performance and application of a multi-wavelength, ground ...performance and application of a...

Meteorologische Zeitschrift Vol 18 No 3 253-265 (June 2009) Open Access Articleccopy by Gebruder Borntraeger 2009

Performance and application of a multi-wavelengthground-based microwave radiometer in intense convectiveweather

PAK WAI CHANlowast

Hong Kong Observatory Hong Kong China

(Manuscript received September 3 2008 in revised form March 20 2009 accepted March 20 2009)

Abstract

Thermodynamic and dynamic quantities such as the K-index and GUSTEX (a wind gust estimate) arecommonly used in the nowcasting of intense convective weather In the past they were derived fromconventional upper-air ascent measurements which were normally only available twice a day In the tropicsthe thermodynamic property of the troposphere could change rapidly in rain episodes and the conventionalupper-air ascent data alone are not sufficient for nowcasting purposes This paper discusses the use of amulti-wavelength ground-based microwave radiometer to provide frequently updated (once every 5 minutes)thermodynamic profiles of the troposphere up to 10 km for the nowcasting of severe convective weather duringa field experiment in Hong Kong in 2004 The accuracy of the radiometerrsquos measurements is first establishedby comparing with the temperature and humidity profiles of upper-air ascents and the integrated water vapourof Global Positioning System (GPS) receivers The humidity profile and K-index from the radiometer in anumber of rainstorm cases are then studied They are found to give useful indications of the accumulation ofwater vapour and the increasing degree of instability of the troposphere before the occurrence of the heavyrain The continuous availability of the thermodynamic profiles from the radiometer also makes it possibleto study the correlation between K-index and the degree of instability of the troposphere In this study thetropospheric instability is expressed in terms of the total number and the rate of lightning strikes within 20km or so from the radiometer It is found to have good correlation with the time-averaged K-index fromthe radiometer during the heavy rain episodes in the field experiment period Furthermore the feasibility ofcombining the thermodynamic profiles from the radiometer and wind profiles given by radar wind profilersin the continuous estimation of wind gusts is studied The estimated and the actual gusts are reasonably wellcorrelated The wind gusts so estimated are found to provide better indications of the strength of squalls witha lead time of about one hour compared with those based on radiosonde measurements Overall speakingthe microwave radiometer is able to provide very useful data in the nowcasting of intense convective weather

Zusammenfassung

Thermodynamische und dynamische Groszligen wie der K-Index und der Boenschatzwert GUSTEX wer-den vielfach bei der Unwettervorhersage benutzt Fruher wurden sie aus ublicherweise zweimal taglichverfugbaren Radiosondenaufstiegsdaten berechnet In den Tropen konnen sich diese Groszligen in Regenepiso-den aber rasch andern so dass zwei Aufstiege pro Tag fur eine Kurzestfristvorhersage nicht ausreichend sindDie vorliegende Arbeit diskutiert die Ableitung thermodynamischer Profile in der Troposphare bis 10 kmHohe mit einem bodengestutzten Multiwellenlangen-Mikrowellen-Radiometer wahrend einer Messkam-pagne in Hong Kong im Jahre 2004 Die Genauigkeit der Messungen wird durch Vergleich mit Temperatur-und Feuchteprofilen aus Radiosondenaufstiegen und mit dem von GPS-Empfangern ermittelten Wasser-dampfsaulengehalt gepruft Anschlieszligend werden Feuchteprofile und Werte des K-Index bei Gewittern ausRadiometer-Daten untersucht Es zeigt sich dass sie brauchbare Hinweise auf die Ansammlung von Wasser-dampf und auf eine wachsende instabile Schichtung der Troposphare vor schweren Gewittern geben Diekontinuierliche Verfugbarkeit von thermodynamischen Profilen aus Radiometer-Daten erlaubt es auch dieKorrelation zwischen dem K-Index und der Instabilitat der Troposphare zu untersuchen Die tropospharischeInstabilitat wird hier uber die Gesamtzahl und die Haufigkeit von Blitzen in ca 20 km Umkreis um dasRadiometer ausgedruckt Es zeigt sich dass dieses Maszlig wahrend der Messkampagne gut mit dem zeitlichgemittelten K-Index aus den Radiometer-Daten korrelierte Weiter wird die Machbarkeit der Verbindungthermodynamischer Profile vom Radiometer mit Windprofilen aus Windprofiler-Daten fur die kontinuier-liche Abschatzung von Boen untersucht Geschatzte und beobachtete Boen erweisen sich als zufriedenstellenkorreliert Die so abgeschatzten Boen geben eine Stunde im Voraus bessere Hinweise auf die Starke vonBoenlinien als solche aus Radiosondendaten Insgesamt kann festgestellt werden dass ein Mikrowellen-Radiometer sehr brauchbare Daten fur die Kurzestfristvorhersage bereitstellt

1 Introduction

In operational weather forecasting it is a common prac-tice to examine the temperature and humidity profiles

lowastCorresponding author Pak Wai Chan Hong Kong Observatory 134A

Nathan Road Kowloon Hong Kong China e-mail pwchanhkogovhk

measured by radiosondes and the derived instability in-dices in nowcasting heavy rain and thunderstorms Ref-erence is also made to the wind gust estimate basedon the temperature humidity and wind measurementsfrom the radiosondes (MCCANN 1994 GEERTS 2001

0941-294820090375 $ 585

DOI 1011270941-294820090375 ccopy Gebruder Borntraeger Berlin Stuttgart 2009

254 PW Chan Intense convective weather Meteorol Z 18 2009

and LEE 2007) However radiosonde measurementsare normally only available twice a day and duringrain episodes thermodynamic properties of the tropo-sphere could change rapidly in between measurementsespecially in subtropical coastal areas like Hong KongMore frequently available thermodynamic informationtogether with wind data up to the upper troposphere (egfrom radar wind profilers CHAN and YEUNG 2003)would greatly benefit severe weather nowcasting

Ground-based microwave radiometers (ROSE andCZEKALA 2003 KNUPP et al submitted) retrieve thetemperature and humidity profiles up to 10 km by mea-suring the radiation intensity at a number of frequencychannels in the microwave spectrum that are dominatedby atmospheric water vapor cloud liquid water andmolecular oxygen emissions These profiles are avail-able nearly continuously at intervals of several min-utes They are mainly used in non-precipitating condi-tions because the radiometer measurements become lessaccurate in the presence of a water film on the outerhousing (radome) of the equipment Moreover the ra-diometer retrieval normally does not include the scat-tering and emissionabsorption effects of rain Recentlythere are some progress in applying rain-effect mitiga-tion methods to the radiometer such as a hydrophobicradome and forced airflow over the radome surface As aresult reasonably accurate thermodynamic profiles canbe obtained at a rainfall rate up to 10 mmhour (egRADIOMETER PHYSICS 2008) However this rainfallrate is still much lower than that observed in typicalheavy rain episodes in the subtropical areas which nor-mally far exceeds 30 mmhour

This paper presents the performance and applica-tion of a multi-wavelength ground-based microwave ra-diometer in intense convective weather during a field ex-periment conducted in Hong Kong in 2004 The equip-ment setup of the experiment is described in Section2 The accuracy of the temperature and humidity pro-files as well as precipitable water vapour (PWV) fromthe radiometer is first assured by comparing with mea-surements from the radiosonde a wind profilerRadioAcoustic Sounding System (RASS) and Global Posi-tioning System (GPS) receivers in Section 3 The per-formance of the radiometer data in a rain episode is de-scribed in Section 4 which suggests that there is rea-sonable degree of correlation between the instability in-dex of the troposphere as calculated from the radiome-ter data and the occurrence of lightning flashes Thisrelationship is studied further in Section 4 Section 5discusses the application of the radiometer data in thenowcasting of squall associated with intense convectiveweather Conclusion of this paper is given in Section 6

2 Equipment setup

The radiometer in this study (similar to that in KNUPP

et al submitted) measures the radiation intensity at

Figure 1 The radiometer has been placed at Hong Kong Observa-

tory (HKO) Headquarters Cheung Chau and Sha Lo Wan for the

periods in parenthesis during the experiment The map shows the

network of surface anemometers (red dots) and radar wind profilers

(green triangles) in Hong Kong The radiosonde station (orange) and

the two GPS receivers mentioned in this paper (blue) are also shown

Height contours are in 100 m

7 oxygen channels (51ndash59 GHz) and 5 water-vapourchannels (22ndash30 GHz) to obtain the temperature andhumidity profiles It is equipped with a hydrophobicradome surface and a blower of heated air over theradome to minimize the effect of rain on radiometermeasurement During the field experiment the radiome-ter was configured to make both zenith scans and eleva-tion scans at fixed angles Following other similar stud-ies (eg HEWISON et al 2004 and KNUPP et al sub-mitted) only the zenith scan data are used in this pa-per to retrieve the thermodynamic profiles of the tropo-sphere up to 10 km which are normally available every5 minutes

Retrieval of temperature and humidity profiles fromthe radiation intensity measurements of the radiometer isaccomplished by neural network methods based on his-torical radiosonde data using a radiative transfer modelto simulate the observations of a radiometer (KNUPP etal (submitted) In this study the neural network wastrained by a high-resolution radiosonde dataset collectedat Kingrsquos Park (location in Figure 1) with a data avail-ability rate of 05 Hz during the ascent of the balloon(which translates to a spatial resolution of around 10 mfor the typical ascent speed) Radiosonde data of a wholeyear (2003) were used in the neural network

The field experiment was conducted in Hong Kongfrom February to June 2004 The radiometer was placedat three locations and its data are compared with thosefrom the radiosonde a wind profilerRASS and GPS re-ceivers (Figure 1) In studying the thermodynamic in-dex given by the radiometer we also refer to the datafrom a lightning location network (MURPHY and CUM-MINS 2000) which detects cloud-to-ground flashes outto about 100 km from Hong Kong

Meteorol Z 18 2009 PW Chan Intense convective weather 255

(a) (b)

Figure 2 The biases and standard deviations of the radiometer measurements with respect to the concurrent radiosonde data The sample

size at each height is about 400

Figure 3 The relative humidity profiles from the radiometer and

the radiosonde at 00 UTC 31 May 2004 showing the wet bias of

the radiometer in the middle troposphere and dry bias in the upper

troposphere (red arrows)

In estimating the wind gust associated with thunder-storms reference is also made to the vertical wind pro-files provided by radar wind profilers (locations in Fig-ure 1) These profilers are of the boundary-layer typeand operate at a frequency of 1299 MHz They measurein two modes viz the low and the high modes The lowmode starts at 116 m above ground with a height incre-ment of 60 m up to about 1500 m the high mode startsat 260 m above ground with a height increment of 200m up to about 9000 m Wind profiles are available every10 minutes

3 Comparison of data from radiometerand other instruments

The quality of the radiometer data is established bycomparison with the measurements from other meteoro-logical instruments in Hong Kong namely radiosondewind profilerRASS and GPS receivers

31 Comparison with radiosondes

Temperature and humidity profiles from the radiometerare compared with those from the radiosonde launchedat Kingrsquos Park Their biases and standard deviations withrespect to the radiosonde measurements are shown inFigure 2 The sample size is about 400 The compari-son results are considered to be statistically significant(a sample size greater than 50) though the compari-son period is quite limited seasonally (just spring andearly summer) The followings are observed about theradiometer data

(a) For temperature there is a warm bias up to about14 km and a cold bias further aloft The standard de-viation increases gradually with height but is gener-ally smaller than about 2 degrees Celsius The rela-tively large values of bias and standard deviation of thetemperature at the lowest levels may be related to theseparation between the radiometer and the radiosondewhich ranges from about 1 km (when the radiometer wasplaced at HKO Headquarters) to more than 20 km (whenthe radiometer was placed at Sha Lo Wan)

(b) For relative humidity there is a wet bias betweenabout 12 km and 58 km and a dry bias higher upThe standard deviation again increases with height butis mostly below 20 Please note that comparison ofrelative humidity data between the radiosonde and theradiometer is more challenging than for instance com-parison of absolute humidity because the errors of bothtemperature and water vapour measurements from theradiometer are taken into consideration Relative humid-ity data are compared here because they are more di-rectly applicable to weather nowcasting in comparisonto absolute humidity

The above results are generally consistent with thoseobtained in other studies of the same kind of ground-based radiometer (eg HEWISON et al (2004) whichstudied radiometer data up to 4 km)

Since the relative humidity data have a greater impacton the application of the radiometer to rain nowcastingwe study the humidity profiles from the radiometer in


Figure 4 Comparison of the K-index from the radiosonde and the

radiometer

more detail First of all the wet bias of the radiometerin the mid-troposphere is related to the occasional oc-currence of an elevated moist layer that is not measuredby the radiosonde (an example in Figure 3) This moistlayer appeared more frequently in the latter part of thefield experiment sometimes close to saturation The oc-currence of this moist layer is not supported by otherobservations (eg appearance of cloud layers) and theavailable analysis of numerical weather prediction mod-els (not shown) It points to a need to review the radia-tive transfer model employed by the radiometer for thesummer monsoonal weather in southern China but thisis outside the scope of the present paper Secondly asregard the dry bias in the upper troposphere (Figure 3)the radiometer basically relies on the climatological pro-file at those altitudes which shows a general decreaseof the relative humidity with height In other wordsthe retrieval is more heavily weighted by climatology athigher levels But sometimes the upper troposphere be-comes more humid than normal in the summer-time dueto the presence of cirrus clouds (eg from the upper-level outflow of a rainband) Retrieval of the humidityprofile at these heights may be improved by incorporat-ing the brightness temperature measurements from geo-stationary meteorological satellites (LILJEGREN 2004)

For nowcasting of rain K-index (KI) is calculatedfrom the temperature and humidity profiles of the ra-diometer

KI = (T850 minus T500) + Td850 minus (T700 minus Td700) (31)

where T is the temperature Td is the dew point andsubscripts are the pressure levels (in hPa) Radiometerdata are given in heights and the conversions to pressurelevels are made by assuming hydrostatic approximationK-indices from the radiometer and the radiosonde are

Figure 5 The bias and standard deviation of the virtual temperature

from the radiometer with respect to the concurrent data from a wind

profilerRASS The sample size at each height is about 800

compared in Figure 4 They are found to correlate wellwith a correlation coefficient of 0856 The radiometertends to give a slightly larger K-index due to the wetbias near 700 hPa which is about 3 km above MSL forsouthern China Nonetheless in view of all the abovecomparison results with the radiosonde the radiometercan provide reasonably accurate temperature and humid-ity profiles for rain nowcasting

KI is considered in this paper because it is com-monly used for assessing convective instability in south-ern China It is a more challenging application of ra-diometer data because the temperature and humidity val-ues at specific pressure levels need to be measured ac-curately Future study would consider using radiome-ter data to calculate vertically integrated quantities suchas convective available potential energy (CAPE) Forinstance FELTZ and MECIKALSKI (2002) have usedground-based atmospheric emitted radiance interferom-eter (AERI) data to calculate CAPE in the monitoring ofsevere storm environment in the US

32 Comparison with wind profilerrass

During the period when the radiometer was placed atSha Lo Wan (Figure 1) its measurements were used tocalculate virtual temperatures for comparison with thosefrom the co-located wind profilerRASS The compari-son results are shown in Figure 5 The sample size isabout 800 and as such the comparison results are con-sidered to be statistically significant (sample size gt50)However the comparison is limited in period seasonally(ie only spring and early summer) Once again there isa warm bias of 05ndash 1 degree Celsius within the bound-ary layer and the standard deviation ranges between 1and 15 degrees All these are consistent with the com-parison results with the radiosonde (Figure 2a) So it ap-pears that the slight warm bias of the radiometer mea-surement in the boundary layer may not only be relatedto the location difference of the radiometer and the ref-erence instrument (radiosonde or wind profilerRASS)


Figure 6 Comparison of PWV (in mm) from the GPS receivers and

the radiometer The two measurements sometimes differ by 10 mm

or more (eg data points encircled

Figure 7 PWV from the radiometer and the GPS receivers as well as

the 1-minute rainfall measured at the radiometerrsquos location (Cheung

Chau) in the morning of 8 May 2004 A doubtful piece of data from

the radiometer is indicated by a red arrow

but could also be an inherent property of the radiometeritself Further study in this area would be required fromco-located measurements of the radiometer and the ra-diosondeRASS over a much longer period of time

33 Comparison with GPS receivers

The precipitable water vapour (PWV) data from the GPSreceiver closest to the location of the radiometer areused in the comparison namely the GPS receiver atStonecutters Island when the radiometer was placed atthe Hong Kong Observatory headquarters and CheungChau and the one at Siu Lang Shui when the radiometerwas moved to Sha Lo Wan (locations all shown in Fig-ure 1) The comparison results are shown in Figure 6The two measurements are correlated well with a cor-relation coefficient exceeding 09 Relatively large dis-

crepancies are occasionally observed between the twomeasurements For instance for the data-points encir-cled in Figure 6 the PWV from the radiometer is largerthan that from the nearest GPS receiver by 10 mm ormore An example of such a case is shown in Figure7 The radiometer was placed at Cheung Chau at thattime and it was co-located with a tipping-bucket raingauge with a resolution of 05 mm It appears from thiscase that when the rain is heavy (at a maximum rateof 05 mmmin = 30 mmhour as also indicated in therainfall estimation from the weather radar [not shown])but not continuous the PWV values from the radiome-ter and the GPS receivers are generally consistent witheach other (eg before 6 am and after 7 am in HongKong time which is 8 hours ahead of UTC) But whenthe rain is heavy and more or less continuous (between6 and 7 am) the PWV from the radiometer exhibitsrather rapid fluctuations that are not observed in the GPSdata The two measurements could differ by more than10 mm at times (eg at 631 am indicated by a red ar-row) To ascertain that this difference is not due to localvariation of rainfall rate we have examined the rainfallmeasurements from a network of rain gauges in HongKong and the weather radar imageries The rainfall ratesare found to be about the same at the locations of theradiometer and the two GPS receivers (not shown) Theradiometer data at these moments are considered to bedoubtful and they will not be used for instance to cal-culate the temporally averaged K-index in Section 42

There could be a couple of reasons for the discrep-ancy in PWV between the radiometer and GPS receiverFirstly for the radiometer a warm bias in the bright-ness temperature is expected from scattering and emis-sionabsorption from rain drops Secondly the radiome-ter information is derived along a vertical beam whereasthe GPS value is derived along various slant paths fromGPS satellites and as such the GPS measurement con-siders a much larger volume Concerning the effect ofrain on the radiometer data further study would be con-ducted to find out any correlation between the PWV dif-ference (GPS - radiometer) and rainfall rate at the timeof the radiometer measurement

4 Application to rain nowcasting

Application of the radiometer to rain nowcasting is firstdemonstrated in a case study The correlation betweenthe K-index derived from the radiometer data and thenumberfrequency of lightning flashes is then discussed

41 A case study

A heavy rain episode in the afternoon of 17 April 2004 isused to demonstrate the application of the radiometer inrain nowcasting Synoptically a trough of low pressurelingered near the south China coast on that day (Fig-ure 8) Weather radar imageries (a snapshot in Figure 9)


Figure 8 Surface isobaric chart at 8 am 17 April 2004

Figure 9 Weather radar imagery at 3 km AMSL at 6 pm 17

April 2004 The range rings have a separation of 100 km The Z-

R (reflectivity in dBZ and rainfall rate in mmhour) relation in use

for the radar is Z = 200lowast

R16

depict a rainband that was tracking basically eastwardsacross Hong Kong between 330 pm and 8 pm bring-ing more than 60 mm of rain to many parts of the ter-ritory Lightning flashes inside Hong Kong are found tobe mainly associated with the radar echoes with a reflec-tivity of at least 44 dBZ at 3 km AMSL (equivalent to arainfall rate of about 20 mmhour based on the Marshall-Palmer relation) The part of the radar echoes reachingthis reflectivity has a typical width (along the directionof movement of the rain band) of around 20 km (Fig-ure 9)

The radiometer shows that with the approach of therainband to Hong Kong the relative humidity increasedgradually at all altitudes (Figure 10) It reached a max-imum at around 2 km where synoptically a fresh tostrong southwesterly airstream brought relatively moistair from the South China Sea to the coast of south-ern China The radiometer also gave a gradual decreaseof the cloud-base height which is consistent with hu-

Figure 10 Time-height plots of the radiometer data between 2 pm

(06 UTC) and 9 pm (13 UTC) of 17 April 2004 The upper panel

shows the temperature and the cloud-base height The middle panel

is the relative humidity The lower one is the liquid water content

(The plots are generated by the software rdquoVizmetrdquo bundled with the

radiometer) The time series of 1-minute rainfall measured at the

Hong Kong International Airport (location in Figure 1) is shown

at the bottom The corresponding time series of lightning flashes is

given in Figure 11

man weather observations During the period of heavyrain the air was near saturation below 3 km or so andthe liquid water content had a maximum value of about49 gm3 Please note that the liquid water content val-ues may be biased due to rain contamination on the ra-diometer measurement As the rainband moved awayfrom Hong Kong the relative humidity decreased andthe cloud base lifted gradually after 11 UTC

The gradual rise of relative humidity at all altitudesas measured by the radiometer is a good indication ofthe imminence of heavy rain Alternatively it can beexpressed in terms of the K-index calculated from theradiometer data (Figure 11) With the approach of therainband to Hong Kong K-index rose steadily It ex-ceeded 35 when heavy rain started to take place and thelightning flashes became active After about 630 pmK-index began to fall rain eased off gradually and therewere no more lightning activities in the vicinity of theradiometer

42 Correlation between k-index and lightning

flashes

K-index is a commonly used tool for assessing thechance of thunderstorms Since radiosondes are nor-mally available twice a day the K-index calculated from


Figure 11 K-index calculated from the radiometer data and the

number of lightning flashes within 20 km from the radiometer in

the afternoon of 17 April 2004

Figure 12 Correlation coefficient between the temporally averaged

K-index and the lightning activity (total number or frequency of

lightning flashes) as a function of the range from the radiometer

radiosonde measurements can only be used in the as-sessment for the next 12 hours or so in the form of rulesof thumb or flow charts in combination with other me-teorological factors (eg COLLIER and LILLEY 1994)The radiometer provides frequently updated temperatureand humidity profiles of the troposphere for calculat-ing K-index and offers a unique opportunity for a morequantitative study of the correlation between K-indexand the degree of tropospheric instability which is re-lated to the occurrence of thunderstorms

During the field experiment it was observed in manyrain episodes that lightning flashes in the vicinity of theradiometer became more frequent when the K-index cal-culated from the radiometer data exceeded about 35 (asin the case of 17 April 2004 Figure 11) For each rainepisode a temporally averaged K-index for the periodwhen K-index reaches 35 or above is calculated Tropo-spheric instability is expressed in terms of the lightningflashes in the following ways

(a) the total number of lightning flashes within acertain range R from the radiometer during the period

when the K-index calculated from the radiometer is atleast 35

(b) the frequency of lightning flashes which equalsthe total number of lightning flashes defined in (a) abovedivided by the period (in hours) when the K-index fromthe radiometer is at least 35

The range R from the radiometer is varied between10 km and 100 km for a statistical study of the correla-tion between the temporally averaged K-index from theradiometer and the tropospheric instability as defined in(a) and (b) above The latter R limit is based on the max-imum detection range of the lightning location networkThe former one is chosen so that there is a reasonablylarge sample of cases (at least 10) for a statistical studyAs R is reduced further the number of rain episodeswith lightning activity within R from the radiometer de-creases dramatically

This paper focuses on the lightning activity associatedwith rainbands in the summer monsoon over southernChina There are altogether 11 cases Rain episodes in-volving isolated thunderstorms (for instance triggeredby prolonged periods of solar heating over inland areasof southern China) are not considered

It is found that the correlation reaches a maximum atboth R = 10 km and 20 km for lightning flash numberwhereas it is the largest at R = 20 km for lightning flashfrequency (Figure 12) The R value of 20 km is consis-tent with the radar observations of the spatial scale ofthe rainbands with active lightning (eg Figure 9) Forboth lightning flash number and frequency the correla-tion with the temporally averaged K-index decreases asR exceeds 20 km It starts to level at R = 60 km as thelightning flash dataset for each rain episode is more orless exhausted

The scatter plots of the temporally averaged K-indexagainst the logarithm of the lightning flash number andfrequency within 20 km from the radiometer are shownin Figure 13 The correlation is relatively high with acorrelation coefficient of about 08 For the lightningflash number the slope of the least-square linear fitis 1183 That means when the temporally averagedK-index is increased by 1 the number of lightningflashes within 20 km from the radiometer will on av-erage increase by a factor of 1011183 = 7 Similarlyfor the frequency of lightning flashes the corresponding

increase will be a factor of 1011276 = 6 Thus the light-ning activity appears to be rather sensitive to the valueof K-index The use of logarithmic scale for lightningactivity follows the results in other studies (eg DEIER-LING and PETERSEN 2008)

As shown in Figure 13 the heavy rain episode on 8May 2004 has the largest values of temporally averagedK-index and the total number of lightning flashes aswell as the second highest frequency of lightning flashesamong all the rain episodes considered It is in fact theonly episode in the rain season of Hong Kong in 2004that necessitated the issuance of the rdquoblack rainstorm


warningrdquo by the Hong Kong Observatory which meansvery heavy rain has fallen generally over Hong Kongexceeding 70 mm in an hour A meso-cyclone was ob-served to pass over the radiometerrsquos location (CheungChau) as analyzed in LI et al (2004)

It should be noted that the above analysis is based on arather limited dataset (a total of 11 rain episodes) in thesummer monsoonal season of a year only (2004) Therobustness of the results is yet to be studied with a largerdataset It has been shown in some studies (eg LANG

and RUTLEDGE 2002) that lightning rate is related toupdraft intensity of thunderstorms The latter may be re-lated to convective indices such as CAPE and KI Thismay explain the correlation between lightning rate andKI as discussed above

5 Nowcasting of squall in intenseconvective weather

The radiometer data are combined with wind measure-ments from the radar wind profilers in the nowcasting ofsqualls associated with intense convective weather Thissection describes how the wind gust estimate (GUS-TEX) is calculated and its application to some thunder-storm cases in Hong Kong

51 Calculation of wind gust estimate

GUSTEX is based on WINDEX developed in MCCANN

(1994) WINDEX is modified in GEERTS (2001) andLEE (2007) and given by

WI = 5[HmRq(L2

max minus 30 + ql minus 2qm)]05 (51)

where WI is the WINDEX in knots Hm is the height ofthe melting layer in km Lmax is the maximum lapse ratefrom the ground to the freezing layer in Kkm ql is themean mixing ratio of water vapour between the groundand 1 km above in gkg qm is the mixing ratio of watervapour at the melting layer in gkg and Rq = ql12

GUSTEX also considers the downward transport ofhorizontal momentum in the upper air It is modified inLEE (2007) and given by

GU = α1WI + α2ρUmax (52)

where GU is GUSTEX in knots Umax is the maximumwind speed between 900 and 500 hPa levels in knotsρ = ρmaxρ990 in which ρmax is the air density at thepressure level at which Umax occurs and ρ990 is the airdensity at 990 hPa level α1 and α2 are constants to bedetermined empirically

In LEE (2007) the empirical constants in equation(52) were determined from the high wind gust casesat the Hong Kong International Airport between 2001and 2005 The upper-air temperature humidity andwind data were taken from the twice-daily radiosonde

measurements Besides determining the empirical con-stants from the least-square linear fit between GUSTEXand the actually measured wind gust from the surfaceanemometer stations the upper bound GUmax and thelower bound GUmin of the wind gust estimate are alsocalculated The relevant equations are summarized be-low

GU = 012WI + 093ρUmax (53)

GUmax = 030WI + 093ρUmax (54)

GUmin = minus005WI + 093ρUmax (55)

In equation (51) if L2max is less than 30 WI is taken as

0 and from equations (53) to (55) GU = GUmax =GUmin in this case

In the present study the zenith mode measurementsby the microwave radiometer are used to retrieve thetemperature and humidity profiles up to 10 km aboveground These profiles are then employed to determinethe thermodynamic parameters in equations (53) to(55) such as the height of the melting layer watervapour mixing ratio lapse rate etc The only compli-cation in this calculation is that the radiometer data areonly available at fixed heights above ground instead offixed pressure levels The height and the pressure arerelated using hydrostatic approximation The maximumwind speed and the corresponding height above groundare obtained by considering the wind profiles from allthe four radar wind profilers in Hong Kong (locations inFigure 1) At a particular moment the maximum windamong the four profilers is selected in the calculation ofGUSTEX as well as its upper and lower bounds Pleasenote the wind data at a particular height from radar windprofilers are obtained from a measuring volume aboutthat height and averaged over a period (10 minutes inthe configuration of the wind profilers in Hong Kong)On the other hand wind data from the radiosonde areobtained in a much smaller sampling volume and nearlyinstantaneous in time

52 Comparison between wind gust estimate

and the actual gust measurements

The performance of the wind gust estimate determinedfrom the above method (viz radiometer and wind pro-filer data) is compared with the actually measured gustby the dense network of surface anemometers in HongKong in 14 thunderstorm episodes between March andJune 2004 A thunderstorm episode is defined as a periodof time (normally a few hours or more) in which thun-derstorm has been reported by the weather observers atHKO Headquarters andor Hong Kong International Air-port In general the GUSTEX itself (eq (53)) could un-derestimate the actual gust measurement at times For


(a) (b)

Figure 13 Scatter plots of the temporally averaged K-index against the number (a) and the frequency (b) of lightning flashes within 20 km

from the radiometer Dates of the rain episodes are marked nearby the data point (in the form of daymonth)

(a) (b)

Figure 14 (a) The scatter plot of the maximum gust estimated from microwave radiometer and radar wind profiler data against the largest

value of the actually measured gust from the surface anemometer stations in the 14 thunderstorm episodes in 2004 of the current study (b)

is the corresponding scatter plot based on the maximum gust estimated from the radiosonde data In (a) and (b) the outliers are encircled

practical applications the upper bound of wind gust es-timate (equation (54)) may be more useful in givingthe weather forecaster an idea about how large the gustcould attain As such GUmax is considered in the com-parison

The scatter plot of the maximum wind gust actuallymeasured by the surface anemometers and the largestvalue of GUmax within an hour prior to the maximumgust occurrence for the thunderstorm episodes understudy is given in Figure 14(a) The two datasets arefound to have good correlation and the latter (GUmax)is only larger than the former (actual gust) by about20 on average from the least-square linear fit Thereare two outlier points in Figure 14(a) (encircled in redin the scatter plot) In both cases the upper-air windsfrom the wind profilers are too strong resulting in over-estimation of the wind gusts

For comparison of performance the correspondingscatter plot based on the GUmax estimated from ra-diosonde measurements is shown in Figure 14(b) In

this plot the upper bound of GUSTEX determined fromthe radiosonde data just prior to the maximum gust oc-currence is used which could be as many as 12 hoursearlier (because there are only two radiosonde ascentsevery day in Hong Kong with both wind and tem-peraturehumidity profiles) Compared to the radiome-terwind profiler plot the radiosonde plot has weakercorrelation and on average the wind gust estimate is63 larger on average There are also two outliers inFigure 14(b) (encircled in red in the scatter plot) In bothcases the upper-air winds measured by the radiosondeare too strong resulting in over-estimation of the windgusts Overall speaking the GUmax estimated from ra-diometerwind profiler data has better performance thanthat based on radiosonde data and could be useful in thenowcasting of wind gust associated with intense convec-tive weather with a lead time of about one hour


(a) (b)

(c)

Figure 15 (a) and (b) are the 256-km radarscope images of Hong Kong at 3 km above mean sea level at 342 and 706 am 8 May 2004

respectively The range rings have a separation of 100 km (c) is the time series of the largest value of the actually measured a wind gust

from the surface anemometers (green) GUmin (blue) GU (red) and GUmax (brown)

53 Case studies

Two convective weather cases within the study period in2004 are discussed in detail in this Section to illustratehow to estimate wind gust based on radiometer data Thefirst case occurred on 8 May 2004 Synoptically a strongsoutherly airflow between a westerly wave to the westof Hong Kong and the subtropical ridge to the east af-fected the Pearl River Estuary (location in Figure 15(a))within the boundary layer At 200 hPa divergent flowoccurred over the south China coast (location in Figure15(a)) ahead of a deep westerly wave From the 256-km radarscope bands of strong echoes with southwest-to-northeast orientation moved across the Pearl RiverEstuary to affect Hong Kong Thunderstorms were re-ported over the territory between 2 and 8 am on thatday The first band of heavy rain affected Cheung Chauat about 342 am (Figure 15(a)) bringing wind gustof 189 ms the largest over Hong Kong The secondband affected Waglan Island at about 704 am (Figure15(b)) bringing the maximum wind gust of 262 ms

The time series of maximum wind gust over the surfaceweather stations in Hong Kong in that morning as wellas the GUSTEX GUmax and GUmin are given in Figure15(c) For the two events of the passage of strong radarechoes the maximum wind gusts lie between the largestvalues of GUSTEX and GUmax within an hour prior toeach event As such the wind gust estimates based onradiometer and wind profilers are considered to be use-ful in nowcasting the actual strength of the gusts

For the two strong gust events (sim342 am and 704am 8 May 2004) the winds picked up gradually fora couple of hours before reaching the maximum val-ues The gradual strengthening of the winds was relatedto the approach of the rain bands As the bands edgedcloser to Hong Kong the southerly winds ahead of thebands strengthened gradually which caused the windsrecorded over the territory to become stronger

It is noted that after the occurrence of squalls at thesurface the wind gust estimates may remain at ratherhigh values such as between 4 and 430 am on 8 May2004 This is because of the appearance of strong winds


(a) (b)

(c)

Figure 16 (a) and (b) are the 256-km radarscope images of Hong Kong at 3 km above mean sea level at 548 and 624 pm 17 April 2004

respectively The range rings have a separation of 100 km (c) is the time series of the largest value of the actually measured wind gust from

the surface anemometers (green) GUmin (blue) GU (red) and GUmax (brown)

at the upper air after the passage of the squall nearthe surface The strong winds may be related to mid-level inflow jet at the rear of the quasi linear convec-tive system or simply westerly jet streams in the mid-dle troposphere which are commonly observed in heavyrain events over southern China Their origin requiresmore detailed study eg mesoscale analysis of the windflow using a local analysis and prediction system andmicroscale analysis of the airflow associated with theconvective storm using multiple Doppler weather radardata

The second case occurred on 17 April 2004 whichhas been discussed in Section 41 On the 256-kmradarscope a band of intense echoes with north-northeastto south-southwest orientation moved across Hong Kongin the afternoon on that day Thunderstorms were re-ported in Hong Kong at 5 and 6 pm At the times shownin Figures 16(a) and (b) high wind gusts were recordedat Cheung Chau (151 ms) and Waglan Island (17 ms)respectively The time series of the highest wind gustin Hong Kong together with the various wind gust es-

timates are given in Figure 16(c) The largest values ofGUSTEX and GUmax within an hour before the highwind gust events provide reasonable estimates on thegusts that could be attained As in the previous case af-ter the occurrence of squalls at the surface the wind gustestimates may remain at rather high values such as af-ter 630 pm on 17 April 2004 This is again because ofthe appearance of strong winds at the upper air after thepassage of the squall near the surface

6 Conclusion

Performance and application of a multi-channel ground-based microwave radiometer in intense convective wea-ther is studied in this paper The radiometer wasequipped with a hydrophobic radome surface and ablower of heated air to mitigate the adverse effect ofa rain-coated radome on the measurement accuracy Itwas found to give reasonably accurate temperature andhumidity profiles as well as precipitable water vapour


(PWV) with reference to the measurements from theradiosonde a wind profilerRASS and GPS receiversHowever some discrepancies between the data from theradiometer and the reference instruments were observed

(a) Temperatures from the radiometer showed awarm bias within the boundary layer

(b) Relative humidity profiles from the radiometertended to show an elevated moist layer at 3 to 4 km Thisresulted in a generally larger K-index as compared withthe value calculated from the radiosonde measurement

(c) Under continuous rain with a rate of 30 mmhouror more PWV from the radiometer exhibited rapid fluc-tuations that were not observed in GPS measurementsand the radiometer data appeared to be doubtful

Despite these discrepancies the radiometer wasfound to provide useful information for rain nowcastingeg increase in relative humidity at all altitudes and theaccumulation of water vapour in the lower troposphereK-index was also calculated from the frequently updatedtemperature and humidity profiles from the radiometerfor nowcasting purpose Heavy rain with active light-ning mostly occurred when K-index stayed above 35 orso

A temporally averaged K-index for the period whenK-index was 35 or above was determined from the ra-diometer data For the summer monsoonal rain in south-ern China it was found to correlate well with the totalnumber and the frequency of lightning flashes within 20km from the radiometer (which was consistent with theradar observations of the spatial scale of the rainbandswith active lightning) The lightning activity turned outto be rather sensitive to the K-index value when K-indexrose by 1 the number and the frequency of lightningflashes would increase by a factor of 7 and 6 respec-tively Continuous availability of the temperature andhumidity profiles of the troposphere from the radiometerenabled a quantitative study of the correlation betweenK-index and the atmospheric instability (expressed interms of the lightning activity) which could not beachieved by using the twice-daily radiosonde measure-ments alone

Application of radiometer and radar wind profilerto the nowcasting of surface high gusts associatedwith intense convective weather was also studied TheGUSTEX originally developed for radiosonde data wasadopted Compared to the twice-daily balloon ascentsthe continuous measurements of temperature humid-ity and wind from the remote-sensing instruments bet-ter captured the rapid evolution of the troposphere inconvective weather conditions As such the wind gustestimates so determined improved the correlation withthe maximum gusts actually measured by the networkof surface anemometer stations in Hong Kong From twocase studies of typical intense convection in the summerthe gust estimates were found to be useful in the now-casting of high winds associated with thunderstorms

The present study is based on the field experimentdata of a ground-based microwave radiometer in HongKong in 2004 only The amount of data involved is ratherlimited A permanent unit of radiometer has recentlybeen set up in Hong Kong and a larger set of dataaccumulated over the years would then be employed toevaluate the application of the microwave radiometer foroperational use

Acknowledgments

The author would like to thank the Lands Department ofthe Government of the Hong Kong Special Administra-tive Region for providing the GPS receiver data in thisstudy The PWV data derived from GPS receivers areproduced in a joint research project between the HongKong Observatory and the Hong Kong Polytechnic Uni-versity Thanks also go to China Light and Power Lim-ited which provides the lightning flash data from thelightning location network in Hong Kong The com-ments from two anonymous reviewers are very muchappreciated for improving the manuscript

References

CHAN PW KK YEUNG 2003 Experimental extension ofthe measurement range of a boundary layer wind profiler toabout 9 km ndash Preprints 12th Symposium on Meteorologi-cal Observation and Instrumentation Amer Meteor Soc

COLLIER C G RBE LILLEY 1994 Forecasting thun-derstorm initiation in north-west Europe using thermody-namic indices satellite and radar data ndash Meteor Appl 175ndash84

DEIERLING W WA PETERSEN 2008 Total lightning ac-tivity as an indicator of updraft characteristics ndash J Geo-phys Res 113 D16210 doi1010292007JD009598

FELTZ WF JR MECIKALSKI 2002 Monitoring high-temporal-resolution convective stability indices using theground-based atmospheric emitted radiance interferometer(AERI) during the 3 May 1999 Oklahoma-Kansas tornadooutbreak ndash Wea Forecast 17 445ndash455

GEERTS B 2001 Estimating downburst-related maximumsurface wind speeds by means of proximity soundings inNew South Wales Australia ndash Wea Forecast 16 261ndash269

HEWISON T C GAFFARD J NASH D RUFFIEUX RNATER H BERGER M PERROUD B ANDRADE 2004Monitoring inversions from ground-based remote sensinginstruments during Temperature Humidity and Cloud pro-filing campaign (TUC) Preprints ndash 8th Specialist Meetingon Microwave Radiometry and Remote Sensing Applica-tions

KNUPP K R WARE D CIMINI F VANDENBERGHE JVIVEKANANDAN E WESTWATER T COLEMAN sub-mitted Ground-Based Passive Microwave Profiling duringDynamic Weather Conditions ndash J Atmos Oceanic Tech-nol

LANG T J SA RUTLEDGE 2002 Relationships betweenconvective storm kinematics precipitation and lightning ndashMon Wea Rev 130 2492ndash2506


LEE OSM 2007 Forecast of strong gusts associated withthunderstorms based on data from radiosonde ascents andautomatic weather stations ndash 21st Guangdong - HongKong - Macao Technical Seminar on Meteorological Sci-ence and Technology Hong Kong China 24ndash26 January2007 (paper in Chinese with English abstract)

LI LO CC LAM WK WONG 2004 The meso-cycloneover Hong Kong on 8 May 2004 Doppler radar observa-tion and numerical simulation Preprints ndash InternationalConference on Mesoscale Convective Systems and HeavyRainfall in East Asia (ICMCS-IV) Chinese Academy ofMeteorological Sciences and China Meteorological Ad-ministration

LILJEGREN J 2004 Improved retrievals of temperatureand water vapor profiles with a twelve-channel radiometerPreprints ndash 8th Symposium on Integrated Observing and

Assimilation Systems for Atmosphere Oceans and LandSurface (IOAS-AOLS) Amer Meteor Soc

MCCANN DW 1994 WINDEX ndash A new index for fore-casting microburst potential ndash Wea Forecast 9 532ndash541

MURPHY MJ KL CUMMINS 2000 Early detection andwarning of cloud-to-ground lightning at a point of inter-est Preprints ndash 2nd Symposium on Environmental Appli-cations Amer Meteor Soc

RADIOMETER PHYSICS GmbH 2008 Operating man-ual of RPG-HATPRO ndash available at wwwradiometer-physicsderpghtmlDownload Radiometerhtml

ROSE T H CZEKALA 2003 Filter bank radiometers for at-mospheric profiling Preprints ndash Sixth International Sym-posium on Tropospheric Profiling Needs and Technolo-gies 261ndash263


and LEE 2007) However radiosonde measurementsare normally only available twice a day and duringrain episodes thermodynamic properties of the tropo-sphere could change rapidly in between measurementsespecially in subtropical coastal areas like Hong KongMore frequently available thermodynamic informationtogether with wind data up to the upper troposphere (egfrom radar wind profilers CHAN and YEUNG 2003)would greatly benefit severe weather nowcasting

Ground-based microwave radiometers (ROSE andCZEKALA 2003 KNUPP et al submitted) retrieve thetemperature and humidity profiles up to 10 km by mea-suring the radiation intensity at a number of frequencychannels in the microwave spectrum that are dominatedby atmospheric water vapor cloud liquid water andmolecular oxygen emissions These profiles are avail-able nearly continuously at intervals of several min-utes They are mainly used in non-precipitating condi-tions because the radiometer measurements become lessaccurate in the presence of a water film on the outerhousing (radome) of the equipment Moreover the ra-diometer retrieval normally does not include the scat-tering and emissionabsorption effects of rain Recentlythere are some progress in applying rain-effect mitiga-tion methods to the radiometer such as a hydrophobicradome and forced airflow over the radome surface As aresult reasonably accurate thermodynamic profiles canbe obtained at a rainfall rate up to 10 mmhour (egRADIOMETER PHYSICS 2008) However this rainfallrate is still much lower than that observed in typicalheavy rain episodes in the subtropical areas which nor-mally far exceeds 30 mmhour

This paper presents the performance and applica-tion of a multi-wavelength ground-based microwave ra-diometer in intense convective weather during a field ex-periment conducted in Hong Kong in 2004 The equip-ment setup of the experiment is described in Section2 The accuracy of the temperature and humidity pro-files as well as precipitable water vapour (PWV) fromthe radiometer is first assured by comparing with mea-surements from the radiosonde a wind profilerRadioAcoustic Sounding System (RASS) and Global Posi-tioning System (GPS) receivers in Section 3 The per-formance of the radiometer data in a rain episode is de-scribed in Section 4 which suggests that there is rea-sonable degree of correlation between the instability in-dex of the troposphere as calculated from the radiome-ter data and the occurrence of lightning flashes Thisrelationship is studied further in Section 4 Section 5discusses the application of the radiometer data in thenowcasting of squall associated with intense convectiveweather Conclusion of this paper is given in Section 6

2 Equipment setup

The radiometer in this study (similar to that in KNUPP

et al submitted) measures the radiation intensity at

Figure 1 The radiometer has been placed at Hong Kong Observa-

tory (HKO) Headquarters Cheung Chau and Sha Lo Wan for the

periods in parenthesis during the experiment The map shows the

network of surface anemometers (red dots) and radar wind profilers

(green triangles) in Hong Kong The radiosonde station (orange) and

the two GPS receivers mentioned in this paper (blue) are also shown

Height contours are in 100 m

7 oxygen channels (51ndash59 GHz) and 5 water-vapourchannels (22ndash30 GHz) to obtain the temperature andhumidity profiles It is equipped with a hydrophobicradome surface and a blower of heated air over theradome to minimize the effect of rain on radiometermeasurement During the field experiment the radiome-ter was configured to make both zenith scans and eleva-tion scans at fixed angles Following other similar stud-ies (eg HEWISON et al 2004 and KNUPP et al sub-mitted) only the zenith scan data are used in this pa-per to retrieve the thermodynamic profiles of the tropo-sphere up to 10 km which are normally available every5 minutes

Retrieval of temperature and humidity profiles fromthe radiation intensity measurements of the radiometer isaccomplished by neural network methods based on his-torical radiosonde data using a radiative transfer modelto simulate the observations of a radiometer (KNUPP etal (submitted) In this study the neural network wastrained by a high-resolution radiosonde dataset collectedat Kingrsquos Park (location in Figure 1) with a data avail-ability rate of 05 Hz during the ascent of the balloon(which translates to a spatial resolution of around 10 mfor the typical ascent speed) Radiosonde data of a wholeyear (2003) were used in the neural network

The field experiment was conducted in Hong Kongfrom February to June 2004 The radiometer was placedat three locations and its data are compared with thosefrom the radiosonde a wind profilerRASS and GPS re-ceivers (Figure 1) In studying the thermodynamic in-dex given by the radiometer we also refer to the datafrom a lightning location network (MURPHY and CUM-MINS 2000) which detects cloud-to-ground flashes outto about 100 km from Hong Kong


(a) (b)


















radiometer



























41 A case study








R16











given in Figure 11




flashes





























WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References



















(a) (b)


















radiometer



























41 A case study








R16











given in Figure 11




flashes





























WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References




















radiometer



























41 A case study








R16











given in Figure 11




flashes





























WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References

































41 A case study








R16











given in Figure 11




flashes





























WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References
























R16











given in Figure 11




flashes





























WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References













































WI = 5[HmRq(L2








GU = 012WI + 093ρUmax (53)










(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References



















(a) (b)



(a) (b)









(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References



















(a) (b)

(c)




53 Case studies






(a) (b)

(c)







6 Conclusion











Acknowledgments


References



















(a) (b)

(c)







6 Conclusion











Acknowledgments


References



























Acknowledgments


References


















performance and application of a multi-wavelength, ground ...performance and application of a...

Documents