTS and used by analysis centers. Data reported over the
GTS contain errors because of errors in the variables, and in their
transcription. Thus, reported temperature and geopotential values are
not compatible among reported radiosonde data. This lack of comparabil-
ity is of concern to all analysis centers since the errors encountered
are, in all liklihood, larger than the true instrument error. This dis-
cussion of precision of reported data is important for that purpose. In
the context of operational upper-air measurements, one must keep in mind
that pressure is an independent variable and by definition, at constant
pressure levels, is exact. The dependent variables are temperature and
relative humidity that vary not only from errors inherent in the sen-
sors, but also from errors due to the radiosonde's position resulting
from the error in the pressure sensor. In a similar fashion, geopoten-
tial precision at fixed pressure surfaces were examined. Here, geopo-
tential is affected only by errors of the temperature measurement.

The tests applied here relate only to the estimated precision of


operationally reported measurements and do not reflect true instrument
precision. Instrument precision is discussed in Section 8.3.

8.2.1 Temperature

Examination of the Australian, Finnish, Indian, and American tem-


perature measurements led to estimates of precision of the reported
variable as shown in Figure 8.1. A description of temperature measure-
ment precision for each of the instruments is given below.

United States - The US radiosonde which uses a relatively large


rod thermistor indicates measurement precision, on average of 0.30 C
between the surface and 100 hPa, 0.32 C to 20 hPa, and decreases signi-
ficantly to 0.89 C between 20 hPa and 10 hPa. The estimated precision
actually reached 1.40 C at the 10 hPa surface. However, the rapid
decrease
page 98

in precision at levels above 20 hPa shown in the figure may be due . more
to the smaller sample size than the actual capability of the sensor.
Nonetheless, the average t.emperature measurement precison between the
surface and 10 hPa was calculated to be 0.42 C, similar to the rmsd
value derived earlier by Schmidlin, et al (1982), who reported using
paired US instruments on a single balloon.

Finland - The Finnish instrument uses a capacitive temperature


sensor which is believed to have a lag response similar to the rod
thermistor (see section 4.0). The Finnish thermistor has a highly
reflective aluminum coating. The temperature measurments have an
estimated precision of about 0.35 C between the surface and 100 hPa, and
0.46 C between surface and 30 hPa, as indicated in Figure 8.1. The pre-
cision gradually decreases until it reaches a value of 2.0 Cat 10 hPa.
However, the sample size is small at these upper altitudes. The average
precision value between surface and 10 hPa was calculated to be 0.67 C.

Australia - The RS4-MK3 radiosonde used the same rod thermistor


that was used with the US radiosonde. Results similar to those of the
US temperature precision should be expected. However, Figure 8.1 indi-
cates that the operational temperature measurements from the Australian
instrument are not as precise as either the US or Finnish instruments.
For example, between the surface and 100 hPa the precision was found to
be 0.46 C, between the surface and 20 hPa was 0.67 C, and the average
precision between the surface and 10 hPa was 0.82 C. However, beginning
at the 150 hPa surface the precision of the Australian temperature meas-
urements gradually worsened and reached 2.30 C at the 10 hPa surface.
It should be remembered that the Australian radiosonde pressures were
not correct for this series of measurements and the temperature measure-
ment precision is affected. Later, it will be shown in the analysis of
simultaneous measurements that the Australian temperatures are quite
comparable to the US temperatures.

India - The results obtained from the original Indian measure-


ments are evaluated as well as the re-submitted Indian data. Figure 8.1
shows precision information only up to 20 hPa since the value produced
at 10 hPa is unusable because of sample size. Between the surface and
100 hPa, the precision of the reported Indian radiosonde data is about
0.86 C. Between the surface and 20 hPa the estimated precision
decreases to about 1.07 C. After reprocessing of the Indian radiosonde
flights the average estimated precision between surface and 100 hPa
improved to 0.66 C, and between surface and 20 hPa, to 0.71 C. The
average precision estimate between surface and 10 hPa was 0.84 C. It is
clear that selecting data at each minute of flight, as opposed to inter-
polating each minute from the significant level selection, improved the
estimated precision of reported Indian temperature measurements.

8.2.2 Geopotential

The height precision of the four instruments were analyzed in a


similar manner. The calculation of geopotential requires temperature
and pressure measurements, thus error in geopotential arises from errors
in these parameters. An analysis of reported geopotential in the
present case is then only an analysis of how the errors in pressure and
temperature measurements enhance or cancel each other, or otherwise
interact to provide valid heights.
page 99

Figure 8.2 depicts the precision of geopotential calculations.


The results shown suggests that the precision of the reported US
radiosonde geopotentials is rather high in relation to the others shown.
This is followed by Australia. The implication of rather good precision
from this instrument seems to be that the pressure sensor has reasonably
good stability in spite of its large mean error. The Finnish radiosonde
seems to have a precision between surface and 100 hPa comparable to that
of Australia, but decreases at levels above 100 hPa. India (original
data) was the least precise of the instruments flown in Phase II, but
the results using the re-submitted data were considerably improved.

The analysis of temperature and geopotential precision is basi-


cally an exercise that complements the results of section 5.0. The pro-
cedure only attempts to prioritize the order of precision of each of the
instruments. This obviously works well for the manner in which India
treated its original and re-submitted data. In all cases, however, the
errors in temperature are correlated with the pressure errors. For this
reason the results may not be entirely valid in spite of being con-
sistent.

8.3 Instrument precision

Instrumental precision can best be derived using simultaneous


measurement techniques and is representative of the radiosonde's actual
measurement capability.

As detailed in previous sections the criteria for simultaneous


comparisons were established in such a way so as to enable direct com-
parison within discrete layers. In estimating precision the uncertain-
ties in the direct sensor measurements must be determined first. These
can than be used to determine how the uncertainties propagate to produce
the precision or uncertainties of derived quantities.

Estimates of precision, following Grubbs (1948), are graphically


displayed in Figure 8.3 for each of the pressure sensors. Between sur-
face and 500 hPa the precision of the Finnish and US pressure sensors
differ by about 0.2 hPa, and between 500 and 50 hPa the measurements are
approximately the same. At levels above 50 hPa the precision of the US
pressure sensor is calculated to be 0.5 hPa less than that of the Fin-
nish sensor. At 10 hPa the precision of the Finnish pressure sensor
goes to zero. This comes about when est( ~e~ ) becomes negative. This
is an occasional condition that requires a more rigorous method of cal-
culation not employed in this report. It is a condition that signifies
high precision, hence, a zero value in the example shown only specifies
that the trend is to high precision rather than to a true value. The
Australian pressure sensor, although biased as discussed in section 7.0,
indicates less precision, i.e., about 2 hPa between surface and 200 hPa
and then is noted to gradually improve to about 0.5 hPa at levels above
30 hPa. Based on the originally reported Indian data, its pressure sen-
sor is the least precise. Considerable improvement is seen when the
improved selection of one-minute levels was employed. Obviously, from
this illustration manual reduction of radiosonde. measurements contri-
butes to instrumental error significantly.

Figure 8.4 shows a very consistent relationship of temperature


sensor precision to each other. Except for the Australian temperature
sensor
.page 100

precision the relationship between all sensors is the same as shown pre-
viously in Figure 8.1. ,The Australian and US temperature sensor preci-
sion is noted to be virtually the same. This is expect.ed since the Aus-
tralian and US sensors are identical and measurements we.re made at the
same time. The Finnish temperature sensor precision continually
decreases with altitude. The Indian sensor apparently is the l.east pre-
cise being about 0.5 C up to 50 hPa and worsens above that level until
it reaches 1.4 C at 10 hPa. In contrast, and as another illustration
that poor treatment or handling of data can affect data quality, the
results calculated from the initial Indian data are shown. Remember
that the original Indian temperatures were interpolated for one-minute
intervals while the final data were read directly at each minute. The
results derived from the two sets of Indian data should provide a cau-
tion to all suppliers and users of reported data that measurement preci-
sion is affected by manual methods of data reduction, and usually will
be less then the instrument's actual capability.

Finally, a precision estimate of geopotentials for the radiosonde


and radar heights indicates that the radar was the most precise, and the
Finnish radiosonde the next precise in reporting heights. Figure 8.5
might be deceiving because it includes all of the observations from the
four time periods. Only Finland's temperature measurements were
corrected during the intercomparison making its geopotentials more con-
sistent with radar geopotentials than those of the other instruments.
The estimated precisions given in Figure 8.5 should be reviewed with the
standard deviations of Figures 6.1 to 6.4 in mind. One apparent reason
for the poor precision shown in the US and Australian results is the
difference that changes in time of day create in geopotential because of
the lack of corrections for their temperature sensor.
page 101
FIg 8.1

PHASE-2 CPL
ALL TIMES

10 / ?" _..X

/
/ ....
/
-/ --
/
~ t
20
t"
/

/
~ .~
I
\ /
¥ />
t I I
I
Cl
a... I I /

..c.
.._,. 50 4 /
X f -~
\ / I .. ..--
w
-0::
::>
t¥ I
(-!1"'

~
w
100 ~ ~
a::: I I 1\
a...

200
,< . \
X \- f>
~Fil
.._._..USA
+--+ AUS

,1 X
I
.
. /
1>
f >E---x IND
~···~OLD

,.'
500
")<
~
1000
0 2 3

TEMP PRECISION (OEG)

Figure 8.1 Temperature sensor precision at reported constant pressure


levels for four radiosonde instruments. The reprocessed Indian data
(IND) are included.
page 102

Fig 8.2

PHASE-2 CPL ·
ALL TIMES

10 t )<
I
I /
/

). ~
/ /
.I
/
20
( ;< ..~

t ,. . .
/
/

.......
tO I :
I
(L
..c. 50 J ?
I I
I
.:
w f ~ ~
a:: I :
::::> I I I
~lOO f X ~
w I I
a::
(L
l I .
/)
. FI 1

200
I
X
I*~
<9-----0
.._ • .....,.USA
+--+ AUS
X .~ >E---K IND
I .
'f..#' ~ .•. _.OLD
I .
,{
500
'X/
I
ll
1000
0 30:. 60 90

GEOP PRECISION <M>

Fi.gure 8. 2 Same as Fig. 8 .1, except for geopotentia1 heights.


page 103

F•o 8.3

PHASE-2 SIMULTANEOUS
All TIMES

10

20

- 1:1
a...
-
..s::. 50 •..
w '· ·.f>
0::.
:::>
I
~ 100 •
w I
0:::
a... • \ <9-----0 FI 1
.to-·__. USA
200 .. +--+ AUS
' '
:>+--~ IND
..., •.. _.OLD
~
/

500 - _···-.
.:. ... ....
··---···- .....
_
-···-··-~
..... -··

1000 ~----------~----------~b-----------~
...... -··· . -··· -···

0 8 12

PRESS PRECISION (hPel

Figure 8.3 Same as Figure 8.1, except for simult"aneous measurements of


pressure.
page 104

Fig 8.4

PHASE-2 SIMULTANEOUS
ALL TIMES

10

}>

20

I
~

I
10
a..
...c 50
/
w
a:::
:::>
~
w
a:::
100
•.
a.. ~ (!)----E) FI 1
\ I ..e.----41. USA
200
* I
I
~
I
+--+ AUS
*-~ IND
I I ~--·~OLD
*\ ~
\ I
\ :
500 Xt
\:

:I
lx
~x
1000 ~----------~----------~----------~
0 2

TEMP PRECIS I ON <DEG>

Figure 8.4 Same as Fig. 8.3, except for temperature.


page 105

Fig 8.5

PHASE-2 SIMULTANEOUS
ALL TIMES

to

20

-ID
a...
~ 50
.._,

w
a:::
::::>
~ 100
w
a:::
a... (9---e) FI 1
A--·-4USA
200 +---+ AUS
>E---x IND
~···~ RDR

500

1000
0 200 400 600 800

GEOP PRECIS I ON CM>

Figure 8.5 Geopotential preclSlon of radiosondes and radar based on


simultaneous measurements.
page 106

9.0 Conclusions

The second phase of the World Meteorological Organization Inter-


national Radiosonde Intercomparison was hosted in the United States by
the National Aeronautics and Space Administration at its Wallops Island,
Virginia facility. Australia, Finland, India, and the United States
participated. The Graw M-60 radiosonde provided by the United Kingdom
also participated on a limited basis. Between 7 February and 13 March,
1985, 100 balloon flights were obtained. Each balloon carried five
radiosondes. The International Organizing Committee had recommended
that the participating Members should, in so far as possible, bring
their most widely used radiosonde systems to Wallops Island. Each par-
ticipant released one radiosonde type, except Finland which occassion-
ally tested a second instrument type when the Graw M60 was not flown.
Thermodynamic and wind data were obtained. Additionally, NASA provided
precision radar tracking of each balloon package flown. The radar data
provided independent altitude and wind data which could be used as a
comparison standard.

The comparison schedule was designed to obtain measurements from


a minimum of four radiosonde instruments during four observation times a
day. Flights were scheduled every three hours beginning at 0900 local
time with the last observation at 1800 local time each day. This latter
time corresponds to nighttime observations. The success criterion esta-
blished by the organizing committee (i.e., 72 balloon flights must reach
50 hPa) was exceeded by a significant amount. More than 50 percent of
all instruments flown (including the Graw M60) exceeded the 10 hPa
level.

Statistical analysis was used to describe mean differences and


standard deviations of the observations. Data were analyzed in pairs
and then only at those levels where measurements from all radiosondes
were present. Analysis of unequal sample sizes were not used simply
because of the uncertainty that would result from atmospheric changes.
Equal observation analysis inhibits this effect.

Comparisons at standard pressure levels provided information on


the data transmitted over the GTS that are received at analysis centers.
These results represent the quality of the data used by the meteorolo-
gist for chart analysis and forecasting. Data used for analysis at the
standard levels contain instrumental errors, observer errors, transmis-
sion errors, as well as errors attendant to data usage. The comparison
of one-minute data (simultaneous comparisons) provides information pri-
marily on the instrumental errors, but will include observer errors if
quality control is not sufficient.
page 107

9.1 Temperature

Comparison of the temperature profiles revealed that the thermis-


tors performed differently given similar atmospheric conditions. Gen-
erally, when the instruments exited clouds the three rod thermistors
reported super-adiabatic lapse rates while the Finnish capacitive sensor
reported an inversion. Historically, super-adiabatic lapse rates have
been observed in the dry air immediately above saturated clouds, and are
considered to be caused by the evaporative cooling of water deposited on
the sensor. During the tests the Finnish sensor reported an inversion
at temperatures below freezing, but at temperatures above freezing the
inversion was not observed. The rod thermistors did not indicate any
difference whether or not the ambient temperature was above or below
freezing. The data examined indicates that unique differences in the
measurements are possible, even when measuring simultaneously.

Comparison of mean temperatures at standard pressure levels


showed that the range of differences experienced by all of the sensors
between the surface and 100 hPa during the daytime was less than 0.5 C.
At night the range of differences increased to about 1.0 C. These tem-
perature difference ranges are not necessarily large; however, they
translate into a range of geopotential differences of about 35 meters
during the day, and about 60 meters at night. Those radiosondes that
used rod thermistors (Australia, India, and the US) reported tempera-
tures warmer than those of the capacitive thin-film thermistor (Finland)
during the day, and reported colder temperatures during the night.

Simultaneous comparison of the temperature measurements indicated


that the differences were about the same as the differences found for
the standard pressure levels, however, the standard deviations of the
differences was about one-half of the standard deviation of the differ-
ences at the constant pressure surfaces. While timing errors could have
been a contributing factor, it is more probable that errors in the pres-
sure cell measurements have a large influence on the variance of tem-
peratures assigned to standard pressure levels.

9.2 Relative Humidity

The relative humidity measurements showed that all of the sensors


(carbon, lithium cloride, and capacitive thin-film) were in good agree-
ment during the initial measurement period in the humid air below the
first cloud layer, and generally reported within two percent of each
other. With the exception of the Indian sensor, the sensors' response
were rapid, especially when the radiosondes entered a cloud or the dry
air immediately above a cloud. However, on some ascensions the humidity
differences between the carbon and capacitive sensors were approximately
10 percent. These differences occurred at low and high humidity values.
The lithium cloride element used by India has. a slower response than
does the carbon elements, and in all data examined there were no compar-
able relative humidity data for the Indian radiosonde at levels above
600 hPa. Even at levels below 600 hPa, the Indian sensor did not exhi-
bit the finer structure observed by the other sensors, however, it did
give
page 108

a mean profile which agreed with the mean profiles of the other sen-
sors~.

9.3 Geopotential

Agreement between radiosonde derived heights and radar measured


heights was relatively good at levels below 100 hPa but differences
became progressively larger above 100 hPa. It was confirmed that the
Finnish and US radiosonde heights relative to radar heights were similar
at local noon and at night. But were different at 0900 and 1500 local
times. Since corrections are applied to the Finnish instrument at all
solar elevation angles, while the US radiosonde does not have correc-
tions applied, it is difficult to understand how any agreement in height
between these two radiosondes could occur. It is important that correc-
tions be derived for the US radiosonde and that they be applied by
either the observing station or the analysis centers.

9.4 Pressure

Flights of the Australian radiosonde early in the intercomparison


indicated that the pressure measurements were reading about 4 hPa too
low. Steps were taken to correct the problem encountered, however, this
could only be accomplished after the completion of the intercomparison.
The problem was traced to the temperature of the standard pressure
transducer used for baroswitch calibration. After a method to correct
the pressure measurements was derived and applied, the overall differ-
ence was reduced to +/- 1 hPa. Once the pressure error was removed, the
temperature and geopotential differences relative to the US radiosonde
also were reduced significantly.

9.5 Other Comments

One important result of the tests with the Indian radiosonde was
that the technique of data level selection tended to give a poor impres-
sion of this radiosonde's capability. The Indian team selected only
significant level data and interpolated the standard level pressures,
temperatures, humidities, and geopotentials. In a similar manner the
one-minute interval data were also interpolated. This procedure
resulted in large variances giving the impression of poor instrument
performance. Fortunately, the Indian team re-submitted their data at
one-minute intervals which improved the results significantly.

Because of the large sample of time correlated measurements it


was believed possible to separate the instrumental measurement errors
from the true variations of the measured parameter. The estimate of
temperature precision for the US sensor turned out to be approximately
0.30 C for the layer up to the 20 hPa level, and about 0.90 C between 20
and 10 hPa. Estimates for the Finnish temperature sensor are 0.35 C for
the layer up to 100 hPa, and 0.46 C in the layer 100 to 30 hPa. Aus-
tralian radiosonde temperature precision was expected to be the same as
that
page 109

of the US temperature sensor, and after compensation for the pressure


sensor error was found to be quite close to the US value. Estimates of
error for the Indian radiosonde were calculated using the original data
reduced from significant levels and again from the re-submitted one-
minute interval data. Between the surface and 100 hPa the temperature
sensor precision determined from the original data submitted was about ·
0.86 C and between the surface and 10 hPa about 1.07 C. After the meas-
urements were re-processed at one-minute intervals the estimated tem-
perature precision up to 100 hPa irnproved to 0.66 C and between surface
and 20 hPa, to about 0.71 C. Thus, significant improvement occurred
after more precise selection of the Indian data took place.

The result of Phase II of the WMO International Radiosonde Inter-


comparison are important for a number of reasons. None of the
radiosondes tested reported identical results. Radiative errors are
large for the rod thermistors as evidenced by their relative temperature
differences at the various times of observation. Intercomparisons can
identify instrumental or calibration defects, such as the pressure error
found with the Australian pressure sensor. Procedural effects on data
quality also are readily identifiable, such as the large variance con-
tained in the Indian radiosonde data because of inappropriate reduction
procedures.

It is clear that intercomparisons are needed more frequently then


the few that have been carried out to date. Major new trials should be
conducted at least once every ten years. More often is desirable, but
from the realistic point of view may not be possible because of the
large cost involved. Future intercomparisons should be held at sites
that contain radar that might be used to provide an independent set of
height and wind data. Computer programs used in intercomparisons should
be standardized to insure that any changes encountered are due to
instruments and not due to other causes.

Finally, certain procedures were developed for Phase I and Phase


II that worked out to be consistent and allowed data to be interchanged.
These procedures and analysis methods, hopefully, will be employed in
future comparisons. The use of similar procedures permits results to be
compared and understood regardless of where the comparison actually
takes place.
page 110

10.0 References

Antikainen, V. and V. Hyvonen, 1983: The accuracy of Vaisala RS 80


radiosonde. Proceedings of Fifth Symposium Meteorological Observations
and Instrumentation. 11-15 April 1983. Toronto, Canada. pp 134-140,

Brousaides, F. J., and J. F. Morrissey, 1973: Residual temperature-


induced humidity errors in the NWS radiosonde. Part 1. AFCRL-TR-73-0214,
Instrumentation Papers, No. 184.

Brousaides, F. J., and J. F. Morrissey, 1974: Residual Temperature-


induced humidity errors in the National Weather Service radiosonde,
Final Report. AFCRL-TR-74-0111, Instrumentation Papers, No. 215.

Delver, A. 1956: A statistical treatment of the Second World Comparison


of Radiosondes (Payerne, 1956). Rep. R. Neth. Meteor. Inst., De Bilt,
Netherlands.

Finger, F. G., and R. M. Mcinturff, 1978: Part 1, Research on compati-


bility of data from radiosondes, rocketsondes and satellites. in The
Compatibility of Upper-Air Data. Technical Note No. 163. World Meteoro-
logical Organization, WMO No. 512.

Federal Meteorological Handbook No. 3, 1981: FMH-3, Radiosonde Observa-


tions, GPO Washington, D.C.

Grubbs, F. E., 1948: On estimating precision of measuring inst·ruments


and product variability. Journ. American Stat. Assoc. Vol. 43, pp 243-
264.

Grubbs, F. E., 1973: Errors of measurement, precision, accuracy and the


statistical comparison of measuring instruments. Technometrics, Vol 15,
No 1. pp 53-66.

Hoehne, W. E., 1980: Precision of National Weather Service Upper Air


Measurements. NOAA Technical Memorandum NWS T & ED-16. 23 pp.

Hooper, A. H., 1975: Volume 1: Studies on Radiosonde Performance. in


Upper-Air Sounding Studies. WMO Technical Note No. 140.

Hooper, A. H., 1986: WMO International Radiosonde Comparison, Phase 1,


Beaufort Park, U. K. 1984. WMO Instruments and Observing Methods Report
No. 28. 118 pp.

Kitchen, M. J., J. Nash, and J. F. Panting, 1985: Evaluation of Tempera-


ture, Pressure, and Geopotential Measurements Obtained During Phase 1 of
the WMO International Radiosonde Comparison. Proceedings Third WMO
Technical Conference on Instruments and Methods of Observation (TECIMO-
III), 8-12 July 1985, Ottawa, Canada.
page 111

Lange, A., 1985: Meteorological observations using Navaid methods. WMO


Technical Note No. 185. WMO, Geneva 44 pp.

Nyberg, A., 1952: On the comparison of radiosonde data in Payerne, May


1950. Comm. Swed. Meteor. Hydrol. Inst., Series B, No 9, Stockholm.

Schmidlin, F. J., J. J. Olivero, and M. S. Nestler, 1982: Can the stan-


dard radiosonde system meet special atmospheric research needs? Geophys.
Res. Ltrs, Vol. 9, No. 9, pp 1109-1112.

Schmidlin, F. J., J. K. Luers, and P. Huffman, 1986: Preliminary esti-


mates of radiosonde thermistor errors. NASA Technical Paper TP 2687, 15
pp.

Spackman, E. A., 1978: Part II: The compatibility and performance of


radiosonde measurements of geopotential height in the lower stratosphere
for 1975-1976. in The Compatibility of Upper-Air Data. WMO Technical
Note No. 163.

WMO, 1986: WMO Catalogue of radiosondes in use by Members, 1986. WMO


Instruments and Observing Methods Report No. 27, Upper-Air Data Compati-
bility. 26 pp.
page 112

Appendix

Participants in Phase II, WMO International Radiosonde Intercomparison

Host:
United States of America
National Aeronautics and Space Administration (NASA)
Goddard Space Flight Center
Wa.llops Flight Facility
Wallops Island, Virginia

Project Leader: K· J. Schmidlin


Site Manager: R. J. Frostrom
Site Coordinator: D. Bruton
Observation Manager: R. Bull
Data Processing Manager: W. Michel
Data Receipt Manager: R. Taylor

Australia

Team Leader: J. Stickland


Z. Svalbe
J. Costello
J. Linton

Finland

Team Leader: V. Antikainen


P. Kuittinen
E. Groendahl
A. Mahkonen
I. Hyvonen

India

Team. Leader: N. Seshadri


S. K. Srivastava
V. P. Verma
B. B. Mandel
page 113

United States

Team Leader: W. Winkert


T. Nazario
E. Bowman
N. Miller
E. Schminke
J. Kotlyn
F. Durrett

Graw (Provided Qy UK)

Team Members: J. Nash


G. Forrester
M. Kitchen

Visitors to NASA Goddard Space Flight Center/Wallops Flight Facility,


Wallops Island, Virginia, during Phase II, WMO International Radiosonde
Intercomparison

International Organizing Committee for WMO Radiosonde Intercomparison

S. Huovila (Chairman), Finland


F. G. Finger, USA
J. Nash, UK
N. Seshadri, India
F. J. Schmidlin, USA
S. Klemm, WMO Secretariat

Host Country (USA)

br. N. Hinners (Director, Goddard Space Flight Center)


Dr. R. Hallgren (Director, National Weather Service)
Dr. R. Curran (Program Office, NASA Headquarters)

You might also like