ice supersaturation in the tropopause region over

Meteorologische Zeitschrift Vol 12 No 3 143-156 (June 2003)c by Gebruder Borntraeger 2003 Article

Ice supersaturation in the tropopause region overLindenberg Germany

PETER SPICHTINGER1 KLAUS GIERENS1 ULRICH LEITERER2 and HORST DIER2

1 Deutsches Zentrum fur Luft- und Raumfahrt Institut fur Physik der Atmosphare Oberpfaffenhofen Germany2 Deutscher Wetterdienst Meteorologisches Observatorium Lindenberg Germany

(Manuscript received January 22 2002 in revised form February 6 2003 accepted April 13 2003)

AbstractThe occurrence of ice-supersaturation layers (in either clear air or in cirrus) over the Meteorological Ob-servatory Lindenberg is investigated for the period February 2000 to April 2001 by means of the humiditytemperature and pressure reports obtained from the Lindenberg corrected RS80A routine radiosonde TheRS80A routine sonde data are corrected on the basis of weekly comparison ascents with Lindenberg researchradiosonde humidity data This research sonde applies the Lindenberg measuring and evaluation techniqueof ldquostandardized frequenciesrdquo We study the frequency of occurrence of ice supersaturation in the tropopauseregion over Lindenberg the vertical distribution of ice-supersaturation layers in the upper troposphere andlowermost stratosphere their situation relative to the tropopause their vertical dimensions their temperaturesand the statistical distribution of relative humidities The mean frequency of occurrence of ice-supersaturationlayers is about 28 Most of them occur within a broad layer extending 200 hPa down from the tropopauseMost events occur in cold air with temperature below 40 C Their vertical extensions can be tted by a pairof Weibull distributions with mean 560 610 m The results are compared with ndings from the MOZAICproject from the SAGE II satellite instrument and with various results from Lidar measurements

ZusammenfassungKorrigierte Routine-Radiosondendaten der Sonde RS80A (Feuchte Temperatur und Druck) werden hin-sichtlich des Auftretens von Eisubersattigung (in klarer Atmosphare sowie in Zirren) uber Lindenberg fur denZeitraum Februar 2000 bis April 2001 ausgewertet Die Korrektur erfolgt anhand wochentlicher Vergleichs-aufstiege mit der Forschungssonde des Meteorologischen Observatoriums Lindenberg Die Forschungssondenutzt das Lindenberger Mess- und Auswerteverfahren der ldquostandardisierten Feuchtefrequenzenrdquo Wir un-tersuchen die Hau gkeit des Auftretens von Eisubersattigung uber Lindenberg die Hohenverteilung derSchichten in denen Eisubersattigung auftritt in der oberen Troposphare und untersten Stratosphare derenLage relativ zur Tropopause deren vertikale Machtigkeit deren Temperaturen sowie die statistischeVerteilung der relativen Feuchte Im Mittel kommt Eisubersattigung in etwa 28 der Sondenaufstiegevor meistens in einem Bereich bis zu 200 hPa unterhalb der Tropopause und die Temperatur liegt meis-tens unterhalb 40 C in den eisubersattigten Luftmassen Die Verteilung der vertikalen Ausdehnung dereisubersattigten Luftmassen lasst sich durch ein Paar von Weibull-Verteilungen anpassen wobei die mit-tlere vertikale Ausdehnung etwa 560 610 m betragt Die Ergebnisse werden mit Resultaten des MOZAIC-Projektes und des SAGE II Satelliteninstrumentes verglichen sowie mit diversen von Lidarmessungengewonnenen Ergebnissen

1 IntroductionCirrus clouds in the cold tropopause region are believedto form mainly by homogeneous freezing of aqueous so-lution droplets (cf HEYMSFIELD and MILOSHEVICH1993 HEYMSFIELD and SABIN 1989 SASSEN andDODD 1988) At temperatures lower than 40 C thisprocess requires airmasses in a state of substantial super-saturation with respect to ice (generally more than 40KOOP et al 2000) Such a high supersaturation can beachieved by either cooling of airmass by water advec-tion or by appropriate mixing of two airmasses (as information of condensation trails) However an airmass

Correspondingauthor Peter SpichtingerDLR Institut fur Physikder Atmosphare OberpfaffenhofenD-82234 Weszligling Germanye-mail peterspichtingerdlrde

that once reaches ice supersaturation does not necessar-ily later reach the freezing threshold and thus there existairmasses that are supersaturated but clear Such regionshave been termed ice-supersaturated regions (ISSRsGIERENS et al 1999) Evidently an airmass containinga cirrus cloud must have gone through an ISSR stageFurther it is possible that a thin or subvisible cirrus oc-curs as a transient phenomenon in an ISSR when icecrystals sediment out of the region while the region it-self remains in an uplifting state

ISSRs have been detected by a variety of hygro-meters during several airborne measurement campaignsThe types of instruments include frost point hygrometer(OVARLEZ et al 2000) tunable diode laser (JENSENet al 1998 VAY et al 2000) and the capacitive sen-

DOI 1011270941-294820030012-01430941-2948030012-0143 $ 0630

c Gebruder Borntraeger Berlin Stuttgart 2003

144 P Spichtinger et al Ice supersaturation in the tropopause region Meteorol Z 12 2003

sor (HELTEN et al 1998 1999) The latter is installedon ve commercial aircraft within the Measurement ofozone by Airbus in-service aircraft (MOZAIC) project(MARENCO et al 1998) The MOZAIC data wereused by GIERENS et al (1999) to determine the fre-quency of occurrence of ISSRs over the northern mid-latitudes and to determine the statistics of relative hu-midity within ISSRs Mean temperature differences of3ndash4 K between ISSRs and their subsaturated environ-ment (ISSRs are colder) could be determined in the samework as well as water vapour mixing ratios in ISSRsthat exceed their counterparts in the neighbouring sub-saturated regions by typically a factor 15 GIERENSet al (2000) related the MOZAIC data to statistics ofoccurrence of subvisible cirrus clouds (WANG et al1996) obtained from data of the Stratospheric Aerosoland Gas Experiment II (SAGE II) on board the EarthRadiation Budget Satellite (ERBS) A close relationbetween ISSRs and subvisible cirrus is probable ac-cording to that work Typical horizontal extensions ofISSRs could be determined from the MOZAIC databy GIERENS and SPICHTINGER (2000) Recently datafrom the Microwave Limb Sounder (MLS) on boardthe Upper Atmospheric Research Satellite (UARS) wereevaluated by SPICHTINGER et al (2002) in order to ob-tain a global version of the humidity statistics inside andoutside ISSRs

Radiosondes are not included in the above list ofinstruments used for measuring humidity in ISSRsThe reason for this is that hygrometers that are opera-tional in current radiosonde types (usually RS80 withA-Humicap by Vaisala) cease to furnish reliable resultsin cold air (T 40 C) or at low absolute humiditiesconditions that are typical for the upper troposphere andbeyond (eg NASH and SCHMIDLIN 1987 ELLIOTTand GAFFEN 1991) However the situation is begin-ning to improve with the advent of the new H-Humicapsensor of Vaisala and the RS90 sonde (ANTIKAINENand JAUHIAINEN 1995) The Lindenberg Observatoryof Deutscher Wetterdienst has developed the method ofldquostandardized frequenciesrdquo (LEITERER et al 1997) so-called FN-method using a modi ed RS90 as a researchradiosonde This method allows to obtain very accuraterelative humidities using the research sonde throughoutthe troposphere and into the lowermost stratosphere Theresearch sonde has been used to correct the routine RS80A radiosondes (LEITERER et al 20021 NAGEL et al2001) Hence this data can be used to examine the up-per troposphere and lowermost stratosphere concerningrelative humidity with respect to ice

As radiosonde reports do not contain any informa-tion about the concentration of ice crystals along theirpath it is not possible to distinguish between cloudyand cloud-free parts of a pro le However evaluatingthe Lindenberg data we nd certain signatures that arecharacteristic of ISSRs and others that are character-

1 aviable from wwwdwddedeFundEObservatorMOLmol3mol3 RS80 korpdf

istic of cirrus clouds (see below in particular 34 and35) Further radiosondes of good quality can be usedto study the vertical extension of ice supersaturated air-masses (ISSR and cirrus) and their situation relative tothe local tropopause Such investigations are the topic ofthe present paper

In the next section the FN-method for the researchsonde and the method for correction of the RS80 A-Humicap pro les (the Lindenberg routine radiosonde 4times daily) will be described Section 3 then presentsand discusses the results These are summarised andconclusions are drawn in Section 4

2 The Lindenberg corrected RS80 A data

21 Fundamental principles of calibrationand corrected RS80 A humidity pro les

Weekly comparison ascents of the routine RS80 A-Humicap radiosonde with the Lindenberg research ra-diosonde (a modi ed type of the RS90 radiosonde) arethe basis of the calibration The research radiosondeuses the Lindenberg measuring and evaluation methodof ldquostandardized frequenciesrdquo (FN-method) (LEITERERet al 1997) as described in Section 211 The derivedcorrection procedure for RS80 A-Humicap data is de-scribed in Section 212

211 The FN-method

A new measuring and evaluation method has been de-veloped at the Meteorological Observatory Lindenbergwhich is suited for using the modi ed RS90 sensorunit as reference humidity sensor during the radiosondeascent The general measuring technique is based onthe capacity dependence of a thin polymer layer in re-lation to the quantity of water molecules which areoozed into the polymer structure The capacity will betransformed into frequencies by a resonant circuit ofthe radiosonde transmitter The technical improvementsof the RS90 sensor unit with reference to the RS80Asensor unit are described in detail by Vaisala authors(ANTIKAINEN and PAUKKUNEN 1994 PAUKKUNEN1995 ANTIKAINEN and JAUHIAINEN 1995) Bothpolymer sensors measure the partial pressure relation ew

esw

(ew partial pressure of water vapour esw saturation par-tial pressure of water vapour over plane liquid water sur-face) or the relative humidity with respect to water evenunder water or ice saturated and supersaturated condi-tions (in ISSR and cirrus) as reported in NAGEL et al2001

The FN-method of ldquostandardized frequenciesrdquo de-scribed in detail in a series of publications (LEITERERet al 1997 LEITERER et al 2002 NAGEL et al2001) enables us to carry out reference humidity sound-ings in comparison to other humidity soundings Theabsolute accuracy of the new measuring and evalua-tion technique is about 1 RH in the temperature

Meteorol Z 12 2003 P Spichtinger et al Ice supersaturation in the tropopause region 145

Table 1 Coef cients for the polynomials of the calibration matrixin Fig 1 as functions of the standardised frequencies FN

U t A FN t2 B FN t C FN

FN A FN B FN C FN00 00008 00384 0486701 00013 00809 685902 00014 01426 1611603 0002 01605 2550904 00023 01721 3636205 00029 01407 4671906 00027 01176 5818607 00025 00986 6979708 00023 00974 8139309 00025 01109 9115910 00027 00982 10055

region of 35 to 70 C The sensor unit consists ofthe F-Thermocap (for temperature) and two alternatelyheated H-Humicaps (for relative humidity RH) whichare heated during 60 s and are used for measurementduring 160 s The measuring frequencies if heating isswitched on are used as additional calibration points at0 RH during the radiosonde ascent Therefore one re-ceives alternate data from both humidity sensors whichmake up one humidity pro le The absolute calibration(standardization) is derived from a so-called ground-check using a calibration box ventilated with 5 ms atsaturation-conditions (100 RH) and room temperature(20 to 25 C) The individual difference DF (in the order

Figure 1 The standardised frequencies FN calibration matrixThe relative humidity U () with respect to water as a func-tion of standardized frequencies FN (solid and dotted polynomialcurveslabelled with their respective FN) and temperature t Thedashed curve represent the saturation with respect to ice

of 2000 Hz) will be measured for each Humicap withan accuracy of 01 Hz a few minutes before the ascentwith DF FH FM 100 (FH measuring frequency ofheated Humicap at 100 RH FM 100 measuring fre-quency of unheated Humicaps at 100 RH) On the ba-sis of this frequency difference DF the so-called indi-vidual standardized frequencies FNi for each measuringpoint i (every 16 s) during an ascent will be calculatedwith

FNiFHi FMi

DF(21)

withFHi individual mean frequency of heated sensor duringthe heated cycle (60 s) equivalent about to 0 RHFMi individual measuring frequency for each measuringpoint i (each 16 s) during the measuring cycle (160 s)

The evaluation algorithm calculates the exact indi-vidual RHUi() for each measuring point i using theseindividual standardized frequencies FNi on the basis ofa FNi universal calibration matrix (see Tab 1 Fig 1)The universal calibration matrix gives the mathemati-cal relation between the standardized frequency FN theRH U () with respect to water and temperature t ( C)The absolute accuracy of the sonde is 1 in RH (NAGELet al 2001)

212 The RS80 A-Humicap correction

It is worth mentioning that the RS80 A-Humicap sen-sor is a polymer humidity sensor which is covered witha rain protection cap Therefore the ventilation of thehumidity sensor is not guaranteed Icing effects on thesensors occur in about 15 percent of cloudy weatherconditions in spite of the rain protection cap (results of360 especially investigated soundings in Lindenberg)But this partly or total sensor icing is not recognized inroutine operation This means that the uncertainty (stan-dard deviation) of the RS80 A-Humicap humidity dataincreases

A special method to recognize icing cases was de-veloped for the Lindenberg station 10393 This methoduses climatological water vapour mixing ratio data ofabout 45 ppmv in the lower stratosphere in a heightof 5 km above the tropopause (for Lindenberg about16 km) If the RH data (expressed in ppmv) in the lowerstratosphere are higher than a threshold (obtained fromresearch and routine RS80 sondes comparisons) the op-erator recognizes this case and cancels all RH data in thetroposphere with temperatures lower than 10 C

Despite of this ldquoicing recognitionrdquo and elimination ofldquoicedrdquo RH data one has to apply the following correctionformula containing a so-called additional RH ground-check-correction and an additional RH temperature de-pendent correction (details are described in LEITERERet al 2002)


Ucorr w Ums w Upsy Ums 0Ums w

Upsy

0 005 t2 0 112 t 0 404 (22)

Ums w Upsy Ums 0Ums wUpsy

Uw t 100 ice 0 005 t2 0 112 t 0 404

Eq 22 has been applied at the station 10393 (Linden-berg) since January 2000Ucorr w is the corrected RS80 RH with respect to waterUms w denotes the measured RS80 RH (with respect towater) corresponding to Vaisala factory calibration andthe dry ground-check (with 0 RH corresponding to theinstructions for use)Upsy stands for measured RH by the psychrometer in theweather screen before the ascent andUms 0 is the measured RH by the RS80 in the weatherscreen before the ascentt is the temperature in degree Celsius ( C)

Uw t 100 iceesi Tesw T

100 (23)

stands for the computed RH with respect to water at theactual temperature t and ice saturationesi and esw are the saturation partial pressure of watervapour at the temperature T over plane ice or liquid wa-ter respectively The saturation curves are taken fromSONNTAG (1994)T 273 16 K + t is the temperature in Kelvin (K)

The quality of humidity standard aerological sound-ings has been improved essentially in recent years ap-plying these additional corrections to the data producedby Vaisala standard product software

Fig 2 shows the improved quality of the correctedRS 80 A-Humicap soundings derived from synchro-nized soundings RS80 A-Humicap and Lindenberg re-search reference sonde in the period July 1999 untilSeptember 2000 at Lindenberg 10393 upper air stationFor the upper troposphere the dry bias (dotted line) ofabout 8 RH near 300 hPa is nearly completely elim-inated (thick line) Only a small dry bias of maximum3 remains in the boundary layer between 1000 and850 hPa In higher levels the bias is smaller than 05relative humidity and in the investigated region of 600ndash100 hPa the mean bias is about 0 3 RH the meanstandard deviation is about 1 9 RH We hope that theremaining dry bias can also be removed but the researchis going on regarding this subject

22 Data selection

For the present study we use the Lindenberg correctedRS80A data for the period February 1 2000 until April30 2001 There are usually four ascents per day at 0006 12 and 18 UTC respectively From these ascentswe obtain pro les of pressure temperature and relativehumidity with respect to liquid water The reported val-ues are 10 s averages and cover a pressure range from

surface pressure to about 100 hPa Additionally we usetropopause pressure and altitude From the total of 1666ascents we had to dismiss 103 because of missing dataor other problems so we have data from 1563 ascentsfor evaluation

Since we are interested in ice supersaturation in theupper troposphere we evaluate only data for pressurelevels above 600 hPa This seems a rather low altitudefor the upper troposphere yet the lowest pressure alti-tude of the tropopause in the dataset is 4827 hPa Therelative humidity with respect to ice RHi is computedfrom Ucorr w according to

RHi Ucorr wesw Tesi T

(24)

A radiosonde crossing a supercooled water cloud re-ports RH 100 which would imply ice supersatu-ration Yet this is neither an ISSR nor a cirrus hencewe use a temperature criterion to avoid to count suchevents This is we only register an event when its bot-tom temperature is lower than 30 C and when simulta-neously its top temperature is lower than 35 C In therange lower than 30 C cloud particles in liquid phaseare very seldom (see PRUPPACHER and KLETT 1997chap 7)

3 Results and discussion

31 Frequency of ice supersaturation

In the following we often use the term ice-super-saturation layer or something similar to denote thoseparts of the radiosonde pro les where RHi 100

Figure 2 a) Mean differences of relative humidities betweencorrected RS80 A-Humicap [URS80CORR] and research referencemethod[UREFERENCE] derived from about 70 synchronized sound-ings RS80 A and Lindenberg research reference sonde (modi edRS90) in the period July 1999 till September 2000 in Linden-berg b) Mean difference of relative humidities between routine ra-diosonde RS80 A [URS80ROUTINE] and research reference method[UREFERENCE] approximate data


Figure 3 Ice-supersaturationlayers over the meteorological station Lindenberg vs date The vertical extension of each event is marked bya gray bar Often there are several layers found within one radiosonde ascent The tropopause pressure is also given (black line)

Thus the notion ice-supersaturation layer includes bothice-supersaturated regions (clear air) and cirrus clouds

Employing the pressure and the temperature criteriawe nd a total of 730 layers with ice supersaturation (iecirrus and or ISSRs) in 437 out of the 1563 ascentsThus we nd ice supersaturation in 28 of the ascentsand often there are more than one ice-supersaturation

layer in one ascent If we allow for a 1 9 RH errorand a bias of 0 3 RH of the humidity measurementsand regard the research radiosonde (FN-method) errorof 1 RH the total number of layers with ice supersat-uration ranges between 641 and 939 and the number ofascents where RHi exceeds 100 at least once rangesbetween 378 (24) and 540 (35)


The cloud atlas of WARREN et al (1986) shows forthe region near Lindenberg average cirrus amounts of17ndash23 and cirrus frequencies of occurrence of 44ndash49 The radiosonde being a point measurement doesnot always cross a cirrus or an ISSR when cirrus cloudsare around Thus it is clear that the radiosonde datamust yield a smaller frequency of occurrence of ice-supersaturation layers than the surface observations ofthe total sky Instead the radiosonde data can be under-stood as random sample measurements and thus theirfrequencies of occurrence of ice-supersaturation layerscan be expected to yield values that resemble the aver-age cirrus amounts from the total sky observations Thisis indeed found here but the values from the radiosondeare slightly larger than those of the cloud atlas probablybecause of the additional inclusion of cloud free ISSRsor subvisible cirrus in the radiosonde data set

The measurements of ice-supersaturation layers foreach date are given in Fig 3 Every such layer ismarked by a gray bar representing its vertical exten-sion Also given on the gure is the pressure of the ther-mal tropopause (black line) and we see immediatelythat most ice-supersaturation layers remain below thetropopause in the upper troposphere (see below) Themultiple ice-supersaturation layers are often separatedby rather thin subsaturated layers It could be that this re- ects the internal (vertical) structure of cirrus and ISSRswhere perhaps by turbulent entrainment of drier ambientair subsaturated zones are produced in an overall super-saturated layer

In Fig 3 we see extended periods of 1ndash2 weekswhere ice supersaturation did not occur in the atmo-sphere over Lindenberg and there are other periodswhere ice supersaturation was frequent The monthlypercentages for February 2000 until April 2001 areshown in Fig 4 The frequencies of occurrence are de- ned here as the number of ascents per month (or sea-son) that contain one or more ice-supersaturation layerdivided by the total number of ascents during the respec-tive period in question Multiple ice-supersaturation lay-ers are not counted here as multiple events Since thedata cover only a little more than one year this guredoes not represent climatological values rather the val-ues represent a snapshot where in uences of weatheranomalies are apparent For instance near minimumfrequency of ice supersaturation occurs in May 2000which month was much warmer and drier relative toclimatological means and in turn the July 2000 wascolder and wetter (DWD 2000) and ice supersatura-tion was reported much more frequent than in eitherJune or August 2000 The frequency of ice supersatu-ration measurements in the latter months agree betterwith our expectation of relatively low summer valuessince it is known that the relative humidity in the uppertroposphere of the mid-latitudes obtains its lowest meanvalues during summer (KLEY et al 2000 cha 332)Due to the anomalies in 2000 the expected annual cyclecannot be seen in the monthly data However when in-

0

5

10

15

20

25

30

35

40

45

F M A M J J A S O N D J F M A

Fre

quen

cy o

f occ

urre

nce

[]

Months 2000 - 2001Figure 4 Monthly frequencies of occurrence in percentage of as-cents with one or more ice-supersaturation layer over Lindenbergfor the period February 2000 until April 2001

tegrating the seasonal means (spring (MAM) 27 sum-mer (JJA) 24 fall (SON) 34 and winter (DJF) 27)a seasonal cycle similar to that of KLEY et al 2000 ndashthough not signi cant ndash might become visible

32 Altitude distribution ofice-supersaturation layers

Evaluating the data shown in Fig 3 we can obtain analtitude distribution of ice-supersaturation layers Forthis purpose we count the appearance of data withRHi 100 on pressure levels from 150 to 600 hPain steps of 50 hPa where each level actually means thecentre of a 50 hPa thick layer Ice-supersaturation layers

Figure 5 Altitude distribution of ice-supersaturation layers aver-aged over the whole period (February 2000ndashApril 2001 thick line)and over the seasons (thin lines see inserted key) Most ISSRs arefound in a broad layer between 200 and 450 hPa with seasonal shiftsof the order 40 hPa


that extend over more than one layer are counted in eachof these layers The annual mean and seasonal distribu-tions are shown in Fig 5 The gure shows that ice su-persaturation is frequent in a broad upper troposphericlayer between 450 and 200 hPa In autumn the ice-supersaturation layers reach their highest altitudes inwinter the lowest The mean altitudes are about 300 hPain summer and fall and 340 hPa in winter and springThis seasonal shift is statistically signi cant (as a t-testshows) although the distributions are rather broad

Ice supersaturation is found in 44 of the pro lesin the 200 hPa layer and in 126 of the pro les in the250 hPa layer in the annual mean These values vary be-tween 24 and 90 for the 200 hPa layer and between100 and 173 for the 250 hPa layer respectively ifwe allow for measurement errors as mentioned in 31

The annual mean values can be compared with thecorresponding values obtained from the humidity evalu-ations of the MOZAIC project of the years 1995ndash1997(GIERENS et al 2000) These data have been evaluatedon a Gaussian grid consisting of 128 cells in zonal di-rection times 64 cells in the meridional direction givinga total of 8192 cells of about 2 8 2 8 In this gridLindenberg (5222 N 1412 E) falls into cell number1670 ( 13 128 6) For this cell the probability to nd ice supersaturation at 250 25 hPa is 26 Unfor-tunately for the corresponding 200 25 hPa layer thereare no MOZAIC data in cell number 1670 The secondcell next to but north of Lindenberg is number 1542There we have ISSR occurrence frequencies of 21 inthe 250 25 hPa layer but only 5 in the 200 25 hPalayer In the upper considered layer the agreement be-tween the Lindenberg and the MOZAIC data is excel-lent but in the 250 hPa layer the Lindenberg data showonly about half the values the MOZAIC evaluation gaveWe do not currently know the reason for this discrep-ancy

In the same way it is possible to compare the occur-rence frequencies of ice-supersaturation layers over Lin-denberg with the occurrence frequencies of subvisiblecirrus clouds obtained from the SAGE II instrument onboard ERBS (WANG et al 1996) The statistics is pre-sented for height levels since this is the coordinate usedin the SAGE II data base A quick view over Fig 6 con- rms what has been found in an earlier comparison ofMOZAIC and SAGE II data namely that we tend to ndsubvisible cirrus rather in regions with higher probabil-ity of ice supersaturation than in drier regions (GIERENSet al 2000) This is plausible However in the gure wesee also that the subvisible cirrus covers a larger altituderange than the ice-supersaturation layers measured overLindenberg We believe that this is not really true andthat this discrepancy originates from the low horizontalresolution of the SAGE II data namely 24 longitude10 latitude Lindenberg is only one point in the hugeSAGE II pixel hence such discrepancies in the data arenot a surprise However the general similarity betweenthe altitude distributions of subvisible cirrus and the ice-

supersaturation layers is pleasant and is furthermore ro-bust against random measurement errors mentioned in31

33 Situation of ice-supersaturation layersrelative to the tropopause

In Fig 3 we could see that ice saturation occurs in mostcases below the tropopause even in meteorological sit-uations where the tropopause pressure is relatively higheg in February 2001 In this section we want to studythe altitude of the ice-supersaturation layers relative tothe tropopause In case of multiple tropopauses we re-late such events to the lowest tropopause

The situation of ice-supersaturation layers relative tothe local tropopause is shown in Fig 7 The result isunambiguous Most layers where the humidity exceedsice saturation remain beneath the tropopause and onlya minor fraction of them (133) extend beyond thetropopause into the lowermost stratosphere Only 62of the ascents report ice supersaturation in the lowermoststratosphere This is consistent with the statistics of icesupersaturation obtained from MOZAIC data (GIERENSet al 1999) Only 2 of the measurements obtained onMOZAIC ights in the lowermost stratosphere indicatedsupersaturation

Fig 8 (left part) shows the altitude distribution of theice-supersaturation layer tops and bottoms relative to thetropopause For this investigation the layer top and bot-tom pressures minus the corresponding tropopause pres-sure are counted in 5 hPa wide pressure bins and thenumber of events in each bin are shown in the gureThere is a sharp spike in the curve for the top pressuresjust above2 the tropopause but the integral under thisspike is rather small compared to the integral numberof ice-supersaturation events that remain completely be-neath the tropopause The gure shows that most cirrusclouds and ISSRs are situated in a broad layer extending200 hPa down from the tropopause A similar investiga-tion was carried out by GOLDFARB et al (2001) for thealtitudes of cirrus cloud tops relative to the tropopauseat the Observatoire de Haute Provence using Lidar dataTheir gure 4 shows most cloud tops at the tropopause(corresponding to our spike in Fig 8) The clouds preferthe upper troposphere region between tropopause and25 km beneath and only a smaller fraction of the cloudtops reach into the lowermost stratosphere Overall thereis a strong similarity between the cloud top altitude dis-tribution relative to the tropopause given by GOLDFARBet al (2001) and the corresponding altitude distributionof ice-supersaturation layers at Lindenberg In Fig 5 aseasonal variation of the altitude distribution is apparentThis is clari ed in the right panel of Fig 8 where we2 This spike about 5 hPa above the tropopause is probably caused

by the temperature dependent lag error of the RS80 A-HumicapFirst evaluation shows the necessity to shift this spike about10 hPa higher so that the spike would actually be about 5 hPa be-low the tropopauseThe nal lag correction for RS80 A-Humicapis subject of research of the Lindenberg research team


Figure 6 Altitude distribution of ice-supersaturationlayers averaged over the whole period (February 2000ndashApril 2001) and frequency ofoccurrence of subvisible cirrus cloud vs altitude as obtained from SAGE II (Wang et al 1996) The four panels represent a) spring b)summer c) autumn and d) winter

Figure 7 Situation of ice-supersaturation layers (bars) over Lindenberg relativ to the local tropopause (solid line in case of multipletropopauses this is the lowest one) The picture is unambiguous Most events stay below the tropopause and only a minor fraction extendsbeyond the tropopause into the lowermost stratosphere

show the seasonal altitude distributions relative to thetropopause (thin lines) together with the annual mean(thick line) expressed as frequency of occurrence (ienumber of ascents with RHi 100 at the considered

pressure level divided by the total number of ascents dur-ing the considered period) The mean pressure distancesfrom the tropopause are (1 standard deviation in brack-ets) 92 (58) for spring 59 (39) for summer 75


Figure 8 Left panel situation of ice-supersaturation relative to the tropopause expressed as pressure of layer top (open squares) andbottom (open circles) minus tropopause pressure The numbers of events are counted in 5 hPa wide pressure bins Right panel Seasonalvariation of the situation of ice-supersaturationrelative to the tropopause The thick line is the average over the whole period while the thinlines are seasonal averages (see inserted key)

(54) for fall 96 (63) for winter and 84 (58) for thewhole year (all in units of hPa) There is a clear sum-mer vs rest-of-the-year difference both in the relativealtitude of the ice-supersaturation layers as well as inthe depth of the atmospheric layer which contains cirrusand ISSRs The layer where cirrus and ISSRs are foundis considerably shallower in summer than in all otherseasons and the cirrus clouds and ISSRs are on averagenearer to the tropopause in summer than during the restof the year A t-test proofs that the summer distributionis signi cantly different from all other seasonal distribu-tions and this statement remains valid if we allow for theerrors in humidity measurements mentioned in 31

34 Vertical extension of ice-supersaturationlayers

The horizontal scale of cirrus clouds can cover some or-ders of magnitude DOWLING and RADKE (1990) es-timate a typical scale of 20ndash30 km but they say that

Figure 9 Weibull-plot of the cumulative frequency distributionF H of the vertical extension H (in metres) of ice-supersaturationlayers (open squares) and two distinct Weibull distribution ts (solidand broken line)

maximum and minimum extents can be larger or smallerthan this by two orders of magnitude The horizon-tal scale of ISSRs is well known neither GIERENSand SPICHTINGER (2000) estimated a typical horizontalwidth of about 150 km and found that there are some-times rather extended ISSRs (up to more than 3000 km)These sizes have been inferred from the MOZAIC dataand it has been stated that as a consequence of a severeselection bias in the data it cannot be excluded that mostISSRs have a much smaller horizontal extension of theorder of only a few km Such a size would agree withhorizontal extensions of subvisible cirrus clouds mea-sured with Lidars at Table Mountain (California seeBEYERLE et al 2001) The MOZAIC data resultingfrom level ights cannot be used to infer the vertical ex-tension of ISSRs But the corrected RS80A data can beused to explore the distribution of the vertical extents ofice-supersaturation layers

The statistical distribution of vertical extension H of ice-supersaturation layers is plotted in Fig 9 on aso-called Weibull plot In order to produce this plotthe data have been binned into 60 m wide classesfor noise reduction then the cumulative frequency dis-tribution F H 1 exp gH p with shape para-meter p and scale parameter g has been calculatedand log log 1 1 F H has been plotted vs logH This way of data presentation has the advantage thata Weibull distribution appears as a straight line withslope p The scale parameter g can be computed fromthe intercept q via g 10q m with m log log e0 362215 Fig 9 shows that the thicknesses of ice-supersaturation layers indeed seem to follow a Weibulldistribution however with different slopes for shallowand thick layers The transition between these regimesoccurs at a thickness of about 1 km (logH 3) Theslope for the shallower layers is p 0 71 whereas thatfor the thicker ones is p 1 If we allow for the hu-midity measurement errors mentioned before the slopefor the shallow layer thickness distribution is hardly af-


fected at all while for the thicker ones it varies between09 and 11

The horizontal sizes of ISSRs and cirrus clouds arealso Weibull distributed (GIERENS and SPICHTINGER2000 BERTON 2000) The shape parameter is p 0 55for ISSRs (GIERENS and BRINKOP 2001) For visiblecirrus the corresponding shape parameters are p 0 5for the horizontal size distribution and p 1 for thethickness distribution (BERTON 2000) The latter re-sult is an indication that the data in our sample withlogH 3 are probably dominated by cirrus cloudsbecause they follow the same distribution type as inBERTONrsquos sample For the same reason we believe thatthe data for logH 3 in our sample are mostly ISSRssubvisible cirrus and perhaps single thin layers of multi-layer cirrus clouds Since there are more data in thelogH 3 branch of our distribution than in the otherthe thinner objects dominate the sample and thick cir-rus is only a minor contributor to the data set

The mean thickness of the ice-supersaturation layersin the radiosonde data is 560 m with a standard devi-ation of 610 m Allowing for the errors mentioned insection 31 the mean thickness ranges between 490 and610 m the standard deviation ranges between 540 and740 m This is considerably less than the typical thick-ness of (visible) cirrus clouds of 15 km (DOWLINGand RADKE 1990) Also the mean thickness of cirrusclouds in the statistical sample used by BERTON (2000)is larger (about 3 km calculated from his tab 3 usingcoef cients from section 52 and tab A4) In a most re-cent compilation of cirrus properties from 10 years ofobservations (SASSEN 2002 tab 23) there is not an en-try giving cloud thickness of less than 1 km The meanthickness of the ice-supersaturation layers over Linden-berg is more similar to the thickness of subvisible cir-rus clouds that have been measured using Lidars (egSASSEN and CHO 1992 WINKER and TREPTE 1998BEYERLE et al 2001 GOLDFARB et al 2001)

35 Statistics of relative humidity

The relative humidity eld is usually very intricate inspace and time since uctuations of both temperatureand absolute humidity are involved in variations of itHowever it turns out that the relative humidity in thetropopause region obeys simple statistical laws overlonger periods of time and larger spatial scales This hasbeen demonstrated using MOZAIC data (GIERENS etal 1999) and data from the Microwave Limb Sounder(MLS SPICHTINGER et al 2002) The statistical lawsare (1) the probability for the relative humidity with re-spect to ice to reach a certain value in the lowermoststratosphere decreases exponentially with this value and(2) the probability for the supersaturation with respect toice Si RHi 100 to reach a certain value in a tro-pospheric ISSR decreases exponentially with this valueExpressed by formulae this reads

PRH u a exp au (lowermost stratosphere) (31)

PSi s b exp bs (troposph supersaturation) (32)

where a b are constants and u s are variables on theRH and Si axes respectively P means probability den-sity However the humidity statistics outside and insidecirrus clouds are fundamentally different the relativehumidity within cirrus clouds turns out to be better de-scribed by either a Gaussian or a Rayleigh distributioncentred at saturation (OVARLEZ et al 2002) Now thequestion arises whether the 15 months of humidity datafrom the station Lindenberg display a similar statisti-cal behaviour although they do not cover a large re-gion and whether both types of humidity distribution(ie GaussRayleigh for data obtained in cirrus and ex-ponential for data from clear air ISSRs) are present inthe sample For this evaluation we have used all RHi-data above 600 hPa without any additional temperature

Figure 10 Statistical distribution (non-normalised) of relative hu-midity with respect to ice RHi () in the troposphere (upper panel)and in the lowermost stratosphere (lower panel) over LindenbergRadiosonde data (solid steps) and several exponential ts (brokenlines) are presented together


Figure 11 Temperature distribution of ice-supersaturation layersover Lindenberg The temperature classes are 2 K wide an eventbelongs to a temperature class T if at least once the temperature Twas measured inside the layer

criterion However there is no change in the shape ofthe distributions if for example we use only data withtemperatures lower than 30 C

Considering rst the tropospheric data (Fig 10 up-per panel) we nd a statistical distribution of Si resem-bling that in the earlier study with the MOZAIC dataie a at distribution for subsaturated values and an ex-ponential distribution for the supersaturated cases Theaverage supersaturation is 6 5 Additionally thereis a distinct bulge around saturation in the distributionthat is the superposition of a noisy Gaussian upon thehumidity distribution Such a bulge is not present in theMLS data and hardly perceivable in the MOZAIC datasince these data were cloud cleared as far as possibleHence the bulge probably stems from radiosonde mea-surements inside cirrus clouds The humidity has relaxedto its equilibrium value (ie saturation) in these stablecirrus clouds and merely some uctuations around satu-ration remain and lead to the width of the bulge (about80 to 110) The exponential distribution of supersat-uration above RHi 110 is characteristic of ISSRsHowever the slope b of the exponential is much steeperfor the Lindenberg data (b 0 21) than in the formerstudies (MOZAIC b 0 058 MLS northern hemi-sphere b 0 051) It is not possible to ascribe this dif-ference to vertical variations since the slope of the ex-ponential is this steep in the Lindenberg data in 8 atmo-spheric layers each 50 hPa thick and centred around 500up to 150 hPa (not shown)

It is more likely that the steep slope in the Linden-berg distribution is caused by the slow response of ra-diosondes to humidity variations especially in cold airIn fact the response time of Vaisala Humicap sensorsincreases exponentially with decreasing temperature at

40 C the time constant of the RS80A sensor is 27 s(MILOSHEVICH et al 2001) whereas the data is trans-mitted every 10 s The slow response means that sharpmaxima in the humidity pro le cannot be resolved The

RHi-distributions in the 8 atmospheric layers mentionedabove tend to show higher supersaturations in the lowerlayers and vice versa ndash consistent with the shorter re-sponse time in these relatively warm layers Althoughthe MOZAIC sensors suffer from a time lag too thiserror source is not as effective as for the radiosonde be-cause a MOZAIC aircraft generally stays in the samealtitude for a period much longer than the time constantof the sensors and the RHi- eld has certainly larger au-tocorrelation scale lengths in the horizontal than in thevertical

Looking at the stratospheric data (Fig 10 lowerpanel) the distribution is different from the earlier dis-tributions of RHi of both MLS and MOZAIC Insteadof getting a constant slope from low humidities (say20) up to highly supersaturated values we rather geta distribution that is similar to the aforementioned tro-pospheric distribution albeit without the bulge at sat-uration in particular there is a kink in the distributionat ice saturation This kink at saturation signi es phys-ical processes acting on the water vapour reservoir indifferent ways on the sub- and supersaturated sides re-spectively and it is most plausible to assume that it isthe presence of clouds in the supersaturated regime thatis responsible for the kink A straight distribution with-out a kink at saturation thus indicates cloud free con-ditions and that was probably the case in most of theMOZAIC and MLS data assigned to the stratospherethat were evaluated earlier On the contrary we mustassume that clouds above the thermal troposphere in- uenced the observed distribution of RHi over Linden-berg Note that the tropopause de nition used for thepresent evaluation is much sharper than those used foreither the MOZAIC data or the MLS data since we havethe exact tropopause pressure available and data imme-diately above the thermal tropopause can be assignedto the stratosphere Cirrus and subvisible cirrus abovethe mid-latitude tropopause have indeed been observedby satellite instruments (eg SAGE II WANG et al1996) or by the Cryogenic Infrared Spectrometers andTelescopes for the Atmosphere (CRISTA SPANG et al2002) Clouds immediately above the tropopause couldtherefore affect the RHi distribution as observed hereThe missing of the bulge around 100 could mean thatthese ice clouds in the lowermost stratosphere do rarelyrelax to equilibrium and that the ice crystals would fallout of the supersaturated layer before the moisture avail-able for condensation is completely used up for crystalgrowth

36 Temperature distribution ofice-supersaturation layers

Evaluating the data shown in Fig 3 we can obtaina temperature distribution for the ice-supersaturationlayers For this purpose we de ne 26 classes of tem-perature between 30 C and 80 C each class 2 Kwide (kT Tmeas T 1K Tmeas T 1K ) An ice-supersaturation layer belongs to a temperature class if at


least once the temperature T has been measured insidethe layer The thicker ones of the ice-supersaturation lay-ers in our data set will normally contribute to more thanone temperature class We obtain a distribution of tem-perature as shown in Fig 11 The major part of the mea-sured temperatures is lower than 40 C (in fact 71of them) the mean value is 46 8 C and the standarddeviation is 84 K This general result is only slightly af-fected by the random errors of humidity measurementsThe fraction of events in which the measured tempera-ture is lower than 40 C ranges between 70 and 75The shape of the distribution is virtually unaltered onlythe mean value and the standard deviation slightly varybetween 46 4 8 4 C and 48 0 8 8 C

4 Summary and conclusions

In the present work we have evaluated radiosonde dataof the meteorological observatory Lindenberg in or-der to determine properties of ice-supersaturation layersand statistical properties of the relative humidity in thetropopause region The term ice-supersaturation layer ishere used to designate a measurement of the relative hu-midity with respect to ice in excess of 100 irrespec-tive of the presence of ice crystals (since cloud infor-mation is not contained in radiosonde pro les) Thus anice-supersaturation layer can either be a cirrus cloud oran ice-supersaturated region (ISSR) that is a supersat-urated but cloud free airmass The data taken with aRS80A radiosonde and corrected by means of a researchsonde cover 15 months from February 2000 to April2001 The following results therefore must not be inter-preted in a climatological sense The main results are1 We nd a total number of 730 ice-supersaturation lay-ers in 437 of 1563 ascents hence the mean frequency ofoccurrence of ice supersaturation over the station Lin-denberg for the tested period is about 28 with seasonal uctuations between 24 (summer) and 34 (fall)These values are consistent but slightly larger than cor-responding cirrus amounts from the cloud atlas of WAR-REN et al (1986)2 Ice supersaturation occurs mostly in a broad uppertropospheric layer between 450 and 200 hPa There is aslight but signi cant seasonal shift of the altitude distri-bution of ice-supersaturation layers the mean altitude isat 300 hPa in summer and fall and descends to 340 hPain winter and spring

The frequencies of occurrence of ice supersaturationfrom the radiosonde data are at 200 hPa in good agree-ment with the corresponding ISSRs frequencies of oc-currence obtained from the MOZAIC data closest toLindenberg At 250 hPa there is a large so far unex-plained discrepancy

The altitude distribution of ice-supersaturation layersover Lindenberg is generally similar to those of subvisi-ble cirrus (from SAGE II data)

3 Ice supersaturation (ISSRs and cirrus clouds) occursmainly below the tropopause only a few events ex-tend into the lowermost stratosphere consistent withthe earlier examination of MOZAIC data Most of theISSRs and cirrus clouds are situated in a broad layer be-tween the tropopause and 200 hPa below it Their an-nual mean pressure distance for the thermal tropopauseis 84 58 hPa the seasonal mean values are 92 58

59 39 75 54 and 96 63 hPa for springsummer fall and winter respectively Similar distribu-tions have been found for cirrus and subvisible cirruswith lidar measurements at the Observatoire de HauteProvence

4 The vertical extensions of ice-supersaturation layersfollows a pair of Weibull distributions with exponentp 0 71 for shallow layers with thickness of less than1 km and p 1 for the thicker ones The mean thick-ness of ice-supersaturation layers is 560 m This value ismore similar to the mean thickness of subvisible cirrusthan to that of visible cirrus However the distribution israther broad with sH 610 m The ice-supersaturationlayers of less than 1 km thickness are probably com-posed of ISSRs subvisible cirrus and perhaps thin strataof multi-layer cirrus clouds whereas the thicker objectsare probably mostly cirrus clouds

5 As in the earlier studies with the MOZAIC and MLSdata we determine the distribution of the relative humid-ity with respect to ice in the upper troposphere In thesubsaturated and the supersaturated range respectivelythe probability of measuring a certain relative humiditydecreases exponentially with the relative humidity butwith different slopes While the slope in the subsaturatedrange is similar to the slope of the distributions foundfrom the MOZAIC data in the supersaturated range theslope is much steeper than the slopes of the distributionsfound from the MOZAIC and MLS data

This is likely the result of the slow response of theRS80A sensor when ascending through cold air Addi-tionally there is a bulge in the distribution at about icesaturation that most probably stems from data obtainedin mature cirrus clouds (where the humidity has relaxedto equilibrium and the ice crystals ceased to grow)

For the stratospheric data we nd a distribution forthe relative humidity with respect to ice that can be ttedby two different exponential distributions for the sub-saturated and supersaturated range respectively Thusthe distribution is quite different from the earlier distri-butions found from MOZAIC and MLS data but moresimilar to the tropospheric distributions (yet without thebulge at saturation) We conjecture that the kink at satu-ration in the distribution is caused by growing ice crys-tals in cirrus immediately above the tropopause Themissing bulge at saturation could indicate that theseclouds rarely reach an equilibrium state (ie the crys-tals fall out of the supersaturated layer before they com-pletely consume the supersaturation for their growth)6 The measured temperatures inside the ice-super-


saturation layers are observed to range from 30 and80 C more than 70 of them have temperatures

lower than 40 C the mean temperature of the ice-supersaturation layers is 46 8 8 4 C7 Random errors and biases in the humidity measure-ments do not change the results qualitatively they onlyslightly affect the calculated mean values and standarddeviations

In this study we have focused on more general as-pects of ice-supersaturation layers but for further studiesa more detailed examination of the several ascents couldgive additional information about the internal structureof these layers Satellite data could help to make the dis-tinction between cirrus clouds and ISSRs

Acknowledgements

We thank two anonymous reviewers making useful sug-gestions that led to a substantial improvement of ourmanuscript

References

ANTIKAINEN V H JAUHIAINEN 1995 Vaisalarsquos newRS90 family of radiosondes ndash Vaisala News 136 9ndash12

ANTIKAINEN V A PAUKKUNEN 1994 Studies on im-proving humidity measurements in radiosondes ndash WMOTech Conf on Instruments and Methods of ObservationGeneva Switzerland WMO Rep 57 137ndash141

BERTON RPH 2000 Statistical distributionsof water con-tent and sizes for clouds above Europe ndash Ann Geophysicae18 385ndash397

BEYERLE G MR GROSS DA HANER NT KJOMEIS MCDERMID TJ MCGEE JM ROSEN H-JSCHAFER O SCHREMS 2001 A lidar and backscattersonde measurement campaign at Table Mountain duringfebruaryndashmarch 1997 Observations of cirrus clouds ndash JAtmos Sci 58 1275ndash1287

DOWLING DR LF RADKE 1990 A summary of physi-cal propertiesof cirrus cloudsndash J Appl Meteorol29 970ndash978

DWD 2000 Witterungsreport Mai 2000 Juli 2000ELLIOTT WP DJ GAFFEN 1991 On the utility of ra-

diosonde humidity archives for climate studies ndash BullAmer Meteor Soc 72 1507ndash1520

GIERENS K S BRINKOP 2001 A model for the horizon-tal exchange between ice-supersaturated regions and theirsurrounding area ndash Theor Appl Climatol 71 129ndash140

GIERENS K P SPICHTINGER 2000 On the size distribu-tion of ice-supersaturated regions in the upper troposphereand the lower stratosphere ndash Ann Geophysicae 18 499ndash504

GIERENS K U SCHUMANN M HELTEN H SMIT AMARENCO 1999 A distribution law for relative humid-ity in the upper troposphere and lower stratosphere derivedfrom three years of MOZAIC measurements ndash Ann Geo-physicae 17 1218ndash1226

GIERENS K U SCHUMANN M HELTEN H SMIT P-H WANG 2000 Ice-supersaturated regions and subvisiblecirrus in the northern midlatitude upper troposphere ndash JGeophys Res 105 22743ndash22753

GOLDFARB L P KECKHUT M-L CHANIN AHAUCHECORNE 2001 Cirrus climatological results fromLidar Measurements at OHP (44 N 6 E) ndash Geophys ResLett 28 1687ndash1690

HELTEN M HGJ SMIT W STRATER D KLEY PNEDELEC M ZOGER R BUSEN 1998 Calibration andperformance of automatic compact instrumentation for themeasurement of relative humidity from passenger aircraftndash J Geophys Res 103 25643ndash25652

HELTEN M HGJ SMIT D KLEY J OVARLEZ HSCHLAGER R BAUMANN U SCHUMANN P NED-ELEC A MARENCO 1999 In- ight intercomparison ofMOZAIC and POLINAT water vapor measurements ndash JGeophys Res 104 26087ndash26096

HEYMSFIELD AJ LM MILOSHEVICH 1993 Homoge-neous ice nucleation and supercooled liquid water in oro-graphic wave clouds ndash J Atmos Sci 50 2335ndash2353

HEYMSFIELD AJ RM SABIN 1989 Cirrus crystal nu-cleation by homogeneousfreezing of solution droplets ndash JAtmos Sci46 2252ndash2264

JENSEN EJ OB TOON A TABAZADEH GWSACHSE BE ANDERSON KR CHAN CW TWOHYB GANDRUD SM AULENBACH A HEYMSFIELD JHALLETT B GARY 1998 Ice nucleationprocesses in up-per tropospheric wavendashclouds during SUCCESS ndash Geo-phys Res Lett 25 1363ndash1366

KLEY D JM RUSSEL III C PHILLIPS (Eds) 2000SPARC assessment of upper tropospheric and stratosphericwater vapour WCRP-113 WMOTD-No 1043 SPARCReport No 2 312 pp (Available from SPARC Of ce BP3 F-91371 Verrieres le Buisson Cedex France)

KOOP T B LUO A TSIAS T PETER 2000 Water ac-tivity as the determinant for homogeneousice nucleation inaqueous solutions ndash Nature 406 611ndash614

LEITERER U H DIER T NAEBERT 1997 Improvementsin radiosondehumiditypro les using RS80RS90 radioson-des of Vaisala ndash Contr Atmos Phys 70 319ndash336

LEITERER U H DIER D NAGEL T NAEBERT D ALT-HAUSEN K FRANKE 2002 Correction method for RS80A-Humicap humidity pro les ndash Internal Report Meteoro-logical Observatory Lindenberg of Deutscher Wetterdienst1ndash19

MARENCO A V THOURET P NEDELEC H SMIT MHELTEN D KLEY F KARCHER P SIMON K LAW JPYLE G POSCHMANN R VON WREDE C HUME TCOOK 1998 Measurement of ozone and water vapor byAirbus in-service aircraft The MOZAIC airborne programan overview ndash J Geophys Res 103 25631ndash25642

MILOSHEVICH LM H VOMEL A PAUKKUNEN AJHEYMSFIELD SJ OLTMANS 2001 Characterizationandcorrection of relative humidity measurements from VaisalaRS80-A radiosondes at cold temperatures ndash J AtmosOceanic Technol 18 135ndash156

NAGEL D U LEITERER H DIER A KATS J REI-CHARD A BEHRENDT 2001 High accuracy humiditymeasurements using the standardized frequency methodwith a research upper-air sounding system ndash Meteorol Z10 395ndash405

NASH J FJ SCHMIDLIN 1987 International RadiosondeIntercomparison Final report WMOTD-195 102 pp

OVARLEZ J P VAN VELTHOVEN G SACHSE S VAYH SCHLAGER H OVARLEZ 2000 Comparison of watervapor measurements from POLINAT 2 with ECMWF anal-yses in high humidity conditions ndash J Geophys Res 1053737ndash3744


OVARLEZ J J-F GAYET K GIERENS J STROM HOVARLEZ F AURIOL R BUSEN U SCHUMANN 2002Water vapor measurements inside cirrus clouds in northernand southern hemispheres during INCA ndash Geophys ResLett 29 1813

PAUKKUNEN A 1995 Sensor heating to enhance reliabilityof radiosondehumidity measurement ndash Preprints 9th Sym-posium on Met Observ and Instr Charlotte NC Ameri-can MeteorologicalSociety Boston MA 65ndash70

PRUPPACHER HR JD KLETT 1997 Microphysics ofClouds and Precipitation ndash Kluwer Dordrecht Netherland954 pp

SASSEN K 2002 Cirrus clouds ndash A modern perspective InDK LYNCH K SASSEN DOC STARR G STEPHENS(Eds) Cirrus Oxford University press Oxford UK pp11ndash40

SASSEN K BS CHO 1992 Subvisual-thin cirrus lidardataset for satellite veri cation and climatological researchndash J Appl Meteorol 31 1275ndash1285

SASSEN K GC DODD 1988 Homogeneous nucleationrate for highly supercooled cirrus cloud droplets ndash J At-mos Sci 45 1357ndash1369

SONNTAG 1994 Advancements in the eld of hygrometryndash Meteorol Z 3 51ndash66

SPANG R G EIDMANN M RIESE D OFFERMANN PPREUSSE L PFISTER P-H WANG 2002 CRISTA ob-servations of cirrus clouds around the tropopause ndash J Geo-phys Res 107 8174

SPICHTINGER P K GIERENS W READ 2002 The sta-tistical distribution law of relative humidity in the globaltropopause region ndash Meteorol Z 11 No 2 83ndash88

VAY SA BE ANDERSON EJ JENSEN GW SACHSEJ OVALEZ GL GREGORY SR NOLF JR PODOL-SKE TA SLATE CE SORENSON 2000 Troposphericwater vapor measurements over the North Atlantic duringthe Subsonic Assessment Ozone and Nitrogen Oxide Ex-periment (SONEX) ndash J Geophys Res 105 3745ndash3756

WANG P-H P MINNIS MP MCCORMICK GS KENTKM SKEENS 1996 A 6-year climatologyof cloud occur-rence frequency from StratosphericAerosol and Gas Exper-iment II observations (1985ndash1990) ndash J Geophys Res 10129407ndash29429

WARREN SG CJ HAHN J LONDON RM CHERVINRL JENNE 1986 Global distribution of total cloudcover and cloud type amounts over land NCARTN-273 +STR available from NTIS US Department of CommerceSpring eld Va 22161

WINKER DM CR TREPTE 1998 Laminar cirrus ob-served near the tropical tropopause by LITE ndash GeophysRes Lett 25 3351ndash3354


sor (HELTEN et al 1998 1999) The latter is installedon ve commercial aircraft within the Measurement ofozone by Airbus in-service aircraft (MOZAIC) project(MARENCO et al 1998) The MOZAIC data wereused by GIERENS et al (1999) to determine the fre-quency of occurrence of ISSRs over the northern mid-latitudes and to determine the statistics of relative hu-midity within ISSRs Mean temperature differences of3ndash4 K between ISSRs and their subsaturated environ-ment (ISSRs are colder) could be determined in the samework as well as water vapour mixing ratios in ISSRsthat exceed their counterparts in the neighbouring sub-saturated regions by typically a factor 15 GIERENSet al (2000) related the MOZAIC data to statistics ofoccurrence of subvisible cirrus clouds (WANG et al1996) obtained from data of the Stratospheric Aerosoland Gas Experiment II (SAGE II) on board the EarthRadiation Budget Satellite (ERBS) A close relationbetween ISSRs and subvisible cirrus is probable ac-cording to that work Typical horizontal extensions ofISSRs could be determined from the MOZAIC databy GIERENS and SPICHTINGER (2000) Recently datafrom the Microwave Limb Sounder (MLS) on boardthe Upper Atmospheric Research Satellite (UARS) wereevaluated by SPICHTINGER et al (2002) in order to ob-tain a global version of the humidity statistics inside andoutside ISSRs

Radiosondes are not included in the above list ofinstruments used for measuring humidity in ISSRsThe reason for this is that hygrometers that are opera-tional in current radiosonde types (usually RS80 withA-Humicap by Vaisala) cease to furnish reliable resultsin cold air (T 40 C) or at low absolute humiditiesconditions that are typical for the upper troposphere andbeyond (eg NASH and SCHMIDLIN 1987 ELLIOTTand GAFFEN 1991) However the situation is begin-ning to improve with the advent of the new H-Humicapsensor of Vaisala and the RS90 sonde (ANTIKAINENand JAUHIAINEN 1995) The Lindenberg Observatoryof Deutscher Wetterdienst has developed the method ofldquostandardized frequenciesrdquo (LEITERER et al 1997) so-called FN-method using a modi ed RS90 as a researchradiosonde This method allows to obtain very accuraterelative humidities using the research sonde throughoutthe troposphere and into the lowermost stratosphere Theresearch sonde has been used to correct the routine RS80A radiosondes (LEITERER et al 20021 NAGEL et al2001) Hence this data can be used to examine the up-per troposphere and lowermost stratosphere concerningrelative humidity with respect to ice

As radiosonde reports do not contain any informa-tion about the concentration of ice crystals along theirpath it is not possible to distinguish between cloudyand cloud-free parts of a pro le However evaluatingthe Lindenberg data we nd certain signatures that arecharacteristic of ISSRs and others that are character-

1 aviable from wwwdwddedeFundEObservatorMOLmol3mol3 RS80 korpdf

istic of cirrus clouds (see below in particular 34 and35) Further radiosondes of good quality can be usedto study the vertical extension of ice supersaturated air-masses (ISSR and cirrus) and their situation relative tothe local tropopause Such investigations are the topic ofthe present paper

In the next section the FN-method for the researchsonde and the method for correction of the RS80 A-Humicap pro les (the Lindenberg routine radiosonde 4times daily) will be described Section 3 then presentsand discusses the results These are summarised andconclusions are drawn in Section 4

2 The Lindenberg corrected RS80 A data

21 Fundamental principles of calibrationand corrected RS80 A humidity pro les

Weekly comparison ascents of the routine RS80 A-Humicap radiosonde with the Lindenberg research ra-diosonde (a modi ed type of the RS90 radiosonde) arethe basis of the calibration The research radiosondeuses the Lindenberg measuring and evaluation methodof ldquostandardized frequenciesrdquo (FN-method) (LEITERERet al 1997) as described in Section 211 The derivedcorrection procedure for RS80 A-Humicap data is de-scribed in Section 212

211 The FN-method

A new measuring and evaluation method has been de-veloped at the Meteorological Observatory Lindenbergwhich is suited for using the modi ed RS90 sensorunit as reference humidity sensor during the radiosondeascent The general measuring technique is based onthe capacity dependence of a thin polymer layer in re-lation to the quantity of water molecules which areoozed into the polymer structure The capacity will betransformed into frequencies by a resonant circuit ofthe radiosonde transmitter The technical improvementsof the RS90 sensor unit with reference to the RS80Asensor unit are described in detail by Vaisala authors(ANTIKAINEN and PAUKKUNEN 1994 PAUKKUNEN1995 ANTIKAINEN and JAUHIAINEN 1995) Bothpolymer sensors measure the partial pressure relation ew

esw

(ew partial pressure of water vapour esw saturation par-tial pressure of water vapour over plane liquid water sur-face) or the relative humidity with respect to water evenunder water or ice saturated and supersaturated condi-tions (in ISSR and cirrus) as reported in NAGEL et al2001

The FN-method of ldquostandardized frequenciesrdquo de-scribed in detail in a series of publications (LEITERERet al 1997 LEITERER et al 2002 NAGEL et al2001) enables us to carry out reference humidity sound-ings in comparison to other humidity soundings Theabsolute accuracy of the new measuring and evalua-tion technique is about 1 RH in the temperature




FN A FN B FN C FN00 00008 00384 0486701 00013 00809 685902 00014 01426 1611603 0002 01605 2550904 00023 01721 3636205 00029 01407 4671906 00027 01176 5818607 00025 00986 6979708 00023 00974 8139309 00025 01109 9115910 00027 00982 10055




FNiFHi FMi

DF(21)









Upsy

0 005 t2 0 112 t 0 404 (22)


Uw t 100 ice 0 005 t2 0 112 t 0 404



100 (23)




22 Data selection





(24)















0

5

10

15

20

25

30

35

40

45


Fre

quen

cy o

f occ

urre

nce

[]


































































Acknowledgements


References












































FN A FN B FN C FN00 00008 00384 0486701 00013 00809 685902 00014 01426 1611603 0002 01605 2550904 00023 01721 3636205 00029 01407 4671906 00027 01176 5818607 00025 00986 6979708 00023 00974 8139309 00025 01109 9115910 00027 00982 10055




FNiFHi FMi

DF(21)









Upsy

0 005 t2 0 112 t 0 404 (22)


Uw t 100 ice 0 005 t2 0 112 t 0 404



100 (23)




22 Data selection





(24)















0

5

10

15

20

25

30

35

40

45


Fre

quen

cy o

f occ

urre

nce

[]


































































Acknowledgements


References











































Upsy

0 005 t2 0 112 t 0 404 (22)


Uw t 100 ice 0 005 t2 0 112 t 0 404



100 (23)




22 Data selection





(24)















0

5

10

15

20

25

30

35

40

45


Fre

quen

cy o

f occ

urre

nce

[]


































































Acknowledgements


References


















































0

5

10

15

20

25

30

35

40

45


Fre

quen

cy o

f occ

urre

nce

[]


































































Acknowledgements


References





































































































Acknowledgements


References









































ice supersaturation in the tropopause region over

Documents