accuracy of analyzed stratospheric temperatures in the winter … · 2015. 8. 10. · accuracy of...

Accuracy of analyzed stratospheric temperatures in the winter Arctic

vortex from infrared Montgolfier long-duration balloon flights

1. Measurements

Jean-Pierre Pommereau,1 Anne Garnier, 1 Bjoern M. Knudsen,2 Ge rard Letrenne,3

Marc Durand,3 Manuel Nunes-Pinharanda,1 Laurent Denis,1 Francois Vial,4

Albert Hertzog,4 and Francesco Cairo5

Received 10 October 2001; revised 3 January 2002; accepted 8 January 2002; published 13 August 2002.

[1] Long series of temperature measurements in the winter Arctic stratosphere have beenobtained from infrared Montgolfier (MIR) long-duration balloon flights in 1997, 1999,and 2000, from which the performance of a variety of meteorological analyses has beeninvestigated. In this first paper of a series of two, the experimental setup and themeasurements are described. First flown in the winter Arctic in 1997, the MIR platformappears well adapted to perform flights for several weeks in the vortex. Flights of 12 and22 days (1997), 7 and 17 days (1999), and 3 and 18 days (2000) have been achieved alongwhich altitude, pressure, and temperature have been sampled every 9–10 min.Comparisons between various independent sensors on the same balloon, and to Vaisalaand Russian radiosondes, have made it possible to evaluate the performances of theinstruments. The accuracy of the altitude/location of the Global Positioning System provedto be ±100 m but that of the pressure sensors only ±2 hPa. The most accurate method forderiving pressure appears to be the use of GPS altitude together with the ECMWF(European Centre for Medium-Range Weather Forecasts) geopotential height. Finally,the temperature was demonstrated to be measured with a precision of ±0.4 K and anaverage bias of less than ±0.5 K, but during nighttime and at an altitude below only28 km. INDEX TERMS: 0341 Atmospheric Composition and Structure: Middle atmosphere—constituent

transport and chemistry (3334); 0350 Atmospheric Composition and Structure: Pressure, density, and

temperature; 0394 Atmospheric Composition and Structure: Instruments and techniques; KEYWORDS:

stratosphere, temperature, meteorological analyses, long-duration balloons

1. Introduction

[2] Large ozone reductions have been reported in theArctic polar stratospheric vortex during the winter in the1990s, qualitatively consistent with heterogeneous activa-tion of chlorine on polar stratospheric clouds (PSC) formingat low temperature. Though generally consistent withobservations, ozone losses are often underestimated bythree-dimensional (3-D) chemical transport models (CTMs),particularly when the minimum temperature is close to thatof the formation of PSCs [Goutail et al., 1999; Chipperfield,1999; Guirlet et al., 2000; Goutail et al., 2002]. Among theuncertainties suggested to explain this underestimation isthe temperature in meteorological analyses used to force theCTM simulations. Indeed, and in contrast to Antarctica,where the temperature is persistently well below that of PSC

formation throughout the winter, the average temperature inthe winter Arctic stratosphere is close to that of theformation of a PSC type Ia (NAT). It falls only episodicallybelow that of type II ice particles condensation. Thepresence of PSCs is thus limited in time and geographicextension. Therefore small errors or systematic biases in thetemperature in the meteorological analyses could lead tolarge errors in predicted ozone loss in this highly nonlinearsystem.[3] In addition, mesoscale processes such as gravity or

orographic waves are represented either only partially or notat all by these analyses of limited grid size. These small-scale waves could result in local cooling events of largeamplitude, where PSC could form [Leutbecher and Volkert,1996; Carlslaw et al., 1998; Dornback et al., 1998].Though such wave-forced cooling is known to occursporadically, its climatology, and therefore its global influ-ence on ozone destruction, is still poorly known.[4] However, testing the ability of a meteorological

model to capture stratospheric temperature and wind atglobal and mesoscales is not straightforward. Most availableradiosonde data are already used by the analysis systems. Inaddition, the number of radiosondes is limited in the Arcticwinter stratosphere for several reasons: (1) reduced number

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D20, 8260, 10.1029/2001JD001379, 2002

1Service d’Aeronomie, CNRS, Verrieres le Buisson, France.2Danish Meteorological Institute, Copenhagen, Denmark.3Centre National d’Etudes Spatiales, Toulouse, France.4Laboratoire de Meteorologie Dynamique, Palaiseau, France.5Istituto de Fisica dell’Atmosfera, CNR, Rome, Italy.

Copyright 2002 by the American Geophysical Union.0148-0227/02/2001JD001379$09.00

SOL 2 - 1

of stations, recently dropped by a factor 4 in Russia foreconomical reasons; (2) balloon bursts at low temperaturenear the tropopause; and (3) high wind speed, which takesthe balloon out of telemetry range when reaching thestratosphere. While comparisons between various analyses[Manney et al., 1996] highlight differences between them,they do not show which is the most accurate. Several studieshave been carried out for evaluating the quality of thetemperature and wind of analyses, using a variety ofmethods. TOVS (TIROS Operational Vertical Sounder)satellite-retrieved temperatures have been used by Swinbankand O’Neill [1994] to evaluate the Meteorological Office(MO) data assimilation model, also compared to radio-sondes byManney et al. [1996]. Naujokat [1994] has shownthat the ECMWF temperature analyses in the winter strato-sphere in 1992 were systematically warmer than radio-sondes and FUB (Freie Universitat Berlin) analyses,which are based on subjective analyses of radiosondemeasurements. The same conclusion was reached by Knud-sen [1996a] for the European Centre for Medium-RangeWeather Forecasts (ECMWF) prior to 30 January 1996 bycomparing radiosondes to ECMWF analyses or first-guessfields (the 6-hour forecast from the previous analysis)assumed to be independent of errors on individual radio-sondes. Pullen and Jones [1997] used data from 28 ozone-sonde stations only partially included in the network andavailable during the winter of 1994/1995 during the Euro-pean campaign SESAME, for evaluating MO analyses.From comparisons between the FUB analyses and theSSU (stratospheric sounding unit) temperature retrievedfrom the TOVS radiance measurements, Pawson et al.[1999] concluded that the FUB data are generally colder,particularly at low temperature, but that the satellite-derivedvalues become colder when the temperature decreases atpressures lower than 50 hPa.[5] In addition, data from the few available long-duration

balloon flights have also been used for comparison toanalyses. Temperature and wind velocity measured along1- to 6-day trajectories of eight ballast-controlled balloonsconducted in 1992–1995 within the Polar Vortex BalloonExperiment (POVORBEX) of the University of Wyominghave been compared to the analysis of ECMWF [Knudsenet al., 1996b]. Past NASA and CNES long-duration bal-loons at a variety of latitudes have also been used forevaluating wind analyses and forecast in the midstrato-sphere by the MO and DAO (Data Assimilation Office)systems [Keil et al., 2001]. Finally, Knudsen et al. [2001]have investigated the quality of wind analyses of a varietyof analyses in the Arctic vortex in 1997 and 1999 from thetrajectories of a part of the balloons described in this paper.[6] Here we report on series of temperature measure-

ments performed along the trajectories of six MontgolfiereInfra Rouge (MIR) long-duration balloons, with twolaunches in each winter of 1997, 1999, and 2000 in thepolar vortex. These launches were made in the frameworkof a European project: the THESEO Lagrangian Experi-ment. The durations of the flights (12 and 22 days in 1997,7 and 17 in 1999, 3 and 18 in 2000), during which thetemperature was sampled every 9 min (1997) or 10 min(1999 and 2000), and the diurnal vertical excursions of theballoons of 16 to 27 km allow investigation of the biasesand the dispersion of the analyses.

[7] The presentation is organized into two companionpapers. The first paper is a description of the experiment andof the flights, followed by an in-depth discussion of theaccuracy of meteorological measurements, a prerequisite forlooking at possible biases in the analyses. The second paper[Knudsen et al., 2002] deals with the comparison of MIRtemperatures with a variety of analyses: ECMWF, MO,DAO, NCEP (National Centers for Environmental Predic-tion), and NCEP/NCAR reanalysis. In situ observations ofgravity and orographic waves during these flights arestudied by Pommereau et al. [1999] and Hertzog et al.[2002], as well as by Knudsen et al. [2002].[8] The manuscript is organized into five sections. After

the Introduction, section 2 describes the MIR system, thevarious payloads flown, and their meteorological sensors.Section 3 describes the flights carried out during the 3 years,including a brief overview of stratospheric meteorology ineach year and of relevant technical or operational incidentsneeded to understand the quality of meteorological data.Section 4 is devoted to an in-depth analysis of precision andaccuracy of altitude/location, pressure, and temperaturemeasurements. The overall achievements and findings aresummarized in section 5.

2. Infrared Montgolfier System

[9] The hot-air MIR balloon was developed by the Bal-loon Division of the Centre National d’Etudes Spatiales(CNES) following a suggestion by Pommereau and Hau-checorne [1979]. It carries a total payload of 50 to 70 kgdepending on the arrangement, made of independent pack-ages separated by 30 to 50 m. The payloads flown in theArctic are as follows: a service gondola, SAMBA (Systemed’Acquisition de Mesures en Ballon), which contains theballoon control systems, meteorological sensors, and insome cases, additional small scientific passenger instru-ments, and a scientific payload of 10 to 40 kg which changedbetween the flights. The payload could be either a SAOZUV-visible spectrometer, a payload named SALSA compris-ing a combination of a SAOZ, a tunable diode laser formethane and a hygrometer, or even a newly developedlightweight meteorological payload RUMBA. As an exam-ple, Figure 1 shows the arrangement of a SAMBA and aSALSA payloads in 1999 and 2000. The SAMBA gondola ishanging 8 m below the bottom of the balloon and the SALSAscientific payload 41 m below. Each gondola is totallyindependent, having its own temperature, pressure, andlocation measurements by GPS and an ARGOS transmitterfor satellite data collection and Doppler location. Othersubsystems are, from top to bottom, the cut-down pyrotech-nic knives, a parachute for soft landing, a radar transponderand a strobe light activated during the initial ascent and thefinal descent, a radar reflector, and at the bottom, a VaisalaRS-80 radiosonde of 3 hours of lifetime for consistencychecks of meteorological sensors during the initial ascent. Inthis arrangement the total load below the balloon was 71 kg.[10] All balloons were launched from ESRANGE, the

facility of the Swedish Space Corporation at Kiruna innorthern Sweden, which also controlled the flights. Clear-ance was obtained from Air Traffic Control (ATC) anddefense authorities of all countries north of 55�N. Acomplementary station run by the Russian Central Aero-

SOL 2 - 2 POMMEREAU ET AL.: STRATOSPHERIC ARCTIC VORTEX TEMPERATURES, 1

logical Observatory was set up at Murmansk for controllingthe flights above Russia. For air traffic safety reasons, theflights were limited to the inside of the vortex and automati-cally terminated if the balloon moved south of 55�N orbelow 140 hPa (13.5 km geometric altitude in the vortex,14.5 km standard aircraft flight level).

2.1. IR Montgolfier

[11] The MIR is a hot air balloon of 45,000 m3 volumemade of aluminized Mylar (IR-absorbent Mylar inside,aluminium of low IR emissivity outside) for its upper partand IR-transparent polyethylene for the bottom. It is heatedby the thermal emission of the Earth and the atmospherealone at night and additionally by the Sun during the day.Though more than 40 flights have been carried out in thetropics [Malaterre et al., 1996], the flights described herewere the first in the polar winter. Since the nighttimeballoon lift is directly proportional to the difference betweenthe brightness temperature of the surface being flown overand the air temperature at flight level, it is particularly suitedto the cold winter vortex. Indeed, in these conditions theatmospheric temperature at 20 km is of the order of �80�Cand that of the ground or clouds is always warmer than�50�, �60�C even in the coldest part of the Arctic (Siberia,Greenland, or the North Pole). This is more than enough toensure a difference of 15�–25�C between the gas and theoutside and to allow the balloon to stabilize between 18 and22 km (60–40 hPa) at night. During daytime, additionalsolar heating causes the balloon to ascend to higher levels,typically around 25–27 km (15 hPa), depending on albedoand solar zenith angle. Since this arrangement does notallow the MIR to leave the ground by itself, an initial lift isprovided by helium, which results in a higher flight levelduring the first day, at 34 km (4–5 hPa). After 2 days thehelium is totally evacuated and the balloon reaches itsnormal hot-air flight level.[12] The main limitation of the system is the occurrence

of stratospheric sudden warming that can occur anytimebetween February and April, depending on the year. If theflight-level temperature warms to �50�/�60�C, the differ-ence with the brightness temperature of the surface belowcould be too small (e.g., above high and dense clouds, theGreenland ice cap, or mountain ranges in Siberia) to providethe necessary lift. The balloon could then drop to altitudeswhere the flight must be terminated. We will see later thatsome significant altitude drops did indeed occur, and in onecase, with the heaviest payload, resulted in the terminationof the flight.

2.2. SAMBA Service Gondola

[13] The role of the SAMBA gondola is threefold. Itallows monitoring of the flights: altitude, location, temper-

Figure 1. (opposite) Schematic of the IR Montgolfiersystem. The SAMBA service payload, hanging 8 m belowthe bottom of the balloon, carries two temperaturemicrothermistors mounted on 1-m-long booms, two pres-sure sensors, and a GPS. The scientific payload 40 m belowalso carries pressure and temperature sensors but of slowersampling rate. An additional Vaisala commercial RS-80radiosonde of 3-hour autonomy is added at the bottom forconsistency checks during the initial nighttime ascent.

0,8

7

2

40

1

1

7

T3

T1 T2

Cut-down

Parachute

SAMBA

Radar Transponder

Strobe Light

Radar reflector

SAOZ/SAMBA

RadiosondeT4

POMMEREAU ET AL.: STRATOSPHERIC ARCTIC VORTEX TEMPERATURES, 1 SOL 2 - 3

ature, and pressure as well as IR flux since 1999. It activatesthe cut down, the strobe light, and the radar transponderrequired by the Air Traffic Control (ATC) if the balloonis flying below 140 hPa or south of 55�N. Alternatively,it terminates the flight after a given duration (between20 and 28 days) to allow the recovery of the payloadover a predetermined area. The data are transmitted tothe ground through the ARGOS satellite data collectionsystem, but there is no remote control. All systems are runautomatically.[14] The location is determined by GPS of Trimble

Navigation. Pressure is measured by two Honeywell sen-sors with different pressure ranges: 0–1000 hPa for theascent sensor and 0–145 hPa, both calibrated in thelaboratory to an accuracy of 1 hPa. Their temperature isalso transmitted for temperature compensation from sensi-tivities measured in a cold chamber. Ambient temperatureis measured by two Veco-aluminized microthermistors of a250-mm diameter, mounted on 1-m-long booms deployedshortly after launch on opposite sides on the gondola. Thethermistors are calibrated in a cold chamber between 170 Kand 320 K, with an accuracy of 0.5 K. All parameters aresampled regularly every 9 min (1997) or 10 min (1999 and2000), increased to once per minute during the initialascent to enable comparison with the radiosonde. TheARGOS system is particularly efficient at high latitudebecause 26 to 28 daily satellite overpasses can be expected,though the best possible total transmission is limited to120 kbits/d.[15] The SAMBA payload is designed to carry additional

small scientific passenger instruments. This possibility wasused once in 1999 and in 2000 to fly an in situ ozone solidstate sensor developed by the University of Cambridge[Hansford et al., 2000] and a backscatter diode laser formeasuring PSC developed by the Italian CNR-IFA [Adrianiet al., 1999]. The total weight of the SAMBA payload is11 kg (28 kg with passengers), including lithium batteriesfor a maximum autonomy of about 30 days.

2.3. SAOZ/SALSA Scientific Payload

[16] The SAOZ payload is an adaptation of the short-duration balloonsonde designed for the remote sensing ofozone, NO2, and OClO profiles by solar occultation UV-Visspectrometry [Pommereau and Piquard, 1994]. In a moresophisticated version, SALSA, flown in 1999 and 2000,carries also a tunable diode laser (TDL) constructed by theNational Physical Laboratory and the University of Cam-bridge (UK) for the in situ measurement of methane[Gardiner et al., 2000] and a Lyman a hygrometer of the

Central Aerological Observatory (Russia) [Yushkov et al.,2000].[17] Additionally, the SAOZ and SALSA payloads carry

a set of meteorological sensors: pressure, temperature, andlocation/altitude by GPS of Trimble Navigation. Meteoro-logical sensors are those of a Vaisala RS-80 radiosonde:temperature of 0.2�C (1 SD) repeatability and a 2.5-s timeconstant and pressure of 0.5 hPa (1 SD) repeatability and±1 hPa accuracy according to the manufacturer. Capacitivemeasurements relative to a reference capacitor and temper-ature compensation are handled by the standard Vaisalaground software installed into the onboard CPU (centralprocessing unit). Calibrations are those of the manufacturer,updated by comparison with a meteorological station priorto launch following the instructions of the manufacturer. Forthe first flight in 1997, the sensors were mounted on the sideof the gondola as for short-duration balloon flights. Sincethis arrangement proved to be noisy for constant levelfloating, a 1-m-long boom was added for the subsequentflights.[18] SAOZ and SALSA have their own ARGOS trans-

mitter and are powered by lithium batteries. To conservepower, the number of measurements is limited to a hundredper day. The CPU and all the subsystems are thus switchedoff and activated by a microcontroller when needed. Sincethe main objective is to measure profiles by solar occulta-tion at sunset and sunrise, the 100 data points are concen-trated at dawn and dusk, with only one measurement every90 min during daytime and nighttime. The temperature andpressure sampling of the SAOZ/SALSA gondolas is thuslimited. The total weight of the SAOZ payload is 28 kg andthat of SALSA 42 kg.

2.4. RUMBA Meteorological Payload

[19] RUMBA is a new lightweight payload designed bythe Laboratoire de Meteorologie Dynamique (LMD) for afuture project using constant level superpressure balloonsover Antarctica, VORCORE [Vial et al., 1995]. It has a GPSof Trimble Navigation for location and altitude, a Paroscien-tific pressure sensor of 0.3 hPa accuracy, and two Vecomicrobead-aluminized thermistors of 250 mm diametermounted onto glass supports hanging 5 m below the gondola.GPS, pressure, and temperatures are sampled every 15 min.In addition, pressure is sampled with a resolution of 0.01 hPaevery minute to allow calculation of dp/dt. As for SAMBA,the sampling of all parameters is accelerated to once/minduring the initial ascent. The calibration of the pressuresensor, including temperature compensation, is that providedby the manufacturer, further controlled in the laboratory

Table 1. Meteorological Sensors and Sampling on Each Type of Gondola

Gondola Meteorological Sensors Sampling

SAMBA GPS, Trimble Navigation 9–10 minpressure, Honeywell (2) (1 min initial ascent)temperature, Veco thermistors (2)

SAOZ/SALSA GPS, Trimble Navigation 110/daypressure, Vaisala Barocaptemperature, Vaisala Thermocap

(1/min SS, SR,1/1h30 daytime, nighttime)

RUMBA GPS, Trimble Navigation 15 minPressure, Parosci.Digiquartz 215A (1/min initial ascent)temperature, Veco thermistors (2)

Radiosonde Vaisala RS-80 10 s (initial ascent)


(precision, 0.02 hPa; accuracy, 0.3 hPa). The thermistors arecalibrated in the laboratory (accuracy 0.3 K). The totalweight of RUMBA for a 30-day autonomy is 9 kg.

2.5. Summary of Meteorological Sensors

[20] Meteorological measurements available on each pay-load and their sampling rates are listed in Table 1. Thecomparison with meteorological analyses will be primarilybased on the measurements of SAMBA, always present at ahigh sampling rate. Those of SAOZ/SALSA and RUMBA,as well as that of the initial ascent radiosonde, will bemainly used for evaluating and checking the performance ofSAMBA.

3. MIR Flights in the Arctic

[21] Because of the automatic cut down of the balloon ifit crosses south of 55�N and the need for low air temper-

ature for the MIR to fly efficiently, it was decided to launchonly inside the vortex. In addition, since solar occultationmeasurements as well as the daytime ascent of the balloonrequire some Sun, it was decided not to launch prior toFebruary 15. Two flights were attempted each year in 1997,1999, and 2000. The payload arrangement, the total weight,the date of launch, the duration of the flight, and the reasonfor the final cut down are summarized in Table 2.

3.1. 1997 Flights

[22] The vortex was exceptionally strong, cold, and longlasting in 1997, with a final breakdown in mid-April [Coyet al., 1997; Naujokat and Pawson, 1998]. It remainedalmost circular and centered around the pole until springwith the minimum temperature being collocated with thecenter of the vortex (barotropic stratosphere). Figure 2shows PV and temperature maps on the 475 K isentrope as

Table 2. Summary of MIR Flights

Flight Payload Weight Launch Duration Cut Down

97.1 SAMBA, SAOZ 46 kg 24 Feb. 12 days erroneous order97.2 SAMBA, SAOZ 46 kg 17 March 22 days to enable recovery99.1 SAMBA, SALSA 71 kg 18 Feb. 7 days P > 140 hPa99.2 SAMBA + pass. 45 kg 19 Feb. 17 days lat. < 55�N00.1 SAMBA + pass. 49 kg 18 Feb. 2 days lat. < 55�N

RUMBA00.2 SAMBA, SALSA 62 kg 18 Feb. 18 days lat. < 55�N

Figure 2. ECMWF potential vorticity (PV) and temperature fields at 475 K (approximately 19 km or 50hPa) at start of the first flight (24 February) and second flight (17 March) and at end of second flight (8April in 1997. Hatched areas represent, respectively, the inner vortex (PV > 48 pvu) and the region wherethe temperature is lower than that of NAT condensation (T < 193 K). A cross shows the balloon location.


calculated from ECMWF analyses at the beginning, themiddle, and the end of the flight period. The crosses showthe location of the balloon on each day.[23] The first launch took place on February 24, when the

vortex drifted above the station. The balloon trajectory, itsaltitude, atmospheric temperature, PV at 475 K, and orog-

raphy are displayed in Figure 3. PV at 475 K at the balloonlocation is compared to that of the vortex edge, calculatedfollowing the method of Nash et al. [1996]. Since thesampling period was 9 min and the average speed of theballoon 24 m/s, the measurements were performed aboutevery 13 km.

Figure 3. (left) Sixteen-day (12 calendar days) trajectory, altitude, temperature, PV at 475 K at thelocation of the balloon compared to the vortex edge, and orography height along the first flight of 1997starting at Kiruna on 24 February and ending over Greenland on 7 March. (right) Same for the 26-day (22calendar days) second flight starting on 17 March and ending over Norway on 8 April.


[24] Liftoff occurred in the evening so as to reach a floataltitude of 34 km before sunrise, thus avoided cumulatingthe lift of helium and solar heating, which would make theballoon climb too fast. The flight lasted for 12 calendar days(16 flight days and 17 nights) until March 8, when the CPUdelivered an erroneous cut-down order over Greenland. Theballoon remained inside the vortex at almost constant PV at475 K at latitude greater than 60�N during its three turnsaround the pole. After one and half day and the loss ofhelium, the daytime flight level became stable at about 25–26 km (13–8 hPa). During nighttime it varied from 20–22km (30–45 hPa) at low temperature (�75�, �80�C) to 14.5km geometric altitude (108 hPa) at warmer temperature(�65�C) on two occasions, above the Baffin Sea andsouthern Greenland.[25] The second launch took place on March 17 at the

inner edge of the vortex (Figure 2). The flight lasted for22 calendar days (26 flight days, 27 nights) until itsplanned automatic cut down when reaching Scandinaviaafter five circumpolar orbits. The instrument landed safelyon April 8 near Trondheim in Norway, where it wasrecovered the following day. The balloon trajectory andits altitude, air temperature, PV at 475 K, and groundaltitude are displayed in the right panels of Figure 3. Onaverage, the daytime altitude was a little higher (26–27 km) than during the first flight because of the higherSun and therefore the larger contribution of albedo duringdaytime. In contrast, the nighttime altitude was lower,often around 17 km, because of the warmer stratosphere.

The trajectory was almost circular and fast (38 m/s onaverage) due to the closeness to the stratospheric jet. Atthe end of the flight the balloon drifted to the vortexedge.[26] On both flights, the SAMBA gondola performed

well, except the unexpected cut down after 12 days ofMIR 1. All meteorological sensors performed also wellthroughout the two flights.

3.2. 1999 Flights

[27] In 1999 the vortex was relatively weak and warm inmid-February. It warmed further and split rapidly by the endof the month [Naujokat, 1999], for reforming later inMarch. Figure 4 shows the evolution of the vortex andthe temperature field at 475 K (ECMWF analyses) at thebeginning, middle, and end of the flights. By mid-Februarythe vortex was already elongated with a cold center on theEuropean side and warmer air on the Siberian side. It splitinto two parts in late February, which moved rapidly tomidlatitudes. A warm anticyclonic area developed over thepole, and the circulation reversed to easterly at the end ofFebruary.[28] Because the ECMWF forecasts showed that the

vortex could split, the two balloons were launched in quicksuccession on two successive days, on 18 and 19 February.Figure 5 shows the trajectories and altitudes of the balloons,air temperature, PV at 475 K, and orography height below.After a fast turn into the small vortex, the second balloonbeing a little deeper inside, the two drifted slowly toward

Figure 4. ECMWF PVand temperature fields at 475 K at start of second flight (19 February) and at theend of the first (27 February) and second flights (8 March) of 1999 (MIR 1, circle; MIR 2, cross).


eastern Siberia in a warm stratospheric area over a coldmountain range. The first and heavier balloon was cut downafter 7 days when it reached its minimum permitted altitudeimmediately before sunrise on February 24. It landed a little

east of Yakutsk where it was recovered a few days later. Thesecond flight survived and turned backward to Scandinaviaand Canada, because of the wind direction reversal after thesudden warming. On March 8 after 17 days of flight, it was

Figure 5. Trajectory of balloons, altitude, air temperature, PV at 475 K, and orography height in 1999(left, MIR 1; right, MIR 2).


automatically cut down after reaching 55�N off the Labra-dor coast, where it was lost.[29] Among the meteorological sensors, the following

three incidents were reported:1. One thermistor failed (T1) on SAMBA in the first

flight because the boom did not deploy correctly.2. Both temperatures in flight 2 were found to be shifted

by the same amount (approximately +3 K) compared to theascent radiosonde, the scientific gondola, as well asindependent radiosondes at Yakutsk. The reason for this isthat the thermistors were not calibrated together with theamplifiers flown, since the original flight models werebroken. The onboard calibration coefficients were thereforewrong. Since the difference was found to be identical forboth thermistors and was consistent with three independentdata sets, a correction factor was calculated and applieduniformly to both sensors.3. The Honeywell float pressure sensor of MIR 2 drifted

by some 10 hPa within the 17-day flight, probably becauseof leakage of the sensor, which did not appear during thetests and was never observed in other flights. Since themeasurements cannot be reliably corrected, they have beenreplaced by ECMWF pressure at the altitude measured bythe GPS at ± 100 m accuracy as described by Knudsen et al.[2001].

[30] Figure 6 shows the evolution of the nighttimetemperature recorded during both flights and its altitudeprofile. The altitude of the minimum temperature wasunusually high (20 hPa) at the beginning of the flights witha steep gradient above. During the next two weeks, thewarming shifted rapidly downward to end up with a mini-mum temperature at 80 hPa by the end of the period. Thesteep vertical temperature gradient is the cause of the largetemperature drop at the beginning of each night during thedescent of the balloon after day 54.

3.3. 2000 Flights

[31] The winter 1999/2000 is one of the coldest winters inthe lower stratosphere in the Berlin data series since 1964/1965 [Naujokat et al., 2000; Manney and Sabutis, 2000]. Inspite of the strong upper stratospheric warming in March,the vortex persisted in the polar region until the end of themonth. The PSC season lasted uninterrupted from mid-November to the first week of February at 30 hPa and fromthe beginning of December to almost mid-March at 50 hPa.[32] Two MIR flights were attempted during the winter.

The first was carrying a SAMBA gondola, including thetwo small scientific passengers and the new RUMBAmeteorological payload; the second carried the SAMBAand SALSA payloads recovered in 1999 from Siberia.[33] The first opportunity to fly within the vortex hap-

pened on February 18, when the vortex located over theBaffin Bay moved for a few days to northern Scandinavia.Since it was predicted to rapidly move over Scandinavia,both balloons were launched on the same day at 1600 and1800 LT. Figure 7 shows the PV and the temperature fieldsat 475 K (ECMWF analyses) on February 18, March 2(when the lowest temperatures were encountered by theballoon), and at the end of the flight on March 8 (whenthe center of the vortex moved to the Barents Sea). Theballoon’s trajectories are displayed in Figure 8 together withtheir altitude, air temperature, potential vorticity at 475 Kand at the vortex edge, and orography height below. Since2–3 days are required for the helium to escape, their altitudedrops from 5 hPa (35 km) at the beginning to an average ofabout 15 hPa (26 km) during daytime and 50–80 hPa (19–17 km) at night afterward. They were both terminated by theautomatic cut down of the payload when reaching 55�N,over Hudson Bay after just 2 days for the first, and after 18calendar days (21 flight days) over Belarus for the second.All payloads have been safely recovered.[34] Two of the meteorological sensors failed on the

second flight: the SALSA temperature sensor broke duringlaunch operations and the SAMBA temperature boomsdeployed late at the end of the ascent.[35] Figure 9 shows the temperature measured along MIR

flight 2. It remained almost constant between 18 Februaryand 6 March with a minimum of �75� ± 6�C around 18–22 km and a steep positive gradient above. However, it wassystematically colder over the European sector, resulting ina peak-to-peak variation of 10�C as the balloon circledaround the vortex. The PSC formation temperature, below�80�C (193 K) at 50 hPa, could be observed on twooccasions: on day 53 (21 February) over Greenland, andon days 62 and 63 (2 and 3 March) at 59�N over southernScandinavia and European Russia. The contribution oflarge-amplitude orographic waves to the cooling in the

Figure 6. (top) Time evolution of nighttime temperaturealong the second flight of 1999. (bottom) Profile (increasingsize of markers with time). In contrast to 1997 the altitudeof the minimum temperature shifted from an unusually highaltitude (20 hPa) in mid-February to 50 hPa during thestratospheric warming by the end of the month.


latter case, well documented by the MIR data, is studied indetail by Hertzog et al. [2002].

4. Accuracy of Meteorological Measurements

[36] The precision and accuracy of the SAMBA data,further used in the comparison with meteorological analy-ses, have been carefully assessed by comparison with allother independent SAOZ, SALSA, RUMBA, and radio-sonde measurements on the same balloon, as well as withupper air soundings of the network along the trajectory ofthe balloons.

4.1. Altitude and Location

[37] Depending on the number of satellites which couldbe received, the GPS is working in three-dimensional (3-D)mode, location and altitude, or 2-D mode, location only,at fixed altitude given by the latest 3-D determination.The GPS information transmitted by two independentgondolas were always found to be consistent within thespecification of 100 m given by the manufacturer when in3-D mode, the mode indication being also transmitted.They were always found consistent also with the ARGOSDoppler location though the latter is less accurate (between500 m and 3 km depending on the viewing angle of thesatellite). Finally, they were also consistent with the geo-potential height of the ascent radiosonde. The 3-D GPSlocation and altitude are therefore basically correct towithin ± 100 m.

[38] The only limitation is the episodic absence of alti-tude readings in the Arctic where the GPS often functions in2-D mode (location only) instead of 3-D, the default beingoften experienced simultaneously by two independent GPSon the same balloon. The reason for this, also frequentlyreported during short-duration flights in the Arctic but not atmidlatitude, is still unclear. No definite correlation wasfound with the temperature of the GPS receiver, nor withgeomagnetic activity. It is likely due to a combination oflow elevation of the viewing angle of some of the satellitesof the constellation, combined with the loss of sensitivityof the receiver at low temperature, the high speed ofthe balloon, and perhaps also to signal attenuation in theauroral zone. The only method of replacing the missing datais to derive the altitude from pressure readings after estab-lishing the relationship between the two. However, asdescribed below, pressure measurements may also experi-ence some problems, resulting in larger-altitude uncertaintyabout altitude.

4.2. Pressure

[39] The performance of the pressure sensors could beevaluated by comparison with pressure derived from GPSaltitude readings and the altitude/pressure relationship ofmodel analyses and, in one case, by the direct comparisonof independent pressure sensors. Examples of both areshown below.[40] Figure 10 shows the difference between the SAMBA

Honeywell and pressure calculated from GPS readings and

Figure 7. ECMWF PVand temperature fields at 475 K in 2000 at start of balloon flights on 18 Februaryon the coldest encountered day (2 March) and at the end of the second flight (6 March). Hatched areas:PV > 48 pvu and T < 193 K.


Figure 8. Balloon trajectories, altitude, air temperature, PV at 475 K, and orography height in 2000(left, MIR 1; right, MIR 2).


the ECMWF geopotential height for the second flight of2000 (Figure 10b), together with the corresponding altitudedifference (Figure 10c) and the altitude of the balloon(Figure 10a). A systematic difference, linearly dependent

on pressure, is detected; it has a magnitude of some 1.4 hPaaverage amplitude (�P = 0.62 + 0.023*P). It translates intoan altitude bias of 200 m around 18 km but of up to 800 maround 26–28 km during daytime. In addition a small

Figure 9. Temperature along the second flight of 2000. Filled Markers: PSC temperature on 21February over Greenland and on 2–3 March over Southern Scandinavia and European Russia.

Figure 10. Comparison of SAMBA pressure measurements along the second flight of 2000 to thatderived from GPS altitude and ECMWF geopotential height: (a) MIR altitude and pressure, (b) differenceof pressure, (c) corresponding height difference (markers, measurements; dotted line, after correction).


residual sensor temperature dependence of �0.05 hPa/10�Ccould also be seen. This is a common feature on all flightswhere either positive or negative differences between theHoneywell sensor and the ECMWF/GPS of up to ± 2 hPaare observed. If a correction is applied (dotted lines inFigures 10b and 10c), the two data sets could be reconciledwithin a residual average standard deviation of 0.66 hPa or115 m. However, errors in both the sensor and the ECMWF/GPS could be responsible, and their respective contributionsare difficult to evaluate.[41] Despite the short (3 days) duration of the first flight

of 2000, the inclusion of an independent Paroscientificpressure sensor in the RUMBA allows identification ofthe respective errors. Figure 11 shows the pressure meas-ured by the two sensors (Figure 11a), their difference withthe GPS/ECMWF pressure (Figure 11b), the differencebetween them (Figure 10c), and the temperature of theHoneywell sensor onboard SAMBA (Figure 11d). Again,a difference could be seen between the Honeywell sensorand the ECMWF of 0.5–1.5 hPa, but in opposite directioncompared to the previously discussed flight, while theindication of the Paroscientific agrees to within ± 0.5 hPa.

Another feature in Figure 11b is that the variation of thedispersion of the difference is very similar for both sensors.The amplitude of the dispersion decreases with altitude inrelation to the pressure change corresponding to the 100 mprecision of the GPS, varying thus from 0.08 hPa during thefirst night and day at 4–5 hPa, to 0.32 hPa on the third andlast night at 20 hPa.[42] This is confirmed by the small dispersion of the

difference between the pressure readings of the two sensorsshown in Figure 11c, though an increase at decreasingaltitude is also evident. In this case, the feature is thoughtto be related to vertical oscillations of the balloon of ± 30 m(± 0.1 hPa at 20 hPa) since the sampling frequency of thetwo sensors is different (10 and 15 min, respectively) andthe measurements thus not in phase.[43] The last feature that could be seen in the comparison

between the two sensors is a pressure shift along the flight,indicative of a differential temperature dependence of thesensors of �0.1 hPa/10�C. The comparison with ECMWF/GPS after day 50.4 at pressure larger than 10 hPa suggeststhat it could come from both sensors: �0.25 hPa/10�C forthe Paroscientfic and �0.35 hPa/10�C for the Honeywell.

Figure 11. Comparison of pressure measured by the SAMBA Honeywell sensor and the RUMBAParoscientific along the 3-day flight of the first balloon of 2000: (a) pressure, (b) difference with GPS/ECMWF, (c) difference between the two sensors, (d) temperature of the SAMBA gauge.


Finally, since the two independent instruments show thesame jump at high altitude (above 10 hPa) at the beginningof the flight, this last error must be attributed to ECMWF,which could be in error by some 0.5 hPa at 4 hPa.[44] In summary, the performances of the Honeywell-

SAMBA pressure readings, which will be used further, areslightly worse than expected. Though the repeatability of0.1 hPa given by the manufacturer is confirmed, theabsolute accuracy of the measurements is ± 2 hPa, partlybecause of calibration errors and partly because of a residualtemperature dependence not totally removed. The simulta-neous GPS altitude measurements in combination with theECMWF pressure/geopotential height relationship allow forcorrection of the data to within 0.5 hPa to 1.5 hPa at 10 to100 hPa, but alternatively, the same accuracy could bereached by directly using the pressure derived from theGPS altitude measurements. The last method is that pre-ferred for 1999 and 2000 when they are compared withmeteorological analyses.

4.3. Temperature

[45] Temperature measurements are difficult on balloonsat float because of the drop of forced convection comparedto the solar heating of the thermistor and to the perturbationwhen in the wake of the gondola, which is heated by theSun. Ideally, the temperature sensor should be small forinertial reasons, but not too small compared to the mean freepath, aluminized and mounted on a transparent mechanicalsupport for minimizing the solar heating, and far enoughfrom the gondola to reduce its radiative and convectiveperturbation. However, these recommendations result infragile sensors that are difficult to launch. Therefore depend-ing on the accepted level of risk as well as the location of thesensor in the flight train during launch operations, differentmountings have been used, ranging from a 1-m-longdeployable boom on the side of the SAMBA gondola tothermistors hanging 5 m below the payload at the bottom ofthe flight train for RUMBA. Since most of the data collectedare those of the SAMBA thermistors, the following dis-cussion concentrates on the evaluation of their performance:[46] An idea about their precision could be obtained by

comparing the simultaneous measurements of the twothermistors at opposite sides of the gondola. In turn, theaccuracy could be checked by comparing the measurementsof independent sensors on the same balloon as well as withradiosondes of the network along the balloon trajectory. Inboth cases an idea of direct or indirect perturbation by solarheating could be obtained by comparing the daytime andnighttime data.4.3.1. Precision and perturbation by solar heating[47] The two thermistors T1 and T2 of the SAMBA

payload are identical and their measurements are independ-ent and simultaneous. The comparison of their readingsalong the flight is therefore a good indicator of the precisionand repeatability of the measurements. As an examplerepresentative of all flights, Figure 12 (top) shows thedifference between T1 and T2 during the 17 days of theMIR 2 flight in 1999. On average, T1 is warmer by 0.39 K,but the difference also shows an average dispersion of± 1.97 K (1 SD). Figure 12 (middle) shows the same dataplotted versus the solar zenith angle (SZA). The dispersionis much larger during daytime (± 2.97 K) than at night

(SZA > 93�), approximately the moment when the balloonstarts to descend rapidly in the evening), when it drops to± 0.3K. The dispersion of SAMBAmeasurements is 10 timeslarger in daytime than at night. In addition, because of thedecreasing convection, the amplitude of the dispersion alsoincreases with altitude, even at night (Figure 12 (bottom)).[48] The question is how much of the dispersion is due to

direct heating of the thermistor and how much to the warmpool air surrounding the aluminized payload, which is up to65�C warmer than the ambient temperature during daytime,according to thermal simulations.[49] Comparing the temperature sensors flown onboard

the SAMBA and RUMBA payloads during the first flight of2000 (see Figure 11 for pressure and altitude) allows betterunderstanding of the origin of the perturbation. Figure 13compares the temperature measured by the sensors of bothpayloads along the 3-day flight. The difference between thetwo SAMBA thermistors mounted along 1-m-long booms(Figure 13b) shows features very similar to that of 1999.The dispersion is much larger during daytime and drops atdecreasing altitude from 2 K at 5 hPa during the first nightto 0.3 K at 22 hPa during the third night. The differencebetween the two RUMBA sensors hanging 5 m below thegondola (Figure 13c) shows the same altitude dependence

Figure 12. Difference between the two SAMBA simulta-neous temperature readings during the second flight of1999: (a) time evolution, (b) solar zenith angle dependence,(c) altitude dependence of nighttime measurements.


of the dispersion but not the daytime increase. Since thethermistors of both payloads are identical, most of thevariance of the daytime SAMBA measurements must beattributed to the indirect perturbation by the gondola.Finally, Figure 13d shows the difference between theSAMBA and the RUMBA sensors displaying the lowesttemperature, (min (T1, T2)). The SAMBA sensors arewarmer during daytime by 4 K at 5 hPa on the first dayand by 2 K at 10 hPa on the second day. Again, since thesensors of both payloads are identical, the differentialheating cannot be attributed to direct solar heating but tothe warm air surrounding the SAMBA payload. Theremaining question is how much warmer are the RUMBAsensors compared to the ambient during daytime? Althougha precise answer to this question cannot be provided inabsence of a reference, the comparison between day andnighttime measurements at the same altitude (i.e., at date50.2 compared to 49.5 and 50.9 compared to 50.4) suggeststhat the heating is small: less than 2 K at 5 hPa and than 1 Kat 10 hPa.[50] In conclusion, the precision of the SAMBA temper-

ature measurements at night and at twilight during the

Figure 13. Comparison of SAMBA and RUMBA temperature measurements during the 3-day firstflight of 2000: (a) temperature, (b) SAMBA T1-T2, (c) RUMBA T1-T2, (d) SAMBA T1-RUMBA T1.

Figure 14. Comparison of temperatures measured duringthe initial ascent by the SAMBA T2 thermistor and theradiosonde on the second flight of 1999 (right, radiosonde;left, difference with T2).


descent of the balloon, at a pressure larger than 10 hPa, is ±0.4 K. Because of the solar heating, they are useless duringdaytime, but largely because being mounted too close to thegondola, rather than the direct heating of the thermistors.Mounting them a few meters below the payload would beenough to provide good measurements during both daytimeand nighttime.4.3.2. Accuracy 1: Comparison withonboard radiosonde[51] As already mentioned, an RS-80 Vaisala radiosonde

was always added to the flight train for comparison duringthe initial nighttime ascent. For 3 hours the sampling rateof the two thermistors of SAMBA was also accelerated to1/50 s (1997) or 1/60 s (1999 and 2000), transmitted to anARGOS ground receiver when still in the telemetry range.As an example, the difference between SAMBAT2 and theradiosonde of flight 1 in 1999, after correction of the lengthbetween the two instruments, is shown in Figure 14. Thepicture is typical of that observed on all flights: (1) analmost constant difference with relatively little scatter(0.03� ± 0.4�C in the present case) between the groundand 28 km (10 hPa); (2) an increasing difference and alarger scatter above, when the sensors are much lessventilated (large Thermocap Vaisala sensor warmer thansmall thermistors). As shown in Table 3, the differencebetween SAMBA and the radiosonde on the other flightsnever exceeded 0.3 K, which is not significant compared tothe quoted accuracy of 0.5 K of each sensor. The MIRtemperature measurements are therefore fully consistentwith those of commercial Vaisala radiosondes largely inuse in the West.4.3.3. Accuracy 2: Comparison to SAOZ/SALSA[52] Though of different sampling rate, the nighttime

SAMBA temperature data could also be compared to thoseof the Thermocap sensor of the SAOZ/SALSA scientificpayload. As an example, Figure 15 shows the difference

between the 142 simultaneous nighttime SAMBA 1 andSAOZ data points during the 22-day flight 2 of 1997. Onaverage, it is + 0.1 ± 0.6 K. Since most SAOZ measure-ments are made during twilight, i.e., the descent of theballoon, and as shown in Figure 4, the temperature increasesfrom 198 K to 215 K during the flight; the comparisoncovers a wide range of temperatures. To the extent that themeasurements are independent, the uncertainty on the biaswould be 0.6/sqrt (142�1) = 0.05 K. The calculated trendalong the flight of �0.01 ± 0.01 K/d is not significant.Similar results were obtained on MIR 1 in 1997, with moredispersion (0.4 ± 2.3 K) because of the bad mounting ofthe SAOZ sensor and on MIR 2 in 1999 (0.5 ± 0.6 K). Thecomparison could not be performed in 2000, since theSAOZ sensor broke during launch operations.[53] Overall, the long-term comparison along the flights

of the SAMBA temperature with that of the Vaisala Ther-mocap flown onboard the SAOZ/SALSA does not show

Table 3. Comparison of Temperature Readings of Various Balloon-Borne Sensors and Radiosondes (SAMBA, Nighttime Only)

Flight Sensor �T ± SD Observation

97.1 SAMBA2-SAMBA1 0.3 ± 0.9SAMBA2-Vaisala 0.0 ± 0.5SAMBA2-SAOZ �0.4 ± 2.3 SAOZ bad mountingSAMBA2-network �0.0 ± 2.6

97.2 SAMBA2-SAMBA1 �0.6 ± 0.4SAMBA2-Vaisala 0.3 ± 0.4SAMBA2-SAOZ 0.1 ± 0.6SAMBA2-network �0.0 ± 2.6

99.1 SAMBA2-SAMBA1 NA SAMBA 1 not deployedSAMBA2-Vaisala 0.0 ± 0.1SAMBA2-SAOZ �0.5 ± 0.6SAMBA2-Russian sonde 0.6 ± 0.5

99.2 SAMBA2-SAMBA1 �0.5 ± 0.3SAMBA2-Vaisala 0.1 ± 0.2SAMBA2-SAOZ NA SAOZ brokenSAMBA2-Russian sonde 0.3 ± 0.6

00.1 SAMBA1-SAMBA2 NA SAMBA2 brokenSAMBA1-Vaisala 0.0 ± 0.2RUMBA1-RUMBA2 0.5 ± 0.4 day and nightVaisala-RUMBA1 0.0 ± 0.2Vaisala-RUMBA2 0.4 ± 0.2SAMBA1-RUMBA1 0.1 ± 0.5 nighttime only, 25 hPa

00.2 SAMBA1-SAMBA2 �0.4 ± 0.4SAMBA2-Vaisala NA SAMBA at float onlySAMBA1-SAOZ NA SAOZ brokenSAMBA1-radiosondes �0.6 ± 1.9 nighttime, <500 km, <1 hour

Figure 15. Difference between temperatures measured atnight by the SAMBA T2 thermistor and the SAOZ VaisalaThermocap sensor along the second flight of 1997.


any significant bias or shift within the 0.5 K accuracy of thecalibration of the sensors.4.3.4. Accuracy 3: Comparison to radiosondes ofthe network[54] The MIR temperature readings have also been com-

pared to radiosondes when the balloon was passing close toa station of the network. Selected stations from which fullascent data have been obtained are Sodankyla in Finland,Thule, and Scoresbysund in Greenland, Ny-Alesund inSvalbard, and Yakutsk in eastern Siberia. However, coloca-tions are infrequent and the altitude range of overlap islimited because of the frequent bursting of the radiosondeballoons in the winter. Thirteen data points have beenobtained for the two flights of 1997, within a distance of1500 km. Though the scatter is quite large (�0.04 ± 2.58 K),no significant departure appears between the two sets ofdata, and more importantly, there is no sign of temporal drift(�0.04 ± 0.05 K/d).[55] A better comparison was possible in 1999 with the

radiosondes of Yakutsk, when the two MIRs were travelingaround the station for 2–3 days. Figure 16 shows thecomparison between the first balloon and the radiosondeon 26 February 1999. In the altitude range of 13–18 kmwhere they do overlap (most of the Yakutsk soundings stopat 18 km), the difference 0.6 ± 0.5 K is not significant. Theresult for the second MIR (0.3 ± 0.6 K) was very similar.There is no indication of a significant difference with theRussian sonde.[56] Finally, eight nighttime colocated measurements

within less than 500 km and 1 hour with the radiosondesof the network could be found on the second flight of 2000.The average difference is �0.6 ± 1.9 K.4.3.5. Summary of temperature accuracy[57] The information available from all comparisons is

summarized in Table 3. Overall, it can be concluded thatnighttime SAMBA temperature measurements at levelsbelow 28 km (10 hPa) are consistent to within the± 0.5 K accuracy of the calibration in the laboratory. Onaverage, the dispersion of the measurements in flight atnight because of the perturbation by the surrounding gon-dola is ± 0.4 K.

5. Concluding Remarks

[58] Two MIRs have been flown each year in 1997, 1999,and 2000 in the Arctic vortex, demonstrating that this type

of vehicle is well adapted for this use even during astratospheric warming event, providing the total weight ofthe payload does not exceed 60 kg for a balloon volume of45,000 m3. It appears that the most difficult constraint toaccommodate is to remain poleward of 55�N, since the edgeof the vortex frequently moves to this latitude in theNorthern Hemisphere. An option could be to launch theballoon farther inside the vortex, but this would requirelaunch operations to be conducted at a higher latitude, sincethe site of ESRANGE is infrequently located in the middleof the vortex.[59] A variety of meteorological measurements has been

conducted along the flights. GPS location and altitude arefound to be accurate within the specification of ± 100 m ofthe system (an even better accuracy of a few meters isexpected after 2001, after the removal of the noise artifi-cially added in the GPS system). However, there are somelimitations in the use of GPS in the Arctic area, where the 3-D information about location and altitude is not alwaysavailable. The accuracy of Honeywell pressure measure-ments of the SAMBA payload (± 2 hPa) is slightly worsethan expected. However, after correction of calibration aswell as of the sensor temperature dependence using the GPSaltitude and the pressure/altitude relationship determinedfrom ECMWF analyses, the accuracy could be improved to0.5–1.5 hPa, respectively, in the 10–100 hPa pressurerange. Since the use of ECMWF is required in both cases,the preferred and easier method is the use of pressuredirectly derived from GPS altitude.[60] The temperature of the SAMBA payload was dem-

onstrated to be measured with an average precision of ± 0.4K and an average bias of less than ± 0.5 K, consistent withthe calibration, but only below 28 km during nighttime.Daytime temperatures are not reliable. It has been demon-strated that the reason for this is not the direct heating of thethermistor, but the radiative warming of air surrounding thegondola. Encouraging results have been obtained withtemperature sensors hanging 5 m below the payload,although these are at an increased risk of breaking atlaunch.[61] Overall, the six MIR flights provide a unique

independent data set of about 1600 points in 1997,1900, in 1999 and 1800 in 2000, over a broad altituderange of 5–140 hPa in a variety of meteorological con-ditions in the winter Arctic vortex. These data will be nowused in the companion paper for evaluating the temper-ature of the ECMWF, MO, DAO, NCEP, and NCEP/NCAR meteorological analyses, which are frequently usedfor simulating the ozone destruction in the polar strato-sphere.

[62] Acknowledgments. The authors thank Pierre Francois, Ericd’Almeida, and Bernard Brioit for their hard work in building the SAOZ,SALSA and RUMBA gondolas, the CNES balloon division who designedand operated SAMBA and the IR Montgolfier, Vaisala OY for providingthe software of the radiosondes, the Swedish Space Corporation for theflights control, ATCs of all Arctic countries for authorizing the long-duration flights to take place, CAO for controlling the flight over Russia,the radiosonde stations for providing their raw data, E. R. Nash for vortexedge calculations and G. O. Braathen for providing ECMWF PV andtemperature maps. The project was supported by the French ProgrammeNational de Chimie de l’Atmosphere (PNCA), the Centre National d’EtudesSpatiales (CNES) and the DG Research of the European Commission(Lagrangian Experiment contract ENV4CT970504).

Figure 16. SAMBAT2 temperature (dots) and radiosonde(solid line) at Yakutsk in Siberia on 26 February 1999.


ReferencesAdriani, A., F. Cairo, L. Pulvurenti, C. Cardillo, M. Viterbini, F. Di Fran-cesco, and J.-P. Pommereau, Aerosol and polar stratospheric clouds ob-servations by laser backscatter-sonde in the frame of the European projectStratospheric Regular Soundings, Ann. Geophys., 17, 1352–1360, 1999.

Carlslaw, K. S., et al., Increased stratospheric ozone depletion due to moun-tain-induced atmospheric waves, Nature, 391, 675–678, 1998.

Carlslaw, K. S., and H. Oelhaf, Polar ozone, in European Research in theStratosphere 1996–2000, EUR 19867, Eur. Comm., St. Jean de Luz,France, 2001.

Chipperfield, M. P., Multiannual simulations with a three-dimensional che-mical transport model, J. Geophys. Res., 104, 1781–1805, 1999.

Coy, L., E. R. Nash, and P. A. Newman, Meteorology of the polar vortex:Spring 1997, Geophys. Res. Lett., 24, 2693–2696, 1997.

Dornback, A., M. Leutbecher, H. Volkert, andM.Wirth, Mesoscale forecastsof stratospheric mountain waves, Meteorol. Appl., 5, 117–126, 1998.

Gardiner, T., I. Howeison, G. Hansford, N. Swann, S. Garcelon, R. L.Jones, and P. Woods, Measurements of stratospheric methane duringTHESEO using a balloon-borne near-infrared laser absorption spectro-meter, in Proceedings of the 5th European Ozone Symposium, EC AirPollut. Res. Rep. 73, Eur. Comm., St. Jean de Luz, France, 2000.

Goutail, F., et al., Ozone depletion in the Arctic during the winters of1993–94 and 1994–95, J. Atmos. Chem., 32, 1–34, 1999.

Guirlet, M., M. P. Chipperfield, J. A. Pyle, F. Goutail, J.-P. Pommereau,P. von der Gathen, and E. Kyro, Modeled arctic ozone depletion in winter1997/1998 and comparison to previous winters, J. Geophys. Res., 105,22,185–22,200, 2000.

Hansford, G. M., et al., Solid state ozone sensors for the future: Lightweight, low power and continuous operation, in Proceedings of the 5thEuropean Ozone Symposium, EC Air Pollut. Rep. 73, pp. 683–686, Eur.Comm., St. Jean de Luz, France, 2000.

Hertzog, A., F. Vial, A. Dornback, S. D. Eckermann, B. M. Knudsen, andJ.-P. Pommereau, In situ observations of gravity waves and comparisonswith numerical simulations during the SOLVE/THESEO 2000 campaign,J. Geophys. Res., 107, 10.1029/2001JD001025, in press, 2002.

Keil, M., M. Heun, J. Austin, W. Lahoz, G. P. Lou, and A. O’Neill, The useof long-duration balloon data to determine the accuracy of stratosphereanalyses and forecasts, J. Geophys. Res., 106, 10,299–10,312, 2001.

Knudsen, B. M., Accuracy of arctic stratospheric temperature analyses andthe implications for the prediction of polar stratospheric clouds, Geophys.Res. Lett., 23, 3747–3750, 1996a.

Knudsen, B. M., J. M. Rosen, N. T. Kjome, and A. T. Whitten, Comparisonof analyzed temperatures and calculated trajectories with long-durationballoon data, J. Geophys. Res., 101, 19,137–19,145, 1996b.

Knudsen, B. M., J.-P. Pommereau, A. Garnier, M. Nunez-Pinharanda,L. Denis, G. Letrenne, M. Durand, and J. M. Rosen, Comparison ofstratospheric air parcel trajectories based on different meteorological ana-lyses, J. Geophys. Res., 106, 3415–3423, 2001.

Knudsen, B. M., J.-P. Pommereau, A. Garnier, M. Nunes-Pinharanda,L. Denis, P. Newman, G. Letrenne, and M. Durand, Accuracy of analyzedstratospheric temperatures in the winter Arctic vortex from infrared Mon-tgolfier long-duration balloon flights, 2, Results, J. Geophys. Res., 107,10.1029/2001JD001329, in press, 2002.

Leutbecher, M., and H. Volkert, Stratospheric anomalies and mountainwaves: A three-dimensional simulation using a multiscale weather pre-diction model, J. Geophys. Res., 23, 3329–3332, 1996.

Malaterre, P., V. Dubourg, G. Letrenne, and M. Durand, The long durationMIR balloon, in Proceedings of the 12th ESA Symposium on Rocket andBalloon Programmes, ESA SP-370, pp. 329–334, Eur. Space Agency,Paris, France, 1996.

Manney, G. L., and J. L. Sabutis, Development of the polar vortex in the1999–2000 Arctic winter stratosphere, Geophys. Res. Lett., 27, 2589–2592, 2000.

Manney, G. L., R. Swinbank, S. T. Massie, M. E. Gelman, A. J. Miller,R. Natagani, A. O’Neill, and R. W. Zurek, Comparison of UK meteor-ological Office and U.S. National Meteorological Center stratosphericanalyses during northern and southern winter, J. Geophys. Res., 101,10,311–10,334, 1996.

Nash, E. R., P. A. Newman, J. E. Rosenfield, and M. R. Schoeberl, Anobjective determination of the polar vortex using Ertel’s potential vorti-city, J. Geophys. Res., 101, 9471–9478, 1996.

Naujokat, B., A synoptic view of the ECMWF stratospheric forecasts,stratosphere and numerical weather prediction, in Proceedings of theECMWF Workshop, Eur. Cent. for Medium-Range Weather Forecasts,Reading, UK, 1994.

Naujokat, B., The winters of 1997/98 and 1998/99 in perspective, in Pro-ceedings of the 5th European Ozone Symposium, edited by Harris, Guir-let, and Amanatidis, pp. 107–110, EC Air Pollut. Res. Rep. 73, Eur.Comm., 1999.

Naujokat, B., and S. Pawson, The unusually cold, persistent vortex inspring 1997, in Proceedings of the 4th European Ozone Symposium,edited by Harris, Kilbane-Dawe, and Amanatidis, pp. 50–53, EC AirPollut. Res. Rep. 60, Eur. Comm., 1998.

Naujokat, B., M. Kunze, and S. Pawson, Winter report 1999/2000, Eur.Ozone Res. Coord. Unit, 2000. (www.ozone-sec.ch.cam.ac.uk)

Pawson, S., K. Kruger, R. Swinbank, M. Bailey, and A. O’Neill, Intercom-parison of two stratospheric analyses: Temperature relevant to polar stra-tospheric cloud formation, J. Geophys. Res., 104, 2041–2050, 1999.

Pommereau, J.-P., and A. Hauchecorne, A new atmospheric vehicle: TheInfra-Red Montgolfier, in Scientific Ballooning, pp. 55–58, Pergamon,New York, 1979.

Pommereau, J.-P., and J. Piquard, Ozone, nitrogen dioxide and aerosolvertical distributions by UV-visible solar occultation from balloons, Geo-phys. Res. Lett., 21, 1227–1230, 1994.

Pommereau, J.-P., P. Cseresnjes, L. Denis, A. Hauchecorne, G. Letrenne,and B. Knudsen, Stratospheric temperature measurements in the winterArctic vortex in 1997 on-board long duration balloons: Comparison toECMWF model and detection of orographic waves, in Proceedings of theEuropean Workshop on Mesoscale Porcesses in the Stratosphere, editedby Carslaw and Amanatidis, pp. 211–216, EC Air Pollut. Res. Rep. 69,Eur. Comm., 1999.

Pullen, S., and R. L. Jones, Accuracy of temperatures from UKMO ana-lyses of 1994/95 in the arctic winter stratosphere, Geophys. Res. Lett., 24,845–848, 1997.

Swinbank, R., and A. O’Neill, A stratosphere-troposphere data assimilationsystem, Mon. Weather Rev., 122, 686–702, 1994.

Vial, F., et al., Strateole, a project to study Antarctic polar vortex dynamicsand its impact on ozone chemistry, Phys. Chem. Earth, 20, 83–96,1995.

Yushkov, V., S. Merkoulov, V. Astakhov, J.-P. Pommereau, and A. Garnier,A Lyman alpha hygrometer for long duration IR Montgolfier during theTHESEO Lagrangian Experiment, in Proceedings of the 5th EuropeanOzone Symposium, EC Air Pollut. Res. Rep. 73, pp. 404–407, Eur.Comm., 2000.

��L. Denis, A. Garnier, M. Nunes-Pinharanda, and J.-P. Pommereau,

Service d’Aeronomie du CNRS, BP3, Route des Gatines, 91371 Verrieres-le-Buisson Cedex, France. ([email protected])B. M. Knudsen, Danish Meteorological Institute, Copenhagen, Denmark.M. Durand and G. Letrenne, Centre National d’Etudes Spatiales,

Toulouse, France.A. Hertzog and F. Vial, Laboratoire de Meteorologie Dynamique,

Palaiseau, France.F. Cairo, Istituto de Fisica dell’ Atmosfera, CNR, Rome, Italy.


accuracy of analyzed stratospheric temperatures in the winter … · 2015. 8. 10. · accuracy of...

Documents